University of Groningen

Optimizing treatment with psychotropic agents through precision drug therapyBerm, Elizabeth Jacoba Johanna

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite fromit. Please check the document version below.

Document VersionPublisher's PDF, also known as Version of record

Publication date:2016

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):Berm, E. J. J. (2016). Optimizing treatment with psychotropic agents through precision drug therapy: It isnot about the mean. Rijksuniversiteit Groningen.

CopyrightOther than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of theauthor(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons thenumber of authors shown on this cover page is limited to 10 maximum.

Download date: 25-02-2021

Optimizing treatment with psychotropic agents

through precision drug therapy –

It is not about the mean

Elizabeth Jacoba Johanna Berm

The studies presented in this thesis were supported by the ZonMW program; ‘ Priority Medicines Elderly.

Printing of this thesis was partially supported by the University of Groningen, the In-stitute for Health Research SHARE, the Graduated School of Science, and “ de Stichting KNMP fondsen”.

Layout and printing: Optima Grafische Communicatie, Rotterdam, The Netherlands

ISBN: 978-90-367-8860-1 (printed version)ISBN:  978-90-367-8859-5 (electronic version)

Copyright, 2016, EJJ Berm

No parts of this this publication may be reproduced or transmitted in any form or by any means, electronically or mechanically by photocopying, recording or otherwise without the written permission of the author or from the publisher holding the copyright of the published articles.

something something promotion something goes here

Optimizing treatment with psychotropic agents through

precision drug therapy

It is not about the mean

Proefschrift

ter verkrijging van de graad van doctor aan de Rijksuniversiteit Groningen

op gezag van de rector magnificus prof. dr. E. Sterken

en volgens besluit van het College voor Promoties.

De openbare verdediging zal plaatsvinden op

vrijdag 17 juni 2016 om 12.45 uur

door

Elizabeth Jacoba Johanna Berm

geboren op 19 augustus 1986 te Nieuwegein

names of people who decide whether this is a good book go herePromotores Prof. dr. B. Wilffert Prof. dr. E. Hak Copromotor Dr. J.G. Maring Beoordelingscommissie Prof. dr. N.C. van de Merbel Prof. dr. H.J. Guchelaar Prof. dr. E. Buskens

Paranimfen: Geraldina Wilhelmina Maria (Lisette) Holtkuile Eva van Doorn

COntEnts

Chapter 1 General introduction 9

Chapter 2 Environmental and genetic variation can influence drug plasma concentrations

21

Chapter 2.1 Switching from smoking to the use of an e-cigarette can increase drug exposure

23

Chapter 2.2 Relation between CYP2D6 genotype, phenotype and therapeutic drug concentrations among nortriptyline and venlafaxine users in old age psychiatry

31

Chapter 3 Validation of dried blood spot sampling for therapeutic drug monitoring of tricyclic antidepressants and venlafaxine

45

Chapter 3.1 A simple dried blood spot method for therapeutic drug monitoring of the tricyclic antidepressants amitriptyline, nortriptyline, imipramine, clomipramine, and their active metabolites using LC-MS/MS

47

Chapter 3.2 Determination of venlafaxine and O-desmethylvenlafaxine in dried blood spots for TDM purposes, using LC-MS/MS

67

Chapter 3.3 A clinical validation study for application of dried blood spots in therapeutic drug monitoring of antidepressants

79

Chapter 4 Cost-effectiveness of precision drug therapy 101Chapter 4.1 Economic evaluations of pharmacogenetic and

pharmacogenomic screening tests: a systematic review. Second update of the literature

103

Chapter 4.2 A model-based cost-effectiveness analysis of routine genotyping for CYP2D6 among older depressed inpatients starting nortriptyline pharmacotherapy

151

Chapter 4.3 Effects and cost-effectiveness of pharmacogenetic screening for CYP2D6 among older adults starting therapy with nortriptyline or venlafaxine: study protocol for a pragmatic randomized controlled trial (CYSCE trial)

175

Chapter 5 General discussion and future perspective 191

summary 205samenvatting (nederlands) 209CV 215Acknowledgements (Dankwoord) 217sHARE publications 219

1General Introduction

General Introduction 11

PERsOnAlizED mEDiCinE

The application of genetic information for the personalization of treatment for differ-ent diseases has gained much attention during the last decade. Personalized medicine is used as a broad concept in which all information about individual characteristics is used for optimization of treatment. These characteristics include, but are not limited to, health factors, drug interactions, biomarkers, genetic variations, and even epigenetic variations as is shown in figure 1 (1). There is no unique definition of the term personal-ized or precision medicine, however it has often been defined as the situation in which drug class and drug dosing are selected according to an individual’s genetic makeup to make drug safer and more effective (2). Frequently, the term personalized medicine and precision medicine are used interchangeable. In addition, when speaking solely of drug therapy, the term precision drug therapy can be used.

Figure 1. Factors of importance in personalized medicine. Novel aspects concern the genetic and in the future maybe also epigenetic variation. From: Journal of Internal Medicine Volume 277, Issue 2, pages 152-154, 26 JAN 2015 DOI: 10.1111/joim.12325

1.2. mEtHODs APPliED in PRECisiOn DRug tHERAPy

Genetic information can be used to predict if a drug would work well in certain patients (i.e. pharmacodynamics) and into information which could predict the metabolism of a drug in certain patients (i.e. pharmacokinetics). For a long time clinicians relied on experience and on techniques like therapeutic drug monitoring (TDM) to monitor the metabolism of drugs. Moreover, due to already well develop bioanalytical methods to

12 Chapter 1

measure plasma drug concentrations, differentiation between patients metabolism related to genetics could already been made without prior knowledge of the underlying genetic variation. This could be referred to as phenotyping. To phenotype a patient, a certain compound with no major pharmacological effect, but a well know metabolism and easy to measure metabolites pattern is given. The ratio of the compound and its metabolites can then be used to predict the metabolic capacity of a certain drug metabolizing enzyme. Compounds which are used for phenotyping can be considered as reference compounds. For example, with the use of caffeine one could test the cy-tochrome P450 CYP1A2 function (3, 4). With the use of phenotyping a differentiation between patients with a slow metabolism, or ultra-rapid metabolism can be made. As such, by solely measuring drug plasma concentrations, predictions about the genetic make-up can be made. As a result of ongoing development and innovation in tech-nique, different phenotypes can nowadays also be predicted by measuring the genetic variations itself instead of measuring the effects of these variation by phenotyping. This genetic information can be determined by collection of blood, saliva, or tissue biopsies (5). Based on this genetic information, a genotype can be determined which can then predict a phenotype. Nowadays, the terms genotype and the predicted phenotype and phenotype (i.e. as determined by a reference compound) are often used interchange-able which can complicate reporting and interpretation of studies about precision drug therapy.

1.3. PHARmACOgEnEtiCs OF PsyCHOtROPiC DRugs

Cytochrome P450 (CYP) are drug metabolizing enzymes and the CYP2D6 and CYP2C19 isoforms are responsible for the metabolism of many psychotropic drugs. There are many different genetic variations of these enzymes between persons and these variation can have a major effect on the clearance of drugs (6, 7). Well-established relationships are the ones between CYP2D6 genotypes and plasma concentrations of the antidepres-sant nortriptyline and venlafaxine (8-11). Depending on the genetic variation in these enzymes, a 4-fold decrease in clearance of nortriptyline up to a 5-fold increase can be observed with respect to the average clearance (10). To compensate for such differences in metabolisms, genotype-based dose adaptations can be made. For drugs like nortrip-tyline and venlafaxine, guidelines for such dose adaptations are available (12). In the Netherlands, pharmacogenetic guidelines are provided by the Royal Dutch Pharmacy Association (KNMP). In addition, international guidelines of the Clinical Pharmacogenet-ics Implementation Consortium (CPIC) are available. The CPIC has currently published >30 guidelines and aims to publish more in the future (13).

General Introduction 13

1.4. nOVEl mEtHODs in tHERAPEutiC DRug mOnitORing

TDM has been used for individualization of pharmacotherapy and its applications have grown over the past 25 years (2). In TDM, drug concentrations are usually measured in blood plasma of the patient and based on reference concentrations a therapeutic range is established (14). When a patients has a drug concentration outside this range, dose adaptations can be made to improve pharmacotherapy. Besides plasma, different body fluids can be used like feces, urine, blood, blood plasma, and even saliva or sweat. The blood plasma is usually collected by venous blood sampling, however novel techniques for less invasive sampling are emerging. One of these techniques is Dried Blood Spot sampling (DBS) (15). DBS samples consist of whole blood which can be collected by fingerprick. With the use of DBS sampling, research and application of precision drug therapy in heterogeneous patient populations and under naturalistic conditions can be achieved. Since DBS samples contain no biohazard risk, the samples can be sent to a laboratory site by regular mail services, reducing logistic costs, patients discomfort and efforts in research and clinical practice (16, 17). Prospective studies about the effects of pharmacogenetic (PGx) testing, often require large numbers of patients to include sufficient genetic variations. This is due to the fact that many polymorphisms which have relevant clinical implications occur only in a small part of the patient population. To obtain a sufficient study size, different study locations can be used, however this will include heterogeneity in the study population. Hence, DBS sampling can be beneficial to centralize TDM analysis in such studies. In addition, digitalization of data collection (e-CRF) with rigid data flow can help to make these studies more accurate and feasible.

1.5. EViDEnCE BAsED PRECisiOn DRug tHERAPy

For drugs that enter the market, thorough assessment of quality, safety, and efficacy takes place. In addition, the cost-effectiveness of drugs is usually assessed in order to achieve reimbursement of drug costs. For diagnostic tools like TDM and PGx testing there are no strict regulations. For some new drugs, like trastuzumab, a drug to treat breast cancer, genetic testing is necessary according to the registration file of the drug (18). For this PGx test reimbursement seems no point of discussion. However, for PGx tests in combination with a drug which has been on the market for decades, evidence about cost-effectiveness will usually not be available (19). For these combinations, PGx tests are usually reimbursed, when according to the treating physician, there is a medical need for this diagnostic test. If critically assessed, such a medical need should be based on recommendations presented in guidelines supported by data from ran-domized controlled trials, observational studies, or expert opinions. Evidence from

14 Chapter 1

several observational studies that assessed associations between CYP2D6 genotype and pharmacotherapy with CYP2D6 metabolized psychotropic drugs, found poor and intermediate metabolizers experience more side-effects (20, 21). Others found a lower efficacy of CYP2D6 metabolized antidepressants among CYP2D6 ultra-rapid metaboliz-ers, but only if co-medication, which can also effect the CYP2D6 function, was taken into account (22). Although it might be reasonable to expect these problems could be prevented by preemptive genotyping, there is not yet any empirical evidence to support this. To create such evidence, large trials are needed (21). To complicate things further, certain co-medication of patients which influence the CYP2D6 function should be care-fully monitored, since CYP2D6 inhibitors can have a significant effect on the CYP2D6 capacity (22). This can lead to a discrepancy between a patient’s genotype and observed phenotype, which is referred to as phenoconversion (23). Nevertheless, demonstrating beneficial effects and associated acceptable cost-effectiveness of preemptive genotyp-ing by use of a randzomized controlled trial (RCT), would likely result in uptake of the PGx test in guidelines. This may increase application in clinical practice and therefore fulfill the expected promises of benefit for patients by precision drug therapy. However, the conduct of a large RCT is expensive and since the drugs are usually already out of patent, funding for these studies is difficult to obtain. Since older patient populations (i.e. 55 years and older) experience a higher burden from adverse drug reactions of antidepressants and non-response towards antidepressants is equal or even higher, this population could serve as a high risk study population to detect possible effects of genotyping (24, 25). Alternatives for a pragmatic RCT can in theory be a modeling approach in which the best available evidence is used to simulate effects and cost-effectiveness of PGx testing (2).

1.6. COst-EFFECtiVEnEss mODEling

Models are well established research tools in the field of pharmacokinetics and dynamics as well as in the field of clinical outcome studies which often include a cost-effectiveness analysis (26). In cost-effectiveness studies, models can be used to either synthesize evi-dence which is not available through prospective studies yet, or to extrapolate results from prospective studies to other countries or to a longer time period. To analyze health and economic outcomes, different model structures are available. The simplest model structure is a decision tree in which patients follow a certain branch and end up in a health state depending on different probabilities to enter a branch of the decision tree (27). Other more sophisticated models are Markov models and discrete event simulation models (28). Depending on the time and complexity of a medical question, the most appropriate model can be selected. Genetic associations with impaired drug treatment

General Introduction 15

outcomes are commonly observed in a minority of the population (29). To extrapolate such findings into cost-effect estimates for larger populations at national or regional level, modeling can be very informative.

1.7. tARgEts OF PRECisiOn AntiDEPREssAnt tHERAPy

Half of the patients does not respond to an antidepressant and around a quarter does discontinue due to side-effects (30-33). To manage non-response and side-effects, a first step is usually TDM directed dose adaptation (34). The next step is switching to another antidepressant type or class (35). Precision drug therapy may increase efficacy of both dose adaptations as well as switching to another antidepressant (36). In addition, a DBS method for TDM of nortriptyline and venlafaxine would be a welcome patient friendly alternative for the current venous sampling. In time, sampling could even be performed at home which would be beneficial for the depressed ambulant patient. There is a need for research which demonstrates cost-effectiveness of pharmacogenetic testing. To obtain reliable effect estimates, a (pragmatic) RCT would be ideal, however modelled estimates could serve as a prior indication (37).

1.8. Aim AnD OutlinE OF tHis tHEsis

This thesis aims to develop DBS analysis of antidepressants and as such facilitate preci-sion drug therapy. In addition, effects of genetic tests on costs and health outcomes will be assessed with a specific focus on CYP2D6 in combination with nortriptyline use among older patients.

Chapter 2.1 starts with an example of environmental influence on drug clearance by the CYP1A2 enzyme. Although the interaction of smoking on CYP1A2 dependent drug clearance has been known for a long time, we wanted to create awareness of changes caused by e-smoking which is increasingly used as an alternative for smoking. In Chap-ter 2.2. the phenomenon of phenoconversion was addressed in a sample of old aged patients. Although the genetic information was limited, useful information could still be extracted. In Chapter 3.1 and 3.2 two method validation studies are described for determination of amitriptyline, nortriptyline, desipramine, imipramine, clomipramine, desmethylclomipramine, venlafaxine and desmethylvenlafaxine in dried blood spots by liquid chromatography-tandem mass spectrometry. In chapter 3.3 problems which are faced when implementing DBS methods, like different concentration in plasma and dried blood spots, hematocrit effects and statistical analysis of bridging studies are

16 Chapter 1

discussed and solutions are formulated. In Chapter 4.1 a review of economic evalua-tions of pharmacogenetics testing is described. This review builds upon two previous reviews to maintain a uniform search strategy and outcome assessment. In Chapter 4.2 a cost-effectiveness study is performed to assess the cost-effectiveness of genotyping for CYP2D6 as routine practice when pharmacotherapy with nortriptyline is initiated. The study population consisted of depressed old aged hospitalized patients. Chapter 4.3. describes a protocol for a pragmatic RCT in which the validated BDS methods are ap-plied to monitor nortriptyline and venlafaxine plasma concentrations in order to assess the effects of genotyping for CYP2D6 when old aged patients start pharmacotherapy with nortriptyline or venlafaxine. The results of all chapters and future perspectives are discussed in Chapter 5.

General Introduction 17

1.9. REFEREnCEs

1. Ingelman-Sundberg M. Personalized medicine into the next generation. J Intern Med. 2015 Feb; 277(2): 152-4.

2. Lesko LJ. Personalized medicine: elusive dream or imminent reality? Clin Pharmacol Ther. 2007 Jun; 81(6): 807-16.

3. Hiemke C, Shams M. Phenotyping and genotyping of drug metabolism to guide pharmaco-therapy in psychiatry. Curr Drug Deliv. 2013 Feb; 10(1): 46-53.

4. Kalow W, Tang BK. Use of caffeine metabolite ratios to explore CYP1A2 and xanthine oxidase activities. Clin Pharmacol Ther. 1991 Nov; 50(5 Pt 1): 508-19.

5. Li A, Meyre D. Jumping on the Train of Personalized Medicine: A Primer for Non-Geneticist Clini-cians: Part 2. Fundamental Concepts in Genetic Epidemiology. Curr Psychiatry Rev. 2014 May; 10(2): 101-17.

6. Li A, Meyre D. Jumping on the Train of Personalized Medicine: A Primer for Non- Geneticist Clini-cians: Part 3. Clinical Applications in the Personalized Medicine Area. Curr Psychiatry Rev. 2014 May; 10(2): 118-32.

7. Kirchheiner J, Nickchen K, Bauer M, Wong ML, Licinio J, Roots I, et al. Pharmacogenetics of anti-depressants and antipsychotics: the contribution of allelic variations to the phenotype of drug response. Mol Psychiatry. 2004 May; 9(5): 442-73.

8. Swen JJ, Nijenhuis M, de Boer A, Grandia L, Maitland-van der Zee AH, Mulder H, et al. Pharma-cogenetics: from bench to byte--an update of guidelines. Clin Pharmacol Ther. 2011 May; 89(5): 662-73.

9. Hermann M, Hendset M, Fosaas K, Hjerpset M, Refsum H. Serum concentrations of venlafaxine and its metabolites O-desmethylvenlafaxine and N-desmethylvenlafaxine in heterozygous carri-ers of the CYP2D6*3, *4 or *5 allele. Eur J Clin Pharmacol. 2008 May; 64(5): 483-7.

10. Dalen P, Dahl ML, Bernal Ruiz ML, Nordin J, Bertilsson L. 10-Hydroxylation of nortriptyline in white persons with 0, 1, 2, 3, and 13 functional CYP2D6 genes. Clin Pharmacol Ther. 1998 Apr; 63(4): 444-52.

11. Whyte EM, Romkes M, Mulsant BH, Kirshne MA, Begley AE, Reynolds CF,3rd, et al. CYP2D6 geno-type and venlafaxine-XR concentrations in depressed elderly. Int J Geriatr Psychiatry. 2006 Jun; 21(6): 542-9.

12. Hicks JK, Swen JJ, Thorn CF, Sangkuhl K, Kharasch ED, Ellingrod VL, et al. Clinical Pharmacogenet-ics Implementation Consortium Guideline for CYP2D6 and CYP2C19 Genotypes and Dosing of Tricyclic Antidepressants. Clin Pharmacol Ther. 2013 Jan 16; 93(5): 402-8.

13. Caudle KE, Klein TE, Hoffman JM, Muller DJ, Whirl-Carrillo M, Gong L, et al. Incorporation of pharmacogenomics into routine clinical practice: the Clinical Pharmacogenetics Implementation Consortium (CPIC) guideline development process. Curr Drug Metab. 2014 Feb; 15(2): 209-17.

14. Oosterhuis B, van Boxtel CJ. Kinetics of drug effects in man. Ther Drug Monit. 1988; 10(2): 121-32. 15. Edelbroek PM, van der Heijden J, Stolk LM. Dried blood spot methods in therapeutic drug moni-

toring: methods, assays, and pitfalls. Ther Drug Monit. 2009 Jun; 31(3): 327-36. 16. Spooner N, Lad R, Barfield M. Dried blood spots as a sample collection technique for the determi-

nation of pharmacokinetics in clinical studies: considerations for the validation of a quantitative bioanalytical method. Anal Chem. 2009 Feb 15; 81(4): 1557-63.

17. Woods K, Douketis JD, Schnurr T, Kinnon K, Powers P, Crowther MA. Patient preferences for capil-lary vs. venous INR determination in an anticoagulation clinic: a randomized controlled trial. Thromb Res. 2004; 114(3): 161-5.

18 Chapter 1

18. The National Health Care institute. FT rapport trastuzumab bij HER2-positief gemetastaseerd maagcarcinoom. https: //www.zorginstituutnederland.nl/binaries/content/documents/zinl-www/documenten/publicaties/publications-in-english/2010/1012-trastuzumab-herceptin-as-adjuvant-treatment-of-her-2-positive-primary-breast-cancer-pharmacotherapeutic-report/1012-trastuzum: 2011.

19. Meckley LM, Neumann PJ. Personalized medicine: factors influencing reimbursement. Health Policy. 2010 Feb; 94(2): 91-100.

20. Laika B, Leucht S, Heres S, Steimer W. Intermediate metabolizer: increased side effects in psychoactive drug therapy. The key to cost-effectiveness of pretreatment CYP2D6 screening? Pharmacogenomics J. 2009 Dec; 9(6): 395-403.

21. Chou WH, Yan FX, de Leon J, Barnhill J, Rogers T, Cronin M, et al. Extension of a pilot study: impact from the cytochrome P450 2D6 polymorphism on outcome and costs associated with severe mental illness. J Clin Psychopharmacol. 2000 Apr; 20(2): 246-51.

22. Gressier F, Verstuyft C, Hardy P, Becquemont L, Corruble E. Response to CYP2D6 substrate antide-pressants is predicted by a CYP2D6 composite phenotype based on genotype and comedications with CYP2D6 inhibitors. J Neural Transm. 2015 Jan; 122(1): 35-42.

23. Shah RR, Smith RL. Addressing phenoconversion: the Achilles’ heel of personalized medicine. Br J Clin Pharmacol. 2015 Feb; 79(2): 222-40.

24. Davies EA, O’Mahony MS. Adverse drug reactions in special populations - the elderly. Br J Clin Pharmacol. 2015 Jan 24.

25. Tedeschini E, Levkovitz Y, Iovieno N, Ameral VE, Nelson JC, Papakostas GI. Efficacy of antide-pressants for late-life depression: a meta-analysis and meta-regression of placebo-controlled randomized trials. J Clin Psychiatry. 2011 Dec; 72(12): 1660-8.

26. Buxton MJ, Drummond MF, Van Hout BA, Prince RL, Sheldon TA, Szucs T, et al. Modelling in eco-nomic evaluation: an unavoidable fact of life. Health Econ. 1997 May-Jun; 6(3): 217-27.

27. Heeg BM, Damen J, Buskens E, Caleo S, de Charro F, van Hout BA. Modelling approaches: the case of schizophrenia. Pharmacoeconomics. 2008; 26(8): 633-48.

28. Brennan A, Chick SE, Davies R. A taxonomy of model structures for economic evaluation of health technologies. Health Econ. 2006 Dec; 15(12): 1295-310.

29. Bradford LD. CYP2D6 allele frequency in European Caucasians, Asians, Africans and their descen-dants. Pharmacogenomics. 2002 Mar; 3(2): 229-43.

30. Trivedi MH, Rush AJ, Wisniewski SR, Nierenberg AA, Warden D, Ritz L, et al. Evaluation of outcomes with citalopram for depression using measurement-based care in STAR*D: implications for clini-cal practice. Am J Psychiatry. 2006 Jan; 163(1): 28-40.

31. Leucht C, Huhn M, Leucht S. Amitriptyline versus placebo for major depressive disorder. Cochrane Database Syst Rev. 2012 Dec 12; 12: CD009138.

32. Rudolph RL, Fabre LF, Feighner JP, Rickels K, Entsuah R, Derivan AT. A randomized, placebo-con-trolled, dose-response trial of venlafaxine hydrochloride in the treatment of major depression. J Clin Psychiatry. 1998 Mar; 59(3): 116-22.

33. Wilson K, Mottram P. A comparison of side effects of selective serotonin reuptake inhibitors and tricyclic antidepressants in older depressed patients: a meta-analysis. Int J Geriatr Psychiatry. 2004 Aug; 19(8): 754-62.

34. Landelijke stuurgroep multidiciplinaire richtlijnontwikkeling in de GGZ. Dutch Guideline Depres-sion (Addendum Elderly) Addendum Ouderen bij MDR derpessie. Utrecht: Trimbos-Instituut; 2008.

General Introduction 19

35. Romera I, Perez V, Menchon JM, Schacht A, Papen R, Neuhauser D, et al. Early switch strategy in patients with major depressive disorder: a double-blind, randomized study. J Clin Psychopharma-col. 2012 Aug; 32(4): 479-86.

36. Gardner KR, Brennan FX, Scott R, Lombard J. The potential utility of pharmacogenetic testing in psychiatry. Psychiatry J. 2014; 2014: 730956.

37. Tunis SR, Stryer DB, Clancy CM. Practical clinical trials: increasing the value of clinical research for decision making in clinical and health policy. JAMA. 2003 Sep 24; 290(12): 1624-32.

2Environmental and genetic variation can infl uence psychotropic drug plasma

concentrations

2.1Overstappen naar het gebruik van een e-sigaret kan de spiegel van medicatie

verhogen

[English] Switching from smoking to the use of an e-cigarette can increase drug

exposure

Elizabeth J.J. Berm, Rianne Ruijsbroek, Anton J.M. Loonen, Kris R. Goethals, Bob Wilffert, Fenneke van Hasselt

Nederlands Tijdschrift voor Geneeskunde. 2015. 26 september;159 (39):A9090.

24 Chapter 2.1

sAmEnVAtting

inleiding: Het cytochroom-P450 type 1A2 (CYP1A2) is verantwoordelijk voor de afbraak van diverse geneesmiddelen, waaronder het antipsychoticum clozapine. Roken van sigaretten geeft inductie van het CYP1A2. Hierdoor hebben patiënten die roken een hogere dosering van het geneesmiddel nodig.Casus: Bij patiënt X werd de dosering clozapine gewijzigd vanwege het vermoeden op een actieve psychose. De clozapine spiegel veranderde echter niet op de wijze waarop verwacht werd. Het bleek dat de patiënt tijdens de doseringswijzigingen was overge-stapt op het gebruik van e-sigaretten en vervolgens weer was begonnen met het roken van sigaretten. Dit gaf een aannemelijke verklaring voor de waargenomen spiegels.Conclusie: Bij een overstap van roken naar e-sigaretten valt de inductie van het CYP1A2 enzym weg. Voorschrijvers van geneesmiddelen die afgebroken worden door CYP1A2 en een beperkte therapeutische breedte hebben, dienen hierop alert te zijn. Een derge-lijke overgang kan een sterke ongewenste verhoging van de geneesmiddelspiegel tot gevolg hebben.

[EnglisH] ABstRACt

introduction: The Cytochrome-P450 subtype 1A2 enzyme (CYP1A2) is responsible for the metabolism of several drugs, including the antipsychotic agent clozapine. Smoking is a known inducer of the CYP1A2 enzyme. Due to this induction, a higher clozapine dosage is given to smokers.Case description: Our patient presented himself with an active psychosis and there-fore clozapine dosing was increased. To our surprise serum-levels of clozapine did not change according to our expectations. It appeared the patient had switched from smok-ing tobacco to using an e-cigarette and then started smoking again. This switch gave a plausible explanation for the observed clozapine serum-levels.Conclusion: When patients switch from smoking to using e-cigarettes prior induction of CYP1A2 disappears. Prescribers should monitor for this phenomenon in smoking patients using medication metabolized by CYP1A2 with a narrow therapeutic window, to avoid unwanted elevations in drug exposure.

Overstap naar e-sigaretten, kan clozapine bloedspiegels beïnvloeden 25

2.1.1. inlEiDing

sinds 2011 is de e-sigaret een populair alternatief voor de normale sigaret. Van het traditionele roken van tabak is bekend dat het leidt tot inductie van het enzym cytochroom-P450 type 1A2 (CyP1A2). in de psychiatrie is dit enzym vooral bekend vanwege het metabolisme van het antipsychoticum clozapine. CyP1A2 is echter ook betrokken bij de afbraak van meer generalistisch gebruikte medicatie, zoals naproxen en theofylline1. in deze casus tonen wij dat het belangrijk is om uzelf te informeren over het rookgedrag van een patiënt.

2.1.2. ziEktEgEsCHiEDEnis

Patiënt X is een 34-jarige negroïde man die na de middelbare school is begonnen bij de luchtmacht. Daar kreeg hij op zijn 19de last van imperatieve hallucinaties en achter-volgingswanen. Hij is gediagnosticeerd met schizofrenie van het paranoïde type. Een periode van problemen met medicatietrouw, misbruik en verslaving van verschillende soorten harddrugs en softdrugs volgde.

2.1.3. BEHAnDEling

De patiënt heeft nog steeds last van hallucinaties. Hij heeft het idee dat mensen hun tong naar hem uitsteken, dat mensen over hem praten en hij heeft een prikkend gevoel in zijn gezicht. Hij wordt behandeld door een psychiater met een maandelijkse injectie van 150 mg paliperidonpalmitaat (een atypisch antipsychoticum) en oraal additioneel clozapine. De patiënt gebruikt geen andere medicatie die de activiteit van het CYP1A2 enzym zou kunnen beïnvloeden. Naast de blijvende psychotische symptomen ervaart de patiënt ook sufheid en moeite om zijn bed uit te komen. Als werkdiagnose worden beide symptomen geduid als passend bij actieve psychose en wordt de dosering clo-zapine verhoogd (tabel 1). In deze periode rookt de patiënt 10-15 sigaretten per dag. Bij het volgende polibezoek geeft de patiënt aan erg suf te blijven, hij heeft vooral ’s morgens veel moeite om uit bed te komen. Aangezien er geen effect is op de psychoti-sche symptomen en in verband met zijn sufheid wordt de dosering weer verlaagd. Twee weken later blijkt echter, ondanks de doseringsverlaging de bloedspiegel verhoogd te zijn. Betrokkene vertelt dat hij tijdens de periode dat zijn dosis verlaagd is en de hierop volgende policontrole drie weken later, overgestapt is naar het gebruik van een e-sigaret. Bij de volgende afspraak na de verlaagde dosering van clozapine blijkt de ervaren sufheid niet minder, daarom wordt er besloten opnieuw een spiegel te bepalen.

26 Chapter 2.1

Hij rookt nu weer zo’n 5 of meer sigaretten per dag en gebruikt de e-sigaret de hele dag door. De spiegel is dan bij een gelijke dosering flink gedaald (tabel 1).

Door de herhaaldelijke spiegelbepalingen kon goed worden geobjectiveerd dat de ervaren sufheid geen verband had met de spiegel van de clozapine. Het is mogelijk te duiden als een negatief symptoom. Er werd gekozen om de dosering op dit niveau te stabiliseren. Op basis van het onverwachte spiegelbeloop werd gevraagd naar verande-ringen in rookgedrag welke een aannemelijke verklaring gaven voor de merkwaardige geneesmiddelspiegels.

Een belangrijke les is om het rookgedrag bij elk consult te bespreken vanwege de mo-gelijke invloed op de geneesmiddelspiegel. Doordat deze patiënt clozapine als additie ontving bij zijn depot en daardoor relatief laag was ingesteld, leidde het overstappen naar e-sigaretten niet tot een potentieel toxische spiegel. Echter, als deze patiënt mo-notherapie met clozapine had ontvangen, had dit mogelijk wel tot een toxische spiegel kunnen leiden.

2.1.4. BEsCHOuwing

Het onverwachte beloop van de clozapine spiegel bij bovengenoemde patiënt maakte ons erop attent dat het noodzakelijk is om naast de ingenomen doseringen ook andere factoren die de afbraak van CYP1A2 kunnen beïnvloeden zoals koorts, bepaalde voe-dingsmiddelen (broccoli) en het rookgedrag uit te vragen bij medicatie die door CYP1A2 wordt gemetaboliseerd1. Dit om onderbehandeling of overdoseringen te voorkomen. Een verhoogde clozapinespiegel kan nadelige gevolgen hebben, zoals sedatie, duizelig-heid, delirium, coördinatiestoornissen en gastro-intestinale hypomotiliteit eventueel leidend tot ileus2. De mortaliteit van een overdosering wordt geschat op 12%3. Do-

tabel 1. Dosering en serumconcentraties van clozapine, en rookgedrag van patiënt A.

tijdstip in weken dosering clozapinein mg per dag

serumconcentratieclozapine in μg/l*†

rookgedrag‡

0 150 10-15

1 200 10-15

3 200 347 10-15

4 100 e-sigaret

5 100 376 e-sigaret

8 100 91 ≥ 5 en e-sigaret

* 12-uursspiegel in ‘steady state’.† Het therapeutisch venster bij monotherapie met clozapine is 350-700 µg/l.2 Het gaat bij patiënt A echter om additietherapie waarvoor geen duidelijke serumspiergelgrenzen beschikbaar zijn.‡ Uitgedrukt als aantal normale sigaretten/dag, tenzij anders aangegeven.

Overstap naar e-sigaretten, kan clozapine bloedspiegels beïnvloeden 27

sisonafhankelijk is het met name bij aanvang van de behandeling aanwezige risico op agranulocytose.

De invloed van roken op het CyP1A2-systeem

Maximale inductie van het CYP1A2 metabolisme vindt plaats bij 7 tot 12 sigaretten per dag. Inductie van CYP1A2 door tabak vindt plaats door binding van de polycyclische aromatische koolwaterstoffen (PAK’s) uit tabaksrook aan de aryl-koolwaterstof-receptor. Ten gevolge hiervan neemt de transcriptionele activatie van het CYP1A2 gen toe, wat resulteert in een toename van de capaciteit van de lever om substraten, zoals PAK’s af te breken. Naast de PAK’s zou ook nicotine inductie kunnen veroorzaken, maar uit onder-zoek blijkt dat humaan CYP1A2 niet beïnvloed wordt door nicotine4. Personen die tabak roken, hebben 40 tot 50% lagere serumconcentraties van CYP1A2 gemetaboliseerde medicatie in vergelijking met niet-rokers. De dosis die gegeven moet worden van deze medicatie om een vergelijkbaar effect te hebben is dus veelal hoger bij rokers dan bij niet-rokers5.

De invloed van e-sigaretten op het CyP1A2-systeem

E-sigaretten worden niet aangestoken en hierdoor worden geen verbrandingsproduc-ten van tabak geïnhaleerd zoals teer en koolmonoxide6. Net zoals in normale sigaretten, kan de hoeveelheid nicotine in de vloeistof voor de e-sigaret sterk variëren7. Nicotine-vervangende producten, zoals e-sigaretten, veroorzaken niet de inductie van CYP1A2 zoals tabaksrook wel doet4,5.

CyP-enzymen buiten de lever

De CYP-enzymen zijn zowel in de lever als cerebraal actief. Zeker voor psychofarmaca kan deze cerebrale CYP-activiteit van belang zijn, met name voor het optreden van bijwerkingen. Ook voor cerebraal CYP1A2 wordt aangenomen dat dit niet geïnduceerd wordt door nicotine en/of alcohol, maar bij andere CYP-enzymen, zoals bijvoorbeeld het CYP2D6 wordt wel inductie verondersteld8.

2.1.5. COnClusiE

Het is belangrijk te weten of, wat en hoe de patiënt rookt. Het roken van sigaretten en shag leidt tot inductie van CYP1A2 dat de geneesmiddelspiegels van o.a. naproxen, theofylline en clozapine verlaagt. Deze inductie valt weg wanneer een patiënt overstapt op een e-sigaret wat kan leiden tot een verhoogde geneesmiddelspiegel. Medicamen-ten die gemetaboliseerd worden door CYP1A2 zijn onder andere te vinden in de Flock-hart tabel1. Voor diverse geneesmiddelen is de spiegel niet van belang voor het effect,

28 Chapter 2.1

echter voor sommige geneesmiddelen gelden nauwe therapeutische grenzen. Wanneer patiënten een dergelijk geneesmiddel, zoals clozapine, gebruiken en overstapen op e-sigaretten of anderszins stoppen of minderen met het roken van tabak is het aan te bevelen om spiegels te controleren

Overstap naar e-sigaretten, kan clozapine bloedspiegels beïnvloeden 29

2.1.6. litERAtuuR

1. Drug interactions: cytochrome P450 drug interaction table. Indiana University School of Me-dicine (2007). 2013. Flockhart DA. (Accessed Maart, 23ste, 2014, at http://medicine.iupui.edu/clinpharm/ddis/table.aspx).

2. van Gool AR, de Jong MH, Verhoeven WMA. Toxische plasmaconcentratie van clozapine bij ont-stekingsprocessen. Tijdschrift voor psychiatrie 2010; 52: 791-796.

3. Novartis Pharma B.V. Leponex. Summary of Product Characteristics. versie: 11 Oktober 2014; . 4. Hukkanen J, Jacob P,3rd, Peng M, Dempsey D, Benowitz NL. Effect of nicotine on cytochrome

P450 1A2 activity. Br J Clin Pharmacol 2011; 72: 836-8. 5. Fankhauser M. Drug interactions with tabacco smoke: implications for patient care. Current

Psychiatry 2013; 12: 12-16. 6. Buisman R, Croes E. Factsheet elektronische sigaretten (e-sigaretten). Utrecht: Nationaal Experti-

secentrum Tabaksontmoediging, 2014. 7. Hahn J, Monakhova YB, Hengen J, et al. Electronic cigarettes: overview of chemical composition

and exposure estimation. Tob Induc Dis 2014; 12: 23,014-0023-6. eCollection 2014. 8. Ivanova SA, Toshchakova VA, Filipenko ML, et al. Cytochrome P450 1A2 co-determines neuro-

leptic load and may diminish tardive dyskinesia by increased inducibility. World J Biol Psychiatry 2015; : 1-6.

9. Nederlandse Vereniging voor Psychiatrie. Multidisciplinaire richtlijn schizofrenie, 2012.

2.2Relation between CYP2D6 genotype, phenotype and therapeutic drug

concentrations among nortriptyline and venlafaxine users in old age psychiatry.

Elizabeth J.J. Berm, Rob M. Kok, Eelko Hak, Bob Wilffert

Accepted, Pharmacopsychiatry (2016)

32 Chapter 2.2

ABstRACt

introduction: To determine relations between drug concentrations and the cytochrome P450-CYP2D6 genotype or phenotype among old aged patients treated with nortripty-line or venlafaxine.methods: A post-hoc analysis of a clinical trial was performed. Patients were grouped into phenotypes according to the metabolite/mother compound ratio. Genotypes were assessed by the CYP2D6 *3 and *4 alleles.Results: Data was available from 81 patients (41 nortriptyline, 40 venlafaxine) with a mean age of 72 years. No phenoconversion from Poor Metabolizers (PM) to Extensive Metabolizers (EM), or vice versa, was found. However, we did find phenoconversion from PM to Intermediate Metabolizers (IM), IM to EM, or vice versa in 36% of observa-tions. Among nortriptyline users, patients with a PM or IM genotype had more supra-therapeutic blood levels, although this did not reach statistical significance. In explor-atory analyses we found men were more likely (RR: 2.4; 95% CI: 1.14- 5.07) to display phenoconversion from an IM genotype to EM phenotype. In addition, compared to non PMs, PMs were found to have higher risk (RR: 1.56 (95% CI: 1.03- 2.37) on non-response, although this was only significant when response was measured on the Hamilton Rating Scale for Depression and not on the Montgomery Åsberg Depression Rating Scale.Conclusion: Patients phenoconversed, but we did not observe phenoconversion from PM to EM, or vice versa. Genotype information could be used as a valuable tool, in addition to therapeutic drug monitoring, to prevent supratherapeutic drug levels of nortriptyline or venlafaxine in old aged patients with a PM genotype.

Relation between genotype and phenotype among old aged patients 33

2.2.1. intRODuCtiOn

The relationship between genotype and phenotype of the polymorphic cytochrome P450 2D6 enzyme (CYP2D6) has been intensively studied over the last decade. The func-tion of this enzyme can be classified into four different drug metabolizer ‘types’. First, the extensive metabolizer (EM) type which corresponds with normal enzyme functionality and concerns the majority of the patient population. Next, the poor metabolizer (PM) type characterized by almost no enzyme activity. Third, the intermediate metabolizer (IM) type, characterized by a decreased, but usually still active, enzyme, and last the ultra-rapid metabolizer type (UM) characterized by an increased enzyme activity. These different patient types can be distinguished by assessment of the metabolic ratio of a drug and its metabolite after administration of a CYP2D6-specific substrate. This is usually referred to as phenotyping. At the same time, these differences can also be predicted by pharmacogenetics tests which is referred to as genotyping.

Nortriptyline and venlafaxine are primarily metabolized by CYP2D6 [1, 2]. To apply knowledge about differences in CYP2D6 capacity, guidelines have been formulated for both nortriptyline and venlafaxine to adapt the dosage based on the CYP2D6 genotype [3, 4]. However, in a study among 865 venlafaxine users, 24% of the patients who were genotyped as EM appeared to have a PM phenotype [5]. This can be due to co-administration of strong CYP2D6 inhibitors, like paroxetine, which is known to induce PM phenotypes among genetically EMs [6]. This phenomenon of discrepancy between genotype and phenotype has been referred to as ‘phenoconversion’ and has provoked discussion whether genotyping can accurately predict the phenotype [5, 7, 8]. In addi-tion, practitioners often refer to therapeutic drug monitoring (TDM) as an accurate and sufficient indicator of the phenotype which renders additional genotyping superfluous. An additional concern is the ageing of the patient population. Although in general it is reported that CYP2D6 function does not decrease in elderly, other CYP enzymes like CYP3A4 or CYP2C19 are associated with a decreased function among old aged patients [9]. In line with this, a decreased clearance of CYP2D6 substrates like flecaïne or desip-ramine has been described in older patients which might be related to a decreased function of these other secondary metabolizing CYP enzymes [10, 11]. Relatively little is known about the genotype-phenotype relationship in old aged populations and due to demographic ageing, this population will increase in the Western civilization.

In this post-hoc analysis we tried to address two research questions: (1) do we observe the phenomenon of phenoconversion in an old aged population, and (2) which geno-types among older patients experience sub-or supratherapeutic drug levels? In addition, two exploratory analysis were performed. First, to study if gender or age were related to phenoconversion. Second, to study if non-response was related to phenotype and/or genotype. This way, more insight is given in the possible added value of genotyping in

34 Chapter 2.2

addition to TDM for optimization of the treatment with nortriptyline and venlafaxine in old aged populations.

2.2.2. mEtHODs

We performed a post-hoc analysis with data from a randomized controlled trial among 81 starters of nortriptyline or venlafaxine aged 60 years or older [12]. In the trial, efficacy and safety of nortriptyline and venlafaxine was compared among inpatients diagnosed with major depressive disorder (according to the DSM-IV criteria). Only patients with a score of ≥20 on the Montgomery Åsberg Depression Rating Scale (MADRS) were included [13]. Patients with or without a prior episode of depression (treated with an an-tidepressant) were included. Patients diagnosed with dementia according to the DSM-IV or a Mini-mental State Examination score of <15 were excluded [14]. Patients were intensively monitored by TDM and blood samples were collected after three, five, and 12 weeks. Both the mother compounds, nortriptyline (NTP) or venlafaxine (VFX) as well as the main CYP2D6 metabolites, OH-nortriptyline and O-desmethylvenlafaxine (ODV), were measured. The genotypes for the CYP2D6 *3 and *4 alleles, were analyzed after patients completed the trial. During the trial, a time-dependent flexible dosage scheme was used. Among NTP users, dosing was adapted based on clinical effects and TDM. Among VFX users, dosing was adapted based on clinical efficacy only. Co-prescription of psychotropic drugs was restricted to oxazepam, temazepam, haloperidol or risperidone. Co-prescription of somatic medication was not restricted. Patients who had been unsuc-cessfully treated with a tricyclic antidepressant or venlafaxine for the studied episode of depression were excluded from the trial.

genotyping

Patients with a *3 or *4 allele were classified as IM and patients with two dysfunctional alleles as PMs (*3/*3; *4/*4; *3/*4). Patients with no dysfunctional alleles were classified as EM. For the determination of the UM genotype, information about gene duplication is needed. This genetic information was not available, and therefore UMs could not be identified.

Phenotyping

We determined the phenotypes according to the metabolite/mother compound ratio after three, five and 12 weeks. A ratio between 0 and 0.5 was categorized as a PM, be-tween 1.5 and 10 as EM [15]. To determine the IM phenotype a ratio between 0.5 and 1.5 was used.

Relation between genotype and phenotype among old aged patients 35

Phenoconversion

The predicted phenotype and the measured phenotype were compared for each TDM sample and the corresponding clinical sensitivity and specificity for CYP2D6 testing was calculated. Patients who had a different phenotype compared to their genotype in the majority of their TDM samples were considered to experience phenoconversion. Phe-noconversion could occur in two directions, being towards more (i.e. IM → EM; PM → IM) or less (EM → IM; IM → PM) metabolic capacity then predicted based on genotype. In an exploratory analysis, relations between gender, age, and phenoconversion were analyzed.

therapeutic drug monitoring

We translated measured plasma concentrations into clinically relevant outcomes being therapeutic (within therapeutic range), subtherapeutic (below lowest level of therapeu-tic range) or supratherapeutic (above highest level of therapeutic range). For NTP we used a therapeutic reference range of 50-150 µg/L. For VFX a range of 100-400 µg/L was used for the sum of VFX and ODV [16]. Note that in the trial the TDM data of VFX was not considered leading in pharmacotherapy, since the dose-effect relation during the time of the trial was highly debated [12].

Response to antidepressant

Twelve weeks after start of treatment response was measured with the MADRS and the Hamilton Rating Scale for Depression (HAM-D; 17 item version) [13, 17]. For both instru-ments, a reduction of at least 50% from baseline score was considered as a response. In an exploratory analysis relations between plasma concentrations and response as well as genotype and response were assessed.

statistics

To test for any significant differences in the occurrence of (sub or supra-) therapeutic levels between the genotypes, Fisher’s Exact test for binomial data was used. To test if phenoconversion was related to gender or age (groups: ≤75 years; >75 years), data was analyzed with logistic regression. In addition relative risk ratios were calculated. To assess the relation between non-response (yes/no) and genotype, genotypes or phe-notypes were collapsed into dichotomies outcomes, normal (EM and IM) or poor (PM) metabolizers. Relative risk ratios were calculated. Analyses were performed with IBM SSPS Statistics, version 23. P-values of <0.05 were considered significant.

36 Chapter 2.2

2.2.3. REsults

For the first TDM measurement (week 3), complete data from 75 out of 81 participants (40 NTP, 35 VFX) was available. Demographics of included patients are shown in table 1. After 5 weeks and 12 weeks, no TDM data was available for 13 and 26 participants, respectively. This resulted in 33 NTP and 35 VFX patient samples after 5 weeks and after 12 weeks in 27 NTP and 27 VFX patient samples. For two patients using nortriptyline (corresponding with six TDM samples) genotype information was missing. Therefore, in total 98 TDM samples were available for NTP and 95 for VFX to compare phenotype with genotype.

table 1. Demographics of included patients.

total of patients included 81

Mean age 72.2 (SD: 7.6)

Gender (female) 59 (73%)

Nortriptyline users 41 (51%)

Venlafaxine users 40 (49%)

Responders according to the MADRS 38 (47%)

Responders according to the HAM-D 36 (44%)

Genotyped 79

*3 and wild type (i.e. IM) 3 (4%)

*4 and wild type (i.e. IM) 28 (35%)

*3/*4; *4/*4 (i.e. PM) 1 (1%) ; 5 (6%)

Number of patients controlled by TDM after week 3, 5 and 12 weeks of treatment

75, 69, 55

HAM-D: Hamilton Rating Scale for Depression; IM: intermediate metabolizer; MADRS: Montgomery Asberg Depression Rating Scale; PM: poor metabolizer; SD: standard deviation; TDM: therapeutic drug monitoring.

Phenoconversion

No phenoconversion from the PM to the EM genotype, or vice versa, was observed for patients treated with NTP or VFX (table 2). Phenoconversion from PM to IM, IM to EM, or vice versa was found in 36% (69 out of 193 comparisons). Among the NTP group this was 28% (27 out of 98 comparisons) and among the VFX group 44% (42 out of 95 com-parisons). Most phenoconversion for NTP concerned patients genotyped as EMs who had an IM phenotype in 28% of their TDM samples (19 out of 69). For VFX most pheno-conversions concerned IM genotyped patients who expressed an EM phenotype in 77% of their TDM samples (34 out of 44). In the logistic regression model, phenoconversion was not related to the age (β: −0.06; p =0.92) of patients, but it was related to gender (β: 1.2; sig. p= 0.03). Among man a higher relative risk ratio of 2.40 (95% CI: 1.14-5.07) on phenoconversion towards a phenotype with more enzyme activity was found (i.e. from

Relation between genotype and phenotype among old aged patients 37

IM→ EM; PM → IM). Phenoconversion towards less enzyme activity (i.e. from EM→ IM or PM; IM→ PM) was not related to either gender or age.

Clinical outcomes

Patients with an IM or PM genotype were more frequently exposed to blood levels above the therapeutic range compared to patients with an EM genotype, although differences were not significant (table 3). After five weeks, differences for NTP were on the border-line of significance (p<0.07). Patients who had blood levels outside the therapeutic range after 12 weeks were more frequently non-responders, compared to patients who had blood levels within the therapeutic range. Based on the HAM-D scale results were significant (Fisher Exact test: p<0.01), but not on the MADRS scale (Fisher Exact test: p<0.16). Interestingly, patients with a drug concentration below the therapeutic range

table 2. Relation between predicted and measured phenotype and corresponding clinical sensitivity and specificity for CYP2D6 *3 and *4 testing to predict PM, IM or EM phenotypes among old aged depressed inpatients for nortriptyline (A) and venlafaxine (B).A.

Predicted phenotype

measured phenotype Pm im Em total

Pm 4 1 0 5

im 3 17 19 39

Em 0 4 50 54

total 7 22 69 98

sensitivity specificity PPV nPV

Pm 0.80 (4/5) 0.97 (90/93) 0.57 (4/7) 0.99 (90/91)

im 0.44 (17/39) 0.92 (54/59) 0.77 (17/22) 0.71 (54/76)

Em 0.93 (50/54) 0.57 (25/44) 0.72 (50/69) 0.86 (25/29)

B.

Predicted phenotype

measured phenotype Pm im Em total

Pm 7 3 0 10

im 1 6 3 10

Em 0 35 40 75

total 8 44 43 95

sensitivity specificity PPV nPV

Pm 0.70 (7/10) 0.99 (84/85) 0.88 (7/8) 0.97 (84/87)

im 0.60 (6/10) 0.55 (47/85) 0.77 (17/22) 0.92 (47/51)

Em 0.53 (40/75) 0.85 (17/20) 0.72 (50/69) 0.33 (17/52)

EM; extensive metabolizer; IM: Intermediate metabolizer; NPV: negative predictive value; PM: Poor metabo-lizer; PPV: positive predictive value.

38 Chapter 2.2

tabl

e 3.

Pro

port

ions

pat

ient

s w

ithin

, bel

ow o

r abo

ve th

e th

erap

eutic

rang

e fo

r eac

h ph

enot

ype

afte

r 3, 5

, and

12

wee

ks o

f tre

atm

ent.

mea

sure

d bl

ood

leve

ls w

ith

resp

ect t

o th

erap

euti

c ra

nge

genotype

wee

k 3

wee

k 5

wee

k 12

With

in

rang

e(%

)

Belo

w (%

)A

bove

(%)

p-va

lue

With

in

rang

e(%

)

Belo

w (%

)A

bove

(%)

p-va

lue

With

in

rang

e(%

)

Belo

w (%

)A

bove

(%)

p-va

lue

nor

trip

tylin

e

All

23 (6

0)6

(16)

9 (2

4)22

(67)

5 (1

5)6

(18)

16 (5

9)4

(15)

7 (2

6)

PM1

(33)

0 (0

)2

(67)

0.18

0 (0

)0

(0)

2 (1

00)

0.07

1 (5

0)0

(0)

1 (5

0)0.

64

IM7

(70)

0 (0

)3

(30)

6 (8

6)0

(0)

1 (1

4)2

(40)

1 (2

0)2

(40)

EM15

(60)

6 (2

4)4

(16)

16 (6

7)5

(21)

3 (1

2)13

(65)

3 (1

5)4

(20)

Venl

afax

ine

All

32 (8

6)1

(3)

4 (1

1)23

(66)

4 (1

1)8

(23)

9 (3

3)1

(4)

17(6

3)

PM3

(100

)0

(0)

0 (0

)1.

002

(67)

0 (0

)1

(33)

0.32

0 (0

)0

(0)

2(10

0)0.

72

IM14

(82)

2 (1

2)1

(6)

12 (8

0)2

(13)

1 (7

)4

(29)

1 (7

)9

(64)

EM15

(88)

2 (1

2)0

(0)

9 (5

3)2

(12)

6 (3

5)5

(45)

0 (0

)6

(55)

EM: e

xten

sive

met

abol

izer

; IM

: Int

erm

edia

te m

etab

oliz

er; P

M: P

oor m

etab

oliz

er.

Relation between genotype and phenotype among old aged patients 39

were mostly responders and patients with blood concentrations above the therapeutic range were more frequently non-responders. In line with these findings, we found patients with a PM genotype, which are more likely to experience high drug concen-trations, had a higher risk on non-response. In contrast, patients with an IM genotype, displayed a response similar to EMs. When PM genotypes were compared to non-PMs, differences in response were again only significant when measured on the HAM-D scale. In addition, comparable outcomes were found for PM phenotypes. Relative risk ratios for non-response of PMs and 95% confidence intervals are shown in table 4.

2.2.4. DisCussiOn AnD COnClusiOn

We assessed an old aged population and did not observe any phenoconversion from the EM genotype to PM phenotype or vice versa. This is in contrast to the findings from Preskorn et al., who found in 24% out of the EM genotyped patients a PM phenotype. These differences could be related to the fact that we incorporated an IM group, which was not the case in the study of Preskorn et al. [5]. Furthermore, patients included in our study did not use potent CYP2D6 inhibitors like paroxetine, fluoxetine or bupropion although usage of other les commonly used potent inhibitors (i.e. cinacalcet, quinidine) was not controlled [18]. For the IM genotype we did observe phenoconversion to PM or EM and vice versa. Among the VFX users we observed most patients with an IM genotype had an EM phenotype. A gender difference was found for this type of phenoconversion (i.e. IM → EM) indicating male subjects have a higher risk. This adds up to the evidence that women have a lower venlafaxine clearance and as such less metabolic reserve to compensate for the reduced CYP2D6 function [19, 20]. On the other hand, these findings indicate that among VFX users, lowering the dosage based on the IM genotype is not indicated in the majority of the patients and especially not among man. This could be due to other metabolic pathways like the formation of N-desmethylvenlafaxine which is catalyzed by the CYP3A4 enzyme [2, 21]. With respect to clinical outcomes, a trend

table 4. Risk estimates for non-response based on PM genotype or phenotype compared to non-PM geno-type or phenotype. Bold relative risk ratios were found significant (based on 95% confidence interval).

Outcomescale

Relative risk ratio

95% confidence interval

p-value

non-responder when Pm genotype MADRS 1.28 0.70- 2.35 0.43

HAM-D 1.56 1.03- 2.37 0.04

non-responder when Pm phenotype MADRS 1.43 0.85- 2.41 0.18

HAM-D 1.67 1.14- 2.43 <0.01

HAM-D: Hamilton Rating Scale for Depression; MADRS: Montgomery Asberg Depression Rating Scale; PM: poor metabolizer.

40 Chapter 2.2

was found indicating patients treated with NTP and a PM genotype experience more supratherapeutic drug concentrations during the first five weeks of treatment, even though intensive TDM is applied. Differences were not significant which indicates a larger samples size is needed. We found no such difference for VFX users during the first five weeks of pharmacotherapy. However, at week 12 most patients had VFX blood levels above the therapeutic range which were relatively more PMs compared to EMs. These numbers are too small to draw any firm conclusions, however our results indicated that not at the start of VFX treatment, but after the first 3 months, patients with less metabolizing capacity become more vulnerable to a relatively high VFX exposure. The PM genotype and phenotype were related to a significant higher risk in non-response based on the HAM-D scale, but not on the MADR scale, although a similar effect was ob-served. These differences are likely related to the small number of PMs, which made the outcomes highly dependent on a different classification of one or two PMs. Although both the MADRS and HAM-D scales are suitable for measuring depression response among elderly, there are some differences in the weight given to certain symptoms of the depression. Compared to the MADRS, the HAM-D is influenced more by items rating sleep, anxiety and somatic symptoms and less by mood changes [22]. These items are more sensitive to somatic symptoms which could occur due to side effects of the antide-pressant like diarrhea, dizziness and dry mouth which are reported to be related to high plasma concentrations of NTP or VFX PMs [23, 24]. It is important, future studies among old aged patients with a larger sample size are conducted to replicate the findings of this exploratory analysis.

The main limitation of this study was the limited genetic information which was avail-able. The assessed mutations (*3 and *4) are the most frequently found mutations in Caucasians, however numerous other mutations have been identified [25]. These muta-tions can also cause decreased enzymatic activity, which we did not test for. In addition, we did not have the information about duplications of the gene, which can cause an increased metabolic activity and can be used to classify UMs. However, in Caucasians, these duplications do only explain one third of the UM phenotypes [25]. The missed mutations for decreased enzymatic activity could explain unjustified classification of genotyped EMs which were phenotyped as IMs, as was observed among the NTP us-ers. Nevertheless, it cannot explain the observed phenoconversion among VFX users (i.e. IM → EM), since missed mutation will lead to a decreased enzyme activity instead of increased activity. Although this could theoretically be related to an UM genotype, which we did not have the genetic information for, the proportion of UMs in Caucasian populations is very small (1-2%) and it is unlikely this would explain our results [25}. After twelve weeks, 26 patients did not complete TDM which was due to 27 patients who were lost to follow-up in the trial [12]. Lost to follow-up was often related to insufficient improvement, which could be related to insufficient plasma levels that could be related

Relation between genotype and phenotype among old aged patients 41

to the genotype. In this analysis we did not assess such influences, because of the small sample size. However, such influences would be interesting and should be studied in larger datasets.

Genotyping was found to reliable predict PM phenotypes of CYP2D6. For a reliable prediction of PM phenotypes, inclusion of an IM group and usage of potent CYP2D6 inhibitors needs to be taken into account. Uncertainty was found among IM genotyped patients and for VFX the IM genotype did not seem to predict lower metabolic activity. In line with this, IM genotype did not influence antidepressant response. Surprisingly, not subtherapeutic, but supratherapeutic drug concentrations were found to be related to non-response. Consequently, PM genotype was found to be related with a higher risk on non-response, although results should be interpreted with caution, due to the post-hoc nature and small sample size of this study. In conclusion, dose adaptation of VFX based on the IM genotype are unlikely to improve clinical outcomes among male patient, which could be related to the frequently observed EM phenotype. Nevertheless, genotype information could be used as a valuable tool, in addition to TDM, to prevent supratherapeutic drug levels of NTP or VFX in old aged patients with a PM genotype.

42 Chapter 2.2

2.2.5. References 1. Breyer-Pfaff U. The metabolic fate of amitriptyline, nortriptyline and amitriptylinoxide in man.

Drug Metab Rev. 2004 Oct; 36(3-4): 723-46. 2. Sangkuhl K, Stingl JC, Turpeinen M, Altman RB, Klein TE. PharmGKB summary: venlafaxine path-

way. Pharmacogenet Genomics. 2014 Jan; 24(1): 62-72. 3. Hicks JK, Swen JJ, Thorn CF, Sangkuhl K, Kharasch ED, Ellingrod VL, et al. Clinical Pharmacogenet-

ics Implementation Consortium Guideline for CYP2D6 and CYP2C19 Genotypes and Dosing of Tricyclic Antidepressants. Clin Pharmacol Ther. 2013 Jan 16; 93(5): 402-8.

4. de Leon J, Armstrong SC, Cozza KL. Clinical guidelines for psychiatrists for the use of pharmaco-genetic testing for CYP450 2D6 and CYP450 2C19. Psychosomatics. 2006 Jan-Feb; 47(1): 75-85.

5. Preskorn SH, Kane CP, Lobello K, Nichols AI, Fayyad R, Buckley G, et al. Cytochrome P450 2D6 phenoconversion is common in patients being treated for depression: implications for personal-ized medicine. J Clin Psychiatry. 2013 Mar 13; 74(6): 614-21.

6. Zourkova A, Hadasova E. Paroxetine-induced conversion of cytochrome P450 2D6 phenotype and occurence of adverse effects. Gen Physiol Biophys. 2003 Mar; 22(1): 103-13.

7. Berm EJ, Risselada AJ, Mulder H, Hak E, Wilffert B. Phenoconversion of cytochrome P450 2D6: the need for identifying the intermediate metabolizer genotype. J Clin Psychiatry. 2013 Oct; 74(10): 1025.

8. Preskorn SH. Dr Preskorn replies. J Clin Psychiatry. 2013 Oct; 74(10): 1025-6. 9. Lotrich FE, Pollock BG. Aging and clinical pharmacology: implications for antidepressants. J Clin

Pharmacol. 2005 Oct; 45(10): 1106-22. 10. Doki K, Homma M, Kuga K, Aonuma K, Kohda Y. CYP2D6 genotype affects age-related decline

in flecainide clearance: a population pharmacokinetic analysis. Pharmacogenet Genomics. 2012 Nov; 22(11): 777-83.

11. Polasek TM, Patel F, Jensen BP, Sorich MJ, Wiese MD, Doogue MP. Predicted metabolic drug clear-ance with increasing adult age. Br J Clin Pharmacol. 2013 Apr; 75(4): 1019-28.

12. Kok RM, Nolen WA, Heeren TJ. Venlafaxine versus nortriptyline in the treatment of elderly de-pressed inpatients: a randomised, double-blind, controlled trial. Int J Geriatr Psychiatry. 2007 Dec; 22(12): 1247-54.

13. Montgomery SA, Asberg M. A new depression scale designed to be sensitive to change. Br J Psychiatry. 1979 Apr; 134: 382-9.

14. Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975 Nov; 12(3): 189-98.

15. Nichols AI, Lobello K, Guico-Pabia CJ, Paul J, Preskorn SH. Venlafaxine metabolism as a marker of cytochrome P450 enzyme 2D6 metabolizer status. J Clin Psychopharmacol. 2009 Aug; 29(4): 383-6.

16. Hiemke C, Baumann P, Bergemann N, Conca A, Dietmaier O, Egberts K, et al. AGNP consensus guidelines for therapeutic drug monitoring in psychiatry: update 2011. Pharmacopsychiatry. 2011 Sep; 44(6): 195-235.

17. Hamilton M. A rating scale for depression. J Neurol Neurosurg Psychiatry. 1960 Feb; 23: 56-62. 18. Drug interactions: cytochrome P450 drug interaction table. Indiana University School of Medicine

(2007). [Internet].; 2013 []. Available from: 19. Unterecker S, Hiemke C, Greiner C, Haen E, Jabs B, Deckert J, et al. The effect of age, sex, smoking

and co-medication on serum levels of venlafaxine and O-desmethylvenlafaxine under naturalis-tic conditions. Pharmacopsychiatry. 2012 Sep; 45(6): 229-35.

Relation between genotype and phenotype among old aged patients 43

20. Sigurdsson HP, Hefner G, Ben-Omar N, Kostlbacher A, Wenzel-Seifert K, Hiemke C, et al. Steady-state serum concentrations of venlafaxine in patients with late-life depression. Impact of age, sex and BMI. J Neural Transm (Vienna). 2015 May; 122(5): 721-9.

21. Mannheimer B, Haslemo T, Lindh JD, Eliasson E, Molden E. Risperidone and venlafaxine metabolic ratios strongly predict a CYP2D6 poor metabolizing genotype. Ther Drug Monit. 2015 Sep 26.

22. Uher R, Farmer A, Maier W, Rietschel M, Hauser J, Marusic A, et al. Measuring depression: compari-son and integration of three scales in the GENDEP study. Psychol Med. 2008 Feb; 38(2): 289-300.

23. Hodgson K, Tansey KE, Uher R, Dernovsek MZ, Mors O, Hauser J, et al. Exploring the role of drug-metabolising enzymes in antidepressant side effects. Psychopharmacology (Berl). 2015 Mar 12.

24. Shams ME, Arneth B, Hiemke C, Dragicevic A, Muller MJ, Kaiser R, et al. CYP2D6 polymorphism and clinical effect of the antidepressant venlafaxine. J Clin Pharm Ther. 2006 Oct; 31(5): 493-502.

25. Bradford LD. CYP2D6 allele frequency in European Caucasians, Asians, Africans and their descen-dants. Pharmacogenomics. 2002 Mar; 3(2): 229-43.

3Validation of dried blood spot sampling for therapeutic drug monitoring of tricyclic

antidepressants and venlafaxine.

3.1A simple dried blood spot method for

therapeutic drug monitoring of the tricyclic antidepressants amitriptyline,

nortriptyline, imipramine, clomipramine, and their active metabolites using

LC-MS/MS

Elizabeth J.J. Berm, Jolanda Paardekooper, Ellen Brummel-Mulder, Eelko Hak, Bob Wilffert, Jan G. Maring

Talanta. 2015. March;134:165-72

48 Chapter 3.1

ABstRACt

introduction: Therapeutic drug monitoring (TDM) of tricyclic antidepressants (TCAs) is considered useful in patients with major depressive disorders, since these drugs display large individual differences in clearance, and the therapeutic windows of these drugs are relatively small. We developed an assay for determination of amitriptyline (ATP), nortriptyline (NTP), imipramine (IMP), desipramine (DSP) clomipramine (CMP) and desmethyl-clomipramine (DCMP) in dried blood spots (DBS).methods: A fast and robust liquid chromatography-tandem mass spectrometry (LC-MS/MS) method was developed and analytically validated for simultaneous determination of ATP, NTP, IMP, DSP, CMP, and DCMP in DBS. Six mm circles were punched out from DBS collected on Whatman DMPK-C paper and mixed with acetonitrile : methanol 1:3 containing the internal standard. The extract was analyzed by LC-MS/MS. Total LC-MS/MS runtime was 4.8 min.Results: The assay was linear in the range 20- 500 µg/L for all compounds. Overall-assay accuracy and precision were < 20% for the lower limit of quantification (LLOQ), except for CMP (CV = 22.3%), and < 15% at other concentrations. The initial LLOQ was 20 µg/L however for CMP and DMCP it was increased to 40 µg/L. The blood volume per spot did not influence the results, but a low hematocrit (≤ 30%) was associated with a > 15% negative bias for all compounds. Punching at the perimeter of the blood spot instead of the center was associated with a positive bias. A good correlation was found between patients plasma and DBS samples of ATP, NTP and DMCP, but not for CMP. In addition, proportional differences were found.Conclusion: This LC-MS/MS method was analytically validated for determination of TCAs in DBS. Future validation will focus on the clinical application of the method.

Determination of tricyclic antidepressants in dried blood spots 49

3.1.1. intRODuCtiOn

Amitriptyline (ATP), nortriptyline (NTP), imipramine (IMP), and clomipramine (CMP) are first choice tricyclic antidepressants (TCAs) which are indicated by Dutch guidelines as part of therapy for anxiety or depressive disorders (1, 2). Therapeutic drug monitoring (TDM) of TCAs and their active metabolites is advised by guidelines, because these drugs display large individual differences in clearance while the therapeutic windows of these drugs are relatively small (3).

In clinical practice conventional TDM is based on venous sampling methods, however, the use of dried blood spot (DBS) sampling and analysis is gaining interest in TDM. DBS sampling originated in newborn screening to obtain an economical and rapid screen-ing method for phenylalanine to detect phenylketonuria (4). DBS sampling has certain advantages compared to conventional venous sampling. First, for DBS sampling blood is collected by a heel or finger puncture which is an advantage for certain vulnerable patient groups as children or elderly patients. Compared to venous sampling, DBS sampling is considered less painful and is found to be more time efficient in TDM of oral anticoagulants (5). Furthermore, DBS samples contain low biohazard risk during trans-port which allows easy sampling at home and less expensive transport by normal postal services (6, 7). In practice, this advantage can be used in different settings to centralize TDM analysis in a safe and easy manner.

Over the last decade, a rising number of DBS assays for different classes of drugs have been reported. So far, antidiabetics, immunosuppressants, analgesics, anti-HIV drugs, antihypertensive drugs, antimicrobial agents, anti-epileptics, and the iron chelator de-ferasirox have been included (8-25). Among the antidepressants, a DBS assay for analysis of IMP has previously been described as well as for the SSRIs fluoxetine, reboxetine, and paroxetine, and for the atypical antidepressant venlafaxine (26-28). However, a DBS as-say for simultaneous determination of TCAs and their main active metabolites has not been reported yet.

We developed an assay for simultaneous determination of ATP, NTP, IMP, and CMP as well as for the active metabolite of CMP and IMP, desmethyl-clomipramine (DCMP) and desipramine (DSP) respectively.

3.1.2. mEtHODs

materials

ATP, NTP, IMP, DSP, CMP, DCMP and promazine (PMZ), all > 99%, were purchased from Sigma Aldrich (Zwijndrecht, The Netherlands). PMZ was used as internal standard. Chemical structures are shown in figure 1. Methanol (Lichrosolv) ammonium acetate

50 Chapter 3.1

(p.a.), acetic acid 100% (p.a.) and trifluoro-acetic acid anhydride (TFAA, p.a.) were pur-chased from Merck (Darmstadt, Germany). Acetonitrile and purified water (ULC/MS grade) were purchased from Biosolve BV (Valkenswaard, The Netherlands). Whatman FTK DMPK-C blood sampling cards were purchased from GE Healthcare (Hoevelaken, The Netherlands). Blank EDTA whole blood was obtained from healthy volunteers for preparation of calibration and quality control samples.

Apparatus

The liquid chromatography-tandem mass spectrometer (LC-MS/MS) consisted of an Accela autosampler and Accela pump connected with a TSQ Quantum Acces tandem mass spectrometer with an electronspray ionization source, from Thermo Scientific (Thermo, Breda, The Netherlands). System control, data acquisition, and data processing were performed using Excalibur 2.0 software from Thermo Scientific (Thermo, Breda, The Netherlands).

Chromatography

Chromatographic separation was performed on a Hypurity Aquastar column, 50 × 2.1 mm, particle size 5 μm (Thermo, Breda, The Netherlands). The separation method was

Amitriptyline Nortriptyline

Imipramine

Clomipramine

Desipramine

Desmethylclomipramine

Promazine (IS)

N

S

NCH3

CH3

Figure 1. Chemical structures of ATP, NTP, IMP, DPM, CPM, DCPM, and PMZ.

Determination of tricyclic antidepressants in dried blood spots 51

similar to our previously described method for DBS analysis of venlafaxine (28). In short, mobile phase A consisted of 2 mL TFAA, 35 ml acetic acid 100%, and 5 g ammonium ac-etate in 1000 ml purified water (pH 3.5). Mobile phase B was purified water, and mobile phase C, acetonitrile. The mobile phase ratio was 5% A and 95% B for two minutes. At t = 2 minutes, the mobile phase ratio changed to 5% of A and 95% of C, in a linear slope over one minute. At t = 3 minutes, the mobile phase ratio reverted to the starting posi-tion, and the column was allowed to equilibrate for one minute (t = 4 min.). Including injection time, total runtime was 4.8 minutes. The flow rate was kept constant at 0.3 ml/min. The respective retention times of the TCAs and IS are given in table 1.

mass spectrometry

Ionization was achieved in the positive electrospray mode. Spray voltage was 3000 V. Sheath and auxiliary gas pressure were set to 40 and 10 (arbitrary units), respectively. Capillary temperature was 375 ºC, and the collision gas (argon) pressure was 1.5 mTorr. The scan modus was set to selective reaction monitoring (SRM) and the mass transitions which were monitored, with their respective lens and collision voltages, are shown in Table 1.

stock and working solutions

Stock solutions of the analytes were prepared by carefully weighting of the compounds which were individually dissolved in methanol to a concentration of 1.00 g/L. 0.5 ml aliquots from all 6 stock solutions were combined in a working solution and diluted with NaCl 0.9% to final concentrations of 10 mg/L of each analyte. The IS stock solution was prepared in methanol as well to a concentration of 0.50 g/L. To obtain the extraction solution, it was diluted with the extraction solvent to a final concentration of 40 µg/L.

table 1. LC-MS/MS conditions for TCAs.

Compound Retention time(min)

Parent mass (m/z)

Daughter mass (m/z)

tube lens Voltage (V)

Collision Energy (eV)

AtP 2.48 278.2 233.3 85 17

ntP 2.42 264.2 233.3 83 15

imP 2.43 281.2 86.3 54 16

DsP 2.35 267.2 72.4 51 15

CmP 2.55 315.2 227.2 59 38

DCmP 2.52 301.1 227.2 52 35

Pmz 2.42 285.1 86.3 56 16

52 Chapter 3.1

Calibration and Quality control samples

Calibration samples and QC samples were prepared independently from a different stock and working solution to control for errors during preparation. Blank EDTA whole blood was spiked to obtain calibration sample concentrations of 20, 50, 100, 200, 375, and 500 µg/L and QC sample concentrations of 20 µg/L (LLOQ), 40 µg/L (QC low), 250 µg/L (QC median), and 400 µg/L (QC high) levels. With a pipette, two droplets of the spiked blood with a total volume of 50 microliters were spotted onto DMPK-C cards. The cards were left to dry overnight at room temperature. The hematocrit (Htc) of the blood which was available from the healthy volunteer was 0.44 (L/L). This blood was used without adjustments of the Htc during the further validation.

sample preparation

During prior method optimization different extraction conditions were tested and it was found that an extraction fluid with a ratio of acetonitrile: methanol 1:3 provided the best extraction recovery of the analytes. The sample preparation method was similar to our previously described method for DBS analysis of venlafaxine (28). In detail, circles of six millimeters were punched out by hand with a DBS puncher (Harris Uni-Core, Whatman, GE Healthcare, Hoevelaken, The Netherlands) and placed into tubes. To the tubes, 250 µL of acetonitrile:methanol 1:3 containing the IS was added, tubes were vortexed briefly and after this, shaken for five minutes. After shaking, the tubes were centrifuged at 16,000 g for 5 minutes, the supernatant was transferred to 0.5 mL auto sampler vials, the vials were closed and 5 µL was injected into the LC-MS/MS by the auto sampler.

Validation

Validation was performed according to the guidelines of the FDA for bio-analytical method validation (29). The guidelines contain no specific regulations for the validations of DBS, however, it is appropriate to estimate the influence of Htc, blood spot volume, and the punch position for the validation of a DBS assay (6, 30). Therefore we analyzed influence of these variables on the analytical bias of the assay as well.

selectivity,matrix effects, and carry over

Selectivity was validated by analyzing six blank samples from different volunteers, which were spotted on DBS cards. Chromatograms of these six blanks were compared to a spiked DBS sample, which contained the LLOQ of all compounds. The method was considered selective if other components of the matrix did not overlap with the peaks from the compounds of interest or the IS. Limits for acceptance for the noise signal from the matrix with respect to the signal of the TCA at LLOQ, was a noise signal of <20% and with respect to the signal of the IS a noise signal of <5%.

Determination of tricyclic antidepressants in dried blood spots 53

To investigate the matrix effect, six post-extraction calibration curves were prepared and compared to calibration curves without matrix. Blank EDTA blood from the volun-teers was used to prepare 15 µL spots on pre-punched paper in a sample tube. After drying the spots were extracted with extraction solution lacking the IS. The supernatant was evaporated to dryness under a stream of nitrogen. The appropriate amounts of TCAs were added and the samples were evaporated to dryness again. Next, the extrac-tion solution containing the IS was added and samples were analyzed. Concentrations were calculated using the analyte area/IS area ratio. All individual slopes of the calibra-tion lines in the different matrices were compared to the slope of the calibration line obtained with the matrix free solution. The quotient of both slopes was calculated to estimate the overall bias resulting from matrix effects. Matrix effects not exceeding 10% are considered as acceptable. (12). The relative matrix effect was calculated as the CV of the six slopes, and this value should not exceed 3-4%, although a limit of 4-5% has been previously reported (31). Our samples were post-extraction spiked (12) which provided the disadvantage that the calculated relative matrix effects did not take into account possible differences in recovery.

Carry over was analyzed by injecting two QC high samples followed by three QC low samples. The difference between the first and the third QC low sample divided by the QC high concentration was calculated as the carry over.

linearity

Calibration samples were prepared as described under ‘sample preparation’. Samples were analyzed on three consecutive days. Calibration plots were made using the peak area ratio (analyte response/IS response) versus the theoretical analyte concentration with a weighting of 1/x. Linearity of the calibration plots from each day were analyzed with linear regression and lack of fit statistics in Excel.

within and between run accuracy and precision

Four concentrations of QC samples were prepared as described above and analyzed in five-fold in three runs (n=15), on consecutive days. Accuracy was estimated by calculat-ing the bias (%) from the theoretical concentration. Variation was analyzed using one-way analysis of variance (ANOVA) to estimate -accuracy and precision.

Absolute recovery

QC samples at the low, median, and high concentration levels were prepared as de-scribed under sample preparation. Samples were compared to blank blood spots, which were post-extraction spiked, as described under matrix effect. Samples were analyzed in five-fold and the average response of the QC samples was compared with that of the post-extraction spiked samples.

54 Chapter 3.1

stability

Long term stability and storage conditions were asses by analyzing six QC low and high samples which were stored in the freezer (−20°C) up to 30 days, in the refrigerator (2 to 8°C), and at room temperature (both up to 90 days). Samples were dried overnight and stored with a desiccant (silica) in a sealed plastic bag. QC samples were analyzed after seven days, 30 and 90 days of storage against a freshly prepared calibration line, which was dried overnight the day before the analyses. If deviations from the nominal concentration were more than 20%, stability was no longer assumed.

To assess stability in the auto sampler (20°C) a QC low and a QC high sample were injected every 4 hours during 48 hours (13 injections in total, including t = 0 h). Results for stability with respect to t = 0 were calculated (peak area ratio analyte/IS) after 12, 24, and 48 hours using linear regression analyses, If deviations were >15% stability was no longer assumed.

Effects of hematocrit

It is known that different Htc levels can influence spotted plasma volume as well as the distribution of the blood spot over the paper. Therefore influence of different Htc levels (range: 0.25 to 0.50) on the accuracy of the assay was tested (6). For calculation of the bias caused by a varying Htc, we chose to normalize the signal (bias = 0%), obtained in a blood matrix with a Htc of 0.45. (30) A Htc of 0.45 was considered close to the expected population mean (32, 33).

Different Htc concentrations were prepared by centrifuging EDTA whole blood to obtain a plasma and erythrocyte fraction. The plasma fraction was spiked with the TCAs to obtain samples with TCA concentrations at QC low and QC high levels. To the spiked plasma, variable volumes of the concentrated erythrocyte fraction and blank plasma were added to obtain a range of Htc concentrations. The reported Htc values are the expected Htc values based on the calculated fractions of erythrocytes and plasma.

Effects of spotted volume and punch position

The Influence of different spotted blood volumes ranging from 20-100 µL at QC low- and QC high- level were tested. Samples were prepared in three-fold. Results were normalized to a spotting volume of 50 µL, and bias with respect to this normalized value was determined.

To assess the influence of the punch position, five additional QC samples of the low and high concentrations were punched out at the center and the edge of the blood spot. Bias of the spots that were punched at the perimeter was calculated with respect to the measured concentrations of the spots punched at the center.

Determination of tricyclic antidepressants in dried blood spots 55

Clinical validation

Patients visiting the laboratory for routine blood analysis of a TCA for TDM purposes, were invited to participate in the clinical validation of the DBS method. If patients gave informed consent, an extra blood sample was collected by finger prick on a DBS card. The plasma samples were analyzed by a fully validated routine LC-MS/MS method, which has proven to be robust and reliable in external quality control programs. The correlation between plasma and DBS samples was assessed with linear regression analysis in Excel. Patient samples sometimes originated from the same patient, however, if this was the case samples were collected on different days. Approval for this study was obtained from the Medical Ethics Committee of Diaconessen Hospital Meppel.

3.1.3.REsults

matrix effects, selectivity and carry over

Visual inspection of the chromatograms indicated there was interference from the blank blood spots for DMCP (figure 2). For the other compounds, no interference was detected. The highest co-eluting peak was observed for DCMP, which was 26.1% of the peak area of the LLOQ. For the other compounds co-eluting peak area was < 20%. For the IS the highest co-eluting peak was 0.2%.

Matrix effects are summarized in table 2. For the active metabolites (NTP, DSP, and DCMP) a moderate negative matrix effect (ranging from −14.2 to −11.3%) and a relative matrix effect ranging from 5.3 to 7.5% was observed. For the parent compounds IMP and CMP, no matrix effects were detected, however for ATP a relative matrix effect of 4.6% was found

No carry over was observed (range: −1.32 to −0.24%).

linearity, recovery, stability, accuracy and precision

Results of linearity, recovery and stability tests are summarized in table 2. The assay was found to be linear from 20-500 µg/L (r2 > 0.990), and no significant lack of fit was found for any of the TCAs.

The recovery was high (> 80%) for all compounds. At the different storage conditions, we observed a positive bias (range: 23- 33%) for all compounds after seven days and one month of storage in the freezer. When stored at room temperature, all compounds were stable for three months, except for the QC high samples of DCMP (bias: −23.8%) which was stable up to one month. After three months of storage in the refrigerator all samples were found stable as well. All TCAs were stable in the auto sampler up to 12 hours. Up to 24 hours, all compounds were considered stable except for CMP.

None of the TCA’s displayed acceptable stability for longer periods as up to 48 hours in the autosampler.

56 Chapter 3.1

0

500

1000

1500

2000

2500

3000

3500

4000

0,0 0,4 0,8 1,21,7

2,12,5

2,93,3

3,7

Det

ecto

r res

pons

e (A

rbitr

ary

Uni

ts)

Time (min)

IMP blanco

DSP blanco

IMP LLOQ

DSP LLOQ

0

500

1000

1500

2000

2500

3000

3500

4000

0,0 0,4 0,8 1,2 1,72,1

2,52,9

3,33,7

Det

ecto

r Res

pons

e (A

rbitr

ary

Uni

ts)

Time (min)

CMP blank

DMCP blank

CMP LLOQ

DMCP LLOQ

0

500

1000

1500

2000

2500

3000

3500

4000

0,0 0,4 0,8 1,21,7

2,12,5

2,93,3

3,7

Det

ecto

r res

pons

e (A

rbitr

ary

Uni

ts)

Time (min)

ATP blanco

NTP Blanco

ATP LLOQ

NTP LLOQ

A

B

C

Figure 2. Chromatograms of LLOQ and blank sample from the volunteer which gave the most interference for ATP and NTP (A), IMP and DSP (B), CMP and DCMP (C).

Determination of tricyclic antidepressants in dried blood spots 57

Within-, between- and overall-run accuracy and precision are summarized in table 3. Preci-sion was within the limits of acceptance, except for the within- and overall-run precision of the LLOQ of CMP. Overall accuracy was within the limits of acceptance for all compounds.

Eff ects of hematocrit, spotted blood volume and punch position

Overall, the Htc eff ects were more pronounced in the QC low samples compared to the QC high samples. Eff ects of Htc on analytical bias are shown in fi gure 3. In the QC low samples, a low hematocrit (≤ 0.30 L/L) was associated with a >15% negative bias for all compounds. Parent compounds (ATP, IMP, CMP) were more sensitive for bias at a low Htc compared to their metabolites (NTP, DSP, DCMP). At higher Htc levels (0.35- 0.40 L/L) a bias > +/− 15% was observed for ATP, CMP, and DMCP.

The eff ects of the spotting volume on analytical bias for the QC low samples are shown in fi gure 4. Diff erent spot volumes induced bias for all compounds and a higher bias was found if spot volume was small (20 µL) or high (100µL). There was no trend observed in the eff ect of the spot volume on the analytical bias.

Punch position infl uenced the result. If the punch position was in the middle of the blood spot, the measured concentration was more accurate. Bias introduced in the QC low samples by punching at the edge of the spot ranged from 9.5 to 23.5%. For the QC high samples, analytical bias was smaller and ranged from 6.1 to 10.9%.

table 2. Validation results of matrix eff ects, specifi city, linearity, recovery and stability.

Validated parameter Results for respective compound

ATP NTP IMP DSP CMP DCMP

matrix eff ects [n=6]

Average bias (%) of slopes with respect to neat solution

−3.8 −14.2 −6.7 −13.5 −4.3 −11.3

Range % bias −10.9 to 2.3

−20.4 to −3.6

−12.7 to −2.7

−19.3 to −1.0

−11.2 to −0.5

−16.3 to −3.1

CV (%) of slopes(relative matrix eff ect)

4.6 7.0 4.2 7.5 4.1 5.3

Overall linearity [n=18]- r2

0.990 0.994 0.994 0.996 0.990 0.990

Recovery (%) (RsD %) [n=5]

- low 81 (14) 101(10) 83(10) 101(9) 83(8) 98 (20)

- medium 104 (14) 123(13) 98 (12) 115(13) 101 (12) 110. (19)

- High 97(5) 113 (6) 94 (5) 108(5) 95 (8) 112 (11)

Auto sampler stability (recovery %) [n=13]

- 12 h 107.0 106.1 106.7 103.1 109.8 106.8

- 24h 114.0 112.1 113.4 106.2 119.6 113.7

- 48 h 128.1 124.1 126.8 112.4 139.1 127.3

58 Chapter 3.1

Clinical validation

Due to limited routine TDM of ATP, only seven patient samples were available for analysis. For one of these samples a concentration below the LLOQ was found in both plasma as well as in DBS, leaving six remaining samples from four different patients for further analysis (figure 5A). For IMP, no samples were available due to lack of routine TDM samples of IMP. The metabolite, DSP is not on the Dutch market as a separate drug, therefore no samples were available for DSP as well. For the other compounds, 12 samples were analyzed from 11 different patients for NTP (figure 5A) and nine different patients for CMP and DCMP (figure 5B). A good correlation was found for ATP (r2 = 0.93), NTP (r2 = 0.92), and DCMP (r2 =0.91), however for CMP the correlation was much lower (r2 = 0.73).

table 3. Validation results of within, between as well as overall run accuracy and precision. n= 15, except for CMP (n= 14 due to IS error).

tCA AtP ntP imP DsP CmP DCmP limits

Repeatability (RsD %)

- llOQ 10.4 8.1 6.7 12.3 20.1 19.7

- low 10.3 6.5 5.9 9.6 14.1 12.7

- medium 6.6 5.1 6.4 5.2 7.7 8.0

- High 4.4 6.2 4.0 4.3 4.8 7.6

intermediate precision (RsD %)

LLOQ< 20%L, M, H< 15%

- llOQ 10.4 8.1 6.7 18.6 22.3 19.7

- low 10.3 6.5 5.9 10.2 14.1 13.4

- medium 8.2 5.1 8.3 5.5 7.9 8.0

- High 6.9 6.2 6.6 4.3 5.9 7.7

Accuracy within-run run 1, 2, and 3 (% bias)

- llOQ −3.2;−3.5; −7.5

−1.2;−2.2;−2.5

2.9;−1.1;3.4

−8.2;−1.6;21.5

−12.7;6.1;13.7

−10.0;−14.6;−21.7

- low −7.8;−6.1;−7.2

−2.7;−1.1;−0.1

−4.2;0.6;−0.7

−6.0;−3.5;4.5

4.7;4.1;15.3

3.3;−5.6;−10.3

- medium −2.1;9.0;0.1

3.3;6.0;5.8

−0.9;11.1;2.0

−0.7;5.1;1.3

2.9;7.6;11.3

−1.4;1.2;−2.6

- High −0.5;8.0;−3.0

6.4;3.9;2.3

−0.6;10.3;1.7

0.3;2.8;−0.7

2.0;7.1;10.6

3.6;3.3;−3.2

Accuracy between-run/overall (bias %)

LLOQ< 20%L, M, H< 15%

- llOQ −4.7 −2.0 1.7 3.9 5.2 −15.4

- low −7.1 −1.3 −1.4 −1.7 8.0 −4.2

- medium 2.3 5.0 4.1 1.9 7.3 −0.9

- High 1.5 4.2 3.8 0.8 6.6 1.2

Determination of tricyclic antidepressants in dried blood spots 59

3.1.4. DisCussiOn & COnClusiOn

An assay to determine concentrations of TCAs in DBS samples was developed and tested against the requirements in the FDA guideline. For ATP, NTP, IMP, and DSP requirements were met, however for CMP and DMCP, the LLOQ of 20 µg/L did not comply with the requirements concerning selectivity and precision. As a result of this validation, we therefore decided to adjust the LLOQ for these compounds to 40 µg/L. It should be noti-fied, that the higher LLOQ of 40 µg/L for DCMP has not been actually tested with respect to the requirements for selectivity, however based on our findings we assume that since the concentration is doubled from 20 to 40 µg/L, the found signal to noise ratio of 26.1% will reduce to a ratio of < 20%. In addition, by using the QC low (40 µg/L) samples of CMP and DMCP as the new LLOQ samples for determination of the precision, the experiment for the precision did no longer contain the 4 concentration levels subscribed by the guidelines. However, based on the results found at 20 µg/L, which were only slightly

-50

-40

-30

-20

-10

0

10

20

30

0,20 0,25 0,30 0,35 0,40 0,45 0,50 0,55

Ana

lytic

al b

ias:

QC

Low

(%)

Htc (L/L)

ATP NTP IMP DSP CMP DCMP

-50

-40

-30

-20

-10

0

10

20

30

0,20 0,25 0,30 0,35 0,40 0,45

Anal

ytic

al b

ias:

QC

Hig

h (%

)

Htc (L/L)

ATP NTP IMP DSP CMPA

B

Figure 3. Effects of Htc on analytical bias (%) at QC low (A) and QC high level (B). For visual purposes, error bars are only shown for (A) DSP and CMP, (B) ATP and IMP. n= 3 for each point, except for QC low Htc = 0.37 [n= 2], due to IS error.

60 Chapter 3.1

deviating for CMP (CV = 22.3%) and within limits for DCMP (CV= 19.7%), we consider this acceptable.

Moderate matrix effects were observed for the active metabolites which could be related to their chemical structure. All active metabolites have a free nitrogen atom which is not present in the parent compounds and IS. This makes the active metabolites more polar and therefore probably more sensitive to ion suppression with respect to the IS (34). Although we observed matrix effects during validation, these effects were considered acceptable, within the scope of TDM purposes.

Dried blood spots could at least be stored up to three months in the refrigerator, and up to one month at room temperature. Auto sampler stability of extracted samples was limited till 12 hour, due to increasing response ratio’s over time. This increase was caused by a decreasing signal from the IS combined with a slightly increasing signal of the analytes. Unfortunately, we were unable to find an adequate explanation for this phenomenon.

The bias caused by variation in Htc, blood spot volume and punch position was quan-tified. Although so far no limits for acceptance have been defined for these variables, it is reasonable to correct for Htc effects if bias exceeds 15% (35). At lower Htc levels, a

-40

-30

-20

-10

0

10

20

30

40

10 30 50 70 90 110

Anal

ytic

al b

ias

(%)

Spot volume (µL)

ATP NTP IMP DSP CMP DCMP

-40

-30

-20

-10

0

10

20

30

40

10 30 50 70 90 110

Anal

ytic

al b

ias

(%)

Spot volume (µL)

ATP NTP IMP DSP CMP DCMP

A

B

Figure 4. Effect of spotting volume on analytical bias (%) at (A) QC Low and (B) QC high . For visual pur-poses, error bars are only shown for (A) CMP and DCMP (B) CMP and IMP. n= 3 for each point .

Determination of tricyclic antidepressants in dried blood spots 61

negative bias was found for all compounds at QC low concentrations. This is assumed to be related to the lower viscosity of the blood at a lower Htc (30). Based on the ex-pected Htc in the average psychiatric patient population a > +/− 15% bias at a Htc of > 0.35 was considered unacceptable. Results suggested such biases could occur at low concentrations of ATP, CMP, and DMCP. Therefore corrections for Htc should be further investigated.

y = 1,0616x + 14,953 R² = 0,9354

y = 0,8117x - 0,718 R² = 0,9212

0

20

40

60

80

100

120

140

160

0 20 40 60 80 100 120 140 160

Plas

ma

(µg/

L)

DBS (µg/L)

ATP

NTP

y = 1,1706x + 14,234 R² = 0,8148

y = 0,7241x - 15,142 R² = 0,932

0

50

100

150

200

250

300

350

400

450

500

0 50 100 150 200 250 300 350 400 450 500

Plas

ma

(µg/

L)

DBS (µg/L

CMP

DCMP

A

B

Figure 5. Correlation between patients plasma and DBS samples for ATP and NTP (A), CMP and DCMP (B). N = 12, except for ATP where N = 6.

62 Chapter 3.1

Blood spot volume did also affect analytical bias, however there was no general trend in these effects for all TCAs. The observed biases are probably related to normal analytical variation. Effects of punch position on analytical bias was limited, however it was found to be more accurate to punch in the middle of the blood spot for all TCAs. If possible, we recommend punching in the middle of the blood spot.

Patient samples indicated a good correlation between plasma and DBS samples for ATP, NTP and DCMP. For CMP the correlation was less pronounced and a correction for Htc should be studied during further validation by measuring Htc in DBS as was recently demonstrated by Capiau et al. (36). Moreover, results suggested a proportional differ-ence in concentrations between plasma and DBS samples. To be able to translate DBS result to plasma values, a correction factor seems therefore indicated. In addition, the assumption of a linear relationship should be tested.

This is the first assay, which describes a DBS method for TDM of common TCAs. Only for IMP, Déglon et al. have previously described a DBS assay, but their report did not include a full validation of the assay (27). We investigated effects of Htc, spotting volume, and punch position, which were not reported before. Several assays for the simultaneous de-termination of TCAs, which make use of blood plasma obtained from venous puncturing are reported (37-39). Compared to these methods, our DBS assay is performing similar in terms of accuracy, and precision, however our LLOQ is somewhat higher. The total chro-matographic run time of our assay is comparable as well. We consider it advantageous that our method does not involve any pretreatment step like solid-phase extraction and is restricted to a few simple extraction steps. Nevertheless, additional bias from varying Htc, blood spot volume and punch position, would make venous puncturing more ac-curate for determination of TCAs in high precision pharmacokinetic research.

For TDM, this DBS assay is analytically validated and it appears a promising alternative sampling method in TCA treated patients. Compared to current methods, reliability of the determination of ATP, CMP, and DCMP might be decreased at lower concentrations, due to Htc effects. However, this assay still provides clinical valuable information, and it has many advantages for the patient as well as for sample logistics, as stated before. Based on the clinical needs at a certain setting, different sampling approaches might be considered. Further clinical validation of this assay is still ongoing to determine possible correction factors for conversion of DBS to plasma values as well as Htc corrections.

This assay is designed to be used for the determination of NTP in a Dutch multicenter trial (CYSCE Trial: ClinicalTrail.gov Identifier NCT01778907), in which the effect of CYP2D6 genotyping combined with TDM on time to reach adequate blood drug levels is investigated, in older patients. The possibility of taking a blood sample by a trained nurse or physician with a DBS, is of great advantage, because it offers more flexibility, which is often a problem in clinical trials.

Determination of tricyclic antidepressants in dried blood spots 63

3.1.5. REFEREnCEs

1. van Avendonk MJ, Hassink-Franke LJ, Terluin B, van Marwijk HW, Wiersma T, Burgers JS. Sum-marisation of the NHG practice guideline ‘anxiety’. Ned Tijdschr Geneeskd. 2012; 156(34): A4509.

2. van Avendonk M, van Weel-Baumgarten E, van der Weele G, Wiersma T, Burgers JS. Summary of the dutch college of general practitioners’ practice guideline ‘depression’. Ned Tijdschr Geneeskd. 2012; 156(38): A5101.

3. Van Brunt N. The clinical utility of tricyclic antidepressant blood levels: A review of the literature. Ther Drug Monit. 1983; 5(1): 1-10.

4. GUTHRIE R, SUSI A. A simple phenylalanine method for detecting phenylketonuria in large popu-lations of newborn infants. Pediatrics. 1963 Sep; 32: 338-43.

5. Woods K, Douketis JD, Schnurr T, Kinnon K, Powers P, Crowther MA. Patient preferences for capil-lary vs. venous INR determination in an anticoagulation clinic: A randomized controlled trial. Thromb Res. 2004; 114(3): 161-5.

6. Edelbroek PM, van der Heijden J, Stolk LM. Dried blood spot methods in therapeutic drug moni-toring: Methods, assays, and pitfalls. Ther Drug Monit. 2009 Jun; 31(3): 327-36.

7. Spooner N, Ramakrishnan Y, Barfield M, Dewit O, Miller S. Use of DBS sample collection to determine circulating drug concentrations in clinical trials: Practicalities and considerations. Bioanalysis. 2010 Aug; 2(8): 1515-22.

8. Aburuz S, Millership J, McElnay J. Dried blood spot liquid chromatography assay for therapeutic drug monitoring of metformin. J Chromatogr B Analyt Technol Biomed Life Sci. 2006 Mar 7; 832(2): 202-7.

9. van der Heijden J, de Beer Y, Hoogtanders K, Christiaans M, de Jong GJ, Neef C, et al. Therapeutic drug monitoring of everolimus using the dried blood spot method in combination with liquid chromatography-mass spectrometry. J Pharm Biomed Anal. 2009 Nov 1; 50(4): 664-70.

10. Wilhelm AJ, den Burger JC, Chahbouni A, Vos RM, Sinjewel A. Analysis of mycophenolic acid in dried blood spots using reversed phase high performance liquid chromatography. J Chromatogr B Analyt Technol Biomed Life Sci. 2009 Nov 15; 877(30): 3916-9.

11. Wilhelm AJ, den Burger JC, Vos RM, Chahbouni A, Sinjewel A. Analysis of cyclosporin A in dried blood spots using liquid chromatography tandem mass spectrometry. J Chromatogr B Analyt Technol Biomed Life Sci. 2009 May 15; 877(14-15): 1595-8.

12. den Burger JC, Wilhelm AJ, Chahbouni A, Vos RM, Sinjewel A, Swart EL. Analysis of cyclosporin A, tacrolimus, sirolimus, and everolimus in dried blood spot samples using liquid chromatography tandem mass spectrometry. Anal Bioanal Chem. 2012 Oct; 404(6-7): 1803-11.

13. Rao RN, Maurya PK, Ramesh M, Srinivas R, Agwane SB. Development of a validated high-through-put LC-ESI-MS method for determination of sirolimus on dried blood spots. Biomed Chromatogr. 2010 Dec; 24(12): 1356-64.

14. Oliveira EJ, Watson DG, Morton NS. A simple microanalytical technique for the determination of paracetamol and its main metabolites in blood spots. J Pharm Biomed Anal. 2002 Jul 31; 29(5): 803-9.

15. Clavijo CF, Hoffman KL, Thomas JJ, Carvalho B, Chu LF, Drover DR, et al. A sensitive assay for the quantification of morphine and its active metabolites in human plasma and dried blood spots using high-performance liquid chromatography-tandem mass spectrometry. Anal Bioanal Chem. 2011 May; 400(3): 715-28.

64 Chapter 3.1

16. Kromdijk W, Mulder JW, Rosing H, Smit PM, Beijnen JH, Huitema AD. Use of dried blood spots for the determination of plasma concentrations of nevirapine and efavirenz. J Antimicrob Che-mother. 2012 May; 67(5): 1211-6.

17. Hoffman JT, Rossi SS, Espina-Quinto R, Letendre S, Capparelli EV. Determination of efavirenz in human dried blood spots by reversed-phase high-performance liquid chromatography with UV detection. Ther Drug Monit. 2013 Apr; 35(2): 203-8.

18. Rao RN, Raju SS, Vali RM, Sankar GG. Liquid chromatography-mass spectrometric determination of losartan and its active metabolite on dried blood spots. J Chromatogr B Analyt Technol Biomed Life Sci. 2012 Aug 1; 902: 47-54.

19. Della Bona ML, Malvagia S, Villanelli F, Giocaliere E, Ombrone D, Funghini S, et al. A rapid liquid chromatography tandem mass spectrometry-based method for measuring propranolol on dried blood spots. J Pharm Biomed Anal. 2013 May 5; 78-79: 34-8.

20. Reddy TM, Tama CI, Hayes RN. A dried blood spots technique based LC-MS/MS method for the analysis of posaconazole in human whole blood samples. J Chromatogr B Analyt Technol Biomed Life Sci. 2011 Nov 15; 879(30): 3626-38.

21. Vu DH, Koster RA, Alffenaar JW, Brouwers JR, Uges DR. Determination of moxifloxacin in dried blood spots using LC-MS/MS and the impact of the hematocrit and blood volume. J Chromatogr B Analyt Technol Biomed Life Sci. 2011 May 1; 879(15-16): 1063-70.

22. Vu DH, Bolhuis MS, Koster RA, Greijdanus B, de Lange WC, van Altena R, et al. Dried blood spot analysis for therapeutic drug monitoring of linezolid in patients with multidrug-resistant tuber-culosis. Antimicrob Agents Chemother. 2012 Nov; 56(11): 5758-63.

23. Vnucec Popov T, Maricic LC, Prosen H, Voncina DB. Development and validation of dried blood spots technique for quantitative determination of topiramate using liquid chromatography-tandem mass spectrometry. Biomed Chromatogr. 2013 Mar 26.

24. Shah NM, Hawwa AF, Millership JS, Collier PS, McElnay JC. A simple bioanalytical method for the quantification of antiepileptic drugs in dried blood spots. J Chromatogr B Analyt Technol Biomed Life Sci. 2013 Apr 1; 923-924: 65-73.

25. Nirogi R, Ajjala DR, Kandikere V, Aleti R, Srikakolapu S, Vurimindi H. Dried blood spot analysis of an iron chelator--deferasirox and its potential application to therapeutic drug monitoring. J Chromatogr B Analyt Technol Biomed Life Sci. 2012 Oct 15; 907: 65-73.

26. Deglon J, Lauer E, Thomas A, Mangin P, Staub C. Use of the dried blood spot sampling process coupled with fast gas chromatography and negative-ion chemical ionization tandem mass spectrometry: Application to fluoxetine, norfluoxetine, reboxetine, and paroxetine analysis. Anal Bioanal Chem. 2010 Apr; 396(7): 2523-32.

27. Deglon J, Thomas A, Cataldo A, Mangin P, Staub C. On-line desorption of dried blood spot: A novel approach for the direct LC/MS analysis of micro-whole blood samples. J Pharm Biomed Anal. 2009 May 1; 49(4): 1034-9.

28. Berm EJ, Brummel-Mulder E, Paardekooper J, Hak E, Wilffert B, Maring JG. Determination of venla-faxine and O-desmethylvenlafaxine in dried blood spots for TDM purposes, using LC-MS/MS. Anal Bioanal Chem. 2014 Apr; 406(9-10): 2349-53.

29. US food and drug administration, bioanalytical method validation [Internet].; 2001. Available from: http://www.fda.gov/downloads/Drugs/Guidances/ucm070107.pdf.

30. Li W, Tse FL. Dried blood spot sampling in combination with LC-MS/MS for quantitative analysis of small molecules. Biomed Chromatogr. 2010 Jan; 24(1): 49-65.

31. Matuszewski BK. Standard line slopes as a measure of a relative matrix effect in quantitative HPLC-MS bioanalysis. J Chromatogr B Analyt Technol Biomed Life Sci. 2006 Jan 18; 830(2): 293-300.

Determination of tricyclic antidepressants in dried blood spots 65

32. Maes M, Van de Vyvere J, Vandoolaeghe E, Bril T, Demedts P, Wauters A, et al. Alterations in iron metabolism and the erythron in major depression: Further evidence for a chronic inflammatory process. J Affect Disord. 1996 Sep 9; 40(1-2): 23-33.

33. Alves de Rezende CH, Coelho LM, Oliveira LM, Penha Silva N. Dependence of the geriatric depres-sion scores on age, nutritional status, and haematologic variables in elderly institutionalized patients. J Nutr Health Aging. 2009 Aug; 13(7): 617-21.

34. Bonfiglio R, King RC, Olah TV, Merkle K. The effects of sample preparation methods on the vari-ability of the electrospray ionization response for model drug compounds. Rapid Commun Mass Spectrom. 1999 Jun; 13(12): 1175-85.

35. Timmerman P, White S, Cobb Z, de Vries R, Thomas E, van Baar B. Update of the EBF recommenda-tion for the use of DBS in regulated bioanalysis integrating the conclusions from the EBF DBS-microsampling consortium. Bioanalysis. 2013 Jul 5.

36. Capiau S, Stove VV, Lambert WE, Stove CP. Prediction of the hematocrit of dried blood spots via potassium measurement on a routine clinical chemistry analyzer. Anal Chem. 2013 Jan 2; 85(1): 404-10.

37. Santos-Neto AJ, Bergquist J, Lancas FM, Sjoberg PJ. Simultaneous analysis of five antidepressant drugs using direct injection of biofluids in a capillary restricted-access media-liquid chromatog-raphy-tandem mass spectrometry system. J Chromatogr A. 2008 May 2; 1189(1-2): 514-22.

38. Hasselstrom J. Quantification of antidepressants and antipsychotics in human serum by precipi-tation and ultra high pressure liquid chromatography-tandem mass spectrometry. J Chromatogr B Analyt Technol Biomed Life Sci. 2011 Jan 1; 879(1): 123-8.

39. de Castro A, Ramirez Fernandez Mdel M, Laloup M, Samyn N, De Boeck G, Wood M, et al. High-throughput on-line solid-phase extraction-liquid chromatography-tandem mass spectrometry method for the simultaneous analysis of 14 antidepressants and their metabolites in plasma. J Chromatogr A. 2007 Aug 10; 1160(1-2): 3-12.

3.2Determination of venlafaxine and O-desmethylvenlafaxine in dried blood

spots for TDM purposes, using LC-MS/MS

Elizabeth J.J. Berm, Ellen Brummel-Mulder, Jolanda Paardekooper, Eelko Hak, Bob Wilffert, Jan G. Maring

Analytical Bioanalytical Chemistry. 2014. April; 406(9-10):2349-53

68 Chapter 3.2

ABstRACt

introduction: Dried blood spot (DBS) sampling and quantitative analyses of many current TDM guided drugs is advantageous because of the minimal invasive sampling strategy. Here, a fast and robust liquid chromatography-tandem mass spectrometry method was developed and analytically validated for simultaneous determination of venlafaxine (VEN) and O-desmethylvenlafaxine (ODV) in DBS.methods: Six mm circles were punched out from DBS collected on Whatman DMPK-C paper and the DBS was extracted with acetonitrile: methanol 1:3. The total run time was 4.8 min.Results: The assay was linear in the range 20-1000 µg/L for both VEN and ODV. Assay accuracy and precision was well within limits of acceptation (LLOQ = 20 µg/L). Normal hematocrit concentrations (0.30-0.50) did not influence the results neither did a normal spot volume (40-80 µL). Punch position at the perimeter instead of the center of the blood spot gave a bias ranging from 2.4-10.4%. Correlation between plasma and spiked DBS samples was high. The concentrations found in spiked DBS samples were higher than those in plasma indicating that a conversion factor for translation of DBS to plasma values is needed.Conclusion: This analytically validated method is suitable for determination of VEN and ODV in DBS and applicable for TDM. The method will be used for TDM of VEN in the Dutch CYSCE multicenter trial (NCT01778907).

Determination of venlafaxine in dried blood spots 69

3.2.1. intRODuCtiOn

Venlafaxine (VEN) is a commonly prescribed antidepressant which is also prescribed and registered for treatment of anxiety and panic disorders. Like several other antidepres-sants, VEN is metabolized by the liver enzyme Cytochrome P450 2D6 and 2C19 into sev-eral metabolites including its main active metabolite O-desmethylvenlafaxine (ODV). The activity of these enzymes is influenced by genetic differences which can cause large inter-individual variation in the drug metabolism and clearance. Additionally, espe-cially in the elderly, plasma concentrations of VEN and ODV may increase substantially, because of aging and the use of CYP 2D6 inhibiting co-medication. Therapeutic drug monitoring is therefore advised.

Several methods for simultaneous determination of VEN and OVD in plasma by liquid chromatography-tandem mass spectrometry (LC-MS/MS) have been reported [1-4]. However, no dried blood spot (DSB) assay has previously been described. DBS sampling is beneficial, because it is a minimal invasive sampling strategy. The blood is obtained by a fingerprick instead of venous puncturing. Additionally, DBS samples can be sent to a laboratory by standard postal service, because the dried sampling cards entail no biohazard risks [5]. These advantages allow patients or their caregivers to collect a blood sample at home. For depressed or other mentally ill patients this can be an important improvement.

Here we report on a novel, validated, and simple DBS assay, for TDM of VEN and ODV, using LC-MS/MS. Method validation included estimation of the correlation between spiked plasma and spiked DBS samples.

3.2.2. mAtERiAl AnD mEtHODs

Chemicals and reagents

VEN, ODV, both >99%, were purchased from Wyeth pharmaceuticals (Hoofddorp, the Netherlands) Promazine (IS) was purchased from Sigma Aldrich (Zwijndrecht, the Neth-erlands). Methanol (Lichrosolv), ammonium acetate (p.a.), acetic acid 100% (p.a.) and trifluoro-acetic acid anhydride (TFAA, GC grade.) were purchased from Merck (Darm-stadt, Germany). Acetonitrile and purified water (ULC/MS grade) were purchased from Biosolve BV (Valkenswaard, the Netherlands). Whatman FTK DMPK-C blood sampling cards were purchased from GE Healthcare (Hoevelaken, the Netherlands). Blank EDTA whole blood was obtained from healthy volunteers. The hematocrit (Htc) of this blood was 0.41.

70 Chapter 3.2

instrumentation

The LC-MS/MS system consisted of an Accela autosampler and UHPLC Accela pump con-nected with a TSQ Quantum Acces tandem mass spectrometer with an electronspray ionization source, obtained from Thermo Scientific (Thermo, Breda The Netherlands). System control, data acquisition, and data processing were performed by Excalibur 2.0 software.

Chromatography and mass spectrometry conditions

A Hypurity Aquastar column, 50×2.1 mm, particle size 5 μm (Thermo, Breda, The Netherlands) was used for chromatographic separation. Mobile phase A consisted of 2 mL TFAA, 35 mL acetic acid 100%, and 5 g ammonium acetate in 1000 mL purified water (pH 3.5). Mobile phase B, was purified water, and mobile phase C consisted out of acetonitrile. The mobile phase ratio was 5% of A and 95% of B during two minutes. After two minutes, the mobile phase ratio changed to 5% of A and 95% of C in a linear slope over one minute. After three minutes, the mobile phase ratio returned to the start-ing position, and the column was allowed to equilibrate during one minute. The total runtime was 4.8 minutes. The flow rate was kept constant at 0.3 ml/min. Retention times of VEN, ODV, and IS are given in table 1. No carry over was observed for either VEN or ODV (−0.18, −0.72%).

Ionization was achieved in the positive electrospray mode. Spray voltage was 3000 V. Sheath and auxiliary gas pressure were set to respectively, 40 and 10 (arbitrary units). Capillary temperature was 375 ºC, and the collision gas (argon) pressure was 1.5 mTorr. The scan modus was set to selective reaction monitoring (SRM) and the different mass transitions, with their respective lens and collision voltages, which were monitored, are showed in Table 1.

sample preparation

Stock solutions of VEN and ODV with a concentration of 1g/L were prepared in metha-nol. From both stock solutions 2.5 mL aliquots were mixed and diluted to a working solution with NaCl 0.9%. Blank EDTA whole blood was spiked with the working solution and completed to the same end volumes with NaCl 0.9%, to obtain calibration sample concentrations of 20, 50, 100, 250, 500, and 1000 µg/L respectively. QC samples were

table 1. LC-MS/MS conditions

Compound Retention time(min)

Parent mass (m/z)

Daughter mass (m/z)

tube lens Voltage (V)

Collision Energy (eV)

VEn 2.10 278.2 121.2 75 28

ODV 1.84 264.2 107.3 74 31

Pmz (is) 2.42 285.1 86.4 56 16

Determination of venlafaxine in dried blood spots 71

spiked to obtain concentrations of respectively, 20 µg/L= LLOQ, 80 µg/L= QC Low, 300 µg/L= QC medium, and 750 µg/L= QC high. Fifty microliter of the spiked blood was spot-ted on DMK-C cards and dried overnight at 20- 25 ºC. With a pipet one bloodspot was carefully made by spotting two small droplets of blood on to the paper. Dried DBS cards were stored at 2-8°C together with a desiccant (silica) bag in a sealed plastic bag. VEN is known to be stable in aqueous solutions and biological matrixes for 57 days at 20°C [6]. Therefore stability of a dry DBS cards at 2-8°C was assumed to be at least 2 months. During method development, DBS punches of six millimetre were extracted with dif-ferent ratios of acetonitrile and methanol. A ratio of 1:3 gave the best recovery results and was therefore used as the extraction fluid. For extraction, 250 µL of extraction fluid containing the IS (final concentration 40 µg/L) was added followed by briefly vortexing and five minutes of shaking. The supernatant was transferred into vials and a volume of 5 µL was injected into the LC-MS/MS.

method validation

Validation was performed according to the guidelines of the FDA for bioanalytical meth-od validation. To assess matrix effects, blank blood spots from six different volunteers were extracted. Concentrations of VEN and ODV equal to the calibration curve were prepared in the blank extracts and analyzed. The individual slopes of the analyte/IS ratio calibration lines from the different matrixes were compared to the slope of a calibra-tion line from the matrix free standard solution. The relative difference between slopes was calculated as the matrix effect. If these differences were <10% no matrix effect was assumed. Additionally, the influence of Htc (ranging from 0.25-0.50 L/L), blood spot volume (ranging from 20-100 µL), and the punch position of the DBS on the assay was validated [7, 8]. Depending on the matrix which is examined, measured drug concentra-tions may differ due to variable binding to cells and proteins [9]. To assess whether there were any differences between plasma and DBS concentrations, 15 samples with spiked concentrations (ranging from 20-1100 µg/L) of VEN or ODV were analyzed by our plasma and DBS assay. Correlation between the two different methods was analyzed by Pear-son’s correlation coefficient and Passing-Bablok regression. Analysis were performed in XLSTAT 2013.4.05.

3.2.3. REsults

Six chromatograms from blank DBS of volunteers were visually inspected and no interfer-ing peaks were observed at the ODV and VEN positions (figure 1) . The highest co-eluting peaks which were found were 1.7, 2.0, and 0.2% for VEN, ODV, and IS, respectively. Matrix interference caused a substantial positive bias for both VEN (14.7%) and ODV (18.4%).

72 Chapter 3.2

However, bias was consistent between the individual samples for both VEN (CV: 3.8%) and ODV (CV 5.6%) . The assay was found linear over the tested concentration range (r-squared > 0.992), no significant lack of fit was found. Recovery of ODV for the QC low samples was 67%, which is somewhat low. Recovery of all other QC samples was > 80%. The stability of VEN and ODV was not affected in the auto sampler up to at least 48 h. After a period of 6 months storage at 2-8 degrees Celsius, no decline of QC samples was observed. The accuracy and precision was high and at all QC levels the CV and the bias was well within the limits of the FDA (table 2).

Analytical bias related to changes in Hct is shown in figure 2. A decrease in Htc was associated with a negative bias for both VEN and ODV. A Htc of ≤ 0.25 resulted in ≥ 15% bias for both VEN and ODV. The punch position appeared to influence the measured concentration. If the punch was made at the perimeter measured concentrations were higher. Analytical bias of the punches at the perimeter was 8.5 and 4.8% (ratio center/parameter: 0.92 and 0.95) for the QC low and high levels of VEN, respectively. For ODV this was 10.4 and 2.4% (ratio center/parameter: 0.91 and 0.98) at QC low and high level, respectively. The punches from the center were closest to the theoretical concentra-tions. Variation in spotting volumes ranging from 20-100 µL resulted in minor biases (range: VEN: −11.4 - 6.4%, ODV: −14.2 - 4.7%). Highest biases were observed for the QC low samples at a spotting volume of either 20 or 100 µL. Linear correlation coefficients between plasma and spiked DBS samples were high (VEN r2: 0.988 ; ODV r2: 0.984). With Passing and Bablok regression, significant proportional biases were found for both VEN and ODV indicating that concentrations found in DBS samples are higher compared to those from plasma samples. The slopes of the regression lines were 0.83 (CI: 0.79 to 0.93) and 0.76 (CI: 0.72 to 0.83) for VEN and ODV, respectively. The intercepts were not

Figure 1. Chromatograms of LLOQ and one of the six blanc DBS from a volunteer

Determination of venlafaxine in dried blood spots 73

significant. These results suggest a conversion factor is needed to translate results from DBS to plasma values. This factor is currently investigated as part of the clinical valida-tion procedure.

3.2.4. DisCussiOn & COnClusiOn

An novel assay for determination of VEN and ODV in DBS was validated and found to be well within the limits of acceptance. Additionally, DBS specific aspects like influence of Htc, spot volume and punch position were assessed and the relation between plasma and DBS levels was established.

table 2. Validation results of within-, between-, and overall-accuracy and precision. N=15, except for VEN QC medium, due to a LC-MS/MS error (sample not injected) [n=14].

VEn ODV limits FDA

Within-run: precision (CV%) LLOQ< 20%L, M, H< 15%

llOQ 12.0 10.8

low 5.8 4.9

medium 5.2 5.3

High 4.9 6.3

Between-run: precision (CV%)

llOQ 0.0 0.0

low 1.9 4.6

medium 4.7 1.3

High 0.0 0.0

Overall precision (CV%)

llOQ 12.0 10.8

low 6.1 6.8

medium 7.0 5.4

High 4.9 6.3

Within-run accuracy run 1, 2 and 3 (bias %)

llOQ 9.9; 7.2; 4.5 7.4; 7.7; 1.8

low 5.1; −0.4; −0.7 7.0, −3.4; 1.5

medium −3.6; −2.2; 5.9 −1.4; −1.1; 3.5

High 0.1; −2.5; 1.1 −0.1; 0.2; 0.0

Between-run accuracy (bias %)

llOQ 7.2 5.6

low 1.3 1.7

medium 0.2 0.4

High −0.4 0.0

74 Chapter 3.2

Considerable matrix interference was observed which might indicate positive biased results due to ion enhancement. However, since the calibration curves are prepared as DBS and a low intraindividual CV was found in QC DBS-samples prepared from blood of six different volunteers, no additional correction is needed. Another consideration can be made about the internal standard promazine . Promazine has no marketing autho-rization on the Dutch market, however it does in some other countries, therefore one might consider using an alternative internal standard. Last, the influence of shipment conditions such as low (−20°C) or high (40°C) temperatures on the DBS stability has not yet been investigated.

-30

-20

-10

0

10

20

30

0,20 0,25 0,30 0,35 0,40 0,45 0,50 0,55

Anal

ytic

al b

ias

(%)

Htc (%)

VEN QC low VEN QC high

-30

-20

-10

0

10

20

30

0,20 0,25 0,30 0,35 0,40 0,45 0,50 0,55

Ana

lytic

al b

ias

(%)

Htc (%)

ODV QC low ODV QC high

A

B

Figure 2. Effects of Htc on analytical bias (%) of VEN (A) and ODV (B) (n=3 for each point, total n=21).

Determination of venlafaxine in dried blood spots 75

We assessed Htc effects and found a small negative bias associated with a lower Htc. This negative relationship was also found in other studies, and is thought to be caused by the lower viscosity of blood at a lower Htc [7]. Our results showed that a correction for Htc is needed if Hct < 0.25, but since such low Hct values are very unlikely in the average psychiatric patient population, no correction for Htc seems indicated. Moderate varia-tion in spotting volume did not influence analytical bias. This is beneficial for the use of the method in clinical practice. The same holds for the punch position. Nevertheless, a maximum of 10% bias was observed when punching at the perimeter of the spot and punching at the center of the spot is therefore advised.

A good correlation between the plasma and DBS samples was found with a consistent higher concentration of VEN and ODV in DBS samples compared to plasma samples. Although not reported for VEN or ODV, these differences could be caused by binding of VEN and ODV to red blood cells which is known for other drugs, like for example tricyclic antidepressants (TCA) [9] and confirmed by our finding concerning TCAs as well (results not published yet). Due to the good correlation between our plasma and DBS method (in vitro), a correction factor to convert DBS values to plasma can probably be used, however still needs to be established by clinical validation results.

In conclusion, we analytically validated a new minimal invasive DBS assay, for TDM purposes. Matrix interference was observed, however consistent between the matrixes and the effect could be compensated by using calibration samples prepared as DBS. Ef-fects from Htc, punch position and spot volume revealed acceptable bias. Furthermore, comparison between plasma and DBS samples revealed that higher concentrations of VEN and ODV were measured in DBS samples, which will be further studied as part of the clinical validation of the assay. Currently, this method is used for TDM in the Dutch CYSCE multicenter trial (ClincialTrail.gov Identifier: NCT01778907) in which the effect of CYP2D6 genotyping combined with TDM on time to reach adequate blood drug levels is investigated in depressed elderly patients.

76 Chapter 3.2

3.2.5. REFEREnCEs

1. Bhatt J, Jangid A, Venkatesh G, Subbaiah G, Singh S (2005) Liquid chromatography-tandem mass spectrometry (LC-MS-MS) method for simultaneous determination of venlafaxine and its active metabolite O-desmethyl venlafaxine in human plasma. J Chromatogr B Analyt Technol Biomed Life Sci 829: 75-81.

2. Liu W, Cai HL, Li HD (2007) High performance liquid chromatography-electrospray ionization mass spectrometry (HPLC-MS/ESI) method for simultaneous determination of venlafaxine and its three metabolites in human plasma. J Chromatogr B Analyt Technol Biomed Life Sci 850: 405-411.

3. Patel BN, Sharma N, Sanyal M, Shrivastav PS (2008) Liquid chromatography tandem mass spec-trometry assay for the simultaneous determination of venlafaxine and O-desmethylvenlafaxine in human plasma and its application to a bioequivalence study. J Pharm Biomed Anal 2008 Jul 15; 47(3): 603-611.

4. Rajasekhar D, Kumar IJ, Venkateswarlu P (2009) Rapid high-performance liquid chromatography-tandem mass spectrometry method for simultaneous measurement of venlafaxine and O-desmethylvenlafaxine in human plasma and its application in comparative bioavailability study. Biomed Chromatogr 23: 1300-1307.

5. Edelbroek PM, van der Heijden J, Stolk LM (2009) Dried blood spot methods in therapeutic drug monitoring: methods, assays, and pitfalls. Ther Drug Monit 31: 327-336.

6. Butzbach DM, Stockham PC, Kobus HJ, Sims N, Byard RW, Lokan RJ, Walker G (2013) Stability of serotonin-selective antidepressants in sterile and decomposing liver tissue. J Forensic Sci 58: 117-125.

7. Li W, Tse FL (2010) Dried blood spot sampling in combination with LC-MS/MS for quantitative analysis of small molecules. Biomed Chromatogr 24: 49-65.

8. Timmerman P, White S, Cobb Z, de Vries R, Thomas E, van Baar B (2013) Update of the EBF recom-mendation for the use of DBS in regulated bioanalysis integrating the conclusions from the EBF DBS-microsampling consortium. Bioanalysis 5: 2129-2136

9. Hinderling PH (1997) Red blood cells: a neglected compartment in pharmacokinetics and phar-macodynamics. Pharmacol Rev 49: 279-295.

3.3A clinical validation study for application of dried blood spots in therapeutic drug

monitoring of antidepressants

Elizabeth J.J. Berm, Blessing Odigie, Maarten J. Bijlsma, Bob Wilffert, Daan J. Touw, Jan G. Maring.

Bioanalysis, 2016, March; 8 (5):413-24.

80 Chapter 3.3

ABstRACt

introduction: A bridging study of plasma and DBS concentrations for therapeutic drug monitoring of antidepressants was performed.methods: Potassium based hematocrit analysis was included. In addition, we defined acceptance criteria based on the differences between individual data points of plasma and DBS concentrations. These criteria were applied to test acceptability of error found in predicted nortriptyline plasma concentrations.Results: Potassium based hematocrit predicted a negative bias for DBS concentrations of amitriptyline, but not for the other compounds. To predict plasma concentrations of antidepressants based on DBS concentrations, a factor of 0.8, 0.65, 0.84, and 0.78 was found for nortriptyline, desmethylclomipramine, venlafaxine, and desmethylvenlafax-ine, respectively.Conclusion: Application of the factor and newly formulated acceptance criteria demon-strated prediction of nortriptyline plasma concentrations based on DBS concentrations.

Clinical validation of dried blood spot analysis 81

3.3.1. intRODuCtiOn

Dried blood spot (DBS) analysis is increasingly described as an alternative sampling method for therapeutic drug monitoring (TDM) (1). This is due to its various advantages, such as patient comfort, more flexibility, simplicity and the lack of biohazard risk of send-ing samples by post (2). Although there are many reports of validated DBS assays, there are few of clinical validations which hamper the application of DBS in clinical practice. For a clinical validation and implementation of a DBS assay, there are issues that need to be addressed.

Firstly, DBS concentrations mirror whole blood concentrations. Therefore, reporting of DBS concentrations is of limited value for clinicians, because most, if not all, thera-peutic reference concentrations are determined in plasma or in serum. To overcome this problem, paired analysis of DBS and plasma samples in patients is needed. With such an approach, the possibility of reporting calculated plasma values based on DBS values is possible and reference values for plasma can be used.

Ideally, the relation between DBS and plasma concentrations is constant. However, there are different aspects in DBS analysis which can influence this relation. An impor-tant aspect is the variation in hematocrit (hct) between patients. The amount of hct influences the viscosity of blood and thereby the spreading of a blood drop on the DBS card. Higher hct values will lead to smaller and more concentrated spots (3, 4). Varia-tion in hct values also affect sample homogeneity, matrix effect, the blood-to-plasma concentration ratio of an analyte, and recovery (5-7). In addition, the influence of hct can be dependent on the type of cards on which the blood spots are collected (8). Thus, deviating hct values may greatly influence analytical results, leading to an unknown fac-tor of uncertainty (9). Various studies have stated that despite the many advantages of DBS, its implementation in TDM is mainly limited by hct effects (10, 11). Recently, Capiau et al. demonstrated that potassium measurement in DBS could serve as a predictor of the hct in venous DBS (11). As such, a possibility to determine hct in patient capillary DBS samples became available to study hct effects in patient samples.

Another issue is the evaluation of the outcomes from clinical validation studies. Pearson’s correlation coefficient could be used to describe the relation between two methods, but there is no clear interpretation available for a “good”, “low”, or “insufficient” correlation coefficient. Moreover, regression methods which either take error in both the (x) and (y) values into account, or make less error distributional assumptions on the (x) and (y) values, are more suitable for analysis of method comparisons (12). Deming (weighted) regression and Passing-Bablok (PB) regression are such regression methods and they could be supplemented with Bland Altman plots (12-16). Although these techniques give a good indication of the average comparability of two methods, they provide no information about deviations on an individual level which might still lead

82 Chapter 3.3

to unwanted errors. This problem is illustrated by Vu et al., who found DBS and plasma analysis of desacetylrifampicin were comparable according to Deming regression, however, according to the authors, correlation between the assay’s was low (r2= 0.69) (17). Therefore, criteria based on single data points would be useful in addition to the previous mentioned techniques. In another clinical validation study de Wit et al., applied a limit of ≤ 25% bias, for single data points. This limit was based on available dosage forms for dose adaptations (18). Although this is a clear limit, generalizability to other compounds seems a problem. Another criterion that can be considered is the proportion of different clinical interpretations based on one method versus the other (18, 19). This comparison gives valuable information, but it is can be biased by a lack of heterogeneity among patient samples. For example, when more patients with concentrations around the lower or upper limit of the reference range are selected, more differences will be observed. Based on the mentioned limitations, there is a need for application of simple and uniform acceptance criteria for validation of bridging studies of DBS and plasma concentrations based on individual data points.

Previously we developed two DBS assays for the analysis of antidepressants. One for determination of amitriptyline (ATP), nortriptyline (NTP), clomipramine (CMP), and des-methylclomipramine (DCMP) and another for the determination of venlafaxine (VEN) and O-desmethyl-venlafaxine (ODV). Both assays were analytically validated according to FDA guidelines (20, 21). In addition, we found indications for a proportional differ-ence between DBS and plasma concentrations. As such, a translation factor is needed to calculate plasma concentrations based on DBS concentrations. In this current study we aimed to establish this factor. We first assessed if the differences found in patient samples were related to variance in hct values that were calculated based on potassium concentrations in capillary DBS. Next, we established the relationship between plasma and DBS concentrations in patient samples and explored the possibility to report cal-culated plasma concentrations. In addition, the comparability of clinical TMD interpre-tations based on plasma or DBS based analysis was assessed. Finally, we propose and apply acceptance criteria for individual data points of calculated plasma concentrations and demonstrate the application of these criteria on an additional set of NTP samples.

3.3.2. mEtHODs

materials potassium analysis

Water (HPLC quality) was purchased from Macron Fine Chemicals (Gliwice, Poland). Potassium chloride was purchased from Merck (Damstadt, Germany). Whatman FTK DMPK-C cards were used for DBS collection and purchased from GE Healthcare (Ho-evelaken, The Netherlands). Potassium ion (K+) concentrations were measured by indi-

Clinical validation of dried blood spot analysis 83

rect potentiometry using a Roche Cobas 6000 analyzer (Roche Diagnostics, Almere, The Netherlands) with technical limits of 1.5 and 10 mM. Hct was measured with a Sysmex XE-5000 hematology analyzer (Sysmex, Etten-leur, The Netherlands). Blank Li-heparin blood for preparation of the calibration and QC samples for potassium analysis was obtained from healthy volunteers.

method potassium analysis

The method for potassium extraction was based on the reported extraction method by Capiau et al. (11). Among some of the collected patient samples not all four spots were of sufficient quality to be analyzed. Therefore, we did not analyze potassium concentra-tions in duplicate as was described by Capiau et al. As a result of prior internal validation studies, parts of the procedure were adapted. A 6 mm DBS punch was extracted with 200 µL KCl (2.5 mM) for two subsequent extractions. During extraction, samples were vortexed for ten seconds and placed in an ultrasonic bath for five minutes. After cen-trifugation, 200µL of the extract was placed into a vial. A second extraction step with 200 µL extraction fluid was performed following the same procedure, however before the samples were placed in the ultrasonic bath, they were left for two hours at room temperature. The extract of the second extraction step was pipetted into the same vial (total 400 µL) for potassium ion analysis. During each run of potassium analysis, four calibration and QC samples were analyzed. Two thirds of the QC samples had to be within the +/− 15% limits of the nominal hct concentrations for acceptance of the analytical results (22). We analyzed DBS potassium concentrations of 20 volunteers (single and duplo data points) and in 26 patients (single data points only) of whom the hct concentration was measured in venous blood by the Sysmex hematology analyzer as well. We analyzed the relationship between potassium concentration and measured hct by least square regression. In addition, we analyzed the calculated and measured hct of the patient samples with PB regression and a Bland Altman comparison. Finally, we analyzed the potassium concentrations in the remaining patient samples to assess the relationship between calculated hct and bias between DBs and plasma concentrations of ADs.

methods antidepressant analysis

ATP, NTP, CMP, DCMP, VEN, and ODV concentrations in DBS and plasma were determined by validated methods which met acceptance criteria of international guidelines. A 6 mm DBS punch was extracted with a acetonitrile:methanol; 1: 3 solution. The DBS methods are described more in depth elsewhere (20, 21). Plasma concentrations were analyzed by a validated routine method which proved to be robust and reliable in external quality control programs. Both the DBS and plasma methods used LC-MS/MS for quantification of compounds and all used promazine as an internal standard.

84 Chapter 3.3

sample collection

Samples were collected between 01-01-2013 and 01-08-2015. After they gave written informed consent, paired samples of capillary DBS (i.e. obtained by fingerprick) and venous plasma were collected from patients who visited the clinic for routine TDM. Ap-proval for obtaining the additional DBS samples was obtained from the Medical Ethics Committee of the Diaconessen Hospital Meppel. Spots were collected by qualified per-sonnel who received written instructions, but no specific training. Instructions included disinfection of the finger, discharge of the first blood drop, and only gentle pressure on the finger to form blood droplets. To fill the preprinted circle on the DBS card, spots com-posed by one or two blood droplets were allowed. The resulting variation in spotting volume was allowed, since a spotting volume ranging from 20-100 µL was found not to introduce any systematic bias at low or high concentrations of the antidepressants (20, 21). After at least two hours of drying at room temperature, samples were placed in a sealed plastic bag and stored with a desiccant at 4°C until analysis. Under these condi-tions, validation samples were found stable for at least eight months and all patient DBS samples were analyzed within five months after the collection date. In addition to the patient samples, we used paired spiked venous DBS and plasma samples to assess the relationship between DBS and plasma concentrations. Blood from a healthy volunteer was used for the preparation of the samples. Samples were spiked at concentrations in the validated range (TCAs: 30-450 µg/L (n=9); VEN/ODV: 100-1100 µg/L(n=12)). Details about the preparation of the spiked samples is described in prior work (21).

Data analysis

Bridging DBS and plasma concentrationsThe deviation (bias %) between the DBS and plasma concentrations of the AD were plot-ted against calculated hct values and visually inspected for a linear relationship caused by hct effects. PB regression was used to test for significant differences between the DBS and plasma method (14). If found significant, the intercept or slope of the regression analysis were considered as translation factors (23). In addition to the relation found in patient samples, we also calculated the ratio of the plasma and venous DBS concentra-tions based on spiked (non-patient) samples.

The difference in clinical interpretation was assessed by comparison of the clinical interpretation based on the measured plasma concentrations and DBS based plasma concentrations. The clinical interpretation was based on Dutch reference ranges (24, 25). The sensitivity of DBS based plasma concentrations was calculated as the amount of identical advices/total advices (19).

Clinical validation of dried blood spot analysis 85

Calculation of limits for acceptance of DBS based calculated plasma concentrationsIn addition to the PB analysis and clinical interpretation, we wanted to assess the error for individual data points and formulate uniform acceptance criteria for this error. To find such acceptance criteria, we used the allowed 15% variance (or 20% between LLOQ and QC low level) in accuracy and precision as is allowed according to FDA and EMA guidelines. In R version 3.0.2 we simulated method validations of two methods which were both on the borderline of FDA and EMA guidelines. A Monte Carlo simulation was performed by random draws from a dataset with a normal distribution with accuracy and precision on the borderline of FDA and EMA guidelines. Next, we calculated the relative differences of single data points between the two methods (A and B) and cal-culated the 95% prediction interval (PI) for this difference. To compare our results with the criteria for incurred sample analysis of the EMA (67% of samples should be within 20% of original value), we also calculated the differences at the 67% PI (26). A total of 10 000 method validations, of each 15 data points, were simulated to calculate the PIs. To demonstrate the application of these additional acceptance criteria, we applied them to an additional dataset of NTP patient samples.

3.3.3. REsults

A total of 162 paired patient samples were included in this validation study. Since samples were collected alongside routine care, different sets of samples were available for the different compounds and calculated hct analysis (table 1). Due to the collection of blood by fingerprick and direct spotting, spots differed in spotting volume and qual-ity (figure 1). Prior validation indicated little bias was observed when spotting volume was varied (20, 21). Therefore, as long as a punch of 6 mm was possible with both the front and backside of the paper saturated, we included the sample in our analysis.

Bias due to variation in hct

Based on least square linear regression the squared correlation coefficients between potassium concentration and hematocrit were 0.53, 0.68, and 0.50 for the single data points of volunteers, duplo data points of volunteers, and single data points of patient samples, respectively. Concentrations of calculated hct in the capillary patient DBS samples were higher than the concentrations found in venous blood (n = 26; mean difference Bland Altman comparison: 0.02; PB analysis: slope: 0.73 (95% CI: 0.50- 1.11); intercept: 0.10 (95% CI: −0.07- 0.20). In figure 2, the relation between calculated hct and the bias between concentrations found in DBS and plasma is shown for ATP and NTP, respectively. For the other compounds see Supplementary file 1. Visual inspection of the plots, suggested that there was no quantifiable linear relationship between the

86 Chapter 3.3

calculated hct and the bias, except a weak relationship for ATP. The relationship between the bias and calculated hct concentration had a squared correlation coefficient of 0.27, for calculated hct values between 0.35 and 0.55 L/L.

Relation between DBs and plasma concentrations.

Results of the analysis of the paired spiked and patient (in vivo) samples are shown in table 2. The regression plots are shown in figure 3. The regression analyses indicated no

A

Figure 1. Example of quality of the samples used for analysis. A. All spots of good quality. B. Spot 2 of insufficient quality, spot 1 and 3 of sufficient quality, spot 4 best quality and used for analysis. C. Backside of the DBS card. Spot 1 and 4 of insuf-ficient quality, spot 3 of sufficient quality, spot 2 of best quality and used for analysis.

B

C

87

3.3.3. Results 1776

A total of 162 paired patient samples were included in this validation study. Since 1777

samples were collected alongside routine care, different sets of samples were 1778

available for the different compounds and calculated hct analysis (table 1). Due to the 1779

collection of blood by fingerprick and direct spotting, spots differed in spotting volume 1780

and quality (figure 1). Prior validation indicated little bias was observed when spotting 1781

volume was varied (20, 21). Therefore, as long as a punch of 6 mm was possible 1782

with both the front and backside of the paper saturated, we included the sample in 1783

our analysis. 1784

1785

A. All spots of good quality. 1786

1787 B. Spot 2 of insufficient quality, spot 1 and 3 of sufficient quality, spot 4 best quality 1788 and used for analysis. 1789

1790

C. Backside of the DBS card. Spot 1 and 4 of insufficient quality, spot 3 of sufficient 1791 quality, spot 2 of best quality and used for analysis. 1792

1793

Figure 1. Example of quality of the samples used for analysis. 1794

1795 Bias due to variation in hct 1796

Based on least square linear regression the squared correlation coefficients between 1797

potassium concentration and hematocrit were 0.53, 0.68, and 0.50 for the single 1798

table 1. Samples included in analyses.

AD used k+ analyzed AD analyzed Paired AD samples

Above llOQ Below llOQ total

AtP 18 ATP 27 2 29

NTP 21 8 90*

ntP 44 NTP 59 2

CmP 42 CMP 38 6 44

DCMP 42 2 44

VEn 23 VEN 23 5 28

ODV 27 1 28

* Out of the 80 samples above LLOQ, 40 random samples were used for regression analyses and the remain-ing samples were used for testing of additional acceptance criteria.AD: antidepressant; ATP: amitriptyline; CMP: clomipramine; DCMP; desmethylclomipramine; LLOQ: lower limit of quantification; NTP; nortriptyline; ODV; desmethyl-venlafaxine VEN: venlafaxine;

Clinical validation of dried blood spot analysis 87

constant bias between plasma and DBS analysis (all intercepts included zero). For ATP, NTP, DCMP, VEN and ODV proportional differences were found between the methods. The differences found in the patient samples were comparable to the results found in the spiked venous samples. The confidence interval around the slope estimate of CMP

A

y = 163,89x - 94,17 R² = 0,27

-60

-40

-20

0

20

40

60

0,25 0,35 0,45 0,55Bia

s (D

BS-

plas

ma/

plas

ma)

(%

)

Htc (L/L)

B

-80-60-40-20

0204060

0,25 0,35 0,45 0,55Bia

s (D

BS-

plas

ma/

plas

ma)

(%

)

Hct (L/L)

Figure 2. Relation between calculated (i.e. potassium based) hct and bias (bias = between DBS and plasma concentrations) of (A) ATP and (B) NTP. For CMP, DCMP, VEN and ODV see Supplementary file 1.

table 2. Relation between antidepressant concentrations in plasma and dried blood samples.

AD Ratio spiked plasma/DBsmean± sD

Ratio in vivo plasma/DBsmean± sD

slope PB(95%-Ci) in vivo plasma-DBs

literature

plasma/whole blood± sD

Ref.

ATP 1.13 ± 0.12 1.27±0.22* 1.22 (1.13-1.34) * 1.45± 0.14 (31)

NTP 0.73 ± 0.07 0.84±0.15 0.80 (0.71-0.89) 0.86± 0.06 (31)

CMP 0.94 ± 0.09 1.15± 0.25 1.15 (0.94-1.38) 0.72-1.07** (34)

DCMP 0.70 ± 0.11 0.66±0.11 0.65 (0.57-0.75) Not found

VEN 0.87 ± 0.06 0.89±0.14 0.84 (0.76-0.92) Not found

ODV 0.80 ± 0.05 0.83±0.09 0.78 (0.71-0.86) Not found

* Note that this relation was influenced by hematocrit of the bloodspots.** Study performed in rats, ratio was dependent on concentration, 0.72 at 5 µg/ml and 1.07 at 0.50 µg/mlAD: antidepressant; ATP: amitriptyline; CI: confidence interval; CMP: clomipramine; DBS: dried blood spot; DCMP; desmethylclomipramine; NTP; nortriptyline; PB: Passing and Bablok regression analysis Ref; refer-ence; SD: standard deviation; ODV desmethyl-venlafaxine VEN: venlafaxine;

88 Chapter 3.3

was much broader compared to confidence intervals of other compounds. Results for the comparison of clinical interpretations based on capillary DBS derived plasma con-centration and measured plasma concentrations are shown in table 3. For TDM of ATP, NTP and VEN, interpretation based on DBS based plasma concentrations were highly comparable (sensitivity >96%) to plasma based advises. For CMP, TDM based on DBS did result in a different clinical interpretation in 25% of the comparisons.

Acceptance criteria for differences between DBs derived plasma concentrations and measured plasma concentrations.

After simulating method validations at an accuracy and precision of 15% or 20% in both methods the following criteria were found:

0

100

200

300

400

500

600

0 100 200 300 400 500 600

DC

MP

foun

d in

pla

sma

(µg/

L)

DCMP found in DBS (µg/L)

Chart Title Slope= 0.65 (0.57-0.75); Intercept= 0 (-15-11)

0

80

160

240

320

400

0 80 160 240 320 400

CM

P fo

und

in p

lasm

a (µ

g/L)

CMP found in DBS (µg/L)

Slope= 1.15 (0.94-1.38); Intercept= -3 (-26 -22)

0

100

200

300

400

500

0 100 200 300 400 500

NTP

foun

d in

pla

sma

(µg/

L)

NTP found in DBS (µg/L)

Slope= 0.80 (0.71-0.89); Intercept= 2 (-7-11)

0

100

200

300

400

500

600

700

800

0 200 400 600 800

OD

V fo

und

in p

lasm

a (µ

g/L)

ODV found in DBS (µg/L)

Slope= 0.78 (0.71-0.86); Intercept= 7 (-7-32)

0

100

200

300

400

500

0 100 200 300 400 500

VE

N fo

und

in p

lasm

a (µ

g/L)

VEN found in DBS (µg/L)

Slope =0.84 (0.76-0.92); Intercept= 2 (-4-10)

0

100

200

300

400

500

0 100 200 300 400 500

ATP

foun

d in

pla

sma

(µg/

L)

ATP found in DBS (µg/L)

Slope = 1.22 (1.13-1.34); Intercept= -1 (-8-9) A

C

E

B

D

F

Figure 3 Passing-Bablok regression of patient DBS and plasma concentrations for (A) ATP, (B) NTP, (C) CMP, (D) DCMP, (E) VEN, and (F) ODV.

Clinical validation of dried blood spot analysis 89

– At an accuracy and precision of 20% in both methods, no more than 5% of the data had a difference between method A and B of >|48%|. This limit should be applied at DBS concentrations ranging from LLOQ to QC low concentrations.

– At an accuracy and precision of 15% in both methods, no more than 5% of the data had a difference between method A and B of >| 36%|. This limit should be applied at DBS concentrations ranging from QC low concentrations and higher.

Histograms of the differences between the single data points of the simulation can be found in Supplementary file 2. In our simulation, 67% of the samples had a bias <25% at an accuracy and precision of 20% (i.e. LLOQ level) and <18% at an accuracy and preci-sion of 15%.

Application of acceptance criteria

We applied the found translation factor for NTP ([DBS]* 0.80 = [plasma]) to another dataset of NTP samples and plotted the remaining differences (figure 4). We found one sample (2.5%) was outside the range for maximum bias, which was within our accep-tance criterion of 5%. As such, we established a translation factor, which proved to give reliable results for calculation of plasma concentrations based on DBS concentrations.

3.3.4. DisCussiOn

To compare plasma and DBS analyses of antidepressants, a large set of paired patient samples was analyzed. We found variance in calculated hct influenced bias between the DBS and plasma assay of ATP, although based on the sample size (n=27) no firm conclusion can be drawn from these results. Nevertheless, calculation of ATP plasma

table 3. Clinical interpretation based on measured plasma concentration and compared to DBS derived plasma concentrations. If significant, the slope of PB regression as reported in table 2 was used to calculate DBS derived plasma concentrations.

AtP&ntP ntP CmP&DCmP VEn&ODV

therapeutic range (µg/l) * 100-300 50-150 200-400 100-400

< llOQ (n) 2 2 2 1

< therapeutic range (n) 14 9 20 4

within therapeutic range (n) 9 44 15 12

> therapeutic range (n) 6 8 9 12

identical interpretation (n) 28 59 33 28

total (n) 29 61 44 28

sensitivity (%) 96.5 96.7 75.0 100.0

* Reference values which were used followed the Dutch guidelines (24, 25). AGPN guideline advices: ATP&NTP: 80-200 µg/L; NTP: 70-170 µg/L; CMP&DCMP: 230-450 µg/L; VEN&ODV: 100-400 µg/L (37) .

90 Chapter 3.3

concentrations based on DBS concentrations becomes more complex. The possible translation factor found with the PB regression might be dependent on the hct range which is covered by the samples. As such, our estimate contains uncertainty. To gain better insight into the correct translation factor, more samples should be collected with a better hct measurement (in duplicate) and better controlled hct of reference samples used for the AD analysis. For the other compounds calculated hct had no quantifiable influence on our results. We found significant proportional differences between DBS and plasma concentrations of NTP, DCMP, VEN, and ODV. The results of the spiked samples were in line with the results from the patient samples which strengthens our findings. The slopes of the PB analysis which were significant were used for calculation of plasma concentrations, based on DBS analysis. When this correction was applied on the DBS concentrations, it was found that the clinical interpretation based on DBS analysis of ATP, NTP, VEN, and ODV were highly comparable to clinical interpretation based on plasma analysis. For these compounds, sensitivity was >96% which was slightly higher then was reported sensitivity of TDM by DBS of antipsychotics (i.e. sensitivity of 92%) (19). However, for TDM of CMP a substantial difference and lower sensitivity was found. To study what is inducing this error, further research is needed.

To take clinical validation one step further, we established acceptance criteria for 95% of single data points and applied these criteria on an additional dataset of NTP. We found our results met these criteria, demonstrating the validity of the DBS assay and the translation factor in vivo. Notably, criteria which were derived from our simulation were in line with the criteria for incurred sample analysis of the EMA (26). However, they were more specific since they include two different criteria, depending on the different level of accuracy and precision at LLOQ or higher concentrations. In addition, they take 95% of the samples into account instead of 67%.

-60

-40

-20

0

20

40

60

0 50 100 150 200 250

Bias

(DBS

der

ived

pla

sma

conc

entra

tion-

pla

sma

conc

entra

tion

/pla

sma

conc

entra

tion)

(%)

NTP plasma concentration (µg/L)

Bias Mean bias 95 % PI for acceptance

Figure 4. Method comparison plots showing the difference between calculated plasma concentrations and measured plasma concentrations of NTP in patient samples.

Clinical validation of dried blood spot analysis 91

In this study potassium based hct analysis was conducted on patient samples using capillary blood obtained by fingerprick. A higher (mean difference Bland Altman com-parison: 0.02) calculated hct was found in capillary DBS (without anticoagulation) when compared to measured hct in venous EDTA blood. This is in contrast to the findings of Capiau et al., who found a lower (mean difference Bland Altman comparison: −0.02) concentration of calculated hct in venous lithium heparin DBS compared to measured hct in venous EDTA blood (11). However, our results were in line with results from Burger et al. who also found higher (mean difference Blant Altman comparison: 0.04) concentra-tions in capillary DBS samples (27). Although these differences could be related to the difference in sample collection, our sample size was too small to draw any firm conclu-sions (28, 29). Nevertheless, we considered the correlation between the measured and calculated hct sufficient to continue our study and assess the relationship between calculated hct and bias between the DBS and plasma assay of the ADs. With respect to this we found no hct effects based on calculated hct, except for ATP. This did not cor-respond with our prior findings in which we found a minor hct effect for all compounds in spiked venous DBS samples (20, 21). Others who applied potassium based analysis of hct effect in venous DBS samples did find a negative bias of up to −30% at a hct level of 0.20 L/L (10). There are different explanations possible for our findings. Firstly, our results could be explained by the small hct range which was studied. All patient samples had a calculated hct of > 0.30 L/L. Nevertheless, we do believe this range is a good repre-sentation of our populations hct (i.e. patients treated with antidepressants), since hct values of < 0.30 L/L are usually found in severely ill patients (23). Secondly, the lack of hct effects could be due to compensation of the negative bias by an increased recovery at lower hct levels. This effect referred to as “recovery bias” was recently demonstrated by Abu-Rabie et al. (7). Such compensation could disappear when whole spot punches are used instead of the sub-punches we used in our assay. However, compounds with a high absolute recovery (> 60%) are less sensitive to recovery bias at varying hct (7). Therefore, recovery bias did probably not have a major influence on our assay since all compounds except ODV had a high absolute recovery (>80%, ODV= 67%) (20, 21). Moreover, any recovery bias was likely already present in our prior validation and is therefore unlikely to explain our current findings. A more reasonable explanation for our findings would be the preparation procedure of the samples. As reported in previous work, the red blood cell fraction was pipetted into the plasma fraction and the hct was calculated based on the pipetted fractions (20, 21). As such, a lower hct could have been obtained than anticipated. This was very recently demonstrated by Koster et al.(30). Therefore, we could have overestimated the hct effect in our prior work. Lastly, our findings could be explained by the moderate correlation found between calculated hct and measured hct. The calculated hct might simply not be precise enough to differentiate between a hct bias of ~ +/− 20% and other influences of bias in capillary patient samples.

92 Chapter 3.3

The results of analysis of both spiked and patient samples showed that with the current sample size and standard deviation there was no significant difference in CMP concen-trations between plasma and DBS. For NTP, DCMP, VEN, and ODV higher concentrations were found in DBS. For ATP, lower concentrations were found in DBS. For ATP, NTP and CMP these results were in line with previous reported estimates (31-35). For DCMP, VEN, and ODV no estimates were found in literature. Nevertheless, our estimates were sup-ported by the consistent results between our spiked and patient samples. It has been reported that tertiary amines like ATP and CMP have more affinity towards plasma than to the red blood cell fraction when compared to their secondary amines metabolites NTP and DCMP, respectively. (31). Our findings were in line with these characteristics.

Our study has its limitations. Although we collected samples for more than two years, the sample size was still relatively small. Due to practical constraints, we could not further increase the inclusion period. According to Passing and Bablok, a sample size of n= 30 would have enough power to detect a slope < 0.87 or > 1.15 under the assumption that the coefficient of variation of both methods was ~7% (36). This assumption was made based on our previous validation work. Indeed, for VEN and ODV, we were able to detect significant proportional differences. However, for CMP we did not reach significance for a slope estimate of 1.15. This could be due to a higher coefficient of variation than found during prior validation. Further validation seems recommended due to the observed biases and relative low clinical comparability. The results based on clinical interpretation of ATP, NTP and VEN were promising. Nevertheless, we only validated the translation fac-tor of NTP on an additional independent dataset which fully guarantees the robustness of this factor. Taking into account the high clinical comparability of the clinical interpre-tation of ATP, VEN, and ODV it is questionable if the efforts to undertake collection of an additional datasets would be necessary for implementation of the DBS assay in clinical practice. Regulatory or international consensus guidelines should be issued to give clear recommendations about the comprehensiveness of clinical validation studies.

We calculated additional acceptance criteria for bridging studies of DBS by a simula-tion of comparison of methods. These comparisons assumed two methods which were both on the borderline of FDA limits. It is important to realize that similar acceptance criteria could be found when simulating two methods of which one is well (e.g. 5% bias) within and one is outside these limits (e.g. 30% bias). This implies a very precise refer-ence method would be able to compensate for substandard performance of a compara-tive method. However, this problem would arise in any current technique for method validation and is not limited to our acceptance criteria. Nevertheless, this underlines the importance of a full analytical validation to ensure bias of both methods meets the FDA and EMA limits.

Clinical validation of dried blood spot analysis 93

3.3.5. COnClusiOns

This study assessed DBS based TDM of antidepressant in clinical practice. For ATP, bias induced by calculated variations in hct influenced the relation between plasma and DBS samples. For all other compounds, these variations had no influence. For NTP, DCMP, VEN, and ODV, a translation factor to calculate plasma concentrations based on DBS concentrations was established. For ATP and CMP, further validation of this translation factors is needed. The agreement between the clinical interpretation based on DBS or plasma concentrations of ATP, NTP and, VEN was high. Acceptance criteria for single data points were calculated. We demonstrated application of these acceptance criteria towards an additional dataset of NTP samples and found our result met these criteria.

3.3.6. FutuRE PERsPECtiVEs

Based on the many analytical validation reports of DBS analysis, many clinical valida-tion studies will likely follow in the near future. Guidelines for requirements of a clinical validation of DBS analysis are needed. Such guidelines should preferably include ac-ceptance criteria for maximum differences between individual data points which are uniform and can be applied in a similar way to all compounds as demonstrated in this study. In addition, guidelines should include acceptance criteria for agreement between the clinical interpretation. This agreement is dependent on the therapeutic range and concentrations of compounds found in the samples. Therefore, samples included in this analysis should, for example, include at least five observations above and five below the therapeutic range to demonstrate the DBS based assay is able to differentiate between clinical interpretations.

94 Chapter 3.3

3.3.7. REFEREnCEs

1. van der Heijden J, de Beer Y, Hoogtanders K, Christiaans M, de Jong GJ, Neef C, et al. Therapeutic drug monitoring of everolimus using the dried blood spot method in combination with liquid chromatography-mass spectrometry. J Pharm Biomed Anal. 2009 Nov 1; 50(4): 664-70.

2. Hoogtanders K, van der Heijden J, Christiaans M, Edelbroek P, van Hooff JP, Stolk LM. Therapeutic drug monitoring of tacrolimus with the dried blood spot method. J Pharm Biomed Anal. 2007 Jul 27; 44(3): 658-64.

3. Mei JV, Alexander JR, Adam BW, Hannon WH. Use of filter paper for the collection and analysis of human whole blood specimens. J Nutr. 2001 May; 131(5): 1631S-6S.

4. Denniff P, Spooner N. The effect of hematocrit on assay bias when using DBS samples for the quantitative bioanalysis of drugs. Bioanalysis. 2010 Aug; 2(8): 1385-95.

5. Li W, Tse FL. Dried blood spot sampling in combination with LC-MS/MS for quantitative analysis of small molecules. Biomed Chromatogr. 2010 Jan; 24(1): 49-65.

6. Rowland M, Emmons GT. Use of dried blood spots in drug development: pharmacokinetic consid-erations. AAPS J. 2010 Sep; 12(3): 290-3.

7. Abu-Rabie P, Denniff P, Spooner N, Chowdhry BZ, Pullen FS. Investigation of different approaches to incorporating internal standard in DBS quantitative bioanalytical workflows and their effect on nullifying hematocrit-based assay bias. Anal Chem. 2015 May 5; 87(9): 4996-5003.

8. Koster RA, Botma R, Greijdanus B, Uges DR, Kosterink JG, Touw DJ, et al. The performance of five different dried blood spot cards for the analysis of six immunosuppressants. Bioanalysis. 2015; 7(10): 1225-35.

9. De Kesel PM, Sadones N, Capiau S, Lambert WE, Stove CP. Hemato-critical issues in quantitative analysis of dried blood spots: challenges and solutions. Bioanalysis. 2013 Aug; 5(16): 2023-41.

10. De Kesel PM, Capiau S, Stove VV, Lambert WE, Stove CP. Potassium-based algorithm allows correc-tion for the hematocrit bias in quantitative analysis of caffeine and its major metabolite in dried blood spots. Anal Bioanal Chem. 2014 Oct; 406(26): 6749-55.

11. Capiau S, Stove VV, Lambert WE, Stove CP. Prediction of the hematocrit of dried blood spots via potassium measurement on a routine clinical chemistry analyzer. Anal Chem. 2013 Jan 2; 85(1): 404-10.

12. Linnet K. Necessary sample size for method comparison studies based on regression analysis. Clin Chem. 1999 Jun; 45(6 Pt 1): 882-94.

13. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986 Feb 8; 1(8476): 307-10.

14. Bablok W, Passing H. Application of statistical procedures in analytical instrument testing. J Automat Chem. 1985; 7(2): 74-9.

15. Wilhelm AJ, Klijn A, den Burger JC, Visser OJ, Veldkamp AI, Janssen JJ, et al. Clinical validation of dried blood spot sampling in therapeutic drug monitoring of ciclosporin A in allogeneic stem cell transplant recipients: direct comparison between capillary and venous sampling. Ther Drug Monit. 2013 Feb; 35(1): 92-5.

16. Jager NG, Rosing H, Schellens JH, Beijnen JH, Linn SC. Use of dried blood spots for the determi-nation of serum concentrations of tamoxifen and endoxifen. Breast Cancer Res Treat. 2014 Jul; 146(1): 137-44.

17. Vu DH, Koster RA, Bolhuis MS, Greijdanus B, Altena RV, Nguyen DH, et al. Simultaneous determi-nation of rifampicin, clarithromycin and their metabolites in dried blood spots using LC-MS/MS. Talanta. 2014 Apr; 121: 9-17.

Clinical validation of dried blood spot analysis 95

18. de Wit D, den Hartigh J, Gelderblom H, Qian Y, den Hollander M, Verheul H, et al. Dried blood spot analysis for therapeutic drug monitoring of pazopanib. J Clin Pharmacol. 2015 Jun 1.

19. Patteet L, Maudens KE, Stove CP, Lambert WE, Morrens M, Sabbe B, et al. Are capillary DBS applicable for therapeutic drug monitoring of common antipsychotics? A proof of concept. Bioanalysis. 2015; 7(16): 2119-30.

20. Berm EJ, Paardekooper J, Brummel-Mulder E, Hak E, Wilffert B, Maring JG. A simple dried blood spot method for therapeutic drug monitoring of the tricyclic antidepressants amitriptyline, nortriptyline, imipramine, clomipramine, and their active metabolites using LC-MS/MS. Talanta. 2015 Mar; 134: 165-72.

21. Berm EJ, Brummel-Mulder E, Paardekooper J, Hak E, Wilffert B, Maring JG. Determination of venla-faxine and O-desmethylvenlafaxine in dried blood spots for TDM purposes, using LC-MS/MS. Anal Bioanal Chem. 2014 Apr; 406(9-10): 2349-53.

22. Guidance for Industry, Bioanalytical Method Validation [Internet].: US Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research (CDER), Center for Veterinary Medicine (CVM), FDA; 2001 [updated 2001, May; cited 17-9-2015]. Available from: http://www.fda.gov/downloads/Drugs/Guidances/ucm070107.pdf.

23. Evans C, Arnold M, Bryan P, Duggan J, James CA, Li W, et al. Implementing dried blood spot sampling for clinical pharmacokinetic determinations: considerations from the IQ Consortium Microsampling Working Group. AAPS J. 2015 Mar; 17(2): 292-300.

24. Dutch Society of Clinical Pharmacist (NVZA). TDM Monograph Tricyclic Antidepressants; 2014. 25. Dutch Society of Clinical Pharmacist (NVZA). TDM Monograph Venlafaxine; 2012. 26. Guideline on bioanalytical method validation. EMEA/CHMP/EWP/192217/2009rev1.corr2 [Inter-

net].: Committee for Medicinal Products for Human Use (CHMP); 2012 [cited 17-9-2015]. Available from: http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2011/08/WC500109686.pdf.

27. den Burger JC, Wilhelm AJ, Chahbouni AC, Vos RM, Sinjewel A, Swart EL. Haematocrit corrected analysis of creatinine in dried blood spots through potassium measurement. Anal Bioanal Chem. 2015 Jan; 407(2): 621-7.

28. Yang ZW, Yang SH, Chen L, Qu J, Zhu J, Tang Z. Comparison of blood counts in venous, fingertip and arterial blood and their measurement variation. Clin Lab Haematol. 2001 Jun; 23(3): 155-9.

29. Daae LN, Halvorsen S, Mathisen PM, Mironska K. A comparison between haematological param-eters in ‘capillary’ and venous blood from healthy adults. Scand J Clin Lab Invest. 1988 Nov; 48(7): 723-6.

30. Koster RA, Alffenaar JW, Botma R, Greijdanus B, Touw DJ, Uges DR, et al. What is the right blood hematocrit preparation procedure for standards and quality control samples for dried blood spot analysis? Bioanalysis. 2015 Feb; 7(3): 345-51.

31. Fisar Z, Fuksova K, Sikora J, Kalisova L, Velenovska M, Novotna M. Distribution of antidepressants between plasma and red blood cells. Neuro Endocrinol Lett. 2006 Jun; 27(3): 307-13.

32. Hulten BA, Heath A, Knudsen K, Nyberg G, Svensson C, Martensson E. Amitriptyline and ami-triptyline metabolites in blood and cerebrospinal fluid following human overdose. J Toxicol Clin Toxicol. 1992; 30(2): 181-201.

33. Kobuchi S, Fukushima K, Aoyama H, Matsuda T, Ito Y, Sugioka N, et al. Effect of oxidative stress on the pharmacokinetics of clomipramine in rats treated with ferric-nitrilotriacetate. Drug Metab Lett. 2011 Dec; 5(4): 243-52.

96 Chapter 3.3

34. Kobuchi S, Fukushima K, Shibata M, Ito Y, Sugioka N, Takada K. Pharmacokinetics of clomipramine, an antidepressant, in poloxamer 407-induced hyperlipidaemic model rats. J Pharm Pharmacol. 2011 Apr; 63(4): 515-23.

35. Dubois JP, Kung W, Theobald W, Wirz B. Measurement of clomipramine, N-desmethyl-clomip-ramine, imipramine, and dehydroimipramine in biological fluids by selective ion monitoring, and pharmacokinetics of clomipramine. Clin Chem. 1976 Jun; 22(6): 892-7.

36. Passing H, Bablok W. Comparison of several regression procedures for method comparison stud-ies and determination of sample sizes. Application of linear regression procedures for method comparison studies in Clinical Chemistry, Part II. J Clin Chem Clin Biochem. 1984 Jun; 22(6): 431-45.

37. Hiemke C, Baumann P, Bergemann N, Conca A, Dietmaier O, Egberts K, et al. AGNP consensus guidelines for therapeutic drug monitoring in psychiatry: update 2011. Pharmacopsychiatry. 2011 Sep; 44(6): 195-235.

Clinical validation of dried blood spot analysis 97

supplementary  file 1. Relation between bias of plasma and DBS concentrations and potassium based hematocrit concentrations. C. CMP; D: DCMP; VEN; ODV.

-80,0

-60,0

-40,0

-20,0

0,0

20,0

40,0

60,0

0,25 0,35 0,45 0,55

bias

bet

wee

n D

BS

and

plas

ma

conc

entr

atio

ns (%

)

Calculated hct (L/L)

-80

-60

-40

-20

0

20

40

60

0,25 0,35 0,45 0,55

Bia

s be

twee

n D

BS

and

plas

ma

conc

entr

atio

ns (%

)

Calculated hct (L/L)

-80

-60

-40

-20

0

20

40

60

0,25 0,35 0,45 0,55

Bia

s be

twen

DB

S an

d pl

asm

a co

ncen

trat

ions

(%)

Calculated hct (L/L)

-80,0

-60,0

-40,0

-20,0

0,0

20,0

40,0

60,0

0,25 0,35 0,45 0,55

Bia

s be

twen

DB

S an

d pl

asm

a co

ncen

trat

ions

(%)

Calculated hct (L/L)

98 Chapter 3.3

supplementary file 2. Histograms of difference between two methods of ~ 450 000 simulated data points based on acceptance limits of methods accuracy and precision of (A) 15% or (B) 20%.

4Cost-eff ectiveness of precision drug therapy

4.1Economic evaluations of pharmacogenetic and pharmacogenomic screening tests: a

systematic review. Second update of the literature

Elizabeth J.J. Berm, Margot de Looff, Bob Wilffert, Cornelis Boersma, Lieven Annemans, Stefan Vegter, Job F.M. van Boven, Maarten J. Postma

PloS One, 2016; January; 11; 11 (1): e0146262.

104 Chapter 4.1

ABstRACt

Objective: Due to extended application of pharmacogenetic and pharmacogenomic screening (PGx) tests it is important to assess whether they provide good value for money. This review provides an update of the literature.methods: A literature search was performed in PubMed and papers published between August 2010 and September 2014, investigating the cost-effectiveness of PGx screening tests, were included. Papers from 2000 until July 2010 were included via two previous systematic reviews. Studies’ overall quality was assessed with the Quality of Health Economic Studies (QHES) instrument.Results: We found 38 studies, which combined with the previous 42 studies resulted in a total of 80 included studies. An average QHES score of 76 was found. Since 2010, more studies were funded by pharmaceutical companies. Most recent studies performed cost-utility analysis, univariate and probabilistic sensitivity analyses, and discussed limita-tions of their economic evaluations. Most studies indicated favorable cost-effectiveness. Majority of evaluations did not provide information regarding the intrinsic value of the PGx test. There were considerable differences in the costs for PGx testing. Reporting of the direction and magnitude of bias on the cost-effectiveness estimates as well as motivation for the chosen economic model and perspective were frequently missing.Conclusions: Application of PGx tests was mostly found to be a cost-effective or cost-saving strategy. We found that only the minority of recent pharmacoeconomic evalua-tions assessed the intrinsic value of the PGx tests. There was an increase in the number of studies and in the reporting of quality associated characteristics. To improve future evaluations, scenario analysis including a broad range of PGx tests costs and equal costs of comparator drugs to assess the intrinsic value of the PGx tests, are recommended. In addition, robust clinical evidence regarding PGx tests’ efficacy remains of utmost importance.

Economic evaluations of pharmacogenetic tests. 105

4.1.1. intRODuCtiOn

Pharmacogenetics and pharmacogenomics investigate the influence of genetic and ge-nomic variations on drug response in individuals (1). The term pharmacogenetics covers the study of single genes, whereas pharmacogenomics is used to describe the study of several genes (1). The abbreviation PGx is used to cover both pharmacogenetics and pharmacogenomics. PGx tests offer great opportunities for personalised medicine, by combining genetic information and corresponding phenotypes (2). Ideally, by applying PGx, the most optimal, tailored pharmacotherapy can be determined, thereby increas-ing the treatment’s overall efficacy and decreasing the incidence of adverse events (1). In the field of oncology it has been shown that for certain therapies the specific genetic variations in cancer cells can affect the drug efficacy and/or adverse events (2). Hence, patients may benefit from the PGx tests by utilising an alternative therapy, or changing the drug dosage (3,4). Therefore, PGx is nowadays often used as a synonym for person-alised medicine, although personalized medicine is a much broader concept (2).

It is likely that an increasing amount of patient-specific genomic information will be available in the near future and this may result in an increased usage of PGx tests which needs evaluation of effects, but also cost effectiveness (5). PGx has the potential to reduce the costs associated with inappropriate, expensive drug treatments and/or serious adverse drug reactions, in particular those that require hospitalisation (3). There-fore, next to optimising health outcomes, PGx tests might be cost-saving (1,5). However, in order to obtain valid, accurate, and relevant estimates of cost-effectiveness, reliable economic studies are required.

Economic evaluations of PGx tests entail some specific difficulties. Often there is no hard clinical evidence regarding the effects of the PGx test on the clinical utility and it is unlikely that such evidence will be available for the use of every genetic variant (6). Furthermore, for PGx tests, compliance and adherence of clinicians to the test results might have an effect on the effectiveness of PGx tests which is hard to incorporate in a cost-effectiveness analysis (7). Differences in costs for the PGx test can be substantial between countries, or even laboratories, and therefore it is advised to include different costs in scenario analysis (7). In addition, the sensitivity and specificity of a PGx test can vary due to different ethnicities studied, or genetic variations analysed.

In the last decade, several reviews investigated economic evaluations of genetic tests (1,8-14). These reviews showed that the level of consistency and quality could be improved. Many original studies lacked a thorough sensitivity analysis and moreover, in general a poor quality of methodology was noticed (9,11). Inconsistencies mainly re-sulted from e.g. lack of clinical evidence, different methodologies as well as statistics and modest heterogeneity among study designs and patient populations (3,9). However, these different methodologies have not been in detail dealt with in previous systematic

106 Chapter 4.1

reviews. Recent developments have led to a bifurcation in the nature of the economic evaluations of PGx testing in, on the one hand, studies assessing the intrinsic value of a test and, on the other hand, studies assessing the value of the test in combination with an active compound. For example in colorectal cancer, the economic value of testing for KRAS as compared to no testing could be considered the “intrinsic value” of the PGx test. KRAS testing before treatment with cetuximab, is found to be dominant (i.e. cost-saving and better) as compared to no prior testing and therefore it is recommended before administration of cetuximab (15). By its uptake in clinical guidelines, a shift in the comparator for future economic evaluations took place, as the combination of cetuximab and KRAS testing became usual care. In future evaluations, the intrinsic value of the KRAS test itself will no longer be assessed, but rather the combination of a drug and its test as compared with a new treatment option. Following this development, a distinction between the nature of economic evaluations of PGx tests is important for a fair comparison of studies.

The objective of this study was to give recommendations for improvement and an update of the literature about PGx tests, taking into account the difference between the intrinsic value of tests themselves and tests embedded into economic evaluations as usual care or best current care. Our new findings were placed in perspective with respect to findings from our previous reviews (1,9). As such, our study links together a period from 2000- September 2014 of PGx testing and pharmacoeconomics.

4.1.2. mEtHODs

A search in PubMed was performed using combinations of the following terms [PubMed search: all fields] and their thesauri variants: [‘cost-effectiveness [including MeSH) ’ OR ‘cost-utility’ OR ‘cost-benefit’ OR ‘cost-minimization’ OR ‘pharmacoeconomics [including MeSH]’] AND [‘pharmaco-genetics’ OR ‘pharmacogenomics [including MeSH]’ OR ‘geno-typing’ OR ‘genetic screening’ OR ‘genetic testing [including MeSH]’ OR ‘genotyped’ OR ‘polymorphism screening’]. These search terms were in line with the terms that were used in previous reviews, performed in 2008 and 2010 (1,9). The search was last updated in October 2014 and studies were included if they were: published between August 2010 and September 2014, peer reviewed, performed on a genetic screening method of the human genome, evaluating economic outcomes, written in English, and the genetic or genomic variations were shown to influence the drug efficacy or drug safety. Articles were first screened on title. If the title was not informative enough to form a decision with respect to these criteria, abstracts were assessed. Additional articles were identified through reference tracking.

Economic evaluations of pharmacogenetic tests. 107

From the selected studies, the following data were extracted: (I) area of disease or patient population, (II) gene(s) analysed by the pharmacogenetic test, (III) the costs of the pharmacogenetic test, (IV) pharmaceutical compound influenced by the genetic varia-tion, (V) type of economic analysis, (VI) type of sensitivity analysis, (VII) time horizon, (VIII) discounting, (IX) perspective, (X) the outcome measurements, and (XI) the funding body. For interpretation of the outcome measure (i.e. cost-effectiveness), the conclusions as reported by the authors were used. Furthermore, an assessment of the papers ‘discussion on the study limitations’ was made. In this assessment, all limitations mentioned by the authors were captured to look for common and uncommon themes. As stated in the previ-ous review, assessment of these points is expected to provide good information for an adequate interpretation of the studies design, reporting, robustness, methodologies used, and statistical analyses performed (9). In addition to these points, which were assessed in our previous reviews, we added (XII) reporting of analytical validity of the PGx test , (XIII) the cost-effectiveness threshold , (XIV) the country which was used for the perspective of the economic evaluation, and (XV) a weighted quality assessment for the studies included by this update. To assess the quality, the Quality of Health Economic Studies (QHES) instru-ment was used (16). This instrument was used to improve generalisability of the results with respect to other reviews performed in the same field (10,13). According to the QHES checklist a score between 0 and 100 was generated. A score of ≥ 75 was considered as a high quality score (11). Two reviewers assessed the quality of the included studies, if results were different, consensus was reached through discussion.

Data from before 2008, and for the period 2008 - July 2010 were retrieved from the two previous reviews by Vegter et al. (2008, 2010) (1,9).

4.1.3. REsults

included studies

Results of the search strategy are provided in a PRISMA flow chart (fig. 1) (17). The PubMed search yielded 4408 hits. Duplicates were removed and out of the remaining 733 articles, 160 were selected for full text assessment. Three articles were identified through reference tracking. After inspection of full texts, 122 studies were excluded, resulting in 38 included studies (15,18-54). Main reasons for exclusion were the type of genetic tests studied (i.e. not related to pharmacogenetics) and review papers. Subse-quently, 42 studies published before July 2010 were added based on the two reviews by Vegter et al. (1,9,55-96), resulting in a final inclusion of 80 studies. Fig. 2A provides an overview of the total of 80 studies, published from 2000 until September 2014, sorted by the type of pharmacoeconomic analysis performed. Most studies were published in 2012 (16%), with a total of 13 publications.

108 Chapter 4.1

type of analysis: intrinsic value or combination with new drug

Cost utility analysis (CUA) was the technique mostly applied, namely in 54 studies (68%). Cost-effectiveness analysis (CEA) was performed in 20 studies (25%), cost minimization analysis (CMA) was used 5 times (6%). Note that for the sake of clarity we explicitly dif-ferentiate between CUA (with results expressed in cost per QALY) and CEA (with other parameters for effectiveness). Before 2008, CEA was the most frequently applied study type. Since 2008, CUA was performed in most of the publications. Some studies directly assessed the intrinsic value of the PGx test (15,18-22,24-28), while other studies applied scenarios in which equal costs and efficacy were assumed for the drug related to the PGx tests and the alternative treatment (23,29). Both were considered cost-effectiveness estimates that provide an indication of the intrinsic value of the PGx test (table 1A).

Records identified through database searching by

combinations of the terms between 2010 and October 2014, human (n

= 4408)

Additional records identified from reference tracking

(n=3)

Unique records (n= 733)

Full-text articles assessed for eligibility

(n = 160)

Excluded based on full text (n = 122)

Reasons: -Not an original research paper (n=64) * of which reviews (n=50) - Not about effect of genetic variance on or related to drug effects (n = 35). - No cost-effect/benefit analysis (n= 16) - Other (n=7) Studies included

(n = 38)

Excluded based on title and abstract (n= 573)

Duplicates removed (n = 3678)

42 studies included from previous reviews

(period 2000- 2010)

80 studies included in review

Figure 1. PRISMA Flow chart of the literature search

Economic evaluations of pharmacogenetic tests. 109

However, we found that the majority of the newly included (i.e. since 2010) CEAs incor-porated a PGx test strategy in combination with a drug and compared this combination to another drug (table 1B). As such, the intrinsic value of the PGx test itself was not assessed. For example, Handorf et al. compared therapy for non-small cell lung cancer with a platinum combination to PGx selected treatment with erlotinib (44). In this analy-sis, the cost-effectiveness of the PGx test itself was not assessed, but the combination of the PGx test and erlotinib. This makes the outcome mainly dependent on the price of erlotinib. The incremental cost-effectiveness ratios (ICER) of base case scenarios are shown in table 1A and B.

Costs of Pgx tests

When comparing the costs of the PGx tests, considerable differences between the costs of the tests were observed. This is not surprising, since technology of genetic testing is developing and costs are likely to be further reduced in the future. This was demonstrated by two studies of which one was performed in 2012 (41) and one in 2013 (40). The costs for this particular screening test, US$72 and £20 respectively, were considerably lower compared to screening costs in studies performed in 2009 and 2010, which ranged from US$175 until US$575, respectively (63,78,96). Another substantial difference was seen in the costs for the CYP2C9 and VKORC1 tests, which ranged from £20 to US$575 (40,78)

sensitivity analysis

Until 2008, only 5 out of the 31(16%) studies performed both a univariate and proba-bilistic sensitivity analysis, whereas since 2008 this were 35 out of the 49 (71%) studies (fig. 2B). One of the major uncertainty factors was the lack of robust clinical evidence for clinical utility of the PGx test itself. Hence, almost all authors had to define assumptions which were sometimes solely based on expert opinion. Other frequently mentioned uncertainty factors were the costs of the PGx tests and their real-world utility and per-formance. As a consequence of the uncertainty around the included genotyping costs, several studies included a costs range in their analysis. For example, Schackman et al. (2013) demonstrated that at a test cost of US$107, genetic testing was not cost-effective. However, at a price of US$10 per test, the PGx test was cost-effective (29). Although vari-ance in genotyping costs was frequently found to have a major impact on the ICER, 11 out of the 38 studies (30%) assessed in this update did not include a range of genotype costs in their sensitivity analysis (S1 Table). With respect to drug costs, some studies showed that drug driven costs did not influence the study’s outcome. For example, Crespin et al. (2011) showed in their sensitivity analysis that even if the costs of the drugs guided by PGx testing dropped substantially, the non-PGx-test-guided drug remained cost-effective (37).

110 Chapter 4.1

tabl

e 1.

Out

com

es a

nd fu

ndin

g so

urce

s of

pha

rmac

o-ec

onom

ic P

Gx

stud

ies

on th

e in

trin

sic

valu

e of

PG

x te

st (A

) or t

reat

men

t com

paris

ons

invo

lvin

g PG

x te

stin

g (B

), pu

blis

hed

betw

een

Augu

st 2

010

and

Sept

embe

r 201

4. N

ote

that

num

bers

wer

e ro

unde

d to

war

ds h

undr

eds.

A 1st A

utho

r (re

fere

nce)

Pgx

test

Ana

lyti

cal

valid

ity

Pgx

test

repo

rted

Out

com

em

easu

reQ

uant

itat

ive

outc

ome

or iC

ER

(us$

)

Cost

eff

ecti

vene

ss

thre

shol

d ($

/Q

Aly

)

Conc

lusi

on

base

d on

ou

tcom

e

Fund

ing

Coun

try

QH

Essc

ore

klan

g20

10 (1

8)21

gen

e as

say

No

QA

LY$1

0,80

0N

o nu

mbe

rCo

st-e

ffect

ive

Teva

Ph

arm

aceu

tical

In

dust

ries

Ltd.

Isra

el83

Bacc

hi20

10 (1

9)21

- gen

e as

say

No

Cost

s$8

00 s

aved

per

pa

tient

n.a.

Cost

-sav

ing

Unk

now

nBr

azil

34

Hal

l20

12 (2

0)21

gen

e as

say

No

QA

LY$8

,900

£20,

00-3

0,00

($

31,2

00-4

6,70

0)Co

st-e

ffect

ive,

th

ough

su

bsta

ntia

l un

cert

aint

y

Acad

emic

re

sour

ces

UK

90

Vand

erla

an20

11 (2

1)21

gen

e as

say

No

QA

LY$4

,400

9 sa

ved

per

patie

nt p

er y

ear

n.a.

Dom

inan

tG

enom

ic H

ealth

, In

c.U

SA31

Verh

oef

2013

(22)

CYP2

C9 a

nd

VKO

RC1

No

QA

LY€2

,700

($32

00)

€20,

000

($24

,200

)Co

st-e

ffect

ive

Euro

pean

gra

ntN

ethe

rland

s87

Don

g20

12 (2

3)H

LA-B

*150

2Ye

sQ

ALY

$29,

800*

$50,

000

Cost

-effe

ctiv

e fo

r Sin

gapo

rean

Ch

ines

e an

d M

alay

s, bu

t not

fo

r Sin

gapo

rean

In

dian

s.

Acad

emic

re

sour

ces

Sing

apor

e63

tiam

kao

2013

(24)

HLA

-B *1

502

Yes

Cost

s98

,600

bah

t ($

3,00

0) s

aved

per

10

0 ca

ses

n.a.

Cost

-effe

ctiv

eN

one

Thai

land

48

Economic evaluations of pharmacogenetic tests. 111

A (c

onti

nued

)

1st A

utho

r (re

fere

nce)

Pgx

test

Ana

lyti

cal

valid

ity

Pgx

test

repo

rted

Out

com

em

easu

reQ

uant

itat

ive

outc

ome

or iC

ER

(us$

)

Cost

eff

ecti

vene

ss

thre

shol

d ($

/Q

Aly

)

Conc

lusi

on

base

d on

ou

tcom

e

Fund

ing

Coun

try

QH

Essc

ore

shir

oiw

a20

10 (1

5)KR

ASN

oQ

ALY

and

LY

GD

omin

ant

n.a.

Dom

inan

tRo

che

Dia

gnos

tics

K.K.

Japa

n90

Blan

k20

11 (2

5)KR

AS a

nd B

RAF

Yes

QA

LYKR

AS

and

BRA

F sa

ves

€3,3

00

($2,

500)

per

pat

ient

n.a.

Cost

sav

ing

Acad

emic

re

sour

ces

Switz

erla

nd83

Behl

2012

(26)

KRAS

and

/or

BRAF

No

LYS

KRA

S al

one

save

s $7

,500

per

pat

ient

Addi

tiona

l BRA

F te

stin

g sa

ves

$1,0

00 p

er p

atie

nt

n.a.

Cost

sav

ing

Acad

emic

re

sour

ces

USA

84

shiff

man

2012

(27)

LPA

No

QA

LY$2

5,00

0N

o nu

mbe

rCo

uld

be c

ost-

effec

tive

Berk

eley

H

eart

Lab,

Inc.

USA

34

Don

nan

2011

(28)

TPM

TYe

sLi

fe m

onth

s- C

ost n

o te

st

CAN

$700

($60

0) p

er

patie

nt- C

ost g

enet

ic te

st

CAN

$1,1

00 ($

900)

pe

r pat

ient

s- W

ith te

st n

o LY

ga

ined

n.a.

Not

cos

t-eff

ectiv

eAc

adem

ic

reso

urce

sCa

nada

77

scha

ckm

an20

13 (2

9)U

GT1

A1Ye

sQ

ALY

$2,0

00,0

00*

$100

,000

Not

cos

t-eff

ectiv

e un

less

ass

ay c

ost

are

low

Acad

emic

re

sour

ces

USA

83

*Aut

hors

ass

umed

equ

al c

osts

and

effi

cacy

for d

iffer

ent p

harm

aceu

tical

sBR

AF,

v-Ra

f mur

ine

sarc

oma

vira

l onc

ogen

e ho

mol

og B

1; H

LA, h

uman

leuk

ocyt

e an

tigen

; HSR

, hyp

er se

nsiti

vity

reac

tion;

ICER

, inc

rem

enta

l cos

t-eff

ectiv

enes

s rat

io; K

RAS,

Ki

rste

n ra

t sa

rcom

a vi

ral o

ncog

ene

hom

olog

; LPA

, lip

opro

tein

-a; L

YG, l

ife-y

ears

-gai

ned;

LYS

, life

-yea

rs-s

aved

; PG

x, p

harm

acog

enet

ic; Q

ALY

, qua

lity-

adju

sted

-life

-yea

r; TP

MT,

thio

purin

e S-

met

hyltr

ansf

eras

e ; U

GT,

UD

P-gl

ucur

onos

yltr

ansf

eras

e.

112 Chapter 4.1

B IC

ER =

ICER

is n

ot re

late

d to

PG

x sc

enar

io

1st A

utho

r (re

fere

nce)

Pgx

test

Ana

lyti

cal

valid

ity

Pgx

test

repo

rted

(y

/n)

Out

com

em

easu

reQ

uant

itat

ive

outc

ome

oriC

ER (u

s$)

Cost

-eff

ecti

vene

ss

thre

shol

d ($

/Q

Aly

)

Conc

lusi

on

base

d on

ou

tcom

e

Fund

ing

Coun

try

QH

Es

scor

e

Olg

iati

2012

(30)

5-H

TTLP

RN

oQ

ALW

- Eur

o A

$1,

100

- Eur

o B

$1,2

00- E

uro

C: $

1,20

0

< 3

times

the

GD

P pe

r cap

itaCo

st-e

ffect

ive

in h

igh-

inco

me

coun

trie

s

Unk

now

nEu

rope

61

serr

etti

2011

(31)

5-H

TTLP

RN

oQ

ALW

$2,9

00$5

0,00

0N

ot c

ost-

effec

tive

Unk

now

nIta

ly87

Reed

2011

(32)

8-14

risk

alle

les

No

QA

LY--

$98,

100

- $10

3,20

0.$1

00,0

00Co

st-e

ffect

ive

Nat

iona

l Can

cer

Inst

itute

USA

83

Dja

lalo

v20

12 (3

3)AP

OE

ε4ye

sQ

ALY

CAN

$38,

000

($32

,700

)N

o nu

mbe

rM

ay b

e ec

onom

ical

ly

attr

activ

e

Acad

emic

re

sour

ces

Cana

da90

kazi

2014

(34)

CYP2

C19

No

QA

LY- E

xten

dedl

y do

min

ated

;- $

52,6

00 ;

- $35

,800

;- $

30,2

00.

$50,

000

may

impr

ove

cost

eff

ectiv

enes

sA

mer

ican

Hea

rt

Ass

ocia

tion

and

acad

emic

re

sour

ces

USA

90

Rees

e20

12 (3

5)CY

P2C1

9N

oCV

E av

oide

d- C

ost s

avin

g- C

ost s

avin

g;- $

2,30

0 ;

- Cos

t sav

ing

No

num

ber

Dom

inan

tU

nkno

wn

USA

64

sori

ch20

13 (3

6)CY

P2C1

9N

oQ

ALY

- AU

S$6,

300

($5,

200)

-- A

US$

22,8

00

($18

,600

)

AUS$

30,0

0-50

,000

($

24,5

00-4

0,80

0)Co

st-e

ffect

ive

Hea

rt F

ound

atio

n of

Aus

tral

iaAu

stra

lia75

Cres

pin

2011

(37)

CYP2

C19

*2Ye

sQ

ALY

$10,

100

$50,

000

Cost

-effe

ctiv

eAc

adem

ic

reso

urce

sU

SA93

lala

2013

(38)

CYP2

C19

*2Ye

sQ

ALY

- Dom

inan

t- D

omin

ant

n.a.

Dom

inan

tAc

adem

ic

reso

urce

sU

SA83

Economic evaluations of pharmacogenetic tests. 113

B (c

ontin

ued)

1st A

utho

r (re

fere

nce)

Pgx

test

Ana

lyti

cal

valid

ity

Pgx

test

repo

rted

(y

/n)

Out

com

em

easu

reQ

uant

itat

ive

outc

ome

oriC

ER (u

s$)

Cost

-eff

ecti

vene

ss

thre

shol

d ($

/Q

Aly

)

Conc

lusi

on

base

d on

ou

tcom

e

Fund

ing

Coun

try

QH

Es

scor

e

Pana

tton

i20

12 (3

9)CY

P2C1

9 *2

No

QA

LY- N

Z$

24,6

00($

19,2

00)

- Dom

inan

t

NZ$

50,0

00

($39

,000

)Co

st-e

ffect

ive

Acad

emic

re

sour

ces

New

Zea

land

75

Pink

2013

(40)

CYP2

C9 a

nd

VKO

RC1

No

QA

LY£1

3,20

0 ($

20,6

00)

£20,

000-

30,0

00($

31,2

00-4

6,70

0)Co

st-e

ffect

ive

Acad

emic

re

sour

ces

UK

93

you

2012

(41)

CYP2

C9 a

nd

VKO

RC1

No

QA

LY- D

omin

ated

by

geno

type

–gui

ded

appr

oach

- $13

,800

per

QAL

Y- D

omin

ated

by

dabi

gatr

an 1

50 m

g

$50,

000

Dab

igat

ran

150

mg

seem

s to

be

cost

-effe

ctiv

e

No

fund

ing

USA

84

you

2014

(42)

CYP2

C9 a

nd

VKO

RC1

No

QA

LY$2

,800

per

QA

LY$5

0,00

0Co

st-e

ffect

ive

Rese

arch

G

rant

s Co

unci

l of

the

Hon

g Ko

ng s

peci

al

adm

inis

trat

ive

Regi

on, C

hina

USA

70

de l

ima

lope

s20

12 (4

3)

EGFR

No

QA

LYD

omin

ant

n.a.

Dom

inan

tA

stra

Zene

ca P

te

Ltd.

Sing

apor

e87

Han

dorf

2012

(44)

EGFR

No

QA

LY$1

10,6

00$1

00,0

00Co

st-e

ffect

ive

OSI

Ph

arm

aceu

tical

s/G

enen

tech

USA

90

zhu

2013

(45)

EGFR

No

QA

LY a

nd

LYG

$57,

000

per Q

ALY

$35,

300

per L

YG<

3 tim

es th

e G

DP

per c

apita

of

Chin

a ($

16,3

00)

Not

cos

t-eff

ectiv

e*Sh

angh

ai H

ealth

Bu

reau

Chin

a90

114 Chapter 4.1

B (c

ontin

ued)

1st A

utho

r (re

fere

nce)

Pgx

test

Ana

lyti

cal

valid

ity

Pgx

test

repo

rted

(y

/n)

Out

com

em

easu

reQ

uant

itat

ive

outc

ome

oriC

ER (u

s$)

Cost

-eff

ecti

vene

ss

thre

shol

d ($

/Q

Aly

)

Conc

lusi

on

base

d on

ou

tcom

e

Fund

ing

Coun

try

QH

Es

scor

e

kauf

2010

(46)

HLA

-B *5

701

No

HSR

avo

ided

$300

(60

days

’ tim

e ho

rizon

)no

num

ber.

Cost

-effe

ctiv

eG

laxo

Sm

ith K

line,

In

c.U

SA87

Ratt

anav

ipa-

pong

2013

(47)

HLA

-B*1

502

Yes

QA

LYEp

ileps

y pa

tient

s:- T

HB2

20,0

00 ($

6,70

0)TH

B32,

522,

000

($98

4,80

0)N

euro

path

ic p

ain

patie

nts:

- TH

B130

,000

($3,

900)

- TH

B35,

877,

00

($10

8,60

0)

THB1

20,0

00

($3,

634)

PGx

test

is

cost

-effe

ctiv

e fo

r neu

ropa

thic

pa

in b

ut n

ot fo

r ep

ileps

y

Acad

emic

re

sour

ces

Thai

land

70

liu

2012

(48)

IL-2

8BN

oQ

ALY

$50,

400

No

num

ber

Not

cle

arAc

adem

ic a

nd

gove

rnm

enta

lU

SA83

gre

eley

2011

(49)

KCN

J11

and

ABCC

8Ye

sQ

ALY

Dom

inan

tn.

a.D

omin

ant

Acad

emic

re

sour

ces

USA

56

Part

han

2013

(50)

KIF6

No

QA

LY$4

5,00

0$1

00,0

00M

ay b

e co

st

effec

tive

Cele

ra

corp

orat

ion

USA

83

Vija

yara

-gh

avan

201

2 (5

1)

KRAS

Yes

LYS

- Cos

t sav

ing;

- Cos

t sav

ing;

- $35

,500

per

LYS.

No

num

ber

Cost

-sav

ing

in

both

US

and

Ger

man

y

Roch

e M

olec

ular

Sy

stem

s, In

c.U

SA a

nd

Ger

man

y75

Hag

aman

2010

(53)

TPM

TN

oQ

ALY

$29,

700

$50,

000

Cost

-effe

ctiv

eU

nkno

wn

USA

64

thom

pson

.20

14 (5

2)TP

MT

No

QA

LYN

egat

ive

ICER

n.a.

Cost

-sav

ing

but

also

hea

lth lo

ssD

epar

tmen

t of

Hea

lth U

KU

K88

Economic evaluations of pharmacogenetic tests. 115

B (c

ontin

ued)

1st A

utho

r (re

fere

nce)

Pgx

test

Ana

lyti

cal

valid

ity

Pgx

test

repo

rted

(y

/n)

Out

com

em

easu

reQ

uant

itat

ive

outc

ome

oriC

ER (u

s$)

Cost

-eff

ecti

vene

ss

thre

shol

d ($

/Q

Aly

)

Conc

lusi

on

base

d on

ou

tcom

e

Fund

ing

Coun

try

QH

Es

scor

e

Pich

erea

u20

10 (5

4)U

GT1

A1 *2

8N

one

utro

peni

a av

oide

d€9

00-1

,100

($11

00-

1300

)N

o nu

mbe

rCo

st-e

ffect

ive

No

spec

ific

finan

cial

sup

port

fo

r thi

s st

udy

Fran

ce84

ABC

C, A

TP-b

indi

ng c

asse

tte

tran

spor

ter

sub

fam

ily C

; APO

E, a

polip

opro

tein

-E; C

YP, c

ytoc

hrom

e P-

450;

EG

FR, e

pide

rmal

gro

wth

fact

or r

ecep

tor;

GPD

: gro

ss d

omes

tic

prod

uct;

HLA

, hum

an le

ukoc

yte

antig

en; H

TTLP

R, s

erot

onin

-tra

nspo

rter

-link

ed p

olym

orph

ic re

gion

; ICE

R, in

crem

enta

l cos

t-eff

ectiv

enes

s ra

tio; K

CNJ,

Pota

ssiu

m in

war

d-ly

-rec

tifyi

ng c

hann

el, s

ubfa

mily

J ;

KRA

S, K

irste

n ra

t sar

com

a vi

ral o

ncog

ene

hom

olog

; LYG

, life

-yea

rs-g

aine

d; L

YS, l

ife-y

ears

-sav

ed; P

Gx,

pha

rmac

ogen

etic

; QA

LW, q

ual-

ity-a

djus

ted-

life-

wee

k; Q

ALY

, qua

lity-

adju

sted

-life

-yea

r; TH

B, T

hai B

aht;

TPM

T, th

iopu

rine

S-m

ethy

ltran

sfer

ase;

UG

T, U

DP-

gluc

uron

osyl

tran

sfer

ase;

UK,

Uni

ted

King

dom

; VK

ORC

, Vita

min

K e

poxi

de re

duct

ase

com

plex

* With

the

gefit

inib

pat

ient

ass

ista

nce

prog

ram

(spo

nsor

ed th

erap

y af

ter fi

rst s

ix m

onth

s) it

mig

ht b

e a

cost

-effe

ctiv

e tr

eatm

ent o

ptio

n.

116 Chapter 4.1

0

2

4

6

8

10

12

14

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

Num

ber o

f stu

dies

pub

lishe

d

Year

CBA

CMA

CEA

CUA

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Type

of s

ensi

tivity

ana

lysi

s pe

rfor

med

(%)

Year

Probabilistic+ uni /multiProbabilistic

MultiplescenariosUnivariate(1-way)Other

None

A

B

Figure 2. Type of outcome analysis (A) and sensitivity analysis (B) of PGx studies from 2000- September 2014. Two studies performed both a CEA and a CUA.* two studies performed both a CEA and a CUACBE: Cost-benefit analysis; CEA: Cost-effectiveness analysis; CMA: Cost-minimization analysis; CUA: Cost-utility analysisCBE: Cost-benefit analysis; CEA: Cost-effectiveness analysis; CMA: Cost-minimization analysis; CUA: Cost-utility analysis.

Economic evaluations of pharmacogenetic tests. 117

genes investigated and analytical validity of Pgx test

In the period from 2000 until 2014, the majority of studies investigated the TPMT gene (17 studies). Among the studies presented in this update (i.e. from August 2010 onwards), most investigations concerned the CYP2C19 screening tests (fig. 3A). We noticed that the CYP2C19 gene was studied in many different scenarios. Some studies assessed if the use of a CYP2C19 independent drug like prasugrel or ticagrelor would be a cost-effective treatment option (35,36,38). Similar results were found for CYP2C9 and VKORC1 testing in combination with warfarin treatment, which was compared with CYP2C9 and VKORC1 independent novel oral anticoagulants, like dabigatran.(41).

In general the analytical validity of PGx tests is high (>95%), nevertheless variation in the analytical validity will in combination with the prevalence of a certain genetic trait determine the predictive value of a test (97). This can be of influence on the cost-effectiveness and was nicely demonstrated by Kauf et al., who identified the negative predictive value as an important input parameter for the costs-effectiveness of HLA-B*5701 genotyping in their sensitivity analysis (46). Nevertheless, underlying assump-tions about the analytical validity of the test were not reported. It was found that only 11 out of the 38 (30%) studies included since 2010, did report underlying analytical validity of the PGx test (table1).

Outcome of the studies

There were 21 out of 80 (26%) studies which concluded that PGx testing was dominant (i.e. resulting in clinical benefits as well as cost-savings). From 2010 onwards, most au-thors concluded that PGx testing was cost-effective, while only four studies concluded that it was not cost-effective (fig. 3B). In the period from 2010 until 2014, several studies provided the specific conditions at which genetic testing might become cost-effective (table 1A and B). For example, Dong et al. (2012) and Rattanavipapong et al. (2013) showed that the PGx tests could be cost-effective depending on either the patient population or the disease (23,47). Due to an imperfect capability of PGx tests to differ-entiate between carriers of a genetic variant, some patients might be misclassified and receive suboptimal treatment. As a result, two studies assessing KRAS and BRAF testing and one study assessing TMPT testing found the PGx testing strategy was cost-saving, but with a small health loss compared to the non-testing strategy (25,26,52).

time horizon and discounting

To capture all costs, savings and effects of an intervention, a lifelong time horizon often seems the best time horizon, although in some scenarios it may be argued that a shorter period is acceptable. Before 2008, 12 out of 26 studies (46%) applied a time horizon of 12 months, and only five (19%) studies applied a lifelong time window. Among the studies published since 2010, there was a broad range in the applied time horizons, from

118 Chapter 4.1

two weeks until lifetime (S1 Table). In addition, out of the 38 studies 17 (45%) applied a lifelong time horizon, six (16%) studies applied a 30 year time horizon of which two studies also used a lifetime horizon. To deal with uncertainty around the appropriated time horizon, different time horizons can be used. We found that six out of 38 studies (16%) varied their time horizons (18,20,26,38,46,49). In this way, insight into short- and long-term outcomes was given.

16%

10%

10%

11% 11%

8%

8%

26% CYP2C19

KRAS (+BRAF)

HLA-B

21 gene assay

CYP2C9 (+ VKORC1)

TMPT

EGFR

unmergeable (other)

55%

16%

13%

11% 5%

Cost-effective

Dominant

Cost-saving

Not cost-effective

Unclear

A

B

Figure 3. Genes analysed (A) and study outcomes (B) of papers published between August 2010 and Sep-tember 2014.

Economic evaluations of pharmacogenetic tests. 119

Discounting was applied for almost all studies published after 2006 that applied a time horizon longer than one year. The majority of the studies used discounting at 3% annually for costs and effects similarly (S1 Table).

Discussion of limitations

The limitations and uncertainties of an economic analysis have to be acknowledged in order to judge the study on its merits. Nearly all of the newly included papers (i.e. published between August 2010 and October 2014) discussed their limitations, uncer-tainties and possible shortcomings of their economic analysis (S2 Table). The topic most commonly discussed was the lack of solid clinical evidence. As a result, many studies had to make assumptions or relied on experts’ opinions. However, as mentioned be-fore, most authors did include different efficacy scenarios in their sensitivity analysis. Not all papers gave clear information about assumptions concerning sensitivity and specificity of the PGx tests, and only three out of 38 (8%) mentioned these assumptions as a limitation (45,49,51). Another typical limitation for the effects of PGx tests is the time in which test results will become available for clinical decision making. In current analyses, test results were often assumed to be immediately available. However, this might be an unrealistic assumption and only one out of 38 studies explicitly mentioned this assumption as a limitation (23). More general limitations were the lack of data with respect to heterogeneity in patient populations, hampering extrapolation of results to patients of different ethnicities, subpopulations and/or country specific populations. Moreover, several papers mentioned the difficulties in extrapolating long-term clinical utility results from the short-term clinical trials and lab studies.

Role of funding

Out of the 78 selected studies, 11 (14%) were funded by pharmaceutical companies. Before 2008, none of the studies was (directly) funded by pharmaceutical companies. In 2008 and 2009 only two (12.5%) out of 16 studies published in that period were funded by a pharmaceutical company (70,71). Between August 2010 and September 2014, 9 (24%) out of the 38 selected studies were funded by pharmaceutical companies (table 1A and B). Regarding outcomes and conclusions all of these studies concluded that PGx tests were dominant, cost-saving, or cost-effective. Among the remaining studies which were funded by other resources, 14% concluded PGx tests were not cost-effective.

Quality assessment

The studies included through this update received a quality score according to the QHES. There was 2% disagreement between the reviewers for which consensus was reached through discussion. The average quality score was 76. The score which was given to each study is shown in table 1A and B. The majority of the studies (71%) were of high quality.

120 Chapter 4.1

On average, studies concerning testing for EGFR received the highest rating and studies about the 21-gene assay received the lowest rating. Some items were scored negative for the majority of the studies. One of them was the item concerning the perspective of the analysis (i.e. societal, health care payers, etcetera). Most studies did not explain why the perspective of the analysis was chosen. In addition, low scoring was received for separate reporting of the short-and long-term outcomes. Most studies did not make such a distinction. Lastly, the direction and magnitude of potential biases were often not discussed.

4.1.4. DisCussiOn

Principal findings

Since 2004, there is an increase in the number of studies evaluating the economic value of PGx tests and this increase accelerated from 2008 onwards. There were not that many economic evaluations of PGx tests available as one might expect given the unravelling of the whole human genome in 2003, the clinical possibilities, and the fast development and decreasing costs of genetic tests (98). This could be related to limited implemen-tation of pharmacogenetic knowledge into daily clinical practice (5). Reasons for this are the uncertainty about clinical relevance, concerns about the availability of genetic data and considerable differences in cost-effectiveness which are found, in particular between countries (5,99,100).

Many studies included through this update did not assess the intrinsic value of the PGx test itself, but compared a PGx test treatment combination with an alternative treat-ment. For example, the tests for CYP2C19 or, both CYP2C9 and VKORC1 were incorporated in several models as the current treatment option in combination with clopidogrel or warfarin treatment, respectively. The alternative treatment options, which were as-sessed in the included studies, were independent of pharmacogenetics and were found to be cost-effective. Therefore, in cost-effectiveness analysis of antithrombotic therapy, there seems an ongoing movement away from pharmacogenetic testing, towards treatment options with compounds that are, so far, considered to be independent of pharmacogenetics.

Before 2008, most analyses were cost-effectiveness analyses, however since 2008 there has been a trend towards the use of cost-utility analyses. Cost-utility analysis is currently considered as the preferred type of analysis for health care choices as is ad-vised in several national guidelines, although other types can be suitable depending on the specific study (101,102). Before 2008, most studies performed only a simple univariate sensitivity analysis, if any (9). We found that more recent studies performed both univariate and probabilistic sensitivity analyses. This combination is also advised in

Economic evaluations of pharmacogenetic tests. 121

several national guidelines for pharmacoeconomic evaluation and is a part of the QHES checklist (16,101,102). As a result of these more comprehensive sensitivity analyses, the quality of economic evaluations appears to be improved over the last decade. However, we found that differences in genotyping costs were not always included in the sensitiv-ity analysis which leaves room for improvement.

Among the new studies included in this update (i.e. since 2010), considerable differ-ences in the length of the time horizon were noticed, varying from two weeks to lifelong. A time horizon shorter than a year was primarily related to the expected relevant clinical outcomes. Interestingly, there were differences in the time horizon between studies in-vestigating the same genes. For example, Kazi et al. and Panattoni et al. applied a lifelong time horizon (34,39), whereas others applied a time horizon of 15 months (35,38). All studies discussed at least some limitations and a lack of robust clinical evidence was most frequently mentioned as an important limitation contributing to uncertainty in the analysis. Although studies frequently made assumptions about analytical validity of PGx tests as well as about the rapid clinical availability, these were often not discussed as limitations or studied in sensitivity or scenario analyses. Besides the analytical validity, the clinical validity of PGx tests is important. In general this is the same as an effect estimate of a PGx test. As such, the clinical validity is correctly embedded in an eco-nomic evaluation. However, this does not apply for all PGx tests, because some PGx tests involve an analysis of multiple genetic variations. These variations can all contribute to a similar genotype prediction and therefore the clinical validity depends on the number and type of variant alleles analysed. For example, CYP2C19 genotyping which depends on general molecular genetic analysis of single nuclear polymorphisms to detect variant alleles. Several studies included in this review studied only the CYP2C19*2 allele (37-39). However, other allele like the *3 allele can also effect CYP2C19 activity and give a similar clinical effect as the *2 allele (103). When more variant alleles are analysed the clinical validity of the PGx test will increase, although some variants are rare and will contribute little. With the ongoing and rapid increase of knowledge about variant alleles, it is impor-tant economic evaluations report the variant alleles on which their assumptions about effects of genotyping (i.e. clinical validity) were based. For the example of CYP2C19, two out of the six studies involving the CYP2C19 gene, did not specify which alleles were included (34,36). This hampers the extrapolation of their findings towards populations of a different ethnicity and other laboratories.

Compared to studies published before 2008, studies were more commonly funded by pharmaceutical companies. Interestingly, all of these sponsored studies concluded that PGx tests were dominant, cost-saving, or cost-effective whereas the few studies which concluded otherwise were financed by academic, governmental, or unknown resources. It is known that studies funded by pharmaceutical companies publish more positive results when compared to studies funded by other resources which is in line with our

122 Chapter 4.1

results (104). These positive biased results are not related towards the quality of the studies, but to the comparison which is made or publication bias (105). For many of the economic evaluations in this review, assumptions about the effect of the genetic test were made. Furthermore, analytical validity was often not included in the model. This leaves room to bias result in favour of a preferred treatment strategy. Therefore, atten-tion should be given toward assumptions about these aspects, especially when studies are funded by pharmaceutical companies or if the funding is not reported.

Most studies included in this review concluded that the application of PGx tests was cost-effective. Yet, the conclusions were not unambiguous, often due to the uncertain-ties in the economic models. Another reason for this was that most of the newly included studies did not assess the intrinsic value of the PGx test itself, but a scenario involving one or more PGx tests. Although such estimates are important for the economic impact of the application of personalized medicine, they do not provide information about the cost-effectiveness of the PGx test itself and therefore outcomes can be different. To improve generalizability between studies, an additional scenario analysis in which equal costs and efficacy of the compared treatment strategies are assumed, to assess the intrinsic value of the PGx tests, could be used. Such an approach would especially be interesting when the comparator drug is under patent and drug costs are likely to decrease in the future. Another aspect creating different conclusions was the genetic variety between study populations. This was clearly described by some papers which mentioned the specific conditions like a specific geographic region or a disease for which the genetic testing was cost-effective. For example, PGx testing for HLA-B*5702 was cost-effective in Singaporean Chinese and Malays, but not in Singaporean Indians (23). In addition, Rattanavipapong et al. (2013) found that HLA-B*5702 testing was cost-effective in epileptic patients, but not in neuropathic pain patients (47).

Comparison with previous literature reviews

For the new studies included in this update (i.e. since 2010), an average QHES-quality score of 76 was found. This was is in line with the results from the review from Wong et al. about pharmacoeconomics of PGx tests. They found an average quality score of 77 (13). However, it was lower compared to a related review of Djalalov et al., who found an average quality score of 90 (10). This is likely due to the subjective nature of some items in the quality assessment. For example, ‘item 3’ asks if the used estimates for the analysis were from the best available source, and is therefore quite sensitive to the interpreta-tion of the reviewer (16). Previous review studies pointed out that the methodology of economic evaluations of PGx tests is often heterogeneous and of insufficient quality (9,11). Assasi et al. found that economic PGx studies which were of low quality (i.e. QHES score of < 50), frequently failed to handle uncertainty, did not inform about the study’s limitations, and did not discuss direction and magnitude of potential uncertainty (11).

Economic evaluations of pharmacogenetic tests. 123

Among the studies included in this update (i.e. since 2010), most of the authors discussed uncertainties and except for the study by Tiamkao et al., all incorporated uncertainty in a sensitivity analysis (24). However, direction and magnitude of potential bias was still not sufficiently discussed and remains a major point for improvement.

limitations of our approach

This review has some limitations. Firstly, we did not include studies from other databases than PubMed or grey literature and we only included English written studies. There-fore some studies might have been missed. In general, studies that are not indexed in MEDLINE or written in English do not have a large impact on reviews’ outcomes (106). Nevertheless, these studies are frequently of lower quality, and therefore the average quality of the studies included in our review might have been slightly overestimated. However, based on the comparison with other reviews we found a slightly lower aver-age quality score. Lastly, publication bias can always influence the findings of a review. Therefore cost-effectiveness of PGx tests could be overestimated.

Recommendations

Based on the quality assessment, reporting of the reasons behind the chosen perspec-tive and the type of economic model can improve the quality of economic evaluations of PGx tests. In addition, reporting of both short-and long-term outcomes and the influ-ence of potential bias, in terms of direction and magnitude on the cost-effectiveness estimates could be improved. Although a substantial and persistent increase in the use of both univariate and probabilistic sensitivity analyses was observed since 2008, there is still room for improvement by using a combination of these techniques instead of one technique, among several studies published since 2008. Publication bias or biased comparators might favour cost-effectiveness of PGx tests. Among studies funded by companies with conflicting interests, the risk on this kind of bias should be critically assessed. In our previous reviews the main limitation that was identified was the un-availability of clinical evidence (1,9). Although this remains an important issue, based on our new findings we can add some recommendations which are more applicable to implementation of PGx tests in clinical practice. First, the clinical validity of a test, i.e. the capability of the test to predict phenotypes with a clinical effect and the analytical validity should be reported as is recommended by the US Academy of Managed Care Pharmacy (107). Both parameters should be included in a sensitivity analysis. In ad-dition, the variant alleles on which these parameters were based should be reported. For some PGx tests, these estimates might be unknown. In this case, a better approach towards this problem would be inclusion of an univariate sensitivity analysis with dif-ferent cut-off values for the analytical and clinical validity of the PGx test. In this way, a minimum for the analytical and clinical validity can be generated. Note that with an

124 Chapter 4.1

increase in analysed alleles and as such the clinical validity, the price for the PGx test usually increases as well. Secondly, different turnaround times in which PGx test results would become available for healthcare professionals after requesting the test would be very informative. If direct availability of the genetic test is assumed, for example in the case of pre-emptive genotyping, this should be clearly stated in the method section. This way, the generalizability of results to other countries where PGx tests are available would improve. In addition, a range of costs for the genetic test should be evaluated in univariate and if applicable probabilistic sensitivity analyses. Furthermore, we rec-ommend the addition of a scenario analysis in which drug costs between comparator groups are equalized to give information about the intrinsic value of the PGx test itself.

4.1.5. COnClusiOn

Application of PGx tests was mostly found to be a cost-effective or cost-saving strategy, although some studies concluded otherwise which underlines the importance of future studies assessing the cost-effectiveness of PGx tests. We found that only the minority of recent pharmacoeconomic evaluations assessed the intrinsic value of the PGx tests. New compounds that are not affected by genetics, are emerging as cost-effective alternatives for pharmacogenetic testing strategies. Over the last decade, there was an increase in the number of studies and in the reporting of quality associated characteristics. Due to rapid development in analytical techniques, reporting of analytical and clinical validity of the assessed PGx test is recommended for future evaluations. Furthermore robust clinical evidence regarding PGx tests’ efficacy is warranted.

Economic evaluations of pharmacogenetic tests. 125

4.1.6. REFEREnCEs

1 Vegter S, Jansen E, Postma MJ, Boersma C. Economic evaluations of pharmacogenetic and genomic screening programs: update of the literature. Drug Development Research 2010; 71: 492-501.

2 Cascorbi I, Bruhn O, Werk AN. Challenges in pharmacogenetics. Eur J Clin Pharmacol 2013 May; 69 Suppl 1: 17-23.

3 Ross S, Anand SS, Joseph P, Pare G. Promises and challenges of pharmacogenetics: an overview of study design, methodological and statistical issues. JRSM Cardiovasc Dis 2012 Apr 5; 1(1): 10.1258/cvd.2012.012001.

4 Garrison LP,Jr, Carlson RJ, Carlson JJ, Kuszler PC, Meckley LM, Veenstra DL. A review of public policy issues in promoting the development and commercialization of pharmacogenomic ap-plications: challenges and implications. Drug Metab Rev 2008; 40(2): 377-401.

5 Johnson JA, Burkley BM, Langaee TY, Clare-Salzler MJ, Klein TE, Altman RB. Implementing per-sonalized medicine: development of a cost-effective customized pharmacogenetics genotyping array. Clin Pharmacol Ther 2012 Oct; 92(4): 437-439.

6 Phillips KA, Ann Sakowski J, Trosman J, Douglas MP, Liang SY, Neumann P. The economic value of personalized medicine tests: what we know and what we need to know. Genet Med 2014 Mar; 16(3): 251-257.

7 Husereau D, Marshall DA, Levy AR, Peacock S, Hoch JS. Health technology assessment and per-sonalized medicine: are economic evaluation guidelines sufficient to support decision making? Int J Technol Assess Health Care 2014 Apr; 30(2): 179-187.

8 Hatz MH, Schremser K, Rogowski WH. Is individualized medicine more cost-effective? A system-atic review. Pharmacoeconomics 2014 May; 32(5): 443-455.

9 Vegter S, Boersma C, Rozenbaum M, Wilffert B, Navis G, Postma MJ. Pharmacoeconomic evalua-tions of pharmacogenetic and genomic screening programmes: a systematic review on content and adherence to guidelines. Pharmacoeconomics 2008; 26(7): 569-587.

10 Djalalov S, Musa Z, Mendelson M, Siminovitch K, Hoch J. A review of economic evaluations of genetic testing services and interventions (2004-2009). Genet Med 2011 Feb; 13(2): 89-94.

11 Assasi N, Schwartz L, Tarride JE, Goeree R, Xie F. Economic evaluations conducted for assessment of genetic testing technologies: a systematic review. Genet Test Mol Biomarkers 2012 Nov; 16(11): 1322-1335.

12 Beaulieu M, de Denus S, Lachaine J. Systematic review of pharmacoeconomic studies of pharma-cogenomic tests. Pharmacogenomics 2010 Nov; 11(11): 1573-1590.

13 Wong WB, Carlson JJ, Thariani R, Veenstra DL. Cost effectiveness of pharmacogenomics: a critical and systematic review. Pharmacoeconomics 2010; 28(11): 1001-1013.

14 Phillips KA, Ann Sakowski J, Trosman J, Douglas MP, Liang SY, Neumann P. The economic value of personalized medicine tests: what we know and what we need to know. Genet Med 2014 Mar; 16(3): 251-257.

15 Shiroiwa T, Motoo Y, Tsutani K. Cost-effectiveness analysis of KRAS testing and cetuximab as last-line therapy for colorectal cancer. Mol Diagn Ther 2010 Dec 1; 14(6): 375-384.

16 Chiou CF, Hay JW, Wallace JF, Bloom BS, Neumann PJ, Sullivan SD, et al. Development and valida-tion of a grading system for the quality of cost-effectiveness studies. Med Care 2003 Jan; 41(1): 32-44.

17 Moher D, Liberati A, Tetzlaff J, Altman DG, PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 2009 Jul 21; 6(7): e1000097.

126 Chapter 4.1

18 Klang SH, Hammerman A, Liebermann N, Efrat N, Doberne J, Hornberger J. Economic implica-tions of 21-gene breast cancer risk assay from the perspective of an Israeli-managed health-care organization. Value Health 2010 Jun-Jul; 13(4): 381-387.

19 Bacchi CE, Prisco F, Carvalho FM, Ojopi EB, Saad ED. Potential economic impact of the 21-gene expression assay on the treatment of breast cancer in Brazil. Rev Assoc Med Bras 2010 Mar-Apr; 56(2): 186-191.

20 Hall PS, McCabe C, Stein RC, Cameron D. Economic evaluation of genomic test-directed chemo-therapy for early-stage lymph node-positive breast cancer. J Natl Cancer Inst 2012 Jan 4; 104(1): 56-66.

21 Vanderlaan BF, Broder MS, Chang EY, Oratz R, Bentley TG. Cost-effectiveness of 21-gene assay in node-positive, early-stage breast cancer. Am J Manag Care 2011; 17(7): 455-464.

22 Verhoef TI, Redekop WK, Veenstra DL, Thariani R, Beltman PA, van Schie RM, et al. Cost-effectiveness of pharmacogenetic-guided dosing of phenprocoumon in atrial fibrillation. Pharmacogenomics 2013 Jun; 14(8): 869-883.

23 Dong D, Sung C, Finkelstein EA. Cost-effectiveness of HLA-B*1502 genotyping in adult patients with newly diagnosed epilepsy in Singapore. Neurology 2012 Sep 18; 79(12): 1259-1267.

24 Tiamkao S, Jitpimolmard J, Sawanyawisuth K, Jitpimolmard S. Cost minimization of HLA-B*1502 screening before prescribing carbamazepine in Thailand. Int J Clin Pharm 2013 Aug; 35(4): 608-612.

25 Blank PR, Moch H, Szucs TD, Schwenkglenks M. KRAS and BRAF mutation analysis in metastatic colorectal cancer: a cost-effectiveness analysis from a Swiss perspective. Clin Cancer Res 2011 Oct 1; 17(19): 6338-6346.

26 Behl AS, Goddard KA, Flottemesch TJ, Veenstra D, Meenan RT, Lin JS, et al. Cost-effectiveness analysis of screening for KRAS and BRAF mutations in metastatic colorectal cancer. J Natl Cancer Inst 2012 Dec 5; 104(23): 1785-1795.

27 Shiffman D, Slawsky K, Fusfeld L, Devlin JJ, Goss TF. Cost-effectiveness model of use of genetic testing as an aid in assessing the likely benefit of aspirin therapy for primary prevention of cardio-vascular disease. Clin Ther 2012 Jun; 34(6): 1387-1394.

28 Donnan JR, Ungar WJ, Mathews M, Hancock-Howard RL, Rahman P. A cost effectiveness analysis of thiopurine methyltransferase testing for guiding 6-mercaptopurine dosing in children with acute lymphoblastic leukemia. Pediatr Blood Cancer 2011 Aug; 57(2): 231-239.

29 Schackman BR, Haas DW, Becker JE, Berkowitz BK, Sax PE, Daar ES, et al. Cost-effectiveness analysis of UGT1A1 genetic testing to inform antiretroviral prescribing in HIV disease. Antivir Ther 2013; 18(3): 399-408.

30 Olgiati P, Bajo E, Bigelli M, De Ronchi D, Serretti A. Should pharmacogenetics be incorporated in major depression treatment? Economic evaluation in high- and middle-income European countries. Prog Neuropsychopharmacol Biol Psychiatry 2012 Jan 10; 36(1): 147-154.

31 Serretti A, Olgiati P, Bajo E, Bigelli M, De Ronchi D. A model to incorporate genetic testing (5-HT-TLPR) in pharmacological treatment of major depressive disorders. World J Biol Psychiatry 2011 Oct; 12(7): 501-515.

32 Reed SD, Scales CD,Jr, Stewart SB, Sun J, Moul JW, Schulman KA, et al. Effects of family history and genetic polymorphism on the cost-effectiveness of chemoprevention with finasteride for prostate cancer. J Urol 2011 Mar; 185(3): 841-847.

33 Djalalov S, Yong J, Beca J, Black S, Saposnik G, Musa Z, et al. Genetic testing in combination with preventive donepezil treatment for patients with amnestic mild cognitive impairment: an explor-atory economic evaluation of personalized medicine. Mol Diagn Ther 2012 Dec; 16(6): 389-399.

Economic evaluations of pharmacogenetic tests. 127

34 Kazi DS, Garber AM, Shah RU, Dudley RA, Mell MW, Rhee C, et al. Cost-effectiveness of genotype-guided and dual antiplatelet therapies in acute coronary syndrome. Ann Intern Med 2014 Feb 18; 160(4): 221-232.

35 Reese ES, Daniel Mullins C, Beitelshees AL, Onukwugha E. Cost-effectiveness of cytochrome P450 2C19 genotype screening for selection of antiplatelet therapy with clopidogrel or prasugrel. Pharmacotherapy 2012 Apr; 32(4): 323-332.

36 Sorich MJ, Horowitz JD, Sorich W, Wiese MD, Pekarsky B, Karnon JD. Cost-effectiveness of using CYP2C19 genotype to guide selection of clopidogrel or ticagrelor in Australia. Pharmacogenom-ics 2013 Dec; 14(16): 2013-2021.

37 Crespin DJ, Federspiel JJ, Biddle AK, Jonas DE, Rossi JS. Ticagrelor versus genotype-driven anti-platelet therapy for secondary prevention after acute coronary syndrome: a cost-effectiveness analysis. Value Health 2011 Jun; 14(4): 483-491.

38 Lala A, Berger JS, Sharma G, Hochman JS, Scott Braithwaite R, Ladapo JA. Genetic testing in patients with acute coronary syndrome undergoing percutaneous coronary intervention: a cost-effectiveness analysis. J Thromb Haemost 2013 Jan; 11(1): 81-91.

39 Panattoni L, Brown PM, Te Ao B, Webster M, Gladding P. The cost effectiveness of genetic test-ing for CYP2C19 variants to guide thienopyridine treatment in patients with acute coronary syndromes: a New Zealand evaluation. Pharmacoeconomics 2012 Nov 1; 30(11): 1067-1084.

40 Pink J, Pirmohamed M, Lane S, Hughes DA. Cost-effectiveness of pharmacogenetics-guided warfarin therapy vs. alternative anticoagulation in atrial fibrillation. Clin Pharmacol Ther 2014 Feb; 95(2): 199-207.

41 You JH, Tsui KK, Wong RS, Cheng G. Cost-effectiveness of dabigatran versus genotype-guided management of warfarin therapy for stroke prevention in patients with atrial fibrillation. PLoS One 2012; 7(6): e39640.

42 You JH. Pharmacogenetic-guided selection of warfarin versus novel oral anticoagulants for stroke prevention in patients with atrial fibrillation: a cost-effectiveness analysis. Pharmacogenetics and genomics 2014; 24(1): 6-14.

43 de Lima Lopes G,Jr, Segel JE, Tan DS, Do YK, Mok T, Finkelstein EA. Cost-effectiveness of epidermal growth factor receptor mutation testing and first-line treatment with gefitinib for patients with advanced adenocarcinoma of the lung. Cancer 2012 Feb 15; 118(4): 1032-1039.

44 Handorf EA, McElligott S, Vachani A, Langer CJ, Bristol Demeter M, Armstrong K, et al. Cost ef-fectiveness of personalized therapy for first-line treatment of stage IV and recurrent incurable adenocarcinoma of the lung. J Oncol Pract 2012 Sep; 8(5): 267-274.

45 Zhu J, Li T, Wang X, Ye M, Cai J, Xu Y, et al. Gene-guided gefitinib switch maintenance therapy for patients with advanced EGFR mutation-positive non-small cell lung cancer: an economic analysis. BMC Cancer 2013 Jan 29; 13: 39-2407-13-39.

46 Kauf TL, Farkouh RA, Earnshaw SR, Watson ME, Maroudas P, Chambers MG. Economic efficiency of genetic screening to inform the use of abacavir sulfate in the treatment of HIV. Pharmacoeco-nomics 2010; 28(11): 1025-1039.

47 Rattanavipapong W, Koopitakkajorn T, Praditsitthikorn N, Mahasirimongkol S, Teerawattananon Y. Economic evaluation of HLA-B*15: 02 screening for carbamazepine-induced severe adverse drug reactions in Thailand. Epilepsia 2013 Sep; 54(9): 1628-1638.

48 Liu S, Cipriano LE, Holodniy M, Owens DK, Goldhaber-Fiebert JD. New protease inhibitors for the treatment of chronic hepatitis C: a cost-effectiveness analysis. Ann Intern Med 2012 Feb 21; 156(4): 279-290.

128 Chapter 4.1

49 Greeley SA, John PM, Winn AN, Ornelas J, Lipton RB, Philipson LH, et al. The cost-effectiveness of personalized genetic medicine: the case of genetic testing in neonatal diabetes. Diabetes Care 2011 Mar; 34(3): 622-627.

50 Parthan A, Leahy KJ, O’Sullivan AK, Iakoubova OA, Bare LA, Devlin JJ, et al. Cost effectiveness of targeted high-dose atorvastatin therapy following genotype testing in patients with acute coronary syndrome. Pharmacoeconomics 2013 Jun; 31(6): 519-531.

51 Vijayaraghavan A, Efrusy MB, Goke B, Kirchner T, Santas CC, Goldberg RM. Cost-effectiveness of KRAS testing in metastatic colorectal cancer patients in the United States and Germany. Int J Cancer 2012 Jul 15; 131(2): 438-445.

52 Thompson AJ, Newman WG, Elliott RA, Roberts SA, Tricker K, Payne K. The cost-effectiveness of a pharmacogenetic test: a trial-based evaluation of TPMT genotyping for azathioprine. Value Health 2014 Jan-Feb; 17(1): 22-33.

53 Hagaman JT, Kinder BW, Eckman MH. Thiopurine S- methyltransferase [corrected] testing in idiopathic pulmonary fibrosis: a pharmacogenetic cost-effectiveness analysis. Lung 2010 Apr; 188(2): 125-132.

54 Pichereau S, Le Louarn A, Lecomte T, Blasco H, Le Guellec C, Bourgoin H. Cost-effectiveness of UGT1A1*28 genotyping in preventing severe neutropenia following FOLFIRI therapy in colorec-tal cancer. J Pharm Pharm Sci 2010; 13(4): 615-625.

55 Blank PR, Schwenkglenks M, Moch H, Szucs TD. Human epidermal growth factor receptor 2 ex-pression in early breast cancer patients: a Swiss cost-effectiveness analysis of different predictive assay strategies. Breast Cancer Res Treat 2010 Nov; 124(2): 497-507.

56 Carlson JJ, Garrison LP, Ramsey SD, Veenstra DL. The potential clinical and economic outcomes of pharmacogenomic approaches to EGFR-tyrosine kinase inhibitor therapy in non-small-cell lung cancer. Value Health 2009 Jan-Feb; 12(1): 20-27.

57 Chou WH, Yan FX, de Leon J, Barnhill J, Rogers T, Cronin M, et al. Extension of a pilot study: impact from the cytochrome P450 2D6 polymorphism on outcome and costs associated with severe mental illness. J Clin Psychopharmacol 2000 Apr; 20(2): 246-251.

58 Cosler LE, Lyman GH. Economic analysis of gene expression profile data to guide adjuvant treat-ment in women with early-stage breast cancer. Cancer Invest 2009 Dec; 27(10): 953-959.

59 Costa-Scharplatz M, van Asselt AD, Bachmann LM, Kessels AG, Severens JL. Cost-effectiveness of pharmacogenetic testing to predict treatment response to angiotensin-converting enzyme inhibitor. Pharmacogenet Genomics 2007 May; 17(5): 359-368.

60 Dendukuri N, Khetani K, McIsaac M, Brophy J. Testing for HER2-positive breast cancer: a system-atic review and cost-effectiveness analysis. CMAJ 2007 May 8; 176(10): 1429-1434.

61 Desta Z, Zhao X, Shin JG, Flockhart DA. Clinical significance of the cytochrome P450 2C19 genetic polymorphism. Clin Pharmacokinet 2002; 41(12): 913-958.

62 Dubinsky MC, Reyes E, Ofman J, Chiou CF, Wade S, Sandborn WJ. A cost-effectiveness analysis of alternative disease management strategies in patients with Crohn’s disease treated with azathio-prine or 6-mercaptopurine. Am J Gastroenterol 2005 Oct; 100(10): 2239-2247.

63 Eckman MH, Rosand J, Greenberg SM, Gage BF. Cost-effectiveness of using pharmacogenetic information in warfarin dosing for patients with nonvalvular atrial fibrillation. Ann Intern Med 2009 Jan 20; 150(2): 73-83.

64 Elkin EB, Weinstein MC, Winer EP, Kuntz KM, Schnitt SJ, Weeks JC. HER-2 testing and trastuzumab therapy for metastatic breast cancer: a cost-effectiveness analysis. J Clin Oncol 2004 Mar 1; 22(5): 854-863.

Economic evaluations of pharmacogenetic tests. 129

65 Furuta T, Shirai N, Kodaira M, Sugimoto M, Nogaki A, Kuriyama S, et al. Pharmacogenomics-based tailored versus standard therapeutic regimen for eradication of H. pylori. Clin Pharmacol Ther 2007 Apr; 81(4): 521-528.

66 Gold HT, Hall MJ, Blinder V, Schackman BR. Cost effectiveness of pharmacogenetic testing for uridine diphosphate glucuronosyltransferase 1A1 before irinotecan administration for metastatic colorectal cancer. Cancer 2009 Sep 1; 115(17): 3858-3867.

67 Hughes DA, Vilar FJ, Ward CC, Alfirevic A, Park BK, Pirmohamed M. Cost-effectiveness analysis of HLA B*5701 genotyping in preventing abacavir hypersensitivity. Pharmacogenetics 2004 Jun; 14(6): 335-342.

68 Kim SK, Jun JB, El-Sohemy A, Bae SC. Cost-effectiveness analysis of MTHFR polymorphism screening by polymerase chain reaction in Korean patients with rheumatoid arthritis receiving methotrexate. J Rheumatol 2006 Jul; 33(7): 1266-1274.

69 Lehmann DF, Medicis JJ, Franklin PD. Polymorphisms and the pocketbook: the cost-effectiveness of cytochrome P450 2C19 genotyping in the eradication of Helicobacter pylori infection associ-ated with duodenal ulcer. J Clin Pharmacol 2003 Dec; 43(12): 1316-1323.

70 Lidgren M, Jonsson B, Rehnberg C, Willking N, Bergh J. Cost-effectiveness of HER2 testing and 1-year adjuvant trastuzumab therapy for early breast cancer. Ann Oncol 2008 Mar; 19(3): 487-495.

71 Lidgren M, Wilking N, Jonsson B, Rehnberg C. Cost-effectiveness of HER2 testing and trastuzumab therapy for metastatic breast cancer. Acta Oncol 2008; 47(6): 1018-1028.

72 Maitland-van der Zee AH, Klungel OH, Stricker BH, Veenstra DL, Kastelein JJ, Hofman A, et al. Pharmacoeconomic evaluation of testing for angiotensin-converting enzyme genotype before starting beta-hydroxy-beta-methylglutaryl coenzyme A reductase inhibitor therapy in men. Pharmacogenetics 2004 Jan; 14(1): 53-60.

73 Marra CA, Esdaile JM, Anis AH. Practical pharmacogenetics: the cost effectiveness of screening for thiopurine s-methyltransferase polymorphisms in patients with rheumatological conditions treated with azathioprine. J Rheumatol 2002 Dec; 29(12): 2507-2512.

74 Meckley LM, Veenstra DL. Screening for the alpha-adducin Gly460Trp variant in hypertensive patients: a cost-effectiveness analysis. Pharmacogenet Genomics 2006 Feb; 16(2): 139-147.

75 Meckley LM, Gudgeon JM, Anderson JL, Williams MS, Veenstra DL. A policy model to evaluate the benefits, risks and costs of warfarin pharmacogenomic testing. Pharmacoeconomics 2010; 28(1): 61-74.

76 Morelle M, Hasle E, Treilleux I, Michot JP, Bachelot T, Penault-Llorca F, et al. Cost-effectiveness analysis of strategies for HER2 testing of breast cancer patients in France. Int J Technol Assess Health Care 2006 Summer; 22(3): 396-401.

77 Oh KT, Anis AH, Bae SC. Pharmacoeconomic analysis of thiopurine methyltransferase poly-morphism screening by polymerase chain reaction for treatment with azathioprine in Korea. Rheumatology (Oxford) 2004 Feb; 43(2): 156-163.

78 Patrick AR, Avorn J, Choudhry NK. Cost-effectiveness of genotype-guided warfarin dosing for patients with atrial fibrillation. Circ Cardiovasc Qual Outcomes 2009 Sep; 2(5): 429-436.

79 Perlis RH, Ganz DA, Avorn J, Schneeweiss S, Glynn RJ, Smoller JW, et al. Pharmacogenetic testing in the clinical management of schizophrenia: a decision-analytic model. J Clin Psychopharmacol 2005 Oct; 25(5): 427-434.

80 Perlis RH, Patrick A, Smoller JW, Wang PS. When is pharmacogenetic testing for antidepressant response ready for the clinic? A cost-effectiveness analysis based on data from the STAR*D study. Neuropsychopharmacology 2009 Sep; 34(10): 2227-2236.

130 Chapter 4.1

81 Priest VL, Begg EJ, Gardiner SJ, Frampton CM, Gearry RB, Barclay ML, et al. Pharmacoeconomic analyses of azathioprine, methotrexate and prospective pharmacogenetic testing for the man-agement of inflammatory bowel disease. Pharmacoeconomics 2006; 24(8): 767-781.

82 Sax PE, Islam R, Walensky RP, Losina E, Weinstein MC, Goldie SJ, et al. Should resistance testing be performed for treatment-naive HIV-infected patients? A cost-effectiveness analysis. Clin Infect Dis 2005 Nov 1; 41(9): 1316-1323.

83 Schackman BR, Scott CA, Walensky RP, Losina E, Freedberg KA, Sax PE. The cost-effectiveness of HLA-B*5701 genetic screening to guide initial antiretroviral therapy for HIV. AIDS 2008 Oct 1; 22(15): 2025-2033.

84 Schalekamp T, Boink GJ, Visser LE, Stricker BH, de Boer A, Klungel OH. CYP2C9 genotyping in acenocoumarol treatment: is it a cost-effective addition to international normalized ratio moni-toring? Clin Pharmacol Ther 2006 Jun; 79(6): 511-520.

85 Sendi P, Gunthard HF, Simcock M, Ledergerber B, Schupbach J, Battegay M, et al. Cost-effective-ness of genotypic antiretroviral resistance testing in HIV-infected patients with treatment failure. PLoS One 2007 Jan 24; 2(1): e173.

86 Siebert U, Sroczynski G, Aidelsburger P, Rossol S, Wasem J, Manns MP, et al. Clinical effectiveness and cost effectiveness of tailoring chronic hepatitis C treatment with peginterferon alpha-2b plus ribavirin to HCV genotype and early viral response: a decision analysis based on German guidelines. Pharmacoeconomics 2009; 27(4): 341-354.

87 Smith KJ, Monsef BS, Ragni MV. Should female relatives of factor V Leiden carriers be screened prior to oral contraceptive use? A cost-effectiveness analysis. Thromb Haemost 2008 Sep; 100(3): 447-452.

88 Tavadia SM, Mydlarski PR, Reis MD, Mittmann N, Pinkerton PH, Shear N, et al. Screening for azathioprine toxicity: a pharmacoeconomic analysis based on a target case. J Am Acad Dermatol 2000 Apr; 42(4): 628-632.

89 van den Akker-van Marle ME, Gurwitz D, Detmar SB, Enzing CM, Hopkins MM, Gutierrez de Mesa E, et al. Cost-effectiveness of pharmacogenomics in clinical practice: a case study of thiopurine methyltransferase genotyping in acute lymphoblastic leukemia in Europe. Pharmacogenomics 2006 Jul; 7(5): 783-792.

90 Veenstra DL, Harris J, Gibson RL, Rosenfeld M, Burke W, Watts C. Pharmacogenomic testing to prevent aminoglycoside-induced hearing loss in cystic fibrosis patients: potential impact on clinical, patient, and economic outcomes. Genet Med 2007 Oct; 9(10): 695-704.

91 Vegter S, Perna A, Hiddema W, Ruggenenti P, Remuzzi G, Navis G, et al. Cost-effectiveness of ACE inhibitor therapy to prevent dialysis in nondiabetic nephropathy: influence of the ACE insertion/deletion polymorphism. Pharmacogenet Genomics 2009 Sep; 19(9): 695-703.

92 Weinstein MC, Goldie SJ, Losina E, Cohen CJ, Baxter JD, Zhang H, et al. Use of genotypic resistance testing to guide hiv therapy: clinical impact and cost-effectiveness. Ann Intern Med 2001 Mar 20; 134(6): 440-450.

93 Welton NJ, Johnstone EC, David SP, Munafo MR. A cost-effectiveness analysis of genetic testing of the DRD2 Taq1A polymorphism to aid treatment choice for smoking cessation. Nicotine Tob Res 2008 Jan; 10(1): 231-240.

94 Winter J, Walker A, Shapiro D, Gaffney D, Spooner RJ, Mills PR. Cost-effectiveness of thiopurine methyltransferase genotype screening in patients about to commence azathioprine therapy for treatment of inflammatory bowel disease. Aliment Pharmacol Ther 2004 Sep 15; 20(6): 593-599.

Economic evaluations of pharmacogenetic tests. 131

95 You JH, Chan FW, Wong RS, Cheng G. The potential clinical and economic outcomes of pharmaco-genetics-oriented management of warfarin therapy - a decision analysis. Thromb Haemost 2004 Sep; 92(3): 590-597.

96 You JH, Tsui KK, Wong RS, Cheng G. Potential clinical and economic outcomes of CYP2C9 and VKORC1 genotype-guided dosing in patients starting warfarin therapy. Clin Pharmacol Ther 2009 Nov; 86(5): 540-547.

97 Daly TM, Dumaual CM, Miao X, Farmen MW, Njau RK, Fu DJ, et al. Multiplex assay for comprehen-sive genotyping of genes involved in drug metabolism, excretion, and transport. Clin Chem 2007 Jul; 53(7): 1222-1230.

98 Antonanzas F, Rodriguez-Ibeas R, Hutter MF, Lorente R, Juarez C, Pinillos M. Genetic testing in the European Union: does economic evaluation matter? Eur J Health Econ 2012 Oct; 13(5): 651-661.

99 Frederix GW, Severens JL, Hovels AM, Raaijmakers JA, Schellens JH. The cloudy crystal ball of cost-effectiveness studies. Value Health 2013 Sep-Oct; 16(6): 1100-1102.

100 Verhoef TI, Redekop WK, van Schie RM, Bayat S, Daly AK, Geitona M, et al. Cost-effectiveness of pharmacogenetics in anticoagulation: international differences in healthcare systems and costs. Pharmacogenomics 2012 Sep; 13(12): 1405-1417.

101 NICE. Guide to the methods of technology appraisal 2013. 2013; Available at: https: //www.nice.org.uk/article/pmg9/resources/non-guidance-guide-to-the-methods-of-technology-appraisal-2013-pdf.

102 Health Care Insurance Board. Dutch pharmacoeconomic guidelines [in Dutch]. 2006; Available at: http://www.zorginstituutnederland.nl/binaries/content/documents/zinl-www/docu-menten/publicaties/publications-in-english/2006/0604-guidelines-for-pharmacoeconomic-research/0604-guidelines-for-pharmacoeconomic-research/Guidelines+for+pharmacoeconomic+research.pdf.

103 Goldstein JA. Clinical relevance of genetic polymorphisms in the human CYP2C subfamily. Br J Clin Pharmacol 2001 Oct; 52(4): 349-355.

104 Lundh A, Sismondo S, Lexchin J, Busuioc OA, Bero L. Industry sponsorship and research outcome. Cochrane Database Syst Rev 2012 Dec 12; 12: MR000033.

105 Lexchin J, Bero LA, Djulbegovic B, Clark O. Pharmaceutical industry sponsorship and research outcome and quality: systematic review. BMJ 2003 May 31; 326(7400): 1167-1170.

106 Egger M, Juni P, Bartlett C, Holenstein F, Sterne J. How important are comprehensive literature searches and the assessment of trial quality in systematic reviews? Empirical study. Health Tech-nol Assess 2003; 7(1): 1-76.

107 Higashi MK, Veenstra DL. Managed care in the genomics era: assessing the cost effectiveness of genetic tests. Am J Manag Care 2003 Jul; 9(7): 493

132 Chapter 4.1

s1 t

able

. Ove

rvie

w o

f pha

rmac

o-ec

onom

ic P

Gx

stud

ies p

ublis

hed

betw

een

Augu

st 2

010

and

Sept

embe

r 201

4 an

alys

ing

the

intr

insi

c va

lue

of a

PG

x te

st (A

), or

com

par-

ing

diffe

rent

trea

tmen

t str

ateg

ies

invo

lvin

g PG

x te

stin

g (B

).

A.

1st A

utho

r (re

fere

nce)

Dis

ease

are

aPg

x te

stPg

x te

st c

osts

Dru

gA

naly

sis

sens

itiv

ity

Ana

lyse

sti

me

hori

zon

Dis

coun

ting

Pers

pect

ive

klan

g20

10 (1

7)ER

+ LN

-

ESBC

21 g

ene

assa

y$3

,460

PGx

test

+ ch

emot

hera

py/n

o ch

emot

hera

pyVs

. che

mot

hera

py/n

o ch

emot

hera

py.

CUA

UV

and

PB30

yrs

and

lif

etim

e3%

Paye

r’s

Bacc

hi 2

010

(18)

ER+ ;

ALN

– ES

BC21

gen

e as

say

$2,2

94PG

x te

st +

che

mot

hera

py/n

o ch

emot

hera

pyvs

. che

mot

hera

py

CMA

UV

Not

m

entio

ned

No

disc

ount

ing

Third

par

ty

paye

r’s

Hal

l20

12 (1

9)ER

+ LN

+

ESBC

21 g

ene

assa

y£2

,000

- £7,

000

£2,5

76PG

x te

st +

che

mot

hera

py/n

o ch

emot

hera

pyvs

. che

mot

hera

py

CUA

UV

and

PB30

yrs

and

lifet

ime

3.5%

Paye

r’s

Vand

erla

an

2011

(20)

ER+ N

+ -HER

2-

ESBC

21 g

ene

assa

y$3

,975

PGx

test

+ c

hem

othe

rapy

/no

che

mot

hera

py v

s. ch

emot

hera

py

CUA

UV

30yr

s3%

Paye

r’s

Verh

oef

2013

(21)

AF

CYP2

C9 a

nd

VKO

RC1

€20-

€16

0Ba

se c

ase:

€40

Phen

proc

oum

onvs

. PG

x te

st +

phe

npro

coum

onCU

AU

V, S

A a

nd P

BLi

fetim

e4%

(cos

t)1.

5% (e

ffect

s)H

ealth

Car

e (P

ayer

s?)

Don

g20

12 (2

2)Ep

ileps

yH

LA-B

*150

2$8

0-$3

80Ba

se c

ase:

$27

0Ca

rbam

azep

ine

and

phen

ytoi

n*CU

AU

V an

d PB

30yr

s3%

Paye

r’s

tiam

kao

2013

(23)

Neu

rolo

gi-c

al

dise

ases

HLA

-B*1

502

3,00

Bah

tCa

rbam

azep

ine

CMA

Non

eU

nkno

wn

No

disc

ount

ing

Paye

r’s

shir

oiw

a20

10 (2

4)m

CRC

KRAS

$220

- $1,

100

Base

cas

e: $

220

Cetu

xim

abCE

A +

CU

AU

V an

d PB

2,5y

rs3%

Paye

r’s

Blan

k20

11 (2

5)m

CRC

KRAS

and

BR

AF€

394

Cetu

xim

abCU

AU

V an

d PB

lifet

ime

3%Pa

yer’s

Economic evaluations of pharmacogenetic tests. 133

A. (

cont

inue

d)

1st A

utho

r (re

fere

nce)

Dis

ease

are

aPg

x te

stPg

x te

st c

osts

Dru

gA

naly

sis

sens

itiv

ity

Ana

lyse

sti

me

hori

zon

Dis

coun

ting

Pers

pect

ive

Behl

2012

(26)

mCR

CKR

AS a

nd/

or B

RAF

KRAS

:$1

12- $

336

Base

cas

e:$2

24KR

AS a

nd B

RAF:

$152

- $45

5Ba

se c

ase:

$30

3

Cetu

xim

abCE

APB

1, 2

, 5 a

nd

10yr

s3%

Not

repo

rted

shiff

man

2012

(27)

CVD

LPA

$100

- $20

0Ba

se c

ase:

$15

0A

spiri

nCU

AU

V an

d PB

10yr

s3.

5%Pa

yer’s

Don

nan

2011

(28)

acut

e ly

mph

obla

stic

le

ukae

mia

TPM

T$8

3- $

414

Base

cas

e: $

380

6-M

erca

ptop

urin

eCE

AU

V an

d PB

3 m

onth

sN

o di

scou

ntin

gSo

ciet

al

scha

ckm

an20

13 (2

9)H

IVU

GT1

A1$1

0 or

$10

7Ba

se c

ase:

$10

7A

taza

navi

rvs

. PG

x te

st +

ata

zana

vir/

daru

navi

r*

CUA

UV

and

PBlif

etim

e3%

Paye

r’s

*Aut

hors

ass

umed

equ

al c

osts

and

effi

cacy

for d

iffer

ent p

harm

aceu

tical

sA

F, at

rial fi

brill

atio

n; B

RAF,

v-Ra

f mur

ine

sarc

oma

vira

l onc

ogen

e ho

mol

og B

1; C

EA, c

ost-

effec

tiven

ess a

naly

sis;

CM

A, c

ost-

min

imiz

atio

n an

alys

is; C

UA

, cos

t-ut

ility

ana

lysi

s;

CVD

, car

diov

ascu

lar d

isea

se; C

YP, c

ytoc

hrom

e P-

450;

ER,

oes

trog

en re

cept

or; E

SBC,

ear

ly s

tage

bre

ast c

ance

r; H

ER2,

Hum

an E

pide

rmal

gro

wth

fact

or R

ecep

tor 2

; HLA

, hu

man

leuk

ocyt

e an

tigen

; KRA

S, K

irste

n ra

t sar

com

a vi

ral o

ncog

ene

hom

olog

; LPA

, lip

opro

tein

-a; (

A)a

xilla

ry L

N, l

ymph

nod

e; m

CRC,

met

asta

tic c

olor

ecta

l can

cer;

MV,

m

ultiv

aria

te; P

B, p

roba

bilis

tic; P

Gx,

pha

rmac

ogen

etic

; TPM

T, th

iopu

rine

S-m

ethy

ltran

sfer

ase;

UG

T, U

DP-

gluc

uron

osyl

tran

sfer

ase;

UV,

uni

varia

te; V

KORC

, Vita

min

K e

pox-

ide

redu

ctas

e co

mpl

ex.

134 Chapter 4.1

B. 1st A

utho

r (re

fere

nce)

Dis

ease

are

aPg

x te

stPg

x te

st c

osts

Dru

gA

naly

sis

sens

itiv

ity

Ana

lyse

sti

me

hori

zon

Dis

coun

ting

Pers

pect

ive

Olg

iati

2012

(30)

Dep

ress

ion

5-H

TTLP

R$1

00- $

300

Base

car

e: $

200

Cita

lopr

am a

nd/o

r bup

ropi

onvs

. PG

x te

st +

cita

lpra

m a

nd/o

r bu

prop

ion.

CUA

UV,

MV,

and

PB

Shor

t ter

mN

o di

scou

ntin

gPa

yer’s

serr

etti

2011

(31)

Dep

ress

ion

5-H

TTLP

R$1

42- $

326

Base

cas

e: $

234

Cita

lopr

am o

r bup

ropi

onvs

. PG

x te

st +

cita

lopr

am/

bupr

opio

n

CUA

UV,

MV,

and

PB

12 w

eeks

No

disc

ount

ing

Paye

r’s

Reed

20

11(3

2)Pr

osta

te

canc

er8-

14 ri

sk

alle

les

$200

- $60

0Ba

se c

ase:

$40

0N

o ch

emot

hera

pyvs

. PG

x te

st +

fina

ster

ide

CUA

UV

and

PBLi

fetim

e3%

Paye

r’s

Dja

lalo

v20

12 (3

3)A

mne

sitc

co

gniti

ve

impa

irmen

t

APO

E ε4

CAN

$150

-CA

N$1

,625

Base

cas

e:

CAN

$325

Stan

dard

of c

are

vs. d

onep

ezil

+ PG

x te

stCU

AU

V an

d PB

30yr

s5%

Soci

etal

kazi

2014

(37)

ACS

CYP2

C19

$100

-$70

0Ba

se c

ase:

$23

5Cl

opid

ogre

l- v

s. pr

asug

rel;

- vs t

icag

relo

r;- v

s. PG

x te

st +

clo

pido

grel

/tic

agre

lor

- vs.

PGx

test

+ cl

opid

ogre

l/ pr

asug

rel

CUA

UV,

SA

, and

PB

Life

time

3%So

ciet

al

Rees

e20

12 (3

4)AC

SCY

P2C1

9$3

10- C

lopi

dogr

el v

s. PG

x te

st +

cl

opid

ogre

l/pra

sugr

el ;

- Pra

sugr

el v

s. PG

x te

st+

clop

idog

rel/p

rasu

grel

;- G

ener

ic c

lopi

dogr

el v

s. PG

x te

st +

ge

neric

clo

pido

grel

/ pra

sugr

el;

- Pra

sugr

el v

s. ge

neric

clo

pido

grel

/pr

asug

rel

CEA

PB15

mon

ths

5%Pa

yer’s

Economic evaluations of pharmacogenetic tests. 135

B. (c

onti

nued

)

1st A

utho

r (re

fere

nce)

Dis

ease

are

aPg

x te

stPg

x te

st c

osts

Dru

gA

naly

sis

sens

itiv

ity

Ana

lyse

sti

me

hori

zon

Dis

coun

ting

Pers

pect

ive

sori

ch20

13 (3

5)AC

SCY

P2C1

9AU

S$47

- Clo

pido

grel

vs.

PGx

test

+

clop

idog

rel/t

icag

relo

r- P

Gx

test

+ c

lopi

dogr

el/t

icag

relo

r vs

. tic

agre

lor

CUA

UV

and

PBLi

fetim

e5%

Hea

lthca

re /

(Pay

ers?

)

Cres

pin

.20

11 (3

6)AC

SCY

P2C1

9*2

$100

- $30

0Ba

se c

ase:

$20

0Ti

cagr

elor

vs.

PGx

test

+ c

lopi

dogr

elCU

AU

V an

d PB

5yrs

3%Pa

yer’s

lala

2013

(38)

ACS

CYP2

C19

*2$6

0-$7

50Ba

se c

ase:

$50

0PG

x te

st +

clo

pido

grel

/ pra

sugr

el- v

s. cl

opid

ogre

l- v

s. pr

asug

rel

CUA

UV

and

PB15

mon

ths

and

10yr

s3%

Paye

r’s

Pana

tton

i20

12 (3

9)AC

SCY

P2C1

9*2

NZ$

175

- Clo

pido

grel

vs.

PGx

test

+

clop

idog

rel/p

rasu

grel

;- P

rasu

grel

vs.

PGx

test

+ cl

opid

ogre

l/pra

sugr

el.

CUA

PBLi

fetim

e3%

Paye

r’s

Pink

2013

(40)

Non

valv

ular

A

FCY

P2C9

and

VK

ORC

1£2

0W

arfa

rinvs

. PG

x te

st+

war

farin

/nov

el o

ral

antic

oagu

lant

s.

CUA

UV

and

PBLi

fetim

e3.

5%Pa

yer’s

you

2014

(41)

AF

CYP2

C9 a

nd

VKO

RC1

$50-

$200

Base

cas

e: $

75W

arfa

rinvs

. PG

x te

st+

war

farin

/nov

el o

ral

antic

oagu

lant

s.

CUA

UV,

MV

and

PB25

yrs

3%Pa

yer’s

you

2012

(42)

AF

CYP2

C9 a

nd

VKO

RC1

$50-

$200

Base

cas

e: $

72PG

x te

st +

war

farin

- vs.

war

farin

;- v

s. da

biga

tran

150

mg;

- vs.

dabi

gatr

an m

g.

CUA

UV,

MV

and

PBLi

fetim

e3%

Paye

r’s

de l

ima

lope

s20

12 (4

3)

NSC

LCEG

FRSG

$190

-SG

$760

Base

cas

e: S

G$3

80St

anda

rd c

hem

othe

rapy

vs. P

Gx

test

+ g

efitin

ib/s

tand

ard

chem

othe

rapy

CUA

UV

and

SALi

fetim

eN

o di

scou

ntin

gPa

yer’s

136 Chapter 4.1

B. (c

onti

nued

)

1st A

utho

r (re

fere

nce)

Dis

ease

are

aPg

x te

stPg

x te

st c

osts

Dru

gA

naly

sis

sens

itiv

ity

Ana

lyse

sti

me

hori

zon

Dis

coun

ting

Pers

pect

ive

Han

dorf

2012

(44)

NSC

LCEG

FR$1

22-$

365

Base

cas

e: $

243

Plat

inum

com

bina

tion

vs. P

Gx

test

+ e

rlotin

ib/p

latin

um

com

bina

tion

CUA

UV

and

MV

12 m

onth

sN

o di

scou

ntin

gPa

yer’s

zhu

2013

(45)

NSC

LCEG

FR$3

81- $

635

Base

cas

e: $

508

Rout

ine

care

vs. P

Gx

test

+ g

efitin

ib/r

outin

e ar

e.CE

A a

nd

CUA

UV

and

PB10

yrs

3%Pa

yer’s

kauf

2010

(46)

HIV

HLA

-B*5

701

$50-

$15

0Ba

se c

ase:

$88

Aba

cavi

r+la

miv

udin

e+ e

favi

renz

vs.P

Gx

test

+a

baca

vir+

lam

ivud

ine+

efav

irenz

/te

nofo

vir+

emtr

icita

bine

CEA

UV

60 d

ays

and

Life

time

No

disc

ount

ing

Paye

r’s

Ratt

ana-

vi

papo

ng20

13 (4

7)

Epile

psy

and

neur

opat

ic

pain

HLA

-B

*150

2TH

B 1,

000

Carb

amaz

epin

e- v

s. PG

x te

st +

carb

amaz

epin

e;- v

s. al

tern

ativ

e (s

odiu

m v

alpr

oate

an

d ga

bape

ntin

.)

CUA

PBLi

fetim

e3.

5%Pa

yer’s

liu

2012

(48)

Chro

nic

hepa

titis

CIL

-28B

$186

-$55

7Ba

se c

ase:

$37

1St

anda

rd th

erap

y (in

terf

eron

+

ribav

irin)

vs. P

Gx

+ st

anda

rd th

erap

y /t

riple

th

erap

y (in

terf

eron

+rib

aviri

n +

prot

ease

inhi

bito

r)

CUA

UV

and

PBLi

fetim

e3%

Soci

etal

gre

eley

2011

(49)

Neo

nata

l di

abet

es

mel

litus

KCN

J11

and

ABCC

8$5

00- $

5,00

0Ba

se c

ase:

KCN

J11:

$70

5 AB

CC8:

$2,

110

Insu

linvs

. PG

x te

st +

sulfo

nylu

rea

CUA

UV

and

PB10

, 20

and

30yr

s3%

Soci

etal

Part

han

2013

(50)

ACS

KIF6

$100

Prav

asta

tinvs

. PG

x te

st +

pra

vast

atin

/at

orva

stat

in

CUA

UV

and

PBLi

fetim

e3%

Paye

r’s

Economic evaluations of pharmacogenetic tests. 137

B. (c

onti

nued

)

1st A

utho

r (re

fere

nce)

Dis

ease

are

aPg

x te

stPg

x te

st c

osts

Dru

gA

naly

sis

sens

itiv

ity

Ana

lyse

sti

me

hori

zon

Dis

coun

ting

Pers

pect

ive

Vija

yara

-gh

avan

20

12 (5

1)

mCR

CKR

AS$2

43- P

anitu

mum

ab v

s. PG

x te

st +

pa

nitu

mum

ab/o

ther

che

mo;

- Cet

uxim

ab v

s. PG

x te

st +

ce

tuxi

mab

/oth

er c

hem

othe

rapy

;- C

ombi

natio

n th

erap

y (c

etux

imab

+ ir

inot

ecan

)vs

. PG

x te

st +

com

bina

tion

ther

apy

/irin

otec

an.

CEA

UV

Life

time

No

disc

ount

ing

Paye

r’s

Hag

aman

2010

(53)

idio

path

ic

pulm

onar

y fib

rosi

s

TPM

T$3

00A

zath

iopr

ine

vs. P

Gx

test

+ az

athi

oprin

e/al

tern

ativ

e

CUA

UV

and

MV

lifet

ime

No

disc

ount

ing

Paye

r’s

thom

pson

20

14 (5

2)Au

toim

mun

e di

seas

eTP

MT

£20

Aza

thio

perin

evs

. PG

x te

st+

azat

hiop

rine/

alte

rnat

ive

CUA

UV

4 m

onth

sN

o di

scou

ntin

gH

ealth

ser

vice

Pich

erea

u20

10 (5

4)m

CRC

UG

T1A1

*28

€71

FOLF

IRI

vs. P

Gx

test

+ F

OLF

IRI/

FOLF

IRI+

CSF

CEA

UV

and

PB2

wee

ksN

o di

scou

ntin

gH

ospi

tal’s

ABC

C, A

TP-b

indi

ng c

asse

tte

tran

spor

ter s

yb fa

mily

C; A

CS, a

cute

cor

onar

y sy

ndro

me;

AF,

atria

l fibr

illat

ion;

APO

E, a

polip

opro

tein

-E; C

EA, c

ost-

effec

tiven

ess

anal

ysis

; CSF

, Co

lony

Stim

ulat

ing

Fact

or ;C

UA

, cos

t-ut

ility

ana

lysi

s; C

YP, c

ytoc

hrom

e P-

450;

EG

FR, e

pide

rmal

gro

wth

fact

or re

cept

or; F

OLF

IRI ,

Fol

inic

aci

d, fl

uoro

urac

il, ir

inot

ecan

; HLA

, hu

man

leuk

ocyt

e an

tigen

; HTT

LPR,

ser

oton

in-t

rans

port

er-li

nked

pol

ymor

phic

regi

on; I

L, in

terle

ukin

e; K

CNJ,

Pota

ssiu

m in

war

dly-

rect

ifyin

g ch

anne

l, su

bfam

ily J

; KIF

, ki-

nase

fam

ily m

embe

r; KR

AS,

Kirs

ten

rat s

arco

ma

vira

l onc

ogen

e ho

mol

og; L

PA, l

ipop

rote

in-a

; mCR

C, m

etas

tatic

col

orec

tal c

ance

r; M

V, m

ultiv

aria

te; N

SCLC

, non

-sm

all c

ell

lung

can

cer;

PB, p

roba

bilis

tic; P

Gx,

pha

rmac

ogen

etic

; SA

, sce

nario

s an

alys

es; T

HB,

Tha

i Bah

t; TP

MT,

thio

purin

e S-

met

hyltr

ansf

eras

e; U

GT,

UD

P-gl

ucur

onos

yltr

ansf

eras

e;

UV,

uni

varia

te; V

KORC

, Vita

min

K e

poxi

de re

duct

ase

com

plex

.

138 Chapter 4.1

supp

lem

enta

ry fi

le 2

. Mai

n lim

itatio

ns w

hich

wer

e di

scus

sed

in th

e pa

pers

pub

lishe

d be

twee

n Au

gust

201

0 an

d Se

ptem

ber 2

014

.

stud

yD

iscu

ssio

n of

lim

itat

ions

Bacc

hiet

al.,

201

0-

The

hypo

thet

ical

nat

ure

inhe

rent

to e

cono

mic

mod

els

- N

ot a

ll di

rect

and

indi

rect

cos

ts w

ere

take

n in

to a

ccou

nt-

The

effec

tiven

ess

of th

erap

y ba

sed

on th

e ris

k pr

edic

tion

of th

e as

say

was

not

con

side

red

- Th

ere

was

var

iabi

lity

in th

e de

finiti

on a

nd in

cide

nce

of fe

brile

neu

trop

enia

- It

was

not

pos

sibl

e to

incl

ude

Braz

ilian

est

imat

es fo

r sta

ge d

istr

ibut

ion

upon

dia

gnos

is o

f bre

ast c

ance

r

Behl

et a

l. 20

12-

No

prog

ress

ion-

free

sur

viva

l out

com

es fo

r the

diff

eren

t str

ateg

ies

was

pro

vide

d-

Lim

itatio

ns in

dat

a fo

r diff

eren

ces

amon

g th

e tr

eatm

ent o

ptio

ns in

qua

lity

of li

fe, a

nd w

as th

eref

ore

not i

ncor

pora

ted

- G

ain

in o

vera

ll su

rviv

al m

ay n

ot h

ave

been

see

n if

the

heal

th im

pact

of t

reat

men

t wer

e ex

pres

sed

in q

ualit

y-ad

just

ed s

urvi

val

- D

iffer

ence

s in

qua

lity

of li

fe a

mon

g pa

tient

s re

ceiv

ing

alte

rnat

e tr

eatm

ents

hav

e no

t bee

n qu

antifi

ed in

a w

ay th

at a

llow

s qu

ality

adj

uste

d su

rviv

al to

be

mod

elle

d-

It w

as n

ot ta

ken

in a

ccou

nt th

at p

atie

nts

coul

d di

scon

tinue

the

trea

tmen

t due

to a

dver

se e

vent

s-

The

anal

ysis

ass

umes

that

all

diffe

renc

es in

sur

viva

l are

due

to a

lack

of r

espo

nse

to c

etux

imab

. Tho

ugh

mut

atio

ns m

ay in

depe

nden

tly p

redi

ct p

rogn

osis

, no

t the

kin

d of

trea

tmen

t-

The

estim

ated

sav

ings

mig

ht b

e ov

eres

timat

ed if

any

par

t of t

he s

urvi

val d

iffer

ence

is d

ue to

mut

atio

ns, a

nd in

depe

nden

t of t

he th

erap

y

Blan

ket

al.2

011

- La

ck o

f clin

ical

dat

a of

a c

lear

ly d

efine

pat

ient

pop

ulat

ion

from

Sw

itzer

land

. Hen

ce, c

linic

al a

nd u

tility

dat

a or

igin

ated

from

few

stu

dies

con

duct

ed o

utsi

de

Switz

erla

nd-

Unc

erta

inty

in tr

ial d

ata

and

pote

ntia

lly li

mite

d tr

ansf

erab

ility

to ro

utin

e cl

inic

al p

ract

ice

popu

latio

ns-

The

qual

ity o

f life

and

util

ity d

ata

allo

wed

to d

iffer

entia

te o

n th

e ba

sis

of tr

eatm

ent,

but n

ot o

f mut

atio

n st

atus

- BR

AF m

utat

ion

seem

ed to

hav

e no

impa

ct o

n re

spon

se to

the

antib

ody,

sug

gest

ing

that

BRA

F m

utat

ion

may

not

hav

e th

e sa

me

pred

ictiv

e va

lue

in fi

rst-

line

and

chem

oref

ract

ory

tum

ours

.

Cres

pin

et a

l.201

1-

The

ultim

ate

effec

tiven

ess

of a

ntip

late

let m

edic

atio

n lik

ely

cann

ot b

e de

term

ined

, bec

ause

the

shor

t dur

atio

n of

the

clin

ical

tria

ls-

The

estim

ates

mig

ht b

e bi

ased

if th

e eff

ectiv

enes

s of

tica

grel

or d

iffer

s af

ter 1

2 m

onth

s of

ther

apy

- Re

sults

may

not

app

ly to

sub

popu

latio

ns e

xtra

cted

from

clin

ical

tria

ls-

The

cost

s fo

r adv

erse

eve

nt m

ight

var

y be

twee

n su

bpop

ulat

ions

due

to e

vent

s th

at s

igni

fican

tly d

iffer

- A

ssum

ptio

ns th

at M

I and

dea

th w

ere

inde

pend

ent,

whi

ch is

like

ly u

nrea

listic

- Re

sults

can

not b

e us

ed to

det

erm

ine

univ

ersa

l cos

t-eff

ectiv

enes

s re

lativ

e to

oth

er v

iabl

e tr

eatm

ent o

ptio

ns fo

r sec

onda

ry p

reve

ntio

n af

ter a

n ac

ute

coro

nary

syn

drom

e

Economic evaluations of pharmacogenetic tests. 139

supp

lem

enta

ry fi

le 2

. (co

ntin

ued)

stud

yD

iscu

ssio

n of

lim

itat

ions

de L

ima

Lope

set

al.

2012

- A

ssum

ptio

ns w

ere

mad

e ab

out l

engt

h an

d tim

e sp

ent i

n tr

eatm

ent b

y ex

trap

olat

ing

from

clin

ical

dat

a, a

nd a

ssum

ing

that

the

time

spen

t in

first

- or

seco

nd-li

ne tr

eatm

ent w

ith c

hem

othe

rapy

of g

efitin

ib w

as e

quiv

alen

t to

acco

unt f

or s

imila

r ove

rall

surv

ival

- Co

st d

ata

is o

btai

ned

from

3 S

inga

pore

onc

olog

y ce

ntre

s, an

d m

ight

not

be

repr

esen

tativ

e fo

r oth

er c

ount

ries

and/

or h

ospi

tals

- Se

vera

l ass

umpt

ions

abo

ut q

ualit

y of

life

val

ues

wer

e m

ade,

hen

ce th

e es

timat

ed IC

ERs

mig

ht b

e bi

ased

- La

ck o

f dat

a to

cap

ture

robu

st in

divi

dual

leve

l var

iatio

n in

trea

tmen

t res

pons

es-

The

stan

dard

of c

are

for fi

rst l

ine

trea

tmen

t has

shi

ft to

incl

ude

pem

etre

xed,

bev

aciz

umab

and

cet

uxim

ab in

firs

t-lin

e tr

eatm

ent a

nd to

incl

ude

pem

etre

xed

as p

oten

tial m

aint

enan

ce o

ptio

n af

ter i

nitia

l che

mot

hera

py-

The

mod

el s

peci

fical

ly c

ompa

res

a st

anda

rd p

ract

ice,

whi

ch in

clud

es n

o EG

FR m

utat

ion

test

ing,

firs

t-lin

e tr

eatm

ent w

ith c

hem

othe

rapy

and

sec

ond-

line

trea

tmen

t with

gefi

tinib

, to

EGFR

test

ing

and

guid

ed th

erap

y ba

sed

on th

e re

sults

of t

he te

st.

- In

the

base

cas

e it

was

ass

umed

that

gefi

tinib

did

not

ben

efit p

atie

nts

with

out a

ctiv

atin

g m

utat

ions

, eve

n be

yond

firs

t lin

e tr

eatm

ent

Dja

lalo

vet

al.2

012

- Th

ere

is li

mite

d ev

iden

ce o

n th

e eff

ectiv

enes

s of

don

epez

il tr

eatm

ent i

n de

layi

ng p

rogr

essi

on fr

om A

MCI

to A

D a

mon

g A

POE

e4 c

arrie

rs.

- Su

rvei

llanc

e co

sts

wer

e ba

sed

on re

sults

from

a s

tudy

that

use

d ol

der p

atie

nts,

and

was

con

duct

ed in

a d

iffer

ent h

ealth

care

sys

tem

- Th

ey d

id n

ot u

sed

som

e da

ta fr

om a

stu

dy, b

ecau

se th

at d

ata

mig

ht h

ave

had

sele

ctio

n bi

ases

that

lim

ited

the

gene

ralis

atio

n of

the

resu

lts-

Adeq

uate

and

wid

ely

acce

pted

crit

eria

for d

iagn

osin

g A

MCI

are

una

vaila

ble,

but

inco

rrec

t dia

gnos

es n

ot in

corp

orat

ed in

the

stud

y. In

corr

ect d

iagn

osin

g m

ight

incr

ease

the

ICER

- Th

ey d

id n

ot in

clud

ed p

atie

nts

that

wer

e no

t see

n an

d di

agno

sed

with

AM

CI b

efor

e th

e de

velo

pmen

t of A

D

Don

get

al.2

012

- Th

e st

udy

assu

mes

that

hea

lth-r

elat

ed Q

oL is

rest

ored

to p

erfe

ct h

ealth

. How

ever

, the

QA

LY g

ains

mig

ht b

e ov

eres

timat

ed if

the

heal

th re

late

d Q

oL is

in

real

ity lo

wer

due

to a

n im

perf

ect r

espo

nse

to d

rugs

, epi

leps

y re

curr

ence

or o

ther

hea

lth p

robl

ems

- Tr

eatm

ent d

ecis

ions

are

in re

ality

influ

ence

d by

man

y fa

ctor

s an

d m

ay s

ubst

antia

lly d

evia

te fr

om th

e in

itial

ass

umpt

ions

- Th

e m

odel

als

o in

clud

es s

ever

al a

dditi

onal

ass

umpt

ions

and

sim

plifi

catio

ns:

•

Som

e pa

tient

sub

popu

latio

ns e

xclu

ded

•

Trea

tmen

t rul

es s

impl

ified

for w

ho re

ceiv

es th

e tr

eatm

ent

•

The

mod

el a

ssum

es th

at V

PA, C

BZ, a

nd P

HT

have

sim

ilar e

ffica

cy.

- Eff

ectiv

enes

s da

ta fr

om a

stu

dy u

sing

Cau

casi

ans,

ethn

ic d

iffer

ence

s ca

nnot

be

rule

d ou

t-

They

ass

umed

that

gen

otyp

ing

resu

lts a

re im

med

iate

ly a

cces

sibl

e, th

ough

this

may

not

hol

d un

iver

sally

Don

nan

et a

l. 20

11-

The

rarit

y of

the

dise

ase

limits

am

ount

and

type

of e

vide

nce

avai

labl

e, th

eref

ore

som

e va

lues

bas

ed o

n ex

pert

opi

nion

- N

o Q

ALY

ana

lysi

s po

ssib

le, d

ue to

lack

of l

itera

ture

on

utili

ty s

core

s fo

r chi

ldre

n w

ith A

LL-

Ther

e w

as s

ome

unce

rtai

nty

in th

e va

lues

use

d fo

r the

uni

t pric

es o

f TPM

T ge

noty

pe a

nd e

nzym

atic

test

s

140 Chapter 4.1

supp

lem

enta

ry fi

le 2

. (co

ntin

ued)

stud

yD

iscu

ssio

n of

lim

itat

ions

Gre

eley

et a

l.201

1-

In th

is s

tudy

it w

as a

ssum

ed th

at th

e te

st w

as 1

00%

sen

sitiv

e an

d sp

ecifi

c, a

lthou

gh th

ere

was

not

acc

ount

ed fo

r fac

tors

that

mig

ht d

imin

ish

the

effec

tive

clin

ical

sen

sitiv

ity a

nd s

peci

ficity

- La

ck o

f dat

a, th

eref

ore

som

e of

the

mod

el’s

assu

mpt

ions

for s

elec

ted

com

plic

atio

ns w

ere

deriv

ed fr

om s

tudi

es p

erfo

rmed

on

patie

nts

with

dia

bete

s ty

pe 2

- D

ue to

a la

ck o

f qua

lity-

of-li

fe d

ata

in n

eona

tal d

iabe

tes,

it w

as a

ssum

ed th

at p

atie

nts

wou

ld e

xper

ienc

e a

utili

ty g

ain

of 0

.1 b

ased

on

surv

ey d

ata

from

di

abet

es ty

pe 2

sub

ject

s-

Due

to in

suffi

cien

t lon

gitu

dina

l dat

a, it

was

ass

umed

that

pat

ient

s w

ith tr

eata

ble

gene

tic d

efec

ts w

ould

rem

ain

resp

onsi

ve to

the

trea

tmen

t ove

r 30y

ears

- Th

e m

odel

did

not

acc

ount

for t

he th

erap

y’s

pote

ntia

l to

impr

ove

neur

odev

elop

men

tal o

utco

mes

Hag

aman

et a

l.201

0-

Ther

e w

ere

no p

ublis

hed

tria

ls th

at s

peci

fical

ly e

xplo

red

thei

r iss

ue; t

here

fore

they

had

to d

raw

dat

a fr

om n

umer

ous

sour

ces.

- Th

e st

udy

assu

med

that

the

low

-dos

e tr

eatm

ent w

ith re

duce

d TP

MT

activ

ity is

the

sam

e as

for f

ull-d

ose

ther

apy

in p

atie

nts

with

nor

mal

TPM

T ac

tivity

- Th

e st

udy

lack

ed d

ata

desc

ribin

g th

e in

cide

nce

of B

MT

in p

atie

nts

with

inte

rmed

iate

TPM

T ac

tivity

- Th

e br

eakd

own

of T

PMT

activ

ity in

pat

ient

s w

ith le

ukop

enia

has

not

bee

n cl

early

elu

cida

ted

- Th

e au

thor

s as

sum

ed th

at c

onse

rvat

ive

ther

apy

was

equ

al in

effi

cacy

to th

e pl

aceb

o ar

m in

ano

ther

stu

dy, i

n w

hich

the

patie

nts

rece

ived

aza

thio

prin

e an

d st

eroi

ds w

ith a

NAC

pla

cebo

. Thi

s as

sum

ptio

n lik

ely

over

estim

ates

the

mar

gina

l cos

t-eff

ectiv

enes

s of

ther

apy

with

aza

thio

prin

e, N

AC a

nd s

tero

ids.

Hal

let

al.

2012

- Th

e au

thor

s co

nduc

ted

an a

naly

sis

of th

e im

med

iate

bud

get i

mpa

ct to

illu

stra

te th

e fin

anci

al im

plic

atio

ns d

urin

g th

e ch

emot

hera

py p

erio

d on

ly-

The

budg

et im

pact

resu

lts d

o no

t cha

ract

eris

e th

e st

reng

th e

vide

nce,

and

do

not c

onsi

der l

ong

term

cos

ts (f

or c

ance

r rec

urre

nce

and

trea

tmen

t tox

icity

, re

lativ

e lif

e ex

pect

ance

and

qua

lity

of li

fe)

- O

nly

the

Onc

otyp

e D

X w

as c

onsi

dere

d, a

nd n

ot th

e re

lativ

e va

lue

and

alte

rnat

ive

test

s-

It is

impo

rtan

t to

reco

gniz

e th

at a

ny m

odel

is a

sim

plifi

catio

n of

real

ity, a

nd th

e m

odel

pre

sent

ed h

ere

is n

o ex

cept

ion.

- It

is c

redi

ble

that

taki

ng in

to a

ccou

nt th

e tr

ansf

erab

ility

of d

ata

from

a U

S tr

ial i

nto

a U

K se

ttin

g w

ould

gen

erat

e ad

ditio

nal u

ncer

tain

ty a

nd e

ven

intr

oduc

e bi

as in

to th

e re

sults

.-

The

use

of d

ecis

ion

aids

suc

h as

Adj

uvan

t Onl

ine!

was

not

spe

cific

ally

inco

rpor

ated

into

the

mod

el-

The

poss

ibili

ty th

at th

e pr

ice

of th

e te

st m

ight

cha

nge

with

the

intr

oduc

tion

of a

ltern

ativ

es w

as n

ot c

onsi

dere

d-

Estim

ates

of l

ong-

term

cos

ts a

nd q

ualit

y of

life

(e.g

., as

soci

ated

with

can

cer r

ecur

renc

e or

car

diac

toxi

city

) are

der

ived

from

hig

her q

ualit

y ev

iden

ce th

an

the

shor

t-te

rm d

ata

sour

ces

(e.g

., co

sts

of to

xici

ty) i

n th

is a

naly

sis

beca

use

they

rely

on

prev

ious

ly p

ublis

hed

dedi

cate

d co

st a

nd q

ualit

y of

life

stu

dies

. The

y ar

e, h

owev

er, s

ubje

ct to

ass

umpt

ions

of d

ata

tran

sfer

abili

ty to

our

pat

ient

pop

ulat

ion

and

requ

ire c

onfir

mat

ion

in a

form

al s

tudy

with

long

-ter

m fo

llow

-up.

- Th

is a

naly

sis

pres

ents

resu

lts o

nly

for t

he a

vera

ge p

atie

nt a

ged

60 y

ears

Economic evaluations of pharmacogenetic tests. 141

supp

lem

enta

ry fi

le 2

. (co

ntin

ued)

stud

yD

iscu

ssio

n of

lim

itat

ions

Han

dorf

et a

l. 20

12-

The

auth

ors

wer

e no

t abl

e to

mod

el th

e fu

ll co

urse

of l

ifetim

e th

erap

y, w

hich

oft

en in

clud

ed s

econ

d- a

nd th

ird-li

ne tr

eatm

ent.

The

inpu

ts fo

r suc

h a

mod

el

wer

e no

t ava

ilabl

e-

A m

odel

-bas

ed a

ppro

ach

was

app

lied,

rath

er th

an a

rand

omis

ed tr

ial.

Ther

efor

e, th

e co

nclu

sion

s de

pend

on

the

valid

ity o

f the

ass

umpt

ions

that

wer

e us

ed to

dev

elop

the

mod

el-

The

auth

ors

wer

e no

t abl

e to

gat

her c

osts

and

effe

cts

for c

laim

s an

d re

gist

ry d

ata

- Th

e cu

rren

t ana

lyse

s w

ere

subj

ect t

o m

odifi

catio

ns in

cos

ts (r

educ

tion

of c

osts

in th

e fu

ture

mak

es th

e st

rate

gies

mor

e fa

vour

able

)

Kauf

et a

l.201

0-

This

ana

lysi

s re

lied

on s

ever

al a

ssum

ptio

ns s

peci

fic to

the

deve

lopm

ent o

f the

AD

VAN

CE m

odel

as

desc

ribed

in th

e te

xt a

nd th

e su

pple

men

tal d

ata.

- Ca

re s

houl

d be

take

n in

the

inte

rpre

tatio

n of

the

resu

lts in

whi

ch s

cree

ning

dom

inat

ed th

e al

tern

ativ

e th

erap

y, b

ecau

se d

iffer

ence

s in

effe

ctiv

enes

s ve

ry

smal

l in

the

dom

inan

t scr

eeni

ng c

ompa

red

with

alte

rnat

ive

stra

tegy

- Th

e re

sults

onl

y ap

ply

to th

ose

patie

nts

for w

hom

aba

cavi

r and

teno

fovi

r are

con

side

red

appr

opria

te tr

eatm

ent a

ltern

ativ

es-

This

ana

lysi

s do

es n

ot c

onsi

der a

ll th

e po

ssib

le b

enefi

ts o

f scr

eeni

ng

Kazi

et a

l. 20

14-

“Est

imat

ed d

iffer

ence

s in

out

com

es b

etw

een

vario

us C

YP2C

19 g

enot

ypes

wer

e la

rgel

y ba

sed

on p

ost h

oc a

naly

ses

of ra

ndom

ized

tria

ls”

- Th

e effi

cacy

and

saf

ety

of p

rasu

grel

and

tica

grel

or w

ere

base

d on

onl

y on

e la

rge,

rand

omiz

ed c

linic

al tr

ial

- Th

e in

dire

ct c

ompa

rison

of t

icag

relo

r with

pra

sugr

el w

as li

mite

d du

e to

str

uctu

ral d

iffer

ence

s in

the

desi

gn a

nd e

xecu

tion

of th

e us

ed c

linic

al tr

ials

- It

was

ass

umed

that

the

clin

ical

out

com

es fr

om th

e PL

ATO

tria

l can

be

tran

slat

ed to

U.S

. pat

ient

s on

low

dos

e as

pirin

ther

apy,

how

ever

this

has

yet

to b

e in

vest

igat

ed-

In o

rder

to d

efine

the

actu

al re

lativ

e co

st-e

ffect

iven

ess,

the

long

-ter

m e

ffect

s of

new

er a

ntip

late

let a

gent

s on

mor

talit

y ra

te h

as to

be

dete

rmin

ed

Klan

get

al.

2010

- Th

e cl

inic

al d

ata

on w

hich

the

anal

yses

wer

e ba

sed

wer

e de

rived

from

a n

on-r

ando

mly

sel

ecte

d sa

mpl

e of

pat

ient

s-

Util

ities

wer

e de

rived

from

Eng

lish

spea

king

lite

ratu

re, t

hus

may

not

fully

cha

ract

eris

e th

e pr

efer

ence

s of

pat

ient

s in

Isra

el-

Valid

atio

n of

the

essa

y w

as b

ased

on

clin

ical

tria

ls c

ondu

cted

in th

e U

S-

Due

to li

mite

d da

ta, t

he a

utho

rs o

mitt

ed s

ome

pote

ntia

l lon

g-te

rm im

plic

atio

ns o

f bre

ast c

ance

r and

its

trea

tmen

ts s

uch

as th

e ris

k of

loca

l rec

urre

nce

and

risk

of s

econ

d pr

imar

y tu

mou

rs a

ssoc

iate

d w

ith c

hem

othe

rapy

- Th

e au

thor

s ex

amin

ed th

e eff

ect o

f the

test

from

the

paye

r’s p

ersp

ectiv

e, h

ence

indi

rect

cos

ts w

ere

not i

nclu

ded

142 Chapter 4.1

supp

lem

enta

ry fi

le 2

. (co

ntin

ued)

stud

yD

iscu

ssio

n of

lim

itat

ions

Lala

et a

l.201

3-

Hea

lthca

re d

eliv

erin

g co

sts

wer

e no

t ava

ilabl

e at

the

indi

vidu

al p

atie

nt le

vel

- Th

e au

thor

s on

ly a

ccou

nted

for t

he im

pact

of t

estin

g fo

r CYP

2C19

*2 a

llele

was

take

n in

acc

ount

, and

the

cost

-effe

ctiv

enes

s of

gen

etic

test

ing

for o

ther

m

utat

ions

was

not

take

n in

acc

ount

- Th

is a

naly

sis

did

not a

ddre

ss th

e ec

onom

ics

of p

late

let r

eact

ivity

test

ing

- Fo

r pat

ient

s w

ith a

prio

r his

tory

of T

IA o

r str

oke,

bod

y w

eigh

t <60

or a

ge >

75 p

rasu

grel

may

not

be

cons

ider

ed o

ptim

al th

erap

y-

It w

as a

ssum

ed th

at th

ere

was

a c

onst

ant m

agni

tude

of b

enefi

t and

a c

onst

ant r

ate

of a

dver

se e

vent

s-

The

auth

ors

assu

med

that

the

rela

tive

risk

of d

eath

in o

ur p

opul

atio

n, c

ompa

red

with

the

gene

ral U

S po

pula

tion,

rem

aine

d fix

ed o

ver t

ime.

- Th

ere

stud

ied

that

foun

d di

ffere

nces

in b

leed

ing

even

ts b

etw

een

CYP2

C19*

2 ca

rrie

rs a

nd w

ild ty

pe-

Alth

ough

gen

etic

test

ing

was

sho

wn

to b

e co

st-s

avin

g co

mpa

red

with

trea

ting

all A

CS p

atie

nts

unde

rgoi

ng P

CI e

mpi

rical

ly w

ith p

rasu

grel

or c

lopi

dogr

el,

the

abso

lute

hea

lth a

nd c

ost d

iffer

ence

s w

ere

smal

l

Liu

et a

l. 20

12-

Beca

use

of a

lack

of e

vide

nce,

str

ateg

ies

invo

lvin

g re

trea

tmen

t with

trip

le th

erap

y af

ter i

nitia

l fai

lure

was

not

take

n in

to a

ccou

nt-

They

did

not

incl

uded

the

effec

t of r

educ

tions

in H

CV tr

ansm

issi

on d

ue to

suc

cess

ful t

reat

men

t-

The

resu

lts a

re li

mite

d to

mon

o-in

fect

ed in

divi

dual

s

Olg

iati

et a

l. 20

12-

It as

sum

es th

at 5

-HTT

LPR

varia

nts

have

the

sam

e di

strib

utio

n an

d eff

ect s

ize

in a

ll Eu

rope

an c

ount

ries.

- Th

e in

fluen

ce o

f 5-H

TTLP

R on

SSR

I res

pons

e is

doc

umen

ted

in ra

ndom

ized

tria

ls, b

ut b

iase

d by

var

ious

lim

itatio

ns-

The

issu

e of

the

tran

sfer

abili

ty o

f the

resu

lts w

as im

plic

itly

addr

esse

d by

con

side

ring

seve

ral r

egio

ns a

t diff

eren

t inc

ome

leve

ls-

Acco

rdin

g th

e au

thor

s it

is a

rgua

ble

that

redu

ctio

n in

hea

lth e

xpen

ditu

res

for a

ppro

xim

atel

y 4%

of n

ew re

spon

ders

und

er p

harm

acog

enet

ic tr

eatm

ent

cann

ot o

ffset

incr

emen

tal c

osts

for g

enet

ic te

st

Pana

tton

iet

al.

2012

- Th

e fin

ding

s ar

e ba

sed

on m

odel

s of

out

com

es ra

ther

than

a ra

ndom

ised

tria

l-

Whe

ther

the

shor

ter d

urat

ion

of c

lopi

dogr

el th

erap

y co

ntrib

uted

to th

e hi

gher

eve

nts

rate

s w

as n

ot c

lear

- Th

e de

finiti

ons

and

clas

sific

atio

n of

adv

erse

eve

nts

diffe

r bet

wee

n N

ew Z

eala

nd h

ospi

tal D

RG d

ata

and

the

TRIT

ON

-TIM

I 38

clin

ical

tria

l-

The

ethn

icity

dat

a ha

s lim

itatio

ns, b

ecau

se m

any

patie

nts

had

herit

age

from

mor

e th

an 1

eth

nica

l gro

up

Economic evaluations of pharmacogenetic tests. 143

supp

lem

enta

ry fi

le 2

. (co

ntin

ued)

stud

yD

iscu

ssio

n of

lim

itat

ions

Part

han

et a

l. 20

13-

The

anal

ysis

was

bas

ed o

n a

post

hoc

sub

stud

y, w

hich

was

not

des

igne

d to

inve

stig

ate

the

cost

-effe

ctiv

enes

s of

KIF

6 te

stin

g-

It w

as a

ssum

ed th

at th

e pa

tient

s w

ith a

maj

or c

ardi

ovas

cula

r eve

nt re

mai

n in

this

sta

te u

ntil

deat

h, a

lthou

gh th

ey h

ave

a hi

gher

risk

of a

sec

ond

even

t-

Card

iova

scul

ar e

vent

rate

s w

ere

extr

apol

ated

bec

ause

dat

a w

as o

nly

avai

labl

e fo

r 2 y

ears

- Th

e es

timat

es o

f diff

eren

tial s

tatin

ben

efit m

ay h

ave

been

ove

rest

imat

ed, b

ecau

se th

e as

soci

atio

n of

KIF

6 w

ith th

e di

ffere

ntia

l red

uctio

n of

CH

D e

vent

ra

tes

from

sta

tin th

erap

y m

ay b

e lim

ited

to h

igh-

dose

ato

rvas

tatin

and

sta

ndar

d do

se p

rava

stat

in-

Ther

e w

as n

o da

ta o

n ev

ent r

ates

for n

on-a

dher

ent p

atie

nts

by K

IF6

carr

ier s

tatu

s, th

eref

ore

the

4-m

onth

eve

nt ra

te fo

r pat

ient

s in

the

plac

ebo

arm

was

us

ed a

s a

prox

y-

The

auth

ors

did

not a

ccou

nted

for a

pos

sibl

e re

latio

nshi

p be

twee

n m

ultip

le e

vent

s an

d ris

k of

dea

th, t

here

fore

the

estim

atio

n of

life

exp

ecta

ncy

may

not

be

acc

urat

e-

The

cost

of s

econ

dary

eve

nts

was

est

imat

e to

be

the

sam

e as

the

cost

s of

a p

rimar

y ev

ent

- “T

he a

cute

cos

ts o

f UA

requ

iring

hos

pita

lizat

ion

is re

pres

ente

d in

the

mod

el a

s a

36-m

onth

cos

t in

light

of t

he a

bsen

ce o

f dat

a on

UA

as

an e

vent

se

cond

ary

to o

ther

car

diov

ascu

lar e

vent

s.”

Pich

erea

uet

al.

2010

- Th

e au

thor

s di

d no

t opt

for a

full

econ

omic

mod

el in

the

anal

ysis

as

the

prim

ary

inte

ntio

n w

as to

eva

luat

e w

heth

er o

r not

UG

T1A

1 ge

noty

pe te

stin

g w

ould

be

an

effici

ent u

se o

f add

ition

al re

sour

ces

from

the

hosp

ital.

- O

nly

few

stu

dies

wer

e us

ed to

bui

ld o

ur m

odel

as

mos

t of t

he a

vaila

ble

data

wer

e no

t ass

ocia

ted

with

com

plet

e in

form

atio

n of

pol

ymor

phis

ms

prev

alen

ce

or F

N in

cide

nce

- Th

e au

thor

s as

sum

ed 1

00%

effi

cacy

of C

SF th

erap

y, th

us in

flatin

g th

e es

timat

es o

f neu

trop

enic

eve

nts

avoi

ded,

if e

ffica

cy is

in fa

ct lo

wer

.-

The

cost

s of

CSF

wer

e no

t inc

lude

d be

caus

e th

is d

rug

was

not

pro

vide

d by

the

hosp

ital

- In

dire

ct a

nd in

tang

ible

cos

ts w

ere

excl

uded

- Iri

note

can

toxi

city

mig

ht a

lso

be a

ffect

ed b

y ot

her p

olym

orph

ism

s, th

ough

this

was

not

take

n in

acc

ount

for t

his

stud

y

Pink

et a

l. 20

13-

The

adju

sted

indi

rect

com

paris

on is

nec

essa

ry to

incl

ude

all p

ossi

ble

trea

tmen

t opt

ions

. How

ever

, thi

s m

ay in

trod

uce

bias

thro

ugh

diffe

renc

es in

tria

l de

sign

, a la

ck o

f acc

ess

to in

divi

dual

pat

ient

dat

a an

d th

e ne

ed to

ext

rapo

late

the

avai

labl

e da

ta fr

om tr

ial t

o lif

etim

e ho

rizon

s.-

The

resu

lts w

ere

obta

ined

from

a n

umbe

r of d

ata

sour

ces

are

diffi

cult

to v

alid

ate

exte

rnal

ly-

Both

the

PKPD

and

the

econ

omic

mod

els

are

para

met

er e

xten

sive

, inc

reas

ing

the

prob

abili

ty th

at s

ome

of th

e va

lues

use

d ar

e in

accu

rate

- Ea

ch o

f the

thre

e st

ages

of t

he m

etho

dolo

gy in

trod

uces

unc

erta

intie

s

Ratt

anav

ipap

ong

et a

l. 20

13-

Due

to th

e ra

rity

of S

JS/T

EN c

ases

, onl

y a

smal

l num

ber o

f pat

ient

s w

as in

clud

ed, r

epre

sent

ing

both

cos

t and

util

ity li

mita

tions

- Th

ere

is n

o su

rvei

llanc

e sy

stem

to q

uant

ify th

e pr

eval

ence

of C

BZ-in

duce

d SJ

S/TE

N in

the

Thai

pop

ulat

ion

- Th

is s

tudy

em

ploy

ed d

ata

from

onl

y on

e st

udy,

whi

ch w

as c

ondu

cted

in a

med

ical

sch

ool i

n Th

aila

nd

144 Chapter 4.1

supp

lem

enta

ry fi

le 2

. (co

ntin

ued)

stud

yD

iscu

ssio

n of

lim

itat

ions

Reed

et a

l. 20

11-

The

impa

ct o

f fina

ster

ide

outs

ide

clin

ical

tria

l set

tings

is u

ncle

ar. I

n th

e PC

PT a

ctiv

e su

rvei

llanc

e fo

r pro

stat

e ca

ncer

con

tinue

d th

roug

hout

the

tria

l, so

it is

co

ncei

vabl

e th

at th

e eff

ect o

f fina

ster

ide

coul

d be

att

enua

ted

amon

g pa

tient

s fo

llow

ed in

rout

ine

prac

tice

- It

is a

lso

uncl

ear w

heth

er th

e tr

eatm

ent e

ffect

s m

easu

red

in th

e PC

PT a

pply

equ

ally

to p

atie

nts

at h

ighe

r ris

k fo

r pro

stat

e ca

ncer

, whe

ther

trea

tmen

t eff

ects

dec

reas

e ov

er ti

me

and

whe

ther

redu

ctio

ns in

the

prev

alen

ce o

f pro

stat

e ca

ncer

lead

to d

ecre

ases

in p

rost

ate

canc

er s

peci

fic m

orta

lity.

- Th

e is

no

esta

blis

hed

thre

shol

d in

the

US

for d

eter

min

ing

whe

ther

an

inte

rven

tion

is c

ost-

effec

tive

- Effi

cien

cy g

ains

wer

e lim

ited

from

a c

ost-

effec

tiven

ess

pers

pect

ive

Rees

e et

al.

2012

- Th

e au

thor

s in

clud

ed b

oth

IMs

and

PMs

for p

rasu

grel

in th

e ge

noty

pe-g

uide

d tr

eatm

ent a

rm, a

lthou

gh a

box

ed w

arni

ng in

the

clop

idog

rel l

abel

onl

y in

clud

es P

Ms.

How

ever

, the

maj

ority

of t

he p

ublis

hed

data

indi

cate

s th

at b

oth

IMs

and

PMs

are

at in

crea

sed

risk.

- Th

e pr

obab

ilitie

s us

ed in

the

base

-cas

e m

odel

s w

ere

obta

ined

from

one

rand

omis

ed tr

ial a

nd s

ubst

udie

s of

that

tria

l. Th

is c

once

rns

age

and

ethn

icity

- Su

bstu

dy a

naly

ses

coul

d ha

ve in

trod

uced

bia

ses

- In

this

stu

dy th

e tr

eatm

ent l

aste

d 15

mon

ths,

whe

ras

the

trea

tmen

t gui

delin

es re

com

men

d an

tipla

tele

t the

rapy

for a

t lea

st o

ne y

ear

- Th

e co

st o

f gen

etic

test

s w

ill p

roba

bly

influ

ence

the

use.

Scha

ckm

anet

al.

2012

- Th

e re

sults

from

a re

tros

pect

ive

anal

ysis

wer

e us

ed, t

houg

h th

ey m

ay n

ot b

e ge

nera

lizab

le to

the

US

or o

ther

pop

ulat

ions

- Th

e au

thor

s as

sum

ed in

the

base

cas

e th

at a

taza

navi

r and

dar

unav

ir ha

t equ

ival

ent e

ffica

cy a

nd c

osts

- Th

e as

sum

ptio

n th

at c

linic

ians

or p

atie

nts

mig

ht p

refe

r to

initi

ate

ataz

anav

ir w

as n

ot c

aptu

red

in th

e ba

se c

ase

QA

LYs

or c

osts

- Th

e ra

nge

of Q

oL e

ffect

s of

hyp

er b

iliru

bina

emia

that

was

con

side

red

may

not

cap

ture

the

full

spec

trum

of c

linic

al s

ituat

ions

- Th

e au

thor

s di

d no

t con

side

r the

pot

entia

l fut

ure

bene

fit o

f UG

T1A1

test

ing

to in

form

pre

scrib

ing

of o

ther

non

-HIV

dru

gs

Economic evaluations of pharmacogenetic tests. 145

supp

lem

enta

ry fi

le 2

. (co

ntin

ued)

stud

yD

iscu

ssio

n of

lim

itat

ions

Serr

etti

et a

l. 20

11-

Ther

e w

as n

o ex

perim

enta

l dat

a ab

out e

ffect

ive

gain

in a

ntid

epre

ssan

t res

pons

e du

e to

pha

rmac

ogen

etic

app

roac

h.-

The

assu

mpt

ions

that

pat

ient

s w

ith s

/s g

enot

ype

trea

ted

with

cita

lopr

am w

ould

hav

e th

e gr

eate

st b

enefi

t fro

m g

enet

ic te

stin

g w

ould

, due

to th

e re

mis

sion

rate

s, im

ply

that

virt

ually

all

the

antid

epre

ssan

t res

pons

e in

the

5-H

TTLP

R s/

s ge

noty

pe g

roup

was

due

to a

pla

cebo

resp

onse

and

non

e to

the

SSRI

ant

idep

ress

ant

- A

t bes

t est

imat

e 50

-60%

of c

hang

e in

resp

onse

out

com

e m

ight

be

due

to a

true

pha

rmac

olog

ic e

ffect

.-

This

stu

dy fo

cuse

d on

ant

idep

ress

ant r

espo

nse

and

side

effe

cts.

How

ever

the

5-H

TTLP

R po

lym

orph

ism

has

als

o be

en fo

und

to m

oder

ate

depr

essi

ve

resp

onse

to e

nviro

nmen

tal s

tres

s-

A lo

ng fo

llow

-up

wou

ld h

ave

been

met

hodo

logi

cally

cor

rect

, but

con

stitu

ted

a se

rious

hur

dle

to b

uild

a re

alis

tic m

odel

. Firs

tly, b

ecau

se o

nly

a m

inor

ity

of p

atie

nts

do n

ot d

rop

out f

rom

trea

tmen

t aft

er a

few

mon

ths.

Seco

ndly

, eve

n in

thos

e w

ho re

mai

n in

trea

tmen

t, ad

here

nce

is s

eldo

m a

t opt

imal

leve

l, an

d th

is m

ay h

ave

nega

tive

cons

eque

nces

for e

ffect

iven

ess

and

cost

. For

a lo

ng-t

erm

ass

essm

ent o

f maj

or d

epre

ssiv

e di

sord

er it

is n

eces

sary

to e

stim

ate

recu

rren

ce ra

te, T

his

is n

ot a

n ea

sy ta

sk, b

ecau

se fo

llow

-up

stud

ies

of d

epre

ssio

n ar

e ch

arac

teriz

ed b

y m

arke

d di

ffere

nces

in te

rms

of d

esig

ns, o

utco

me

defin

ition

s an

d cr

ude

mea

sure

s of

pha

rmac

othe

rapy

- Th

e in

fluen

ce o

f 5-H

TTLP

R on

SSR

I res

pons

e is

doc

umen

ted

in ra

ndom

ised

tria

ls, b

ut b

iase

d by

var

ious

lim

itatio

ns-

Less

cle

ar e

vide

nce

cam

e fr

om n

atur

alis

tic s

tudi

es-

The

real

effe

ct is

not

yet

est

ablis

hed

- Th

e au

thor

s po

sted

that

sen

sitiv

ity to

5-H

TTLP

R va

riant

s w

as e

quiv

alen

t for

all

SSRI

s, th

ough

rece

nt s

tudi

es fo

und

subt

le d

iffer

ence

s-

The

mod

el d

id n

ot a

ccou

nt fo

r rec

ent d

isco

verie

s th

at c

hang

ed th

e st

ruct

ure

and

func

tion

of th

e 5-

HTT

LPR

poly

mor

phis

m.

- In

ord

er to

sim

plify

the

asso

ciat

ion

betw

een

5-H

TTLP

R va

riant

s an

d an

tidep

ress

ant r

espo

nse,

sec

ond-

orde

r int

erac

tions

with

gen

der a

nd li

fe-e

vent

s w

ere

not f

eatu

red

- In

form

atio

n w

as m

issi

ng re

gard

ing

cost

s to

car

egiv

er o

r fam

ily m

embe

rs a

nd p

sych

othe

rapy

- Th

e us

ed ty

pica

l sta

rtin

g do

ses

for S

SRI t

reat

men

t mig

ht b

e su

bopt

imal

and

inte

rfer

e w

ith th

e as

sess

men

t of c

linic

al re

spon

se-

The

impa

ct o

f ant

idep

ress

ant t

reat

men

t on

suic

idal

risk

was

not

feat

ured

- Th

e re

sults

are

onl

y pr

ovis

iona

l bec

ause

key

ass

umpt

ions

rega

rdin

g ga

in in

ant

idep

ress

ant r

espo

nse

and

redu

ctio

n in

sid

e-eff

ect b

urde

n ar

e sp

ecul

ativ

e an

d no

t sup

port

ed b

y em

piric

al fi

ndin

gs

Shiff

man

et a

l. 20

12-

The

auth

ors

wer

e un

able

to m

easu

re o

r find

a p

ublis

hed

estim

ate

of th

e ris

k of

str

oke

even

ts a

ssoc

iate

d w

ith th

e 2

varia

nts

of th

e LP

A ge

ne in

wom

en.

Ther

efor

e, th

ey a

ssum

ed th

at th

e in

crea

sed

stro

ke ri

sk in

wom

en w

as th

e sa

me

as th

e in

crea

sed

MI r

isk

estim

ated

in m

en-

It w

as a

ssum

ed th

at fr

eque

ncy

of th

e LP

A ris

k al

lele

s is

unc

hang

ed a

t diff

eren

t lev

els

of th

e FR

S.-

Ther

e w

as n

o so

urce

doc

umen

ting

the

timin

g of

GI b

leed

ing

even

ts th

at o

ccur

aft

er in

itiat

ion

of a

spiri

n th

erap

y, th

eref

ore

it w

as a

ssum

ed th

at G

I ble

edin

g ev

ents

wou

ld o

ccur

in th

e fir

st y

ear a

fter

initi

atio

n of

ther

apy

- Th

ere

wer

e no

repo

rts

of a

n as

soci

atio

n be

twee

n th

ese

LPA

varia

nts

and

the

risk

of C

VD in

pop

ulat

ions

of n

on-E

urop

ean

ance

stry

146 Chapter 4.1

supp

lem

enta

ry fi

le 2

. (co

ntin

ued)

stud

yD

iscu

ssio

n of

lim

itat

ions

Shiro

iwa

et a

l. 20

10-

The

auth

ors

estim

ated

the

BSC

cost

s, w

hich

wer

e ba

sed

on th

e fe

e-fo

r-se

rvic

e co

sts

of d

aily

opi

oid

use

alth

ough

som

e pa

tient

s do

not

use

them

, or l

ess

freq

uent

ly-

Ther

e w

as n

ot d

ata

avai

labl

e fo

r Jap

anes

e ut

ility

sco

res,

so C

anad

ian

scor

es w

ere

extr

apol

ated

Soric

het

al.

2013

- Co

st b

etw

een

heal

th s

yste

ms

are

diffe

rent

and

resu

lt m

ay n

ot a

pply

to o

ther

hea

lthca

re s

yste

ms

- N

o co

mpa

rison

with

pra

sugr

el o

r gen

otyp

e gu

ided

dos

ing

incl

udin

g al

tern

ativ

e ge

noty

pes

like

Ultr

a Ra

pid

met

abol

izer

s.-

It w

as a

ssum

ed th

at c

lopi

dogr

el o

r tic

agre

lor w

ere

used

for 1

2 m

onth

s af

ter A

CS a

nd th

en s

topp

ed.

Thom

pson

et a

l. 20

14-

The

stud

y’s

sam

ple

may

not

be

gene

raliz

able

to a

ll po

tent

ial p

atie

nts

who

will

be

offer

ed T

PMT

geno

typi

ng in

clin

ical

pra

ctic

e.-

The

stud

y tim

e ho

rizon

sho

uld

have

bee

n lo

nger

to c

aptu

re a

ny p

oten

tial l

ong-

term

cos

ts a

nd b

enefi

ts-

The

pers

pect

ive

of th

e st

udy

mig

ht b

e a

limita

tion,

bec

ause

it d

oes

not t

ake

into

acc

ount

the

cost

s be

yond

hea

lth c

are

reso

urce

use

Tiam

kao

et a

l. 20

13-

This

stu

dy w

as b

ased

on

revi

ews

of li

tera

ture

, so

the

deta

ils c

ould

be

slig

htly

diff

eren

t and

cou

ld re

sult

in d

iscr

epan

cies

- Co

sts

of tr

eatm

ent o

f oth

er c

ompl

icat

ions

aft

er d

isch

argi

ng th

e pa

tient

, re-

adm

issi

on o

r out

-pat

ient

follo

w u

p w

ere

not i

nclu

ded

Vand

erla

anet

al.

2011

- Th

e es

timat

e of

the

net r

educ

tion

in c

hem

othe

rapy

ass

ocia

ted

with

ass

ay te

stin

g w

as b

ased

on

a si

ngle

pub

lishe

d st

udy

for t

he g

ener

al N

+(1-

3)/E

R+

popu

latio

n an

d on

a s

epar

ate

anal

ysis

for t

he >

65-y

ear a

ge g

roup

- Th

e m

odel

use

d an

age

dis

trib

utio

n re

pres

enta

tive

for t

he U

S po

pula

tion

estim

atin

g th

e in

cide

nce

of b

reas

t can

cer,

thou

gh th

e m

odel

was

app

lied

to a

po

tent

ially

you

nger

man

aged

car

e po

pula

tion

- A

pay

er’s

pers

pect

ive

was

app

lied,

and

indi

rect

cos

ts w

ere

excl

uded

- Cl

inic

al tr

ial d

ata

was

use

d to

est

imat

e co

sts

of tr

eatm

ent,

hosp

italis

atio

n an

d m

edic

atio

n as

soci

ated

with

adv

erse

eve

nts,

yet t

he c

osts

are

oft

en h

ighe

r in

a no

n-co

ntro

lled

sett

ing

- Th

e au

thor

s m

odel

led

chan

ge in

che

mot

hera

py u

se o

nly

amon

g th

ose

with

low

recu

rren

t sco

re re

sults

in o

rder

to p

rovi

de a

con

serv

ativ

e es

timat

e of

the

bene

fits

gain

ed fr

om th

e as

say’

s us

e in

this

pop

ulat

ion

- Th

e pl

an c

osts

of c

hem

othe

rapy

dru

gs a

nd s

uppo

rtiv

e ca

re m

ay h

ave

been

ove

rest

imat

ed-

Unp

ublis

hed

surv

ey d

ata

was

use

d to

est

imat

e pr

actic

e pa

tter

ns o

f che

mot

hera

py-r

elat

ed s

uppo

rtiv

e ca

re-

It w

as e

stim

ated

that

90%

of t

he c

ance

rs w

ould

be

non-

HER

2 ov

erex

pres

sing

, alth

ough

this

mig

ht b

e un

cert

ain

- Th

e es

timat

ed 2

.7%

exc

ess

risk

amon

g pa

tient

s re

ceiv

ing

chem

othe

rapy

may

be

an o

vere

stim

ate,

bec

ause

it w

as b

ased

on

data

that

incl

uded

ra

diot

hera

py a

nd h

orm

onal

trea

tmen

t-

The

mod

el d

id n

ot a

ddre

ss th

e us

e of

the

assa

y in

the

popu

latio

n w

ith th

e cu

rren

t gre

ates

t use

- Th

e m

odel

ass

umed

that

can

cer a

nd a

ll-ca

use

mor

talit

y w

ere

estim

ated

bas

ed o

n na

tiona

l sta

tistic

s bu

t did

not

inco

rpor

ate

one

addi

tiona

l ben

efit o

f the

on

coty

pe D

X as

say

(if th

e re

curr

ent s

core

is k

now

n, p

atie

nts’

risk

of c

ance

r-re

late

d de

ath

is s

trat

ified

into

3 c

ateg

orie

s)-

Ther

e is

gre

at v

aria

bilit

y in

the

valu

e th

at p

atie

nts

attr

ibut

e to

che

mot

hera

py tr

eatm

ent

Economic evaluations of pharmacogenetic tests. 147

supp

lem

enta

ry fi

le 2

. (co

ntin

ued)

stud

yD

iscu

ssio

n of

lim

itat

ions

Verh

oef

et a

l. 20

13-

Lack

of d

ata

conc

erni

ng th

e eff

ectiv

enes

s of

gen

otyp

ing

in p

henp

roco

umon

pat

ient

s fr

om c

linic

al tr

ials

.-

Cost

of g

enet

ic te

st-

Surr

ogat

e en

d po

int (

INR)

- Po

ssib

le c

orre

latio

n be

twee

n pa

ram

eter

s w

hich

wer

e in

depe

nden

tly v

arie

d in

the

prob

abili

stic

sen

sitiv

ity a

naly

sis

Vija

yara

ghav

anet

al.

2012

- Th

e an

alys

is w

as re

stric

ted

to th

e us

e of

EG

FR in

hibi

tors

in s

econ

d an

d su

bseq

uent

line

s of

ther

apie

s-

It w

as n

ot ta

ken

in a

ccou

nt th

at re

cent

find

ings

indi

cate

d th

at u

se o

f cet

uxim

ab in

che

mot

hera

py-r

efra

ctor

y co

lore

ctal

can

cer p

atie

nts

is a

ssoc

iate

d w

ith

long

er o

vera

ll su

rviv

al c

ompa

red

to p

atie

nts

with

oth

er K

RAS-

mut

ated

tum

ours

- D

ue to

a la

ck o

f dat

a on

util

ities

for p

atie

nts

with

mCR

C re

ceiv

ing

ther

apy,

no

qual

ity o

f life

adj

ustm

ents

wer

e in

corp

orat

ed in

the

mod

el-

Ther

e is

lim

ited

data

on

the

prec

ise

sens

itivi

ty a

nd s

peci

ficity

of t

he K

RAS

test

, the

refo

re th

e au

thor

s es

timat

ed a

95%

sen

sitiv

ity a

nd 1

00%

spe

cific

ity

You

et a

l. 20

12-

The

mod

el p

roje

cted

life

-long

eve

nts

base

d on

key

fact

ors

deriv

ed fr

om a

2 y

ears

clin

ical

tria

l-

The

cost

item

s w

ere

limite

d to

the

reso

urce

s of

ant

icoa

gula

tion

ther

apy

and

rela

ted

com

plic

atio

ns

You

JH,

2014

- Lo

ng-t

erm

eve

nts

wer

e pr

ojec

ted

by u

sing

sho

rt-t

erm

clin

ical

tria

l dat

a-

Mon

itorin

g of

new

dat

a re

gard

ing

NO

ACs

is re

quire

d to

upd

ate

the

deci

sion

mod

el-

The

cost

item

s w

ere

limite

d to

reso

urce

s of

ant

icoa

gula

tion

ther

apy

and

rela

ted

com

plic

atio

ns-

A T

TR o

f 60%

was

ass

umed

, whi

ch m

ight

not

be

appl

icab

le to

clin

ics

with

hig

her T

TR-

In th

e Co

umaG

en-II

tria

l pat

ient

s w

ith v

ario

us w

arfa

rin in

dica

tions

wer

e m

ixed

, whi

ch c

ould

resu

lt in

unc

erta

inty

in th

e es

timat

ion

of g

enot

ype-

guid

ed

dosi

ng e

ffect

iven

ess

in A

F pa

tient

s in

this

stu

dy.

Zhu

et a

l. 20

13-

The

curr

ent a

naly

sis

did

not e

valu

ate

the

cost

-effe

ctiv

enes

s of

gefi

tinib

mai

nten

ance

trea

tmen

t for

the

who

le c

ohor

t with

out E

GFR

gen

otyp

ing

beca

use

this

dat

a w

as n

ot a

vaila

ble

- Th

e pr

esen

t mod

el d

id n

ot in

clud

e ot

her E

GFR

-tar

gete

d ag

ents

use

d as

mai

nten

ance

trea

tmen

ts, s

uch

as e

rlotin

ib, t

o as

sess

the

incr

emen

tal c

ost-

effec

tiven

ess

in c

ompa

rison

with

gefi

tinib

bec

ause

no

head

-to-

head

tria

l dat

a ar

e cu

rren

tly a

vaila

ble.

- A

bud

get i

mpa

ct a

naly

sis

for t

he a

dditi

on o

f gefi

tinib

mai

nten

ance

trea

tmen

t on

soci

ety

was

not

con

duct

ed-

The

curr

ent a

naly

sis

inco

rpor

ated

PSF

and

OS

data

aft

er c

ance

r pro

gres

sion

from

diff

eren

t tria

ls.

- So

me

mod

el in

puts

wer

e ob

tain

ed fr

om li

tera

ture

pub

lishe

d ab

road

due

to a

lack

of C

hine

se c

linic

al d

ata

- Th

e se

nsiti

vity

and

spe

cific

ity o

f diff

eren

t gen

otyp

ing

faci

litie

s w

as n

ot a

ccou

nted

- In

ord

er to

sim

plify

the

eval

uatio

n, o

ther

adj

uvan

t the

rapi

es w

ere

excl

uded

148 Chapter 4.1

s3. PRimsA checklist

section/topic # Checklist itemReported on page #

titlE

Title 1 Identify the report as a systematic review, meta-analysis, or both. 1

ABstRACt

Structured summary

2 Provide a structured summary including, as applicable: background; objectives; data sources; study eligibility criteria, participants, and interventions; study appraisal and synthesis methods; results; limitations; conclusions and implications of key findings; systematic review registration number.

2

intRODuCtiOn

Rationale 3 Describe the rationale for the review in the context of what is already known.

3-4

Objectives 4 Provide an explicit statement of questions being addressed with reference to participants, interventions, comparisons, outcomes, and study design (PICOS).

3-4

mEtHODs

Protocol and registration

5 Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide registration information including registration number.

4-5

Eligibility criteria

6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale.

4-5

Information sources

7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched.

4-5

Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated.

4-5

Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis).

4-5

Data collection process

10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

4-5

Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made.

4-5

Risk of bias in individual studies

12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.

No meta-analysis was performed

Summary measures

13 State the principal summary measures (e.g., risk ratio, difference in means). No meta-analysis was performed

Synthesis of results

14 Describe the methods of handling data and combining results of studies, if done, including measures of consistency (e.g., I2) for each meta-analysis.

No meta-analysis was performed

Economic evaluations of pharmacogenetic tests. 149

Page 1 of 2

section/topic # Checklist itemReported on page #

Risk of bias across studies

15 Specify any assessment of risk of bias that may affect the cumulative evidence (e.g., publication bias, selective reporting within studies).

No meta-analysis was performed

Additional analyses

16 Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, meta-regression), if done, indicating which were pre-specified.

No meta-analysis was performed

REsults

Study selection 17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram.

5

Study characteristics

18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations.

5-13

s1 and 2

Risk of bias within studies

19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12).

No meta-analysis was performed

Results of individual studies

20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.

No meta-analysis was performed

Synthesis of results

21 Present results of each meta-analysis done, including confidence intervals and measures of consistency.

No meta-analysis was performed

Risk of bias across studies

22 Present results of any assessment of risk of bias across studies (see Item 15).

No meta-analysis was performed

Additional analysis

23 Give results of additional analyses, if done (e.g., sensitivity or subgroup analyses, meta-regression [see Item 16]).

No meta-analysis was performed

DisCussiOn

Summary of evidence

24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).

14

Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias).

16

Conclusions 26 Provide a general interpretation of the results in the context of other evidence, and implications for future research.

16

FunDing

Funding 27 Describe sources of funding for the systematic review and other support (e.g., supply of data); role of funders for the systematic review.

Online submission info

From: Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Med 6(6): e1000097. doi:10.1371/journal.pmed1000097

4.2A model-based cost-effectiveness analysis of routine genotyping for CYP2D6 among

older depressed inpatients starting nortriptyline pharmacotherapy

Elizabeth J. J. Berm, Judith J. Gout-Zwart, Jos Luttjeboer; Bob Wilffert, Maarten J. Postma

Submitted

152 Chapter 4.2

ABstRACt

introduction: Genotyping for CYP2D6 has the potency to predict differences in metabo-lism of nortriptyline. This information could optimize pharmacotherapy. We determined the costs and effects of routine genotyping for old aged Dutch depressed inpatients.methods: With a decision-tree, we modelled the first 12 weeks of nortriptyline therapy. Direct costs of genotyping, hospitalization, therapeutic drug monitoring and drugs were included. Based on genotype, patients could be correctly, sub-, or supratherapeutically dosed. Improvement from sub- or supratherapeutically dosed patients to correctly dosed patients was simulated, assuming that genotyping would prevent under- or overdosing of patients. In the base case, this improvement was assumed to be 35%. A probabilistic sensitivity analysis (PSA) was performed to determine uncertainty around the incremen-tal cost-effectiveness ratio (ICER).Results: In the base case analysis, costs for genotyping were assumed €200 per test with a corresponding ICER at €1 333 000 per QALY. To reach a €50 000 per QALY cut-off, genotyping costs should be decreased towards €40 per test. At genotyping test costs < €35 per test, genotyping was dominant. At test costs of €17 per test there was a 95% probability that genotyping was cost-effective at €50 000 per QALY.Conclusions: CYP2D6 genotyping was not cost-effective at current genotyping costs at a €50 000 per QALY threshold, however at test costs below €40, genotyping could be costs-effective.

A model based cost-effectiveness analysis for CYP2D6 genotyping 153

4.2.1. intRODuCtiOn

Major depressive disorder (MDD) is a disease with a significant burden of disease in the aging European population (1). Antidepressants can be used in the treatment of MDD. There are different classes of antidepressants, with the tricyclic antidepressants (TCAs) representing a treatment option typically initiated after unsuccessful treatment with a selective serotonin reuptake inhibitor (SSRI) (2). Based on the most favorable side-effect profile, nortriptyline is the TCA of first choice among older patients as advised by different guidelines (3,4). In contrast to most SSRIs, TCAs display a clinically relevant dose-effect relationship and therefore therapeutic drug monitoring (TDM) is strongly recommended (5).

Nortriptyline is metabolized by the polymorphic cytochrome P450 2D6 (CYP2D6) enzyme (6). Generally, its polymorphisms are classified into four different phenotypes: poor metabolizer (PM), intermediate metabolizer (IM), extensive metabolizer (EM) or ultra-rapid metabolizer (UM)(7). Notably, EMs represent the most prevalent group in Caucasian populations. Compared to the EMs, PMs and IMs have a decreased activity of the CYP2D6 enzyme, whereas UMs have an increased activity of this enzyme. As a result of this variance in enzymatic activity, patients display large variations in plasma concentrations despite similar drug dosages (8-10).

In addition to TDM, to further improve pharmacotherapy with nortriptyline, routine testing for CYP2D6 has been proposed to facilitate dose adaptations in an early stage of pharmacotherapy (11). Especially in aged patients, genotyping might be beneficial, be-cause patients above 60 years are more frequently exposed to higher plasma concentra-tions of nortriptyline (12). Although the literature is not completely unambiguous, these higher plasma concentrations might relate to more and/or more severe adverse drug reactions (ADRs) (13,14). Besides a reduction in ADRs, better efficacy of nortriptyline can be expected in patients whose plasma concentrations are within therapeutic range, since sub- or supratherapeutic plasma concentrations of nortriptyline can reduce the efficacy of the drug (5,15,16). Potentially related to reductions in ADRs and an increased efficacy, suggestions have been made that CYP2D6 genotyping can reduce psychiatric hospitalization costs (11,17-20). For schizophrenic patients who used CYP2D6 depen-dent antipsychotic agents, this was recently demonstrated in Denmark (21).

To facilitate dose finding of nortriptyline with the use of genetic information, specific guidelines for dose adaptations of nortriptyline have recently become available (22). Indeed, in some secondary psychiatric care facilities in the Netherlands, genotyping is already implemented as care as usual (23). However, no information concerning the cost-effectiveness of routine genotyping at the start of nortriptyline treatment among depressed hospitalized patients is available. We constructed a pharmacoeconomic model, to assess cost-effectiveness of routine genotyping. As safe and fast dose find-

154 Chapter 4.2

ing is considered particularly important among severely depressed aged patients, we designed the model to simulate a Dutch hospitalized population of 60 years and older.

4.2.2. mEtHODs

Design

To evaluate the cost-effectiveness of CYP2D6 genotyping, a decision-tree was built in Microsoft Excel version 2010. Within this model, two virtual cohorts of 1000 depressive patients with an age of 60 years or older and treated with nortriptyline were simulated. In one cohort patients were genotyped whereas in the other patients received care as usual. Subsequently, costs and health outcomes of both cohorts were compared.

According to the Dutch guideline on depression (addendum elderly), patients should be titrated towards a dose of 75 mg per day which takes approximately 12 days (4). At this dose, plasma concentrations should be evaluated. Therefore, we assumed that in the model, plasma concentrations of all patients were evaluated after 12 days. As a result of this evaluation, patients were labeled to have either received a correct dosage, suboptimal dosage, supratherapeutic dosage or had discontinued therapy. Patients who were not optimally dosed received a dose adaptation and a second evaluation of plasma concentration. Patients who were still incorrectly dosed after this second evaluation received one more dose adaptation and control of plasma concentration and were assumed to be correctly dosed afterwards. Patients who discontinued nortriptyline pharmacotherapy entered the next step of treatment (i.e. tranylcypromine 40 mg) which was initiated after a wash out period of 14 days (2). The structure of the decision-tree is shown in figure 1. For both the “care as usual” and “genotyping” cohorts the same structure was used, although the input variables were different.

model Characteristics

All patients in the simulated treatment cohorts were hospitalized in a psychiatric hospi-tal for an episode of severe depression which was followed by discharge. Patients were structured by their genotypes. For simplification of the model, different genotypes led to specific assumptions. First, PMs and IMs could only receive a dose that was either correct or too high. In addition, after dose adjustment, dosing could not become incor-rect in an opposite way. For example, when an EM was dosed too high, the dose could not become too low after adjustment. An UM could only receive a dose that was either correct or too low.

As mentioned before, most patients had an EM genotype, however distribution of genotypes is dependent on the ethnicity of a population. For this study, we used fre-quencies that are reported for European Caucasian populations (24,25).

A model based cost-effectiveness analysis for CYP2D6 genotyping 155

time horizon

It was assumed that CYP2D6 genotype information would only optimize dose finding of nortriptyline during the titration phase. Therefore, effects were limited to the timespan of the titration phase. According to expert opinion, this phase is usually completed within three evaluations. As such, we chose a time horizon of maximally three evalua-tions. To assess the time it would take for three evaluations to occur, we used guideline recommendations and retrospective (2009-2014) anonymous TDM data from the

nortriptyline starter

PM

correct quit

continue

too high quit

continue

correct quit

continue

too high quit

continue correct

IM

correct quit

continue

too high quit

continue

correct quit

continue

too high quit

continue correct

EM

correct quit

continue

too high quit

continue

correct quit

conintue

too high quit

continue correct

too low quit

continue

correct quit

continue

too low quit

continue correct

UM

correct quit

continue

too low quit

continue

correct quit

continue

too low quit

continue correct

evaluation 1 evaluation 2 evaluation 3

nortriptyline starter

PM

correct quit

continue

too high quit

continue

correct quit

continue

too high quit

continue correct

IM

correct quit

continue

too high quit

continue

correct quit

continue

too high quit

continue correct

EM

correct quit

continue

too high quit

continue

correct quit

conintue

too high quit

continue correct

too low quit

continue

correct quit

continue

too low quit

continue correct

UM

correct quit

continue

too low quit

continue

correct quit

continue

too low quit

continue correct

.

evaluation 1 evaluation 2 evaluation 3

Figure 1. Model structure for the treatment of major depressive disorder with nortriptyline during the first 12 weeks. Patients who receive a dosage which led to therapeutic plasma concentrations (i.e. “correct”) did not receive a next dose evaluation. Patients who quit nortriptyline pharmacotherapy entered a wash out period with a disutility of 0.20 for 14 days. After this period, pharmacotherapy with tranylcypromine was initiated for the remaining time in the model. PM= poor metabolizer, IM= intermediate metabolizer, EM= extensive metabolizer, UM= ultra-rapid metabolizer.

156 Chapter 4.2

Clinical Pharmacy of the Diaconessen Hospital Meppel/Hoogeveen, the Netherlands (unpublished results; see Sup 1). First, based on the advices in guidelines, all patients received the first evaluation 12 days after start of therapy (4). Second, based on the TDM data, a second or third evaluation took place after 31 and 38 days. For a total of three evaluations, this resulted in a 12 weeks’ time horizon of the model. In the model it was assumed that after this period, patients were either correctly dosed or had discontinued nortriptyline pharmacotherapy.

Effects of genotyping

For economic evaluations of pharmacogenomics the analytical and clinical validity of the genetic test has to be defined (26). The analytical validity (i.e. ability of the test to differentiate between carriers of different alleles) depends on the method which is used, but is in general known to be very high: >99% for CYP-enzymes (27). In the model it was therefore assume that the analytical validity was 100%. The clinical validity of CYP2D6 testing (i.e. the ability of the test to differentiate between clinical characteristics, like a PM or UM phenotype) depends on the number and type of variant alleles which are ana-lyzed. In the model, genetic variation in the CYP2D6*2, *3, *4, *5, alleles, and duplications of the gene were used for the clinical effects. These effects were based on the prediction of the PM and UM phenotype among a European Caucasian population. In a pharma-cokinetic model it has been reported that when CYP2D6 phenotype predictions are based on genetic variations in these alleles 78.5% of PMs would have supratherapeutic nortriptyline plasma concentrations at a normal daily dose (28). In the pharmacokinetic model dosing of nortriptyline was adapted based on these CYP2D6 alleles As a result, a 51.1% instead of 78.5% of the patients had supratherapeutic plasma concentrations in the PM group. This corresponded to a reduction of ~35% (1- (51.1/78.5)= 0.35) of inadequately dosed patients (28). This reduction was used in the base case of our model for the PM group in the genotyping cohort. For the UM group, an improvement from 60.2% to 39.3% in suboptimal dosed patients was reported which almost equals the effect found in the PM group (i.e. a 35% reduction in inadequately dosed patients). This effect was also included in the model. In the base case no effect for IMs or EMs were assumed. In a scenario analysis we did include effects for IMs (see sensitivity analysis).

Depending on their genotype, PMs and UMs in the ‘genotyping’ cohort received a dif-ferent start dosage compared to patients in the ‘care as usual’ cohort. Dose adaptations were based on the advices given in the Dutch guidelines: 40% of standard dose for PMs and 160% of standard dose for UMs (29). A detailed dosing schedule can be found in Sup.2.

A model based cost-effectiveness analysis for CYP2D6 genotyping 157

Outcome measures

Better efficacy as well as a reduction of ADRs can be expected in patients who have nortriptyline plasma concentrations within the therapeutic range as compared to out-side (5). Both affect quality of life. For ADRs, reduced utilities (ranging from: −0.12 to −0.01) among patients who experience ADRs of antidepressants have been reported (30). For efficacy, a difference in utility ranging from 0.22 to 0.45 between untreated and treated depression has been described (30-34). In addition, a difference in utilities between responders and non-responders to antidepressant of 0.20 has been reported (35). In the model we included a disutility of 0.04 for experiencing an ADR when dosed supra-therapeutically (36). The reported difference of 0.20 between non-responders and responders was used as a disutility for subtherapeutic dosed patients, assuming that patients who were dosed too low, would not respond optimally to nortriptyline pharmacotherapy. Among patients who discontinued nortriptyline pharmacotherapy, a similar disutility of 0.20 was included for the wash out period of 14 days, before the next treatment step with tranylcypromin was initiated. In the model, these disutility’s were linked to the proportions of patients with sub, or supratherapeutic dosages, or to a wash out period for each day patients were in a certain ‘health state’.

For the first evaluation we assumed correctly dosed patients were patients with a nortriptyline plasma concentration within the therapeutic range. Proportions of thera-peutic, sub-, and supratherapeutic plasma concentrations were extracted from literature (28). After the first weeks of treatment, the effects in terms of antidepressant response, start to improve patient’s utility (37). At this stage, in addition to plasma concentration, drug dosing is adapted based on clinical efficacy and ADRs. Therefore, in the model, the proportion of patients who were incorrectly dosed at the second evaluation was not based on the plasma concentration, but on the proportion of patients who had a third TDM request in the clinical pharmacy. As such, it was assumed that patients who were controlled by TDM for a third time were not correctly dosed before (Sup.1). Notably, the TDM data had no information about differences between genotypes. Therefore, conservatively, we assumed similar proportions of incorrectly dosed patients among all genotypes.

In addition to an improved quality of life, a reduction in hospitalization days was in-cluded in the model. Hospitalization of US patients with MDD and a decreased CYP2D6 metabolic capacity was found to be 13% longer than the average (7.8 instead of 6.9 days) (19). Among the Dutch population, average hospitalization duration for MDD (from 2007 to 2011) of patients aged 65 or older was 28.6 days (38). To translate US-results into a Dutch setting, we assumed that based on genotype guided dosing, a proportion-ally similar reduction of 3.7 hospitalization days from the average (0.13*28.6= 3.7) was achieved among eligible patients, by shifting them from suboptimal to therapeutically dosed. As a result, patients who were correctly dosed after the first evaluation were

158 Chapter 4.2

table 1. Main model inputs and variables of deterministic and probabilistic sensitivity analysis

Variable Base case

Range deterministic sensitivity analysis(min- max)

PsA distribution (parameters)

source Old age specific?

Effect of genotyping (i.e. improvement of incorrectly dosed to correctly dosed) (%)

35 31- 38 β (253, 470) (28) No

Incorrectly dosed until first evaluation (%)

Poor metabolizer too high 76 73- 78 β (735, 239) (28) No

intermediate metabolizer too high 56 53- 59 β (572, 443)

Extensive metabolizer too high 12 n.a. n.a.

Extensive metabolizer too low 25 n.a. n.a.

ultra-rapid metabolizer too low 57 54- 60 β (593, 455)

Incorrectly dosed until second evaluation (%)

All genotypes 43 34- 53 β (40, 53) Sup.1 Yes

Additional days after start at which an evaluation took place

Dose evaluation 1 12 5-19 γ (11, 1.06) (4) Yes

Dose evaluation 2 31 Dependents on evaluation 1

Dependent on evaluation 1

Sup.1 Yes

Dose evaluation 3 38 Fixed Fixed Sup.1 Yes

Genotype (% of population)

Poor metabolizer(2 dysfunctional alleles)

8 6- 10 Uniform (24) No

intermediate metabolizer(1 dysfunctional allele)

11 9- 13 Uniform

Extensive metabolizer 79 Dependents on other genotypes

n.a.

ultra-rapid metabolizer(duplication of the gene)

2 1-2 Uniform

Hospitalization

Average duration inpatient care (days) 28.6 +/− 25% γ (61, 0.47) (19,38) Yes

shorter hospital stay when correctly dosed (%)

13 0- 26 β (40, 53) (19) No

Patients who discontinue therapy

After first dose evaluation (%) 22 12- 33 β (11, 39) (56) No

After second dose evaluation (%) 8.5 2- 17 β (3, 37)

Disutility’s

if dosed to high 0.04 0.01- 0.12 PERT (0, 0.04, 0.13) (36) No

if dosed to low 0.20 0.12- 0.28 β (19, 76) (35)

if discontinue nortriptyline therapy 0.20 0.12- 0.28 β (19, 76) assumed

A model based cost-effectiveness analysis for CYP2D6 genotyping 159

discharged from hospitalization after 24.9 days (28.6-3.7= 24.9). The remaining patients (i.e. suboptimal dosed patients) were assumed to be hospitalized for 31.6 days to realize the weighted average of 28.6 days as registered. Main model inputs are shown in table 1.

Costs

Our evaluation was performed from a health-care insurance payers perspective and only direct medical costs were included in the model (table 2). This perspective was chosen, due to a lack of data concerning effects of genotyping on indirect costs and on the relatively modest productivity loss to be expected in a population of 60 years and older. Costs and utilities were not discounted, since our model had a time horizon of 12 weeks. Costs were expressed in 2014 price levels. Drug costs were collected from the website of the Dutch National Health Care Institute (39) and costs for genotyping and TDM were obtained from the Dutch Healthcare Authority (40). Hospitalization costs and costs for consultations with a psychiatrist were obtained from guideline prices listings and con-verted to costs in 2014 using the appropriate deflators (41,42). Drug and TDM costs were included in hospitalization costs and calculated separately after patients were dismissed from hospital stay. Based on an average hospital stay of 28.6 days all patients were as-sumed to have a first dose evaluation during hospitalization which entailed no extra costs (38). For a second or third evaluation additional costs for monitoring were counted. These costs included costs for TDM as well as ¼ of a psychiatric consult for interpretation of the nortriptyline plasma concentration and, if applicable, dose adaptations. Costs for genotyping of CYP2D6 were based on costs for qualitative DNA amplification at €38 per amplification (22). Assuming genotyping would involve testing for five amplifications, costs would be €190 per CYP2D6 test. Note that test costs can be different between labs, due to different methods used for genotyping (i.e. TaqMan analysis, Roche AmpliChip

table 2. Drug costs included in the model.

type of costs Costs in 2014

Genotyping (5 DNA amplifications) € 188.20

TDM (per measurement) € 23.11

Hospitalization (per day) € 255.62

Ambulant contact with psychiatrist € 190.62

Nortriptyline

10 mg € 0.08

25mg € 0.15

50mg € 0.29

Tranylcypromin

40 mg € 0.96

160 Chapter 4.2

CYP450, whole DNA sequencing). Furthermore, we assumed genotype information would be available before or at the moment of start of nortriptyline treatment.

Cost-effectiveness

The incremental cost-effectiveness ratio (ICER) was calculated by comparing costs and quality adjusted life years (QALYs) between the two cohorts. In addition break even test costs were calculated.

Deterministic, probabilistic and scenario sensitivity analysis

Input variables for the deterministic sensitivity analysis are shown in table 1. The outcomes of the deterministic sensitivity analysis was reflected in Tornado diagrams. Threshold analysis was performed on all variables to calculate break even test costs and maximal test costs to reach the €50 000 per QALY cut off (43).

For further assessment of the uncertainty, a probabilistic sensitivity analysis (PSA) was performed. The distributions and parameters of the variables are shown in table 1. In addition, the probability of cost-effectiveness at €50 000 per QALY was calculated for various genotyping test costs (43).

Two scenario analysis were performed. First, a scenario in which the proportions of in-correctly dosed patients at the second evaluation did differ per genotype was included. For this differentiation, the same estimates as for the first evaluation were used (i.e. % incorrect evaluation 1= % incorrect evaluation 2). In the second scenario analysis dose adaptations for IMs (60% of standard dose) were included (29). No data was reported about supratherapeutically dosed IMs or clinical effects of genotyping in the pharma-cokinetic model which was used for the PM and UM genotypes. Therefore, we took the average of the EM and PM group and as such 56% of the IMs was supratherapeutically dosed. Clinical effects of IM phenotype predictions is unknown, but we assumed an similar effect of 35% improvement among incorrectly dosed patients. With respect to clinical validity, It has been reported that 38% of CYP2D6 IMs are falls positive (44). This means, these patients would be dosed to low when dose adaptations would be made based on genotype. Therefore, in this scenario analysis another branch was included in the decision tree for IM patients who were subtherapeutic dosed. As a result, 38% of the IM patients who were incorrectly dosed in the genotyping cohort, were dosed subtherapeutic instead of supratherapeutic. ICERs, QALY loss, break even test costs and test costs to reach the €50 000 per QALY were calculated for both scenarios.

A model based cost-effectiveness analysis for CYP2D6 genotyping 161

4.2.3. REsults

Base case

In the base case scenario an ICER of €1 333 000 per QALY was found. If test costs were €40, an ICER of €50 000 per QALY was found and when test costs were €35 per test, test costs equaled costs savings (i.e. break even). At lower costs, genotyping would become the dominant strategy. A summary of the model outcomes is shown in table 3. The difference in QALY loss were small. Genotyping resulted in a QALY gain of 0.12. The majority of costs concerned hospitalization costs in the first weeks of treatment. In the genotyping cohort, costs for inpatient care and follow-up were lower. According to the model, genotyping could save €32 294 of inpatient care costs, €2 195 for TDM costs, and €244 of drug costs. However, due to genotyping costs, total costs in the genotyping cohort were higher.

table 3. Outcomes of the model for 1000 patients per cohort during 12 weeks of pharmacotherapy in the base case scenario.

Care as usual cohort

genotyping cohort

Difference iCER (€) max. test costs for €50000/QAly (€)

Base case

Patients with 24.9 days of inpatient care (correctly dosed)(n)*

483 447 36

Days of inpatient care (mean) 28.60 28.47 −0.13

Patients who stopped therapy (n) 248 247 −1

Costs of care (€) 7 374 826 7 340 091 −34 734

Costs CYP2D6 genotyping (€) 0 188 200 188 200

total costs (€) 7 374 826 7 528 292 153 466

QAly loss 4.57 4.46 0.12

1 333 148 40

scenario analysis; including genotype based differences for incorrectly dosed patients at evaluation 2

total costs (€) 7 375 299 7 528 218 152 918

QAly loss 4.50 4.34 0.15

999 622 43

scenario analysis; including dose adaptations for ims

total costs (€) 7 374 826 7 497 172 122 346

QAly loss 4.57 4.52 0.05

2 380 626 68

* Suboptimal dosed patients were hospitalized for 31.5 days.

162 Chapter 4.2

Deterministic, probabilistic and scenario sensitivity analysis

The results of the deterministic sensitivity analyses are shown in figure 2. Reduction in the duration of hospitalization among correctly dosed patients emerged as the most important parameter. When no reduction of inpatient care was assumed, test costs should be €2 to reach the €50 000 per QALY cut off, or €8 to break even. The proportion of patients discontinuing nortriptyline pharmacotherapy in the first weeks of therapy markedly influenced the ICER. When more patients continued nortriptyline pharmaco-therapy, genotyping became more effective. With respect to quality of life, the utilities

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140

% incorrect check 1 UM% stops at check 2

disutility when discontinue nortriptylinedisutility dosed too low

% incorrect check 2% incorrect check 1 IM

% incorrect check 1 PMtiming check 1 and 2

% UMdisutility dosed too high

% IM% improvement (effect of genotyping)

% PMmean days of inpatient care

% stops at check 1% improvement hospitalization

Break even test costs €50 000/QALY (€/test)

range min range max

0 10 20 30 40 50 60 70 80 90 100 110 120 130

timing check 1 and 2% stops at check 2

% incorrect check 1 UM% incorrect check 2

% UM% incorrect check 1 IM

% incorrect check 1 PM% IM

% improvement (effect of genotyping)% PM

% stops at check 1mean days of inpatient care

% improvement hospitalization

Break even test costs (€/test)

range min range max

A

B

Figure 2. Results of deterministic sensitivity analyses for genotyping test costs to reach the €50 000/QALY cut-off (A) or to break even with costs savings (B).

A model based cost-effectiveness analysis for CYP2D6 genotyping 163

gained by preventing too high dosing had more influence on the ICER compared to the utilities gained by preventing patients who were dosed to low, even though there was more utility loss counted per case for the latter (0.20 vs. 0.04).

The PSA revealed that in the base case scenario (i.e. at current genotyping costs) a probability applied of <1% for genotyping to be cost-effective (€50 000 per QALY). If genotyping costs were €17 a probability of >95% genotyping to be cost-effective ap-plied (€50 000 per QALY). Cost-effectiveness acceptability curves of the maximal CYP2D6 test costs to obtain an ICER of €50 000 per QALY and breakeven test costs are shown in figure 3.

The first scenario analysis indicated that when proportions of incorrectly dosed pa-tients at the second evaluation equaled those of the first evaluation, this would increase the test costs to reach the €50 000 per QALY cut-off with €3 (i.e. €43 vs. €40 per test in the base case). This scenario analysis had no influence on break even test costs. In the

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

0 25 50 75 100 125 150 175 200 225 250

Prob

abili

ty

CYP2D6 genotyping costs (€/test)

Probability genotyping costs equal €50 000/QALY cut off

Probability genotyping costs equalcosts savings

Figure 3. Probability on cost-effectiveness at €50 000 per QALY (solid line) or probability of test costs to break even with cost-savings (broken line) shown at different CYP2D6 genotyping test costs.

164 Chapter 4.2

scenario analysis which included dose adaptations for IM, test costs to reach the €50 000 per QALY cut-off or break even were identified at €68 or €66, respectively, due to more costs savings by reduced inpatient care of correctly dosed IMs. Although the costs in the genotyping cohort were lower in this scenario, a QALY loss among false positive IMs was introduced by genotyping. This was due to a larger utility loss when dosed sub- instead of supratherapeutic. As a result, incremental utility gain in the genotyping cohorts was reduced with 0.07 towards a remaining QALY gain of 0.05 in the genotyping cohort. Consequently, a higher ICER of €2 400 000 was found. If the proportion of falls posi-tive IMs was decreased towards 26% (instead of 38%) an ICER, similar to the base case scenario was found.

4.2.4. DisCussiOn

Our cost-effectiveness study assessed possible economic implications of genotyping for CYP2D6 among older hospitalized and depressed patients who start nortriptyline therapy. The results were mainly dependent on genotyping costs. At a price of €190 per test, we found genotyping for CYP2D6 unlikely to be cost-effective. In the PSA, a probability of <1% was found at the €50 000 per QALY threshold. However, when genotyping costs were <€35 per test, genotyping was found to be dominant. Within countries costs for CYP2D6 genotyping can be different between labs. For example, in the UK costs of £30 (€39) and £500 (€645) are reported whereas in US costs of $439 (€363) are found, respectively (45). In these labs, the variant alleles studied in our model are usually included and amended with analysis of other alleles. Based on our model, purchasing these tests for routine genotyping of patients who start nortriptyline phar-macotherapy will unlikely be cost-effective. Furthermore, broader genotyping packages exist, for example including other genes like CYP2C19 and CYP2C9. For such packages, even higher test costs of $914 (€1106) are reported (46). Nevertheless, many patients aged > 60 years will receive polypharmacotherapy. For these patients, information about different genes might be beneficial and cost-effective due to multiple drug-gene interactions. Pharmacogenetic tests which asses these multi-drug and gene interaction can be key towards cost-effectiveness of genotyping, although the generalizability of these studies will be challenging (47,48). Besides genotyping costs, the main param-eters which influenced the ICER were the improvement in the duration of hospitalization among correctly dosed patients, mean duration of hospitalization, and the proportion of patients who discontinued nortriptyline pharmacotherapy. In scenario analysis, inclu-sion of dose adaptations for IMs was found to increase the costs savings, but decrease QALY gain, due to false positive IM genotypes which resulted in more subtherapeutic dosed patients. Based on this model, it was found that inclusion of dose adaptations for

A model based cost-effectiveness analysis for CYP2D6 genotyping 165

IMs does decrease the cost-effectiveness of genotyping. To obtain a similar ICER as in the base scenario a better correlation than currently reported, between genotype and phenotype of IMs, is needed.

Our analysis has some limitations. Firstly, the model was based on some assumptions of which the most important one was the genotyping-facilitated effect in 35% of pa-tients transferring from sub- or supra-therapeutically dosed to therapeutically dosed. Except for the pharmacokinetic modeling study, a description of such estimates was not found (28). It is important to obtain better estimates about the effects of genotyping to increase the reliability of the established cost-effectiveness of genotyping. In addi-tion, we only included genotype specific variation in the estimate of the proportion of incorrectly dosed patients at the first evaluation and not at the second evaluation, due to a lack of data. In scenario analyses, inclusion of such a differentiation was found to have almost no influence on the breakeven test costs and only a minor increase in the maximum test costs for the €50 000 per QALY cut-off. Secondly, the relation between ADRs and high plasma concentrations of nortriptyline is maybe not as straightforward as in our model (49). Nevertheless, among antidepressants, tricyclic antidepressants like nortriptyline have a well-accepted dose-effect relationship and ADRs can give a lower quality of life (5,30). Moreover, the upper limit of the therapeutic range is not solely there due to the occurrence of ADRs. Patients with a supratherapeutic dose often display non-response, irrespective of ADRs, and therefore it is important to maintain the upper limit of the therapeutic range (9). Therefore, a disutility of 0.04 for patients with supratherapeutic plasma concentration as included in our model, was considered reasonable. Lastly, we assumed patients who were dosed to low would not be dosed to high after dosed adjustments, and vice versa for patients who were dosed to high. This is an oversimplification of the reality, however, considering the relative wide range of the therapeutic window of nortriptyline is it a fair assumption for most of the patients. In addition, when not applying this simplification, the effects of including other branches in the model, would have had little effect on our estimates due to the small number (n <50) of patients with deviating genotypes who were still incorrectly dosed at the second evaluation.

Similar approaches as sketched here for depressed patients can be found in other disease areas. In a trial based analysis of genotyping for thiopurine methyltransferase (TMPT) in combination with azathioprine treatment it was found that clinicians con-tinued to give a low dosing although the genotype information suggested otherwise (50). Therefore, we did not assume an effect of genotyping among the EMs, however if such an effect can be demonstrated, this will markedly influence the ICER in favor of the genotyping strategy.

There are over 100 alleles reported which influence the CYP2D6 phenotype. The preva-lence of the different alleles depends on the ethnicity of a population (24). Consequently,

166 Chapter 4.2

the relation between genotypes and phenotypes can be different between populations when tested for the same set of alleles and is likely to increase when a test includes more alleles (51). According to their methods, Jornil et al. only included variation explained by four alleles and duplications of the gene in their estimates (8). In our model, we included testing and therefore costs for these alleles. However, testing for additional alleles is likely to improve the linkage between genotypes and phenotypes (i.e. clinical sensitivity and specificity of the test). Therefore the 35% reduction of incorrectly dosed patients which we included as the effect of genotyping might be a slight underestimation. Nev-ertheless, testing for more alleles will likely, but not necessarily, increase costs as and will need additional economic evaluation. Beside clear differences like costs, differences between countries and labs in methods for analyses of the CYP2D6 gene and the result-ing analytical and clinical accuracy should be considered when extrapolating our results to other regions or countries.

Two studies assessing clinical utility of genotyping for CYP2C9 in combination with warfarin therapy found contradicting results. Consequently, the authors emphasized the importance of the moment in time of genotyping and effect measurement (52,53). For warfarin, methods to correct for differences in metabolism between patients like TDM, gender, and age based algorithms are well developed. Therefore, effects of geno-typing depend on existing monitoring methods and the study design. Moreover, effects would predominantly be observed in the first weeks of therapy. With our model, we tried to capture such little effects in the first weeks of nortriptyline pharmacotherapy which might be challenging to find in observational studies. We assumed an optimal situation in which genotype information was available before start of treatment. This is an essential assumption, because if a lab is not able to provide the information at this moment in time, effects of genotyping will be diluted and the ICER will increase. Matchar et al., analyzed the possible effect, but not cost-effectiveness of genotyping for CYP2D6 when a selective serotonin reuptake inhibitors (SSRIs) was indicated (54). They used fluoxetine as a model SSRI to be influenced by CYP2D6 polymorphisms. Similar to our model, the authors had to make substantial assumptions. In addition, they used an expert opinion to predict the clinical effects of genotyping in relation to treatment response. As a consequence, one of their main recommendations was to conduct a prospective study to assess the real clinical effects of genotyping. The same holds true for our model, however our assumptions were based on a thorough pharmacokinetic study (28). In addition we performed an extensive sensitivity analysis to give insights into the uncertainty. Therefore we consider our estimates valuable for decision makers.

A model based cost-effectiveness analysis for CYP2D6 genotyping 167

4.2.5. COnClusiOn

A cost-effectiveness analysis on effects of genotyping for the CYP2D6 enzyme in older Dutch inpatients treated with nortriptyline was performed. The effects of genotyping were esti-mated with a decision tree. At current costs, genotyping was not cost-effective. Our finding suggests genotyping would be cost-effective (i.e. €50 000 per QALY) when genotyping costs were decreased towards €40 per test. When genotyping costs were decreased towards €35 per test, genotyping was found to be the dominant treatment option.

168 Chapter 4.2

4.2.6. REFEREnCEs

1 Mathers CD, Loncar D. Projections of global mortality and burden of disease from 2002 to 2030. PLoS Med 2006 Nov; 3(11): e442.

2 Rush AJ, Trivedi MH, Wisniewski SR, Nierenberg AA, Stewart JW, Warden D, et al. Acute and longer-term outcomes in depressed outpatients requiring one or several treatment steps: a STAR*D report. Am J Psychiatry 2006 Nov; 163(11): 1905-1917.

3 Reuben D, Herr K, Pacala J, Pollock B, Potter J, Semla T. Depression. Geriatrics at your fingertips. 16th ed.: American Geriatrics Society; 2014.

4 Landelijke stuurgroep multidiciplinaire richtlijnontwikkeling in de GGZ. Dutch Guideline Depres-sion (Addendum Elderly) Addendum Ouderen bij MDR derpessie. 2008.

5 Hiemke C, Baumann P, Bergemann N, Conca A, Dietmaier O, Egberts K, et al. AGNP consensus guidelines for therapeutic drug monitoring in psychiatry: update 2011. Pharmacopsychiatry 2011 Sep; 44(6): 195-235.

6 Breyer-Pfaff U. The metabolic fate of amitriptyline, nortriptyline and amitriptylinoxide in man. Drug Metab Rev 2004 Oct; 36(3-4): 723-746.

7 Ingelman-Sundberg M. Genetic polymorphisms of cytochrome P450 2D6 (CYP2D6): clinical consequences, evolutionary aspects and functional diversity. Pharmacogenomics J 2005; 5(1): 6-13.

8 Dalen P, Dahl ML, Bernal Ruiz ML, Nordin J, Bertilsson L. 10-Hydroxylation of nortriptyline in white persons with 0, 1, 2, 3, and 13 functional CYP2D6 genes. Clin Pharmacol Ther 1998 Apr; 63(4): 444-452.

9 Montgomery SA, Braithwaite RA, Crammer JL. Routine nortriptyline levels in treatment of de-pression. Br Med J 1977 Jul 16; 2(6080): 166-167.

10 Gressier F, Verstuyft C, Hardy P, Becquemont L, Corruble E. Response to CYP2D6 substrate antide-pressants is predicted by a CYP2D6 composite phenotype based on genotype and comedications with CYP2D6 inhibitors. J Neural Transm 2015 Jan; 122(1): 35-42.

11 Kropp S, Lichtinghagen R, Winterstein K, Schlimme J, Schneider U. Cytochrome P-450 2D6 and 2C19 polymorphisms and length of hospitalization in psychiatry. Clin Lab 2006; 52(5-6): 237-240.

12 Unterecker S, Riederer P, Proft F, Maloney J, Deckert J, Pfuhlmann B. Effects of gender and age on serum concentrations of antidepressants under naturalistic conditions. J Neural Transm 2013 Aug; 120(8): 1237-1246.

13 Kok RM, Aartsen M, Nolen WA, Heeren T. The course of adverse effects of nortriptyline and venla-faxine in elderly patients with major depression. J Am Geriatr Soc 2009 Nov; 57(11): 2112-2117.

14 Kin NM, Klitgaard N, Nair NP, Amin M, Kragh-Sorensen P, Schwariz G, et al. Clinical relevance of serum nortriptyline and 10-hydroxy-nortriptyline measurements in the depressed elderly: a multicenter pharmacokinetic and pharmacodynamic study. Neuropsychopharmacology 1996 Jul; 15(1): 1-6.

15 Preskorn SH, Dorey RC, Jerkovich GS. Therapeutic drug monitoring of tricyclic antidepressants. Clin Chem 1988 May; 34(5): 822-828.

16 Kraus RP, Diaz P, McEachran A. Managing rapid metabolizers of antidepressants. Depress Anxiety 1996-1997; 4(6): 320-327.

17 Laika B, Leucht S, Heres S, Steimer W. Intermediate metabolizer: increased side effects in psychoactive drug therapy. The key to cost-effectiveness of pretreatment CYP2D6 screening? Pharmacogenomics J 2009 Dec; 9(6): 395-403.

A model based cost-effectiveness analysis for CYP2D6 genotyping 169

18 Chou WH, Yan FX, de Leon J, Barnhill J, Rogers T, Cronin M, et al. Extension of a pilot study: impact from the cytochrome P450 2D6 polymorphism on outcome and costs associated with severe mental illness. J Clin Psychopharmacol 2000 Apr; 20(2): 246-251.

19 Ruano G, Szarek BL, Villagra D, Gorowski K, Kocherla M, Seip RL, et al. Length of psychiatric hospitalization is correlated with CYP2D6 functional status in inpatients with major depressive disorder. Biomark Med 2013 Jun; 7(3): 429-439.

20 Chen S, Chou WH, Blouin RA, Mao Z, Humphries LL, Meek QC, et al. The cytochrome P450 2D6 (CYP2D6) enzyme polymorphism: screening costs and influence on clinical outcomes in psychia-try. Clin Pharmacol Ther 1996 Nov; 60(5): 522-534.

21 Herbild L, Andersen SE, Werge T, Rasmussen HB, Jurgens G. Does pharmacogenetic testing for CYP450 2D6 and 2C19 among patients with diagnoses within the schizophrenic spectrum reduce treatment costs? Basic Clin Pharmacol Toxicol 2013 Oct; 113(4): 266-272.

22 Hicks JK, Swen JJ, Thorn CF, Sangkuhl K, Kharasch ED, Ellingrod VL, et al. Clinical Pharmacogenet-ics Implementation Consortium Guideline for CYP2D6 and CYP2C19 Genotypes and Dosing of Tricyclic Antidepressants. Clin Pharmacol Ther 2013 Jan 16; 93(5): 402-408.

23 Kootstra-Ros JE, Van Weelden MJ, Hinrichs JW, De Smet PA, van der Weide J. Therapeutic drug monitoring of antidepressants and cytochrome p450 genotyping in general practice. J Clin Pharmacol 2006 Nov; 46(11): 1320-1327.

24 Bradford LD. CYP2D6 allele frequency in European Caucasians, Asians, Africans and their descen-dants. Pharmacogenomics 2002 Mar; 3(2): 229-243.

25 van Schaik RHN, van Fessem MAC, Schek PW, Lindemans J. CYP2D6 genotypen in de Nederlandse populatie, bepaald met de Roche AmpliChip CYP450 (in Dutch). Nederlands Tijdschrift voor Klinische Chemie en Laboratoriumgeneeskunde 2006; 31: 234-235.

26 Higashi MK, Veenstra DL. Managed care in the genomics era: assessing the cost effectiveness of genetic tests. Am J Manag Care 2003 Jul; 9(7): 493-500.

27 Daly TM, Dumaual CM, Miao X, Farmen MW, Njau RK, Fu DJ, et al. Multiplex assay for comprehen-sive genotyping of genes involved in drug metabolism, excretion, and transport. Clin Chem 2007 Jul; 53(7): 1222-1230.

28 Jornil J, Jensen KG, Larsen F, Linnet K. Risk assessment of accidental nortriptyline poisoning: the importance of cytochrome P450 for nortriptyline elimination investigated using a population-based pharmacokinetic simulator. Eur J Pharm Sci 2011 Oct 9; 44(3): 265-272.

29 Dutch Pharmacogenetics Working Group Guideline for nortriptyline and CYP2D6. 2011; Avail-able at: https: //www.pharmgkb.org/guideline/PA166104961.

30 Revicki DA, Wood M. Patient-assigned health state utilities for depression-related outcomes: dif-ferences by depression severity and antidepressant medications. J Affect Disord 1998 Feb; 48(1): 25-36.

31 Sapin C, Fantino B, Nowicki ML, Kind P. Usefulness of EQ-5D in assessing health status in primary care patients with major depressive disorder. Health Qual Life Outcomes 2004 May 5; 2: 20.

32 Lapid MI, Piderman KM, Ryan SM, Somers KJ, Clark MM, Rummans TA. Improvement of quality of life in hospitalized depressed elderly. Int Psychogeriatr 2011 Apr; 23(3): 485-495.

33 Caruso R, Rossi A, Barraco A, Quail D, Grassi L, Italian FINDER study group. The Factors Influencing Depression Endpoints Research (FINDER) study: final results of Italian patients with depression. Ann Gen Psychiatry 2010 Jul 29; 9: 33-859X-9-33.

34 Sobocki P, Ekman M, Agren H, Runeson B, Jonsson B. The mission is remission: health economic consequences of achieving full remission with antidepressant treatment for depression. Int J Clin Pract 2006 Jul; 60(7): 791-798.

170 Chapter 4.2

35 Greenhalgh J, Knight C, Hind D, Beverley C, Walters S. Clinical and cost-effectiveness of electro-convulsive therapy for depressive illness, schizophrenia, catatonia and mania: systematic reviews and economic modelling studies. Health Technol Assess 2005 Mar; 9(9): 1-156, iii-iv.

36 Perlis RH, Patrick A, Smoller JW, Wang PS. When is pharmacogenetic testing for antidepressant response ready for the clinic? A cost-effectiveness analysis based on data from the STAR*D study. Neuropsychopharmacology 2009 Sep; 34(10): 2227-2236.

37 Gildengers AG, Houck PR, Mulsant BH, Pollock BG, Mazumdar S, Miller MD, et al. Course and rate of antidepressant response in the very old. J Affect Disord 2002 May; 69(1-3): 177-184.

38 Statistic Netherlands (CBS). Hospitalization; gender, age, diagnosed ICD9. 2014; Available at: http://statline.cbs.nl/StatWeb/publication/?DM=SLNL&PA=71860ned&D1=4,6&D2=0&D3=5-6&D4=290&D5=26-30&HDR=T,G1&STB=G2,G3,G4&VW=T.

39 Medical Costs. 2014; Available at: www.medicijnkosten.nl. 40 Cost table DBC-careproducts and other products, January, 1st, 2014. Available at: http://www.nza.

nl/98174/149057/Tarieventabel-DBC-zorgproducten-en-overige-producten-per-1-januari_2014.xls.

41 Consumersprices; inflation since 1963. 2015; Available at: http://statline.cbs.nl/Statweb/publication/?DM=SLNL&PA=70936ned&D1=0&D2=610,623,636,649,662,675&HDR=T&STB=G1&VW=T.

42 Tan SS, Bouwmans CA, Rutten FF, Hakkaart-van Roijen L. Update of the Dutch Manual for Costing in Economic Evaluations. Int J Technol Assess Health Care 2012 Apr; 28(2): 152-158.

43 Rozenbaum MH, Sanders EA, van Hoek AJ, Jansen AG, van der Ende A, van den Dobbelsteen G, et al. Cost effectiveness of pneumococcal vaccination among Dutch infants: economic analysis of the seven valent pneumococcal conjugated vaccine and forecast for the 10 valent and 13 valent vaccines. BMJ 2010 Jun 2; 340: c2509.

44 Rebsamen MC, Desmeules J, Daali Y, Chiappe A, Diemand A, Rey C, et al. The AmpliChip CYP450 test: cytochrome P450 2D6 genotype assessment and phenotype prediction. Pharmacogenom-ics J 2009 Feb; 9(1): 34-41.

45 Fleeman N, Martin Saborido C, Payne K, Boland A, Dickson R, Dundar Y, et al. The clinical ef-fectiveness and cost-effectiveness of genotyping for CYP2D6 for the management of women with breast cancer treated with tamoxifen: a systematic review. Health Technol Assess 2011 Sep; 15(33): 1-102.

46 Brixner D, Biltaji E, Bress A, Unni S, Ye X, Mamiya T, et al. The effect of pharmacogenetic profiling with a clinical decision support tool on healthcare resource utilization and estimated costs in the elderly exposed to polypharmacy. J Med Econ 2015 Oct 19: 1-16.

47 Biskupiak JE, Biltaji E, Bress A, Unni S, Ye X, Yu B, et al. Value Assessment for Genetic Testing of Drug Variation In An Elderly Population. Value Health 2015 Nov; 18(7): A747.

48 Altar CA, Carhart JM, Allen JD, Hall-Flavin DK, Dechairo BM, Winner JG. Clinical validity: Combi-natorial pharmacogenomics predicts antidepressant responses and healthcare utilizations better than single gene phenotypes. Pharmacogenomics J 2015 Oct; 15(5): 443-451.

49 Hodgson K, Tansey KE, Uher R, Dernovsek MZ, Mors O, Hauser J, et al. Exploring the role of drug-metabolising enzymes in antidepressant side effects. Psychopharmacology (Berl) 2015 Mar 12.

50 Thompson AJ, Newman WG, Elliott RA, Roberts SA, Tricker K, Payne K. The cost-effectiveness of a pharmacogenetic test: a trial-based evaluation of TPMT genotyping for azathioprine. Value Health 2014 Jan-Feb; 17(1): 22-33.

51 LLerena A, Naranjo ME, Rodrigues-Soares F, Penas-LLedo EM, Farinas H, Tarazona-Santos E. Inter-ethnic variability of CYP2D6 alleles and of predicted and measured metabolic phenotypes across world populations. Expert Opin Drug Metab Toxicol 2014 Nov; 10(11): 1569-1583.

A model based cost-effectiveness analysis for CYP2D6 genotyping 171

52 Pirmohamed M, Burnside G, Eriksson N, Jorgensen AL, Toh CH, Nicholson T, et al. A randomized trial of genotype-guided dosing of warfarin. N Engl J Med 2013 Dec 12; 369(24): 2294-2303.

53 Kimmel SE, French B, Kasner SE, Johnson JA, Anderson JL, Gage BF, et al. A pharmacogenetic versus a clinical algorithm for warfarin dosing. N Engl J Med 2013 Dec 12; 369(24): 2283-2293.

54 Matchar DB, Thakur ME, Grossman I, McCrory DC, Orlando LA, Steffens DC, et al. Testing for cytochrome P450 polymorphisms in adults with non-psychotic depression treated with selective serotonin reuptake inhibitors (SSRIs). Evid Rep Technol Assess (Full Rep) 2007 Jan; (146)(146): 1-77.

55 Lesko LJ. Personalized medicine: elusive dream or imminent reality? Clin Pharmacol Ther 2007 Jun; 81(6): 807-816.

56 Roberts RL, Mulder RT, Joyce PR, Luty SE, Kennedy MA. No evidence of increased adverse drug reactions in cytochrome P450 CYP2D6 poor metabolizers treated with fluoxetine or nortriptyline. Hum Psychopharmacol 2004 Jan; 19(1): 17-23.

172 Chapter 4.2

332 patients

264 patients

100 patient returened for 2nd TMD within 3 months after 31 (SD ±21) days

43 returned for 3rd TDM within 3 months

after 38 (SD ±24) days

26 returned after 3 months for 3rd

TDM 31 no 3rd TDM

18 returned after 3 months for 2nd

TDM 146 no 2nd TMD

68 patients identified as periodic TDM and excluded

supplementary file 1. Data used to identify time horizon. Based on retrospective (2009-2014) collected TDM data from the Clinical Pharmacy of the Diaconessen Hospital Meppel/Hoogeveen, the Netherlands 100 out of 264 (38%) patients aged ≥ 60 years received TDM for the second time and 43 out of these 100 (43%) patients received TDM for a third time.

supplementary file 2: Drug dosing schedules used to calculate differences in drug costs. Dose adjustment were based on the guidelines from the Dutch pharmacogenetics working group.

Care as usual cohort

genotype starting dose(mg per day)

Dose after first evaluation(mg per day)

Dose after second evaluation(mg per day)

Pm 75 If too high: 50 If too high: 25

im 75 If too high: 50 If too high: 25

Em 75 If too low: 100If too high: 50

If too low: 125If too high: 25

um 75 If too low: 100 If too low: 125

Genotyping cohort

genotype starting dose (mg per day)

Dose after first evaluation(mg per day)

Dose after second evaluation(mg per day)

Pm 30 If too high: 25 If too high: 10

im 75 If too high:50 If too high: 25

Em 75 If too low: 100If too high: 50

If too low: 125If too high: 25

um 125 If too low: 150 If too low: 175

Genotyping cohort (as included in scenario analysis with dose adaptations for IMs)

genotype starting dose (mg per day)

Dose after first evaluation(mg per day)

Dose after second evaluation(mg per day)

Pm 30 If too high: 25 If too high: 10

im 50 If too low: 75If too high: 30

If too low: 100If too high: 25

Em 75 If too low: 100If too high: 50

If too low: 125If too high: 25

um 125 If too low: 150 If too low: 175

4.3Effects and cost-effectiveness of

pharmacogenetic screening for CYP2D6 among older adults starting therapy with

nortriptyline or venlafaxine: study protocol for a pragmatic randomized controlled trial

(CYSCE trial)

Elizabeth J. J. Berm, Eelko Hak, Maarten J. Postma, Marjolein Boshuisen, Laura Breuning, Jacobus R. B. J. Brouwers, Ton Dhondt, Paul A. F. Jansen,

Rob M. Kok, Jan G. Maring, Rob van Marum, Hans Mulder, Richard C. Oude Voshaar, Arne J. Risselada, Harry Venema, Liesbeth Vleugel, Bob Wilffert

Trials. 2015 January 31;16:37

176 Chapter 4.3

ABstRACt

introduction: Nortriptyline and venlafaxine are commonly used antidepressants for treatment of depression in older patients. Both drugs are metabolized by the poly-morphic cytochrome P450-2D6 (CYP2D6) enzyme and guidelines for dose adaptations based on the CYP2D6 genotype have been developed. The CYP2D6 Screening Among Elderly (CYSCE) trial is designed to address the potential health and economic value of genotyping for CYP2D6 in optimizing dose-finding of nortriptyline and venlafaxine.methods: In a pragmatic randomized controlled trial, patients diagnosed with a major depressive disorder according to the DSM-IV and aged 60 years or older will be recruited from psychiatric centers across the Netherlands. After CYP2D6 genotyping determined in peripheral blood obtained by finger-prick, patients will be grouped into poor, inter-mediate, extensive, or ultrarapid metabolizers. Patients with deviant genotype (that is poor, intermediate or ultrarapid genotype) will be randomly allocated to an intervention group in which the genotype and dosing advice is communicated to the treating physi-cian, or to a control group in which patients receive care as usual. Additionally, an ex-ternal reference group of patients with the extensive metabolizer genotype is included. Primary outcome in all groups is time needed to obtain an adequate blood level of the antidepressant drug. Secondary outcomes include adverse drug reactions measured by a shortened Antidepressant Side-Effects Checklist (ASEC), and cost-effectiveness of the screening.Discussion. Results of this trial will guide policy-making with regard to pharmacogenet-ic screening prior to treatment with nortriptyline or venlafaxine among older patients with depression.trial registration. ClinicalTrials.gov: NCT01778907; registration date: 22 January 2013.

Study protocol of the CYSCE trial 177

4.3.1. intRODuCtiOn

Depressive disorder is a chronic disease with a considerable impact on mental and physical health as well as quality of life (1). The 1-month prevalence of major depressive disorder (MDD) in the Dutch population (aged 55 to 85 years) is estimated around 2.0% (2). From a meta-analysis (studies included from 1999 to 2009) a higher point prevalence estimation of 7.2% has been reported for late life depression (75+ years) (3). Among all mental diseases, MDD is ranked as a very disabling disease (4) and depressive disorders are projected to be the largest cause for disability by 2030 in high-income countries (5).

MDD can be treated with antidepressants, which are classified into different groups: selective serotonin reuptake inhibitors (SSRIs), tricyclic antidepressant (TCAs) and others (for example,. selective serotonin-noradrenaline reuptake inhibitors (SNRIs), mirtazap-ine, monoamine oxidase inhibitors). Although overall efficacy estimates are virtually similar between these drug groups, adverse drug profiles differ (6). Worldwide, clinical guidelines usually advise first-step treatment with an SSRI because of a more favorable adverse effect profile compared to other antidepressants (for example, fewer cardiac side-effects and often lower drug anticholinergic side-effects) (7,8). If the chosen SSRI is not effective, switching to a TCA, preferably nortriptyline, or an SNRI, mostly venlafaxine, is an option (7).

Venlafaxine and nortriptyline are both metabolized by the polymorphic cytochrome P450 (CYP2D6) iso-enzyme (8). Its polymorphic differences are generally classified into four different phenotypes: poor metabolizers (PM), intermediate metabolizers (IM), extensive metabolizers (EM), and ultrarapid metabolizers (UM). With respect to EM, PM and IM have a decreased enzymatic activity whereas UM have a higher activity, which is described in more depth elsewhere (9). As a result of the variation in metabolizing capacity, as well as differences in gender, age, co-medication, and comorbidity among patients, different pharmacokinetics are observed (9,10). These differences can be monitored by therapeutic drug monitoring (TDM), which helps to find the right dosage for an individual patient. TDM of nortriptyline is ‘strongly recommended’ and for venla-faxine it is recommended especially during start of therapy (11). Prior and in addition to TDM, information about the metabolic CYP2D6 activity could further improve this dose-finding. Since it is possible to detect polymorphic differences of CYP2D6 based on genetic material, it is not dependent on the actual antidepressant intake (patient compliance) or collection time of the sample, like TDM samples are. For nortriptyline and venlafaxine genotyped-based dose adjustment guidelines have been formulated (10,12,13). In the Netherlands, the Royal Dutch Pharmacist Association (KNMP) has made such guidelines available for current clinical practice (14,15).

Despite the large number of studies that assessed the relationship between phar-macogenetics and pharmacokinetics of antidepressants, implementation of phar-

178 Chapter 4.3

macogenetic knowledge into clinical practice is still scarce. The hypothesized clinical improvements facilitated by genotype-based dose adjustments in addition to current TDM, like improved efficacy and prevention of adverse drug reactions, are still a matter of controversy and discussion (16,17). There is a need to test this hypothesis in daily clinical practice. Therefore, we designed a pragmatic multicenter randomized controlled trial to determine the effects of a pharmacogenetic screening for CYP2D6 on the time needed to obtain adequate blood levels of nortriptyline or venlafaxine. Since older persons are more vulnerable for adverse drug reactions, beneficial effects of genotype adjusted dosing are expected to be more apparent in this population. Therefore, the trial is conducted among older depressed patients starting with nortriptyline or venlafaxine. The cost-effectiveness of this screening will be assessed to support decision-making on the potential implementation of screening for CYP2D6 genotype in daily clinical prac-tice. The trial is designed to allow reporting according to the CONSORT guidelines (18).

4.3.2. mEtHOD/DEsign

study design

The study is a multicenter randomized controlled trial across multiple old age psychiatry and geriatric mental health care institutions in the Netherlands. Patients will be recruited by their treating physician or a specialized trial nurse. The study consists of two parts and for both parts separate written informed consent will be obtained. The first part is a basic genotype screening study in which each eligible patient (see description in Participants below) starting with nortriptyline or venlafaxine will be asked permission for genotyping. Patients with a PM, IM and UM genotype will be selected for participa-tion in the second (trial) part of the study. After giving the second informed consent, the trial entails a follow-up period of 6 weeks or longer in order to monitor the blood levels and adverse reactions of either nortriptyline or venlafaxine. A random selection of patients with an EM genotype will be allocated to an additional reference group (Figure 1). Recruitment is expected to be completed within approximately 3 years. If selected for the trial part of the study, eligible patients are invited for a baseline visit. At baseline, 2 weeks, and 4 weeks after baseline blood samples are collected to estimate the blood level of the drug by a ‘Dried Blood Spot’ (DBS) method. Additionally, questionnaires concerning adverse drug reactions, quality of life, productivity and health care use, and severity of depression are completed. Patients who do not complete dose-finding after 6 weeks will be followed for additional 2-week measurements until dose-finding is com-pleted. These additional 2-week samples will be collected only for those patients who had a dose change 3 weeks or less prior to the moment the sample was collected, since it can take up to 2 to 3 weeks for PMs of nortriptyline to reach steady state concentration

Study protocol of the CYSCE trial 179

(19). When the follow-up period is finished for an individual patient, the genotype infor-mation will be supplied to the treating physician. Ethical approval was obtained by an Independent Ethics Committee (RTPO-Leeuwarden-NL; file number: NL40925.099.12) and the study will be conducted in accordance with the Declaration of Helsinki.

Participants

Patients 60 years or older and diagnosed with a major depression according to the Diagnostic and Statistical Manual of Mental Disorders, fourth edition text revision (DSM-IV TR) (20), criteria (code: 296.2x or 296.3x) diagnosed by the treating physician, are eligible for inclusion. Patients should be starting with either nortriptyline or venlafaxine and competent to understand the two separate informed consent procedures. Patients with known liver cell damage, which in clinical practice often serves as a proxy for a poor he-patic function (aspartate aminotransferase and alanine aminotransferase (ASAT/ALAT) or gamma-glutamyl transferase (γ-GT) ≥ twice the maximal reference value), impaired renal function (eGFR < 30 mL/minute) in combination with venlafaxine use, or currently using drugs influencing blood levels of nortriptyline or venlafaxine are excluded for trial participation. Patients using terbinafine, ketoconazole, voriconazole, kinidine, propafenon, cimetidine, fluoxetine, paroxetine, bupropion, duloxetine, sertraline, abirateron, cinacalcet, rifampicine, and ritanovir are excluded for trial participation. For

 

  Genotype patients

Deviating genotypes

(PM/IM/UM)

Randomization

Intervention group (DG-I) n=75

Control group (DG-C) n=75

Normal genotype

(EM)

Select controls depending on the

number in randomization group

EM control group (NG-C) n=75

Not included

IC  1  

IC  2  

IC Informed consent PM Poor Metabolizer IM Intermediate Metabolizer EM Extensive Metabolizer UM Ultra-rapid Metabolizer DG-I Deviating genotype- Intervention arm DG-C Deviating genotype – Control arm NG-C Normal genotype – Control arm

   

Figure 1. Flowchart for inclusion of patients

180 Chapter 4.3

these drugs, interactions can be expected based on pharmacy interaction monitoring software (G-standaard, Z-index BV, The Hague, The Netherlands), summary of product characteristics, and Flockhart’s interaction table (21).

CYP2D6 genotyping

In the first part of the study, a DBS sample for genotyping will be taken at the time of prescription of nortriptyline or venlafaxine by means of a finger-prick (22). The sample will be sent to the Pharmacogenetic lab of the Wilhelmina Hospital Assen (Assen, the Netherlands), for analysis. Single nuclear polymorphism (SNPs) will be assessed for the *3, *4, *5, *6, *10, *17, *41 alleles, as well as duplications of the CYP2D6 gene. Presence of an allele that completely lacks enzyme activity (*3, *4, *5, *6) paired with another allele completely lacking enzyme activity results in a PM phenotype prediction and in an IM phenotype prediction if paired with an allele with decreased enzyme activity (*10, *17, *41). Presence of only one allele lacking enzyme activity, or two alleles with decreased enzyme activity also results in an IM phenotype prediction. Duplications of the CYP2D6 gene results in an UM phenotype prediction unless it is combined with any allele with reduced or lacking enzyme activity. Duplications combined with any allele with reduced or lacking enzyme activity results in exclusion of the patient from follow-up because no adequate dosing advice can be given for these genotypes. After inclusion, genotype information will be available for randomization within 6 to 9 days.

Baseline

Baseline assessment will include a severity of depression score measured by the Montgomery-Asberg Depression Rating Scale and measurement of symptoms which are followed-up as adverse events during the further trial (23). In addition, comorbidities and drug use other than nortriptyline and venlafaxine will be collected by the Short Form Health and Labor Questionnaire (24).

Randomization and allocation

After the genotype is determined, patients with a PM, IM, or UM genotype are selected and randomly allocated by computer to the ‘deviating genotype-intervention arm’ (DG-I) or ‘deviating genotype-control arm’ (DG-C) for participation in the main study (Figure 1). An additional sample of patients with the EM genotype is allocated to a third ‘normal genotype- control’ arm (NG-C). The PM, IM and UM genotypes for CYP2D6 are expected to be found in approximately 30% of the population (9). Therefore, it is expected that patients with the EM genotype will be found more frequently than patients with a devi-ating genotype. To prevent any potential time-dependent bias, the selection of patients with an EM genotype allocated to the NG-C arm is dependent on the number of patients in the other trial arms.

Study protocol of the CYSCE trial 181

Deviating genotype-intervention arm (Dg-i)

The DG-I arm includes patients with a PM, IM, or UM genotype. The specific genotype accompanied by dosing advice is directly communicated to the treating physician. This should preferably take place 14 days after inclusion. The dosing advice is based on the genotype and drug only, meaning some standard advice depending on these two variables is given by an automatically generated message (see Supplementary file 1).

Deviating genotype-control arm (Dg-C) and normal genotype-control arm (ng-C)

The DG-C arm includes patients with a deviating genotype and the NG-C arm will in-clude patients with a normal genotype. In contrast to patients in the intervention arm, the treating physician will not be informed about the genotype of patients in one of the control arms. Although methodologically there is a different approach for deviating and normal genotype patients, for the physician, patients will appears as one control group. This ensures that patients in the control group can have any genotype as in the normal population.

Outcomes

Primary outcomePrimary outcome will be the time needed to reach adequate blood levels of nortripty-line or venlafaxine, which will be determined by a DBS sample (25,26). The use of DBS for therapeutic drug monitoring is described in more depth elsewhere (27). Blood samples will be sent to the Laboratory for Drug Analysis & Toxicology of the Diaconessen Hospital (Meppel, the Netherlands). Advised therapeutic ranges follow the current Dutch Clinical Pharmacy guidelines (NVZA), indicating for nortriptyline levels between 50 to 150 μg/L and for venlafaxine + O-desmethylvenlafaxine between 100 to 400 μg/L. We defined adequate drug levels as (1) being within the therapeutic range and (2) no dose changes during the previous 3 weeks.

Secondary outcomeSecondary outcomes will be adverse drug reactions measured by a shorter and modified version of the Antidepressant Side-Effect Checklist (ASEC) (28), quality of life measured by the EuroQol 5D (EQ5D) (29), and productivity and cost of health care use measured by the Trimbos/iMTA questionnaire for Cost associated with Psychiatric Illness (TiC-P) (24). Severity of depression will be measured to control for possible differences between groups related to the severity of depression instead of the genotype-based interven-tion, by the Quick Inventory of Depressive Symptomatology Self-Reported Question-naire (QIDS-SR) (30). Secondary outcomes will all be measured by a telephone interview from baseline till endpoint, every 2 weeks, except for the side-effects. Side-effects will

182 Chapter 4.3

be assessed by the treating physician and will additionally be collected before start of the treatment. In this way we will be able to correct for symptoms of depression or other underlying physical diseases, which are often perceived by the patient as side-effects of the antidepressant (28).

sample size calculation

Based on clinical experience in the research setting it is expected that on average it takes 4 weeks for patients with a deviating genotype to reach adequate serum drug levels. It is hypothesized that time to reach adequate levels will be reduced by 50% (2 weeks) if we provide the screening information to the physician within 6 to 9 days. To obtain 90% power to detect a 50% reduction from 4 to 2 weeks (SD 3 weeks) with a type-1 error of 2.5% to account for multiple testing (within trial and externaltrial), a minimum of 48 patients per arm is needed. To account for loss to follow-up we intend to include at least 75 patients per arm.

statistical analysis and report

Reporting and analysis of the data from this trial will be in accordance with the CONSORT 2010 guidelines (18).

BaselineBaseline characteristics will be presented using descriptive statistics including means or medians for continuous variables and percentages for categorical variables. Analytical statistics to estimate difference between the group of patients who were lost to follow-up versus patients who completed the trial as well as between the different study arms will be determined by t-tests or non-parametric alternatives for continuous variables and Chi-square tests for categorical variables.

Primary analysisThe primary analysis will assess the mean time needed to obtain adequate drug levels. The first time drug blood levels are within the therapeutic range and dose is not in-creased or decreased in the following 3 weeks is the moment that is considered as the time needed to obtain adequate drug levels. Mean differences in time to reach adequate drug levels between trial groups will be estimated using the t-test or a non-parametric alternative if not normally distributed. Analysis of variance (ANOVA) or Kruskal-Wallis test will be applied to test for potential differences between the trial groups and the external reference group.

Study protocol of the CYSCE trial 183

Secondary analysesSecondary analysis will focus on statistical differences in median number and severity scores of side-effects of the drug. Since the side-effect profile is different for nortriptyline and venlafaxine, these analyses will be drug-specific.

Cost-effectiveness analysesEffects and costs will be determined from a societal perspective including both indirect and direct costs. Data on health care associated resource use will be collected and ef-fects on productivity and quality of life (EQ5D) will be investigated. To determine the incremental cost-effectiveness ratio for intervention versus control strategy from a societal perspective, guideline unit costs will be linked to these resource use data. Incre-mental cost-effectiveness will be calculated using state-of-the-art methods including uncertainty analysis (bootstrapping, Fieller’s estimates), scenario analysis, (probabilistic) sensitivity analysis and presentation in cost-effectiveness acceptability curves.

4.3.3. DisCussiOn

To the best of our knowledge, this is the first pragmatic randomized controlled trial designed to test whether CYP2D6 polymorphism genotyping among older depressed patients starting with nortriptyline or venlafaxine in clinical practice adds value to regular care. Results of this trial will generate evidence as to whether routinely-based genotype testing can reduce the time needed for dose-finding at the start of treatment and as a result reduce adverse drug reactions.

A possible limitation of this study is that the genotype information is given to the physician after treatment with the antidepressant has already been started for several days. For the purpose of this trial, it would be ideal to give the genotype information before treatment has started. However, in clinical practice due to the urgency of the disease, time between initiating a therapy with nortriptyline or venlafaxine and actual start is limited. Waiting for the genotype information before starting pharmacotherapy could delay pharmacotherapy and is, therefore, unwanted. Nevertheless, an effect of genotype is still expected, because at start of treatment a low dose is generally pre-scribed. When the treating physician has to evaluate therapy and to decide if the dosage should be increased, the genotype information will be available (31).

Another limitation of this study is the study domain represented by a heteroge-neous group of severely depressed patients, since the disease shows much variation in symptomatology, severity and co-existing psychiatric diseases like anxiety disorders and physical diseases, especially in secondary care (32). However, a strict selection of patients who suffer from depression without the presence of certain comorbid physical

184 Chapter 4.3

or psychiatric disorders would exclude a lot of patients from this study and limit ap-plicability of the results to daily practice. Therefore, inclusion is not limited by severity of depression, co-existing psychiatric diseases or other comorbidities. This is a common characteristic of pragmatic trials and gives results which are better generalizable to actual clinical practice (33).

Furthermore, the drug concentration-effect relationship for venlafaxine is still a sub-ject of debate (34). This is illustrated by different advised therapeutic windows. Some refer to a therapeutic window between 250 to 750 μg/L, whereas recent guidelines advie a therapeutic window between 100 to 400 μg/L, which is also incorporated in current Dutch guidelines (11,35). Nevertheless, current guidelines advise a smaller therapeutic window, indicating the need for a more accurate drug dosing. To facilitate this, genotyp-ing may play an important role.

Strengths of this study are the random allocation of subjects to receive genetic in-formation that avoids incomparability of groups, strict protocolled design with regular measurements, and a large number of patients to detect statistically significant differ-ences. We also included an external control arm (NG-C) including patients with the EM genotype. This ensures physicians do not know the genotype of control patients or that it deviates from EM.

The overall impact of CYP2D6 screening may be underestimated in this trial, since the genotype is a lifelong characteristic and is, therefore, affecting the metabolism of many other drugs than the drugs studied in this trial. We only look at a short drug-specific time effect in this study; however, it is not unlikely that patients might benefit from their genotype information during further pharmacotherapy.

trial status

The recruitment of patients started in spring 2013 and is expected to continue until spring 2016.

Acknowledgements

This study was funded by a grant from ‘Netherlands Organization for Health Research and Development (ZonMw, project number: 113102002) as a part of the ‘Priority Medi-cines in Elderly Program’. We acknowledge the work by Inge van Doornik, MSc, in sup-porting the organization of the logistics and implementation of the trial. We thank René Schutte, Ellen Brummel-Mulder, and Jolanda Paardekooper for their help in organizing the laboratory facilities during the trial.

Study protocol of the CYSCE trial 185

4.3.4. REFEREnCEs

1. Brown PJ, Roose SP. Age and anxiety and depressive symptoms: the effect on domains of quality of life. Int J Geriatr Psychiatry. 2011; 26(12): 1260–6.

2. Beekman AT, Copeland JR, Prince MJ. Review of community prevalence of depression in later life. Br J Psychiatry. 1999; 174: 307–11.

3. Luppa M, Sikorski C, Luck T, Ehreke L, Konnopka A, Wiese B, et al. Age- and gender-specific preva-lence of depression in latest-life - systematic review and meta-analysis. J Affect Disord. 2012; 136(3): 212–21.

4. Murray CJ, Vos T, Lozano R, Naghavi M, Flaxman AD, Michaud C, et al. Disability-adjusted life years (DALYs) for 291 diseases and injuries in 21 regions, 1990–2010: a systematic analysis for the global burden of disease study 2010. Lancet. 2012; 380(9859): 2197–223.

5. Mathers CD, Loncar D. Projections of global mortality and burden of disease from 2002 to 2030. PLoS Med. 2006; 3(11): e442.

6. Kok RM, Nolen WA, Heeren TJ. Efficacy of treatment in older depressed patients: a systematic review and meta-analysis of double-blind randomized controlled trials with antidepressants. J Affect Disord. 2012; 141(2–3): 103–15.

7. Reuben D, Herr K, Pacala J, Pollock B, Potter J, Semla T. Depression. Geriatrics at your fingertips. 16th ed. American Geriatrics Society; 2014.

8. Landelijke stuurgroep multidiciplinaire richtlijnontwikkeling in de GGZ. Dutch Guideline Depres-sion (Addendum Elderly) Addendum Ouderen bij MDR derpessie. 2008.

9. Ingelman-Sundberg M. Genetic polymorphisms of cytochrome P450 2D6 (CYP2D6): clinical con-sequences, evolutionary aspects and functional diversity. Pharmacogenomics J. 2005; 5(1): 6–13.

10. Kirchheiner J, Nickchen K, Bauer M, Wong ML, Licinio J, Roots I, et al. Pharmacogenetics of anti-depressants and antipsychotics: the contribution of allelic variations to the phenotype of drug response. Mol Psychiatry. 2004; 9(5): 442–73.

11. Hiemke C, Baumann P, Bergemann N, Conca A, Dietmaier O, Egberts K, et al. AGNP consensus guidelines for therapeutic drug monitoring in psychiatry: update 2011. Pharmacopsychiatry. 2011; 44(6): 195–235.

12. de Leon J, Armstrong SC, Cozza KL. Clinical guidelines for psychiatrists for the use of pharmaco-genetic testing for CYP450 2D6 and CYP450 2C19. Psychosomatics. 2006; 47(1): 75–85.

13. Hicks JK, Swen JJ, Thorn CF, Sangkuhl K, Kharasch ED, Ellingrod VL, et al. Clinical pharmacogenet-ics implementation consortium guideline for CYP2D6 and CYP2C19 genotypes and dosing of tricyclic antidepressants. Clin Pharmacol Ther. 2013.

14. Swen JJ, Nijenhuis M, de Boer A, Grandia L, der Zee AH M-v, Mulder H, et al. Pharmacogenetics: from bench to byte - an update of guidelines. Clin Pharmacol Ther. 2011; 89(5): 662–73.

15. Swen JJ, Wilting I, de Goede AL, Grandia L, Mulder H, Touw DJ, et al. Pharmacogenetics: from bench to byte. Clin Pharmacol Ther. 2008; 83(5): 781–7.

16. Kirchheiner J, Seeringer A, Viviani R. Pharmacogenetics in psychiatry - a useful clinical tool or wishful thinking for the future? Curr Pharm Des. 2010; 16(2): 136–44.

17. Hodgson K, Tansey K, Dernovsek MZ, Hauser J, Henigsberg N, Maier W, et al. Genetic differences in cytochrome P450 enzymes and antidepressant treatment response. J Psychopharmacol. 2014; 28(2): 133–41.

18. Schulz KF, Altman DG, Moher D. CONSORT Group. CONSORT 2010 statement: updated guidelines for reporting parallel group randomised trials. PLoS Med. 2010; 7(3): e1000251.

186 Chapter 4.3

19. Dalen P, Dahl ML, Bernal Ruiz ML, Nordin J, Bertilsson L. 10-Hydroxylation of nortriptyline in white persons with 0, 1, 2, 3, and 13 functional CYP2D6 genes. Clin Pharmacol Ther. 1998; 63(4): 444–52.

20. American Psychiatric Association. Diagnostic and statistical manual of mental disorders (4th ed., text rev.). Washington, DC: Author; 2000.

21. Flockhart DA. Drug interactions: cytochrome P450 drug interaction table. Indiana University School of Medicine (2007). 2013. Available at: http://medicine.iupui.edu/clinpharm/ddis/table.aspx. Accessed 8 August 2014.

22. de Boer T, Wieling J, Meulman E, Reuvers M, Renkema G, den Daas I, et al. Application of dried blood spot sampling combined with LC-MS/MS for genotyping and phenotyping of CYP450 enzymes in healthy volunteers. Biomed Chromatogr. 2011; 25(10): 1112–23.

23. Montgomery SA, Asberg M. A new depression scale designed to be sensitive to change. Br J Psychiatry. 1979; 134: 382–9.

24. Hakkaart-Van Roijen L, van Straten A, Donker M, Tiemens B. Manual Trimbos/iMTA questionnaire for costs asscociated with psychiatric ilness (TiC-P) (in Dutch). 2002.

25. Berm EJ, Brummel-Mulder E, Paardekooper J, Hak E, Wilffert B, Maring JG. Determination of venla-faxine and O-desmethylvenlafaxine in dried blood spots for TDM purposes, using LC-MS/MS. Anal Bioanal Chem. 2014; 406(9–10): 2349–53.

26. Berm EJJ, Paardekooper J, Brummel-Mulder E, Hak E, Wilffert B, Maring JG. A simple dried bloodspot method for therapeutic drug monitoring of the tricyclic antidepressants amitriptyline, nortriptyline, imipramine, clomipramine, and their active metabolites using LC-MS/MS. Talanta Advance online publication. Doi: 10.1016/j.talanta.2014.10.041.

27. Edelbroek PM, van der Heijden J, Stolk LM. Dried blood spot methods in therapeutic drug moni-toring: methods, assays, and pitfalls. Ther Drug Monit. 2009; 31(3): 327–36.

28. Uher R, Farmer A, Henigsberg N, Rietschel M, Mors O, Maier W, et al. Adverse reactions to antide-pressants. Br J Psychiatry. 2009; 195(3): 202–10.

29. EuroQol - a new facility for the measurement of health-related quality of life. The EuroQol Group. Health Policy. 1990; 16(3): 199–8.

30. Rush AJ, Trivedi MH, Ibrahim HM, Carmody TJ, Arnow B, Klein DN, et al. The 16-item quick inven-tory of depressive symptomatology (QIDS), clinician rating (QIDS-C), and self-report (QIDS-SR): a psychometric evaluation in patients with chronic major depression. Biol Psychiatry. 2003; 54(5): 573–83.

31. Jurgens G, Jacobsen CB, Rasmussen HB, Werge T, Nordentoft M, Andersen SE. Utility and adoption of CYP2D6 and CYP2C19 genotyping and its translation into psychiatric clinical practice. Acta Psychiatr Scand. 2012; 125(3): 228–37.

32. Skoog I. Psychiatric disorders in the elderly. Can J Psychiatry. 2011; 56(7): 387–97. 33. March JS, Silva SG, Compton S, Shapiro M, Califf R, Krishnan R. The case for practical clinical trials

in psychiatry. Am J Psychiatry. 2005; 162(5): 836–46. 34. Sangkuhl K, Stingl JC, Turpeinen M, Altman RB, Klein TE. PharmGKB summary: venlafaxine path-

way. Pharmacogenet Genomics. 2014; 24(1): 62–72. 35. Wille SM, Cooreman SG, Neels HM, Lambert WE. Relevant issues in the monitoring and the toxi-

cology of antidepressants. Crit Rev Clin Lab Sci. 2008; 45(1): 25–89.

Study protocol of the CYSCE trial 187

supplementary file 1. Advice based on the Dutch KNMP’s pharmacogenetic guidelines, specified for this specific study purpose with the help of experts in the field. Description: this file contains the advice that was communicated to the treating physicians during the study period.

Genotype Nortriptyline Venlafaxine

Pminterventiongroup

During genotyping a deviating genotype was found. This leads to classification of the patient as a Poor Metabolizer (PM). A PM has a lower capacity of the CYP2D6 enzyme which can cause an increase in nortriptyline plasma concentrations.Advice: give a 50% lower dosage with respect to normal dosing.Additional information:For a PM dosing between 50 and 75 mg /day should give blood levels within the therapeutic window. However, differences between individuals can be significant. Dosing above 75 mg/day gives a substantial chance of blood levels exceeding the therapeutic window. The half-live of the drug is extended. Keep in mind that it can take longer before a steady state period is reached (12-21 days, normal 7-8 days).

During genotyping a deviating genotype was found. This leads to classification of the patient as a Poor Metabolizer (PM). A PM has a lower capacity of the CYP2D6 enzyme which can cause an increase in venlafaxine (V) plasma concentrations and a decrease of its active metabolite o-desmethylvenlafaxine (ODV). There is evidence suggesting a reduced efficacy among depressed patients with this genotype.Advice: In case of response combined with adverse drug reactions: lower dosing to 75% of normal dosing. This should give a sum of V+ODV equal to that of an extensive metabolizer (EM). However, differences between individuals are significant and the ratio V/ODV remains different from an EM. If dosing is further decreased it is unknown to which extend efficacy is maintained.

iminterventiongroup

During genotyping a deviating genotype was found. This leads to classification of the patient as an Intermediate Metabolizer (IM). An IM has a lower capacity of the CYP2D6 enzyme which can cause an increase in nortriptyline plasma concentrations.Advice: give a 75% lower dosage with respect to normal dosing.Additional information:For a IM dosing 75 mg /day should give blood levels within the therapeutic window. However, differences between individuals can be significant. Dosing above 75 mg/day should preferably be guided by TDM. The half-live of the drug is probably extended. Keep in mind that it can take longer before a steady state period is reached (9-21 days, normal 7-8 days).

During genotyping a deviating genotype was found. This leads to classification of the patient as an Intermediate Metabolizer (IM). An IM has a lower capacity of the CYP2D6 enzyme which can cause an increase in venlafaxine (V) plasma concentrations and a decrease of its active metabolite O-desmethylvenlafaxine (ODV). Kinetic profiles show an increase in V+ODV concentration of 1-22%. There is insufficient evidence to give a dose adaptation based on literature. The advice is to control the sum of V + ODV by TDM and adapt dosing if indicated.Although there is no dosing advice available, the patient remains in the trial. This is to collect more data concerning elderly and the IM genotype.

188 Chapter 4.3

supplementary file 1. (continued)

Genotype Nortriptyline Venlafaxine

uminterventiongroup

During genotyping a deviating genotype was found. This leads to classification of the patient as an Ultrarapid Metabolizer (UM). A UM has an increased capacity of the CYP2D6 enzyme which can cause a decreased nortriptyline plasma concentration and an increased plasma concentration of metabolites (E-10-OH-nortriptyline).Advice: Increase dosing to 150%. The metabolite of nortriptyline (OH-nortriptyline) might be cardiotoxic. Be alert to changes in the EKG.Additional information:For an UM dosing of 100 mg can give therapeutic concentrations. To reach adequate therapeutic levels dosing up to 300 mg/day can be necessary. However, differences between individuals are significant. The half-life of the drug is decreased. Keep in mind steady state is reached earlier (4-7 days, normal 7-8 days).

During genotyping a deviating genotype was found. This leads to classification of the patient as an Ultrarapid Metabolizer (UM). A UM has an increased capacity of the CYP2D6 enzyme which can cause a decreased venlafaxine plasma concentrations and an increased plasma concentration of the active metabolite O-desmethylvenlafaxine.Advice: Be alert to low sum plasma concentrations of V and ODV. If indicated by TDM increase dosing up to 150% of normal dosing.

Em/Pm/im/umcontrol group

The patient is not included in the intervention group. Therefore we ask you to continue care as usual. Knowing a patient is in the control group does not give you any information concerning the genotype of the patient, since patients with either a decreased, increased and normal metabolism are included in this group.

5General Discussion

General discussion 193

gEnERAl DisCussiOn

In this thesis knowledge of environmental and genetic drug interactions, advances in bioanalytical technology and economic evaluations were integrated to explore the value of precision drug therapy in depression. These aspects were all combined in the design of a pragmatic randomized controlled trial which is currently being conducted among old age depressed patients.

5.1. EnViROnmEntAl AnD gEnEtiC inFluEnCEs On PsyCHOtROPiC DRug EXPOsuRE

In the first chapter of this thesis we aimed to illustrate environmental and genetic influ-ences on drug exposure. An observation of a different phenotype than predicted by genotype is referred to as phenoconversion (1). In this light Chapter 2.1 can be consid-ered as a typical example of phenoconversion induced by environmental factors like smoking. In the described case report, a patient was treated with clozapine which is metabolized by cytochrome P450 1A2 (CYP1A2). CYP1A2 it is known to be induced by smoking (2). As such, genetically extensive metabolizers who smoke, frequently display phenoconversion towards an ultra-rapid metabolizer phenotype. Not only co-medica-tion, but also changes in behavior like switching to e-cigarettes can have an effect on enzyme induction which can result in a different phenotype which was demonstrated in the case-report. This could lead to substantial changes in drug exposure within the same patient over time.

To further study the phenomenon of phenoconversion, we were interested if aging could be considered as an environmental factor to provoke phenoconversion towards a reduced metabolism of CYP2D6. In Chapter 2.2. an old aged population (>60 years) was studied and surprisingly no support for this hypothesis was found with regard to the use of and exposure to venlafaxine. On the contrary, phenoconversion was directed towards an increased metabolism and it was found that for venlafaxine men had a higher risk on this type of phenoconversion. This could be related to the fact that venlafaxine is partly metabolized by other enzymes than CYP2D6, such as CYP3A4 which could compensate for a loss in CYP2D6 function (3). As for nortriptyline, which is more dependent on CYP2D6 than venlafaxine, we found some phenotypes with less metabolic capacity than would have been expected based on genotype, among intermediate metabolizers. However, due to the post-hoc nature of this study, we cannot rule out that other mutations than the CYP2D6 *3 and *4 alleles, like the *5, *6, *10, *41, etc. which were not addressed in this study could have played a role in these patients. These other, “untested mutations” could also explain the decreased metabolism observed in some of the patients.

194 Chapter 5

The fact that phenoconversion occurred, does not imply genotyping is misleading or of less value, but results should be interpreted by trained professionals. These pro-fessionals should incorporate co-medication and environmental factors like gender and smoking in their dosing recommendations. To achieve precision drug therapy, it is important that young clinicians who are now trained with knowledge of a growing amount of genetic variations, continue to look at the patient and talk about behavioral changes like compliance, smoking, caffeine intake, grapefruit juice, etc (4). These tasks might become complicated and decision support, which could be incorporated in electronic prescription software and updated based on current knowledge, might be used to facilitate precision drug therapy in practice. To monitor and study the influence all mentioned variables, therapeutic drug monitoring (TDM) remains of undoubted importance.

5.2. nEw sAmPling stRAtEgiEs FOR tHERAPEutiC DRug mOnitORing OF tRiCyCliC AntiDEPREssAnts AnD VEnlAFAXinE

To facilitate TDM, a Dried Blood Spot (DBS) method was developed for six common tricy-clic antidepressants, venlafaxine and desmethylvenlafaxine (ODV). The same extraction method was used for all compounds and the methods were validated according to FDA and EMA guidelines (5, 6). As is required by these guidelines, accuracy and precision, selectivity, matrix effects, carry over, recovery, linearity, and stability were assessed and the outcomes are described in Chapter 3.1 and 3.2. In addition, we studied some DBS-specific characteristics like spotting volume, punch position and different hematocrit concentrations (7). For the clinical interpretation of DBS concentrations, a comparison between plasma and DBS concentrations was made. We found all guidelines require-ments were met, except for the selectivity of the method for desmethylclomipramine (DCMP) which was a little too low as well as the precision for clomipramine (CMP). We choose for a pragmatic solution and increased the lower limit of quantification (LLOQ) for both compounds towards 40 instead of 20 µg/L. Since the lower limit of the thera-peutic range for the sum of CMP and DCMP is 200 µg/L we felt a LLOQ of 40 µg/L was still sufficient for TDM purposes. With respect to DBS specific characteristics, like variation in hematocrit, we found a trend towards a negative bias at a lower hematocrit for all compounds as is reported in many studies for various compounds (8-11). Although the results did not indicate major bias at physiological hematocrit concentrations, the bias observed for amitriptyline, desmethylclomipramine, venlafaxine, and desmethylvenla-faxine indicated the need for further examination. Spotting volume and punch position did not introduce any substantial bias. In theory, drugs can be distributed equally over plasma and whole blood and if so, DBS concentrations mirror plasma concentrations.

General discussion 195

However, for most compounds this distribution is not a 1:1 distribution and in line with this we found that DBS concentrations of the antidepressants differed from plasma concentrations. Therefore a clinical validation study in which we compared paired patients samples of DBS and plasma concentrations was performed which is described in Chapter 3.3. Not only the relationship between these two matrixes, but also the influence of different hematocrit concentrations on this relationship was assessed. To measure the hematocrit in patient samples, we used potassium as a derivative. A large set of patient samples was analyzed and we applied a Passing and Bablok regression to test the comparability of the DBS and plasma method (12, 13). We found significant proportional differences between the DBS and plasma concentrations which was in line with our findings in the analytical validation. To correct for the proportional difference, the slope of the regression lines can be used as a correction factor (14, 15). We did not find a quantifiable hematocrit effect, except for ATP in the patient samples. Especially for venlafaxine and desmethylvenlafaxine these findings where unexpected. The most simple explanation for this difference was the preparation procedure of the different hematocrit concentrations in the first validation studies. This procedure was recently found to give lower hematocrit concentrations than anticipated by calculations based on the different fractions of plasma and red blood cells (16). As a result, our first study could have overestimated the influence of hematocrit. Besides this simple explanation, it might be that more complicated unknown influences are at play, which compensate the negative hematocrit bias. Such factors could be related to the recovery, which can be influenced by hematocrit (17). As we did not study these influences, strict coherence towards our sample preparation and analysis procedure is required. For example, the use of a 4 mm punch instead of the validated 6 mm punch could give, hypothetically, a major influence on the assays accuracy, because it might be more sensitive to variations in hematocrit.

The field of DBS analysis has been a field of research for more than a decade, however, more and more issues which could introduce bias are being identified (17-19). As a con-sequence, validation of a DBS method becomes a comprehensive task and it is unlikely all laboratories will be equipped for this task in the future. On the other hand, one could argue that it is not necessary to know all the variations which are ongoing in a certain assay as long as the outcome is of sufficient quality. To facilitate this quality, we studied an additional validation step in Chapter 3.3. For this step, a new set of patient samples was used and we applied the correction factor to calculate plasma concentrations of nortriptyline based on DBS concentration. Biases between the calculated and measured plasma concentrations were calculated for individual data points. For the individual biases we calculated a limit of acceptance based on a Monte Carlo simulation. In this simulation we compared in silico results of 10 000 method comparisons to calculate a 95% predicting interval for acceptance. This principle is illustrated in figure 1. Note that

196 Chapter 5

the regression line of example one is closer to the line of unity than in example 2, but the error of the individual points of example 1 is unacceptable whereas the error of the individual points of example 2 is acceptable. If the error of individual points is taken into account, one can be confident that possible “unknown” bias in the DBS assay will not result in unacceptable bioanalytical performance according to the FDA and EMA requirements.

Current quality of TDM is maintained by external quality control programs in which laboratories participate. Although the Dutch authority (i.e. Dutch Association for Qual-ity Assessment in Therapeutic Drug Monitoring and Clinical Toxicology, KKGT) which is conducting quality control programs for TDM is taking steps towards development of such programs for DBS methods, DBS specific programs are not yet in place (21). Until the field of DBS analysis is also covered by such programs, comprehensive analytical and clinical validation studies of DBS methods as we illustrated in Chapter 3 remain of utmost importance.

Example 1 does not meet additional criterium Example 2 does meet additional criterium

-80-60-40-20

020406080

0 50 100 150

Bias

cal

cula

ted

plas

ma

conc

entra

ion

(%)

ATP plasma concentration (µg/L)

95 % PI for acceptance Bias ATP

0

30

60

90

120

150

0 30 60 90 120 150

ATP

foun

d in

pla

sma

(µg/

L)

ATP found in DBS (µg/L)

-80-60-40-20

020406080

0 50 100 150

Bias

cal

cula

ted

plas

ma

conc

entra

ion

(%)

ATP plasma concentration (µg/L)

95 % PI for acceptance Bias ATP

0

30

60

90

120

150

0 30 60 90 120 150

ATP

foun

d in

pla

sma

(µg/

L)

ATP found in DBS (µg/L)

Figure 1. Demonstration of additional validation criteria.

General discussion 197

5.3. EVAluAting EFFECts AnD COsts AssOCiAtED witH tHE APPliCAtiOn OF gEnEtiC tEsting tO imPROVE PHARmACOtHERAPy

In Chapter 4.1 an update on a systematic review concerning cost-effectiveness of pharmacogenetic and pharmacogenomic screening (PGx) tests was given. In order to enable integration of outcomes with data from previous reviews, the same search strategy was used as in the previous reviews (22, 23). We found that the number of studies was doubled from 2010 till 2014. Furthermore, the quality was improved with regard to more comprehensive sensitivity analyses and reporting of study limitations. This is likely due to the availability and better adherence to guidelines on the conduct of cost-effectiveness analyses such as the CHEERS statement which was published in 2013 (24). When the added papers were studied in more detail, it appeared that most of the reported incremental cost-effectiveness ratios (ICERs) were not giving specific informa-tion about the cost-effectiveness of the PGx test itself, but rather gave an analysis of a comparison of (drug) therapies including genetic tests. As such, the ICER was mainly dependent on the costs of the comparator drug and gives no information about the added value or costs of the PGx tests (i.e. the intrinsic value of the PGx test). To facilitate a clear as possible comparison, a distinction between studies that gave information about the intrinsic value of the PGx tests and studies that did not was made. In this way, more insight was given into the comparability of the studies. Furthermore, general and PGx specific recommendations for improvement of economic evaluations of PGx tests were formulated. As the amount of PGx tests will likely increase in the future, our recommendations together with those of others could be used for the formulation of guidelines for cost-effectiveness analysis of PGx tests (25). Most studies which were included in the review involved genetic testing in the field of oncolytic and antithrom-botic treatment. In the field of psychotropic medicine, two studies were found which evaluated a screening for a serotonin transporter (5-HTTLPR) to select either citalopram or bupropion for treatment of major depressive disorder. In the first study, the ICER for genotype guided selection of antidepressant was found not cost-effective at a $50,000 per quality adjusted life year (QALY) cut-off (26). In the second study, 5-HTTLPR screen-ing was only costs-effective (ICER<3 times the gross domestic product per capita) for high income countries (27). No studies assessing cost-effectiveness for screening for CYP enzymes in combination with psychotropic agents were found. Nevertheless, initiatives are undertaken in the Netherlands to perform genotyping for CYP2D6 and CYP2C19 enzymes, when antidepressants are used (28). To either support or discourage such screening initiatives, a cost-effectiveness analysis was performed in chapter 4.2 to assess cost-effectiveness of a routine screening for CYP2D6. We identified nortriptyline as the best study example, because of the evidence which is available concerning the dose-

198 Chapter 5

effect relationship and evidence for a major influence of genetic variation in CYP2D6 activity on plasma concentrations of nortriptyline (29).

A decision tree model was constructed to assess the cost-effectiveness of genotyping for CYP2D6 at the start of nortriptyline treatment. An old aged (> 60 years) depressed in-patient population was chosen. This patient population was chosen, because of a higher burden of adverse drug reactions and a possible reduction in hospitalization duration which could result in more health gains and lower health care costs (30, 31). We found at current costs (i.e. €190 per test) of genotyping, genotyping was not a cost-effective (i.e. €50 000 per QALY) strategy to follow, unless the price for genotyping was reduced toward €40 per test. On the other hand, if genotyping costs would be less than €35 per test, genotyping was found to be a dominant treatment option (i.e. cost savings and better health outcomes). The analysis contained limitations, which should be consid-ered when interpreting the results. The most important limitation is the assumption that genotyping can prevent sub- or supra-therapeutic dosing. Although many studies have confirmed sub- or supra-therapeutic dosing occurs in deviating genotypes (i.e. poor, intermediate, or ultra-rapid metabolizers) (32-36), it has not been shown, that geno-typing could prevent such dosing. In addition, under naturalistic conditions, in which TDM guided flexible dosing is applied, plasma concentrations might be regulated to such an extent that variations in plasma concentrations between genotype groups are minimalized. This is supported by the findings of Hodgson et al. who found no relation between genotype and adverse drug reactions after eight weeks of nortriptyline phar-macotherapy in patients of different ages (i.e. 19- 72 years) under naturalistic conditions (37). They did find a significant relation between adverse drug reactions and nortripty-line plasma concentrations. Therefore, if there are beneficial effects of genotyping, one might expect that they occur at the very start of treatment when no TDM results are available yet. Moreover, they should be studied in a naturalistic setting in which TDM based dose adaptations are applied. In the model, effects were incorporated in a very short time span, i.e. the first twelve weeks of pharmacotherapy and in addition to TDM to give the most realistic simulation of effects.

To establish effects, which were assumed in the pharmacoeconomic model, of genotyping for CYP2D6 under naturalistic conditions, genotyping should ideally be studied in a randomized controlled trial (RCT). The design of such a trial is described in Chapter 4.3. Two CYP2D6 substrates (i.e. nortriptyline and venlafaxine) were selected for this study. In the Cyp SCreening Elderly trial (CYSCE), genotype information is given to the treating psychiatrist within 10- 15 days of treatment. The patients are followed until six till eight weeks of treatment and various outcomes like adverse drug reactions (including severity), quality of life, and blood plasma concentrations are measured. Plasma concentrations are measured by DBS sampling. Unfortunately, mental care faced extensive reorganization in the period 2012-2014 and as a result inclusion of patients in

General discussion 199

a trial, in which timing is of major importance, was taking more efforts than anticipated. Nevertheless, by the time this thesis is written, one third of the anticipated patients completed the CYSCE trial. Since sufficient power to detect any differences in outcomes is of utmost importance, it was decided to extent the inclusion period by one year. With the CYSCE trial, questions about the capability of genotyping to optimize treatment with the selected antidepressants under naturalistic conditions can be answered by a RCT in the near future.

5.4. FutuRE PERsPECtiVE

DBS sampling for monitoring plasma concentrations of antidepressants is not only a patient friendly alternative for venous sampling, but it could also serve to centralize drug analysis. As a result, logistics of (scientific) studies can be optimized. This was demonstrated by the application of the described DBS methods for TDM of nortriptyline and venlafaxine in the CYSCE trial. In the trial samples from ten different locations are collected by even more health care workers (i.e. psychiatrists, medical doctors, nurses). All samples are analyzed and reported in a uniform way and in this way heterogeneity between centers can be reduced. By the time this thesis is written >175 DBS samples were collected for TDM of nortriptyline or venlafaxine in the CYSCE Trial. Although not all DBS samples which were collected were of perfect quality (i.e. repeatedly only one out of four spots on a card was suitable for analyses), > 98% could be analyzed which can be considered as a successful application of DBS sampling. The timing of DBS sampling was found quite challenging, like for example sampling of nortriptyline which should be performed ~12 hours after nortriptyline intake. This might not be a specific problem for DBS sampling as this is also a requirement for venous sampling. Nevertheless, atten-tion towards this aspect when implementing a DBS method is important. In addition, control measures to assure the right timing of the sample should be in place. When tak-ing these considerations into account, DBS sampling could facilitate patients to obtain a TDM sample by themselves and at home, which would likely reduce TDM costs and patients discomfort. In addition to TDM, genetic information will likely become part of psychotropic pharmacotherapy. In the CYSCE trial most health care professionals were positive about genetic testing and wanted to apply pharmacogenetic information for optimization of pharmacotherapy of their patients. Whether the CYSCE trial will show an added value of CYP2D6 screening is still unknown. However, there is a willingness and interest from the field as well as patients to start implementing genetic knowledge. This is supported by the findings from a Canadian group who found 80% of physicians thought pharmacogenetics will be applied in the future of mental health care (38). Nev-ertheless, communication about when and how genetic information should be given

200 Chapter 5

and in what kind of format might be different between different settings and hospitals. Local (clinical) pharmacies could start to facilitate these aspects, together with mental health care institutions within their regions. In some Dutch regions around Rotterdam and Assen, such initiatives are already undertaken. Note that a (clinical) pharmacist does not have to perform the genetic testing him-/herselve, however (s)he could be the responsible party to give sufficient education, logistic support and interpretation of pharmacogenetic information based on the national pharmacists guidelines (39, 40). Such efforts, can be the key towards cost-effectiveness of precision drug therapy with regard to psychotropic treatment. As a result, in the hopefully near future, precision drug therapy with psychotropic agents will become care as usual.

General discussion 201

5.5. REFEREnCEs

1. Preskorn SH, Kane CP, Lobello K, Nichols AI, Fayyad R, Buckley G, et al. Cytochrome P450 2D6 phenoconversion is common in patients being treated for depression: implications for personal-ized medicine. J Clin Psychiatry. 2013 Mar 13; 74(6): 614-21.

2. Zevin S, Benowitz NL. Drug interactions with tobacco smoking. An update. Clin Pharmacokinet. 1999 Jun; 36(6): 425-38.

3. Mannheimer B, Haslemo T, Lindh JD, Eliasson E, Molden E. Risperidone and venlafaxine metabolic ratios strongly predict a CYP2D6 poor metabolizing genotype. Ther Drug Monit. 2015 Sep 26.

4. Koren G. Personalized Medicine-Disregarding the Obvious Analysis of trends among papers published on “personalized medicine”. Ther Drug Monit. 2015 Mar 9.

5. Guidance for Industry, Bioanalytical Method Validation [Internet].: US Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research (CDER), Center for Veterinary Medicine (CVM), FDA; 2001 [updated 2001, May; cited 17-9-2015]. Available from: http://www.fda.gov/downloads/Drugs/Guidances/ucm070107.pdf.

6. Guideline on bioanalytical method validation. EMEA/CHMP/EWP/192217/2009rev1.corr2 [Inter-net].: Committee for Medicinal Products for Human Use (CHMP); 2012 [cited 17-9-2015]. Available from: http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2011/08/WC500109686.pdf.

7. Edelbroek PM, van der Heijden J, Stolk LM. Dried blood spot methods in therapeutic drug moni-toring: methods, assays, and pitfalls. Ther Drug Monit. 2009 Jun; 31(3): 327-36.

8. Vu DH, Koster RA, Alffenaar JW, Brouwers JR, Uges DR. Determination of moxifloxacin in dried blood spots using LC-MS/MS and the impact of the hematocrit and blood volume. J Chromatogr B Analyt Technol Biomed Life Sci. 2011 May 1; 879(15-16): 1063-70.

9. Vu DH, Koster RA, Bolhuis MS, Greijdanus B, Altena RV, Nguyen DH, et al. Simultaneous determi-nation of rifampicin, clarithromycin and their metabolites in dried blood spots using LC-MS/MS. Talanta. 2014 Apr; 121: 9-17.

10. Rao RN, Raju SS, Vali RM, Sankar GG. Liquid chromatography-mass spectrometric determination of losartan and its active metabolite on dried blood spots. J Chromatogr B Analyt Technol Biomed Life Sci. 2012 Aug 1; 902: 47-54.

11. Wilhelm AJ, den Burger JC, Vos RM, Chahbouni A, Sinjewel A. Analysis of cyclosporin A in dried blood spots using liquid chromatography tandem mass spectrometry. J Chromatogr B Analyt Technol Biomed Life Sci. 2009 May 15; 877(14-15): 1595-8.

12. Passing H, Bablok W. Comparison of several regression procedures for method comparison stud-ies and determination of sample sizes. Application of linear regression procedures for method comparison studies in Clinical Chemistry, Part II. J Clin Chem Clin Biochem. 1984 Jun; 22(6): 431-45.

13. Bablok W, Passing H. Application of statistical procedures in analytical instrument testing. J Automat Chem. 1985; 7(2): 74-9.

14. Evans C, Arnold M, Bryan P, Duggan J, James CA, Li W, et al. Implementing dried blood spot sampling for clinical pharmacokinetic determinations: considerations from the IQ Consortium Microsampling Working Group. AAPS J. 2015 Mar; 17(2): 292-300.

15. de Wit D, den Hartigh J, Gelderblom H, Qian Y, den Hollander M, Verheul H, et al. Dried blood spot analysis for therapeutic drug monitoring of pazopanib. J Clin Pharmacol. 2015 Jun 1.

16. Koster RA, Alffenaar JW, Botma R, Greijdanus B, Touw DJ, Uges DR, et al. What is the right blood hematocrit preparation procedure for standards and quality control samples for dried blood spot analysis? Bioanalysis. 2015 Feb; 7(3): 345-51.

202 Chapter 5

17. Abu-Rabie P, Denniff P, Spooner N, Chowdhry BZ, Pullen FS. Investigation of different approaches to incorporating internal standard in DBS quantitative bioanalytical workflows and their effect on nullifying hematocrit-based assay bias. Anal Chem. 2015 May 5; 87(9): 4996-5003.

18. Koster RA, Botma R, Greijdanus B, Uges DR, Kosterink JG, Touw DJ, et al. The performance of five different dried blood spot cards for the analysis of six immunosuppressants. Bioanalysis. 2015; 7(10): 1225-35.

19. Koster RA, Alffenaar JW, Botma R, Greijdanus B, Uges DR, Kosterink JG, et al. The relation of the number of hydrogen-bond acceptors with recoveries of immunosuppressants in DBS analysis. Bioanalysis. 2015 Aug; 7(14): 1717-22.

20. Patteet L, Maudens KE, Stove CP, Lambert WE, Morrens M, Sabbe B, et al. Are capillary DBS applicable for therapeutic drug monitoring of common antipsychotics? A proof of concept. Bioanalysis. 2015; 7(16): 2119-30.

21. Robijns K, Koster RA, Touw DJ. Therapeutic drug monitoring by dried blood spot: progress to date and future directions. Clin Pharmacokinet. 2014 Nov; 53(11): 1053,014-0197-3.

22. Vegter S, Jansen E, Postma MJ, Boersma C. Economic evaluations of pharmacogenetic and genomic screening programs: update of the literature. Drug Development Research. 2010; 71: 492-501.

23. Vegter S, Boersma C, Rozenbaum M, Wilffert B, Navis G, Postma MJ. Pharmacoeconomic evalua-tions of pharmacogenetic and genomic screening programmes: a systematic review on content and adherence to guidelines. Pharmacoeconomics. 2008; 26(7): 569-87.

24. Husereau D, Drummond M, Petrou S, Carswell C, Moher D, Greenberg D, et al. Consolidated Health Economic Evaluation Reporting Standards (CHEERS) statement. Value Health. 2013 Mar-Apr; 16(2): e1-5.

25. Husereau D, Marshall DA, Levy AR, Peacock S, Hoch JS. Health technology assessment and per-sonalized medicine: are economic evaluation guidelines sufficient to support decision making? Int J Technol Assess Health Care. 2014 Apr; 30(2): 179-87.

26. Serretti A, Olgiati P, Bajo E, Bigelli M, De Ronchi D. A model to incorporate genetic testing (5-HT-TLPR) in pharmacological treatment of major depressive disorders. World J Biol Psychiatry. 2011 Oct; 12(7): 501-15.

27. Olgiati P, Bajo E, Bigelli M, De Ronchi D, Serretti A. Should pharmacogenetics be incorporated in major depression treatment? Economic evaluation in high- and middle-income European countries. Prog Neuropsychopharmacol Biol Psychiatry. 2012 Jan 10; 36(1): 147-54.

28. Kootstra-Ros JE, Van Weelden MJ, Hinrichs JW, De Smet PA, van der Weide J. Therapeutic drug monitoring of antidepressants and cytochrome p450 genotyping in general practice. J Clin Pharmacol. 2006 Nov; 46(11): 1320-7.

29. Dalen P, Dahl ML, Bernal Ruiz ML, Nordin J, Bertilsson L. 10-Hydroxylation of nortriptyline in white persons with 0, 1, 2, 3, and 13 functional CYP2D6 genes. Clin Pharmacol Ther. 1998 Apr; 63(4): 444-52.

30. Ruano G, Szarek BL, Villagra D, Gorowski K, Kocherla M, Seip RL, et al. Length of psychiatric hospitalization is correlated with CYP2D6 functional status in inpatients with major depressive disorder. Biomark Med. 2013 Jun; 7(3): 429-39.

31. Davies EA, O’Mahony MS. Adverse drug reactions in special populations - the elderly. Br J Clin Pharmacol. 2015 Jan 24.

32. Laika B, Leucht S, Heres S, Steimer W. Intermediate metabolizer: increased side effects in psychoactive drug therapy. The key to cost-effectiveness of pretreatment CYP2D6 screening? Pharmacogenomics J. 2009 Dec; 9(6): 395-403.

General discussion 203

33. Chou WH, Yan FX, de Leon J, Barnhill J, Rogers T, Cronin M, et al. Extension of a pilot study: impact from the cytochrome P450 2D6 polymorphism on outcome and costs associated with severe mental illness. J Clin Psychopharmacol. 2000 Apr; 20(2): 246-51.

34. Chen S, Chou WH, Blouin RA, Mao Z, Humphries LL, Meek QC, et al. The cytochrome P450 2D6 (CYP2D6) enzyme polymorphism: screening costs and influence on clinical outcomes in psychia-try. Clin Pharmacol Ther. 1996 Nov; 60(5): 522-34.

35. Hicks JK, Swen JJ, Thorn CF, Sangkuhl K, Kharasch ED, Ellingrod VL, et al. Clinical Pharmacogenet-ics Implementation Consortium Guideline for CYP2D6 and CYP2C19 Genotypes and Dosing of Tricyclic Antidepressants. Clin Pharmacol Ther. 2013 Jan 16; 93(5): 402-8.

36. Kraus RP, Diaz P, McEachran A. Managing rapid metabolizers of antidepressants. Depress Anxiety. 1996-1997; 4(6): 320-7.

37. Hodgson K, Tansey KE, Uher R, Dernovsek MZ, Mors O, Hauser J, et al. Exploring the role of drug-metabolising enzymes in antidepressant side effects. Psychopharmacology (Berl). 2015 Mar 12.

38. Walden LM, Brandl EJ, Changasi A, Sturgess JE, Soibel A, Notario JF, et al. Physicians’ opinions following pharmacogenetic testing for psychotropic medication. Psychiatry Res. 2015 Oct 30; 229(3): 913-8.

39. Swen JJ, Nijenhuis M, de Boer A, Grandia L, Maitland-van der Zee AH, Mulder H, et al. Pharma-cogenetics: from bench to byte--an update of guidelines. Clin Pharmacol Ther. 2011 May; 89(5): 662-73.

40. Swen JJ, Wilting I, de Goede AL, Grandia L, Mulder H, Touw DJ, et al. Pharmacogenetics: from bench to byte. Clin Pharmacol Ther. 2008 May; 83(5): 781-7.

Summary 205

summARy

For the application of precision drug therapy, physicians have an increasing set of tools which can be used to individualize drug therapy. Pharmacogenetics information is one of these tools. One of the applications of pharmacogenetic information is prediction of drug metabolizing capacity. Nevertheless, for precision drug therapy both internal factors, like pharmacogenetics, and external variations need to be considered.

As was shown in chapter 2, precision drug therapy is important, because several en-vironmental and genetic variations can influence drug exposure. That these differences in exposure are substantial is illustrated in Chapter 2.1. in which an unexpected rise in plasma concentration of clozapine was found when the dose of clozapine was actually decreased. The observed discrepancy was likely due to a change in smoking behavior of the patient. In Chapter 2.2. the relation between genetic differences in the drug me-tabolizing enzyme cytochrome P450 2D6 (CYP2D6) and observed CYP2D6 activity was studied. Age (60-75 years vs.>75 years) or gender were not identified as risks factor for phenoconversion towards a lower metabolic activity. However, male subjects treated with venlafaxine had more change on phenoconversion towards a higher metabolic activity. As a result, most genetically IMs who used venlafaxine, had an EM phenotype. Based on these findings, it can be concluded, that old age does not have to be considered as an additional reason for dose adaptations among IMs treated with venlafaxine. On the other hand, PM phenotypes were found to be reliably predicted based on genotype. In addition, a lower response to the antidepressants was found on the Hamilton Rating Scale for Depression for PMs, although this did not reach significances when measured on the Montgomery Åsberg Depression Rating Scale. It was concluded that genotyping for CYP2D6 can be used as a reliable tool in old age psychiatry to predict PM phenotypes which could be related to a higher risk for non-response. A lower dose should be given to these patients and their plasma concentrations should be carefully monitored to ensure that they are within the therapeutic range.

Plasma concentrations of nortriptyline and venlafaxine are usually monitored in blood collected by a venipuncture. To individualize TDM, a more patient friendly alternative is described in this thesis. This alternative included dried blood spot (DBS) sampling in which droplets of blood are obtained by a fingerprick and spotted onto paper. A small piece of this paper is analyzed to assess the concentrations of the anti-depressants. Besides patients comfort, DBS have more advantages. The stability of the analyte is usually high and samples can be sent with normal post services due to the low biohazard risks of dried blood. Consequently, patients do not have to travel to a laboratory, because samples can be collected at the psychiatrist office or at home. In Chapter 3.1. and 3.2. results of analytical validation are described for DBS analysis with

206 Chapter 5

the use of liquid chromatography-tandem mass spectrometry (LC-MS/MS). Methods for six tricyclic antidepressants (TCAs), venlafaxine (VFX) and O-desmethylvenlafaxine (ODV), were validated. Validation was performed according to requirements of guidelines and except for clomipramine (CMP) and desmethylclomipramine (DCMP) criteria were met. For CMP and DCMP the lower limit of quantification was increased from 20 to 40 µg/L to comply with the guidelines criteria. In chapter 3.3. results of a clinical validation of DBS analyses is described. Paired plasma and DBS samples were collected from 162 patients. Evidence for proportional differences between plasma and DBS concentrations, which were identified in chapter 3.1. and 3.2., were confirmed in chapter 3.3. To be able to link DBS to plasma concentrations a translation factor was calculated based on Passing and Bablok regression analysis. DBS concentrations were found related to plasma concentrations by a factor of 1.22, 0.8, 0.65, 0.84, and 0.78 for amitriptyline (ATP), nortriptyline (NTP), DCMP, VFX, and ODV, respectively. For CMP, no reliable factor could be identified, due to substantial bias in the analysis. Evidence was found that the factor of ATP was sensitive to variations of the hematocrit of the DBS which was analyzed by potassium based hematocrit analysis. Further research is needed for quantification of the translation factors of ATP and CMP. We applied the translation factors on the datasets and we compared the clinical interpretation of the plasma and DBS derived plasma concentrations. We found interpretation was in agreement for 96%, 97%, 75% and 100% of the observations for the sum of ATP & NTP, NTP, the sum of CMP & DCMP and the sum of VEN & ODV, respectively. Again, the results for the sum of CMP & DCMP were inadequate compared to the other compounds. To give a clear cut-off for differences between plasma and DBS derived plasma concentrations, criteria were cal-culated based on guidelines limits for accuracy and precision of bioanalytical methods. A Monte Carlo simulation was performed to simulate 10,000 method comparison stud-ies on the borderline of acceptance criteria. We found that based on the bias between plasma and DBS derived plasma concentrations of individual data points no more than 5% of the samples should have a bias >36% at concentrations above the quality control low (QCL) level. For data points with concentrations between the LLOQ and QCL, no more than 5% of the samples should have a bias > 48%. These criteria were applied on additional samples of nortriptyline and criteria were met. As such, TDM of NTP by DBS was validated according to these new criteria. Criteria for acceptance of clinical method comparison studies should be issued by bioanalytical guidelines, these criteria should include clinical comparability and biases of individual data points. For the later, limits presented in this thesis can be used.

In chapter 4 of this thesis, economic evaluations of pharmacogenetic testing were studied. In chapter 4.1. a systematic review is described to study outcomes, qual-ity and shortcomings of economic evaluations of pharmacogenetics testing. The last decade an increase in the number of studies and in the reporting of quality associated

Summary 207

characteristics was observed. To improve future evaluations, scenario analysis includ-ing a broad range of PGx tests costs and equal costs of comparator drugs to assess the intrinsic value of the PGx tests, were recommended. In addition, the lack of robust clini-cal evidence regarding PGx tests’ efficacy was identified as an important limitation of economic evaluations.

Cost-effectiveness analyses of genotyping for CYP enzymes in combination with antidepressants were not found in the review. Nevertheless, the relation between exposure of nortriptyline and differences in CYP2D6 genotypes are well-established. Guidelines for dose recommendations are issued by international and national institutes and should be considered when pre-emptive genotyping is performed. However no guidance is available whether such an approach would be cost-effective. Therefore, a cost-effectiveness analysis was performed to assess if routine genotyping, in addition to TDM, for all patients who start nortriptyline pharmacotherapy in a clinical setting would be cost-effective. Genotyping costs were determined based on screening for five muta-tions (~€200 per test) of which clinical effects were reported based on a pharmacokinetic modelling study. It was found that routine genotyping is not a costs-effective strategy (€50 000 per quality adjusted life year gained), unless genotyping costs are reduced towards €40 per test. Interestingly, it was found that including dose adaptations for IMs, would decrease the cost-effectiveness of genotyping due to the relatively high risk that patients with an IM genotype express an EM phenotype. As such, a dose which is too low would be given to these patients which was assumed to result in a delayed response to the antidepressant and therefore a lower quality of life. This lower quality of life resulted in a lower cost-effectiveness. Our estimates contain substantial uncertainty, mainly due to the lack of hard clinical evidence that genotyping can indeed prevent sub- or supratherapeutic plasma concentrations of nortriptyline. It is important that effects of genotyping are studied in a study design with a small change on biased outcomes, for which a randomized controlled trial (RCT) is currently still considered the best study design. Therefore a RCT was designed and is conducted. The design of the Cyp Screen-ing Elderly (CYSCE) trial is described in chapter 4.3. Inclusion of patients started in 2013. While this thesis is written, inclusion of patients in the trial is still ongoing. Results of this trial will help to inform clinicians and decision makers about the added value of genotyping for CYP2D6 among elderly patients who start nortriptyline and venlafaxine pharmacotherapy.

In conclusion, different efforts to optimize pharmacotherapy with psychotropic drugs were described in this thesis. DBS analysis was implemented for TDM of nortrip-tyline and venlafaxine in the CYSCE trial and as such centralized TDM of a national multicentre trial was realized. Based on a pharmacoeconomic model it was concluded that genotyping of CYP2D6 in addition to TDM was not cost-effective in nortriptyline pharmacotherapy at current genotyping costs. It was cost-effective at test costs of

208 Chapter 5

€40 per test. However, these estimates contain substantial uncertainty. Results of the CYSCE trial, which will be available in the near future will help to reduce this uncertainty. With ongoing developments like lower genotyping costs due to improved technology, better knowledge of the relation between genotype and phenotypes and multi-gene-approaches, precision drug therapy of psychotropic medicine facilitated by genetic information is within reach. Although, future research is needed to study bottlenecks which hamper implementation. Moreover, harmonization is needed on a regional and national level about which patients should be genotyped, for what kind of variations, when genotyping should be performed and how to document the results.

Samenvatting 209

sAmEnVAtting

Voor het toepassen van individuele farmacotherapie, hebben medici een toenemend aantal gegevens tot hun beschikking waaronder farmacogenetica. Met behulp van farmacogenetica kan onder andere een voorspelling worden gedaan over de enzymca-paciteit van patiënten voor het omzetten van geneesmiddelen in metabolieten. Echter, voor individuele farmacotherapie zijn zowel externe als interne factoren (waaronder farmacogenetica) van belang. Beide dienen door de arts in overweging genomen te worden. In hoofdstuk 2.1. wordt het belang van individuele farmacotherapie en waak-zaamheid voor verandering in externe factoren geïllustreerd aan de hand van een casus. In de casus wordt een onverwachte toename in de plasmaconcentratie van het antipsychoticum clozapine geobserveerd aan de hand van spiegelcontroles. Terwijl de dosering van het geneesmiddel verlaagd wordt, neemt de geneesmiddelspiegel toe. Het blijkt dat deze discrepantie waarschijnlijk veroorzaakt wordt door veranderingen in het rookgedrag van de patiënt. Vervolgens wordt in hoofdstuk 2.2. dieper ingegaan op de relatie tussen genetische verschillen in het cytochroom P450 2D6 enzym (CYP2D6) en geobserveerde activiteit van dit enzym. Wanneer deze twee niet met elkaar in overeenstemming zijn, wordt wel gesproken van fenoconversie. We vonden dat leeftijd (60-75 jaar vs. >75 jaar) en geslacht geen verhoogt risico gaven op fenoconversie in de zin van minder geobserveerde enzym activiteit dan voorspeld. Echter, mannen die behandeld werden met venlafaxine, hadden meer kans op een hogere geobserveerde enzym activiteit dan op basis van genetische eigenschappen zou worden verwacht. Dit betroffen vooral observaties waarbij de genetische eigenschappen een ‘intermediate metabolizer’ (IM) fenotype voorspelden. Hiervan uitgaande, zouden bij patiënten met het IM genotype, doseringsaanpassingen voor venlafaxine op basis van een hogere leeftijd niet geïndiceerd zijn. Anderzijds konden ‘poor metabolizer’ (PM) fenotypes be-trouwbaar voorspeld worden op basis van het genotype. Daarbij werd er ook een lagere response op het antidepressivum gevonden onder zowel PM genotypes als fenotypes. Dit verschil was significant op basis van de ‘Hamilton Rating Scale for Depression’, maar niet wanneer dit op de ‘Montgomery Åsberg Depression Rating Scale’ gemeten werd. Concluderend kan gesteld worden dat genotyperen voor CYP2D6 onder ouderen betrouwbare informatie geeft over het PM fenotype, maar niet over het IM fenotype. PMs lopen mogelijk een risico op een lagere response. Het is daarom belangrijk dat geneesmiddelspiegels van PMs nauwkeuring worden gemonitord, zodat deze binnen het therapeutische venster blijven.

Huidige monitoring van geneesmiddelspiegels vindt plaats door middel van bloed wat via een venapunctie verkregen wordt. Voor het individualiseren van deze monitoring wordt in dit proefschrift een patiëntvriendelijker alternatief beschreven.

210 Chapter 5

Dit alternatief betreft bloedafname door middel van een vingerprik waarbij zogeheten ‘dried blood spots’ (DBS) worden verzameld op een stukje papier. Een klein stukje van dit papier met daarop de gedroogde druppels bloed kan worden geanalyseerd om de concentratie van verschillende antidepressiva te bepalen. Naast dat DBS meer patiënten comfort bieden, zijn er ook andere voordelen aan verbonden. Zo is de stabiliteit van het geneesmiddel vaak hoog, waardoor er een langere tijd tussen bloedafname en analyse beschikbaar is. Daarbij brengt gedroogd bloed geen risico op besmettingen met zich mee en daarom kan het via de reguliere post verstuurd worden. Dit maakt dat patiënten niet meer naar een prikpunt hoeven toe te gaan voor bloedafname, maar dat dit bij de psychiater of bij de patiënt thuis kan plaatsvinden. In hoofdstuk 3.1 en 3.2. worden de resultaten van de analytische validatie van een DBS methode voor de analyse van tricyclische antidepressiva evenals venlafaxine (VFX) en O-desmethylvenlafaxine (ODV) beschreven. De methode maakt gebruik van vloeistofchromatografie gecombineerd met massaspectrometrie. De validatie werd uitgevoerd ten opzichte van internationale richtlijnen. Voor de stoffen amitriptyline, nortriptyline, imipramine, desipramine, VFX en ODV, werd er aan de in richtlijnen gestelde criteria voldaan, maar voor clomipra-mine (CMP) en desmethylclomipramine (DCMP) moest de kwantificatielimiet omhoog worden bijgesteld van 20 naar 40 µg/L. In hoofdstuk 3.3. worden de resultaten van de klinische validatie van de DBS methoden beschreven. Hiervoor werden 162 gepaarde plasma en DBS monsters van patiënten verzameld en geanalyseerd. De in hoofdstuk 3.1. en 3.2. gevonden aanwijzingen op een proportioneel verschil tussen plasma en DBS concentraties werd met deze analyse bevestigd. We vonden dat voor het uitdrukken van DBS als plasma concentratie, een correctiefactor nodig was van respectievelijk: 1.22, 0.80, 0.65, 0.84 en 0.78 voor amitriptyline (ATP), nortriptyline (NTP), DCMP, VFX en ODV. Deze factoren werden berekend op basis van ‘Passing and Bablok’ regressie analyse. Voor CMP kon geen betrouwbare factor worden berekend door de hogere mate van spreiding in de uitkomsten. Verder onderzoek moet uitwijzen wat de oorzaak is van deze spreiding. Op basis van gemeten kaliumconcentraties van de DBS, werden aanwijzingen gevonden dat de gevonden factor voor ATP mogelijk beïnvloed wordt door variatie van het hematocriet in het bloed. Het wordt daarom aanbevolen verder onderzoek van deze factor uit te voeren in een grotere patiëntengroep. De gevonden correctiefactoren wer-den vervolgens toegepast op de gegevens waarna de klinische interpretatie van analyse gebaseerd op plasma en DBS vergeleken werden. Voor de somspiegel van ATP & NTP, NTP, en de somspiegel van VEN & ODV kwamen de adviezen overeen in >96% van de gevallen. Wederom bleven de resultaten voor de somspiegel van CMP & DCMP, met een overeenkomst van 75% achter ten opzichte van de overige geneesmiddelen. Om een duidelijke grens te stellen voor wat een acceptabel verschil tussen plasma en op DBS gebaseerde plasmaconcentraties zou zijn, werden er met behulp van een ‘Monte Carlo’ simulatie grenzen berekend op basis van de grenzen voor juistheid en precisie, welke

Samenvatting 211

vermeld staan in internationale richtlijnen. In de simulatie werden 10.000 methoden vergelijkingen gesimuleerd om tot een uitkomst te komen. Een verschil van >36% voor individuele verschillen tussen plasma en op DBS gebaseerde plasmaconcentraties werd gevonden in niet meer dan 5% van de monsters. Dit betrof de grens voor concentraties welke boven het concentratieniveau van de kwaliteitscontrole op lage concentraties liggen. Voor concentraties tussen de kwantificatielimiet en het lage concentratieniveau, werd een verschil van >48% voor individuele verschillen tussen plasma en op DBS gebaseerde plasmaconcentraties in niet meer dan 5% van de monster gevonden. Deze grenzen werden toegepast op additionele monsters van NTP en de uitkomsten van de verschillen tussen plasma en op DBS gebaseerde plasmaconcentraties vielen binnen de grenzen. Het wordt aanbevolen dat criteria voor acceptatie van methoden vergelijkingen die gebruikt worden voor ‘therapeutic drug moniotoring (TDM)’ worden opgenomen in richtlijnen. Deze criteria zouden duidelijkheid moeten geven over de mate van klinische vergelijkbaarheid en de verschillen tussen gemeten concentraties. Voor dit laatste kun-nen de criteria die in dit proefschrift berekend werden, worden gebruikt.

In hoofdstuk 4 van dit proefschrift worden economische evaluaties van farmacogene-tische testen bestudeerd. Hoofdstuk 4.1. van dit proefschrift beschrijft een systematische review met als doel uitkomsten, kwaliteit en tekortkomingen in kaart te brengen van reeds gerapporteerde economische evaluaties. Het bleek dat in het afgelopen decen-nium het aantal studies en de rapportage van kwaliteitsindicatoren was toegenomen. Om toekomstige studie te verbeteren werd aanbeveling gedaan tot het opnemen van een brede range aan kosten voor de farmacogenetische test in de analyse. Daarbij werd aanbeveling gedaan om gelijke kosten op te nemen voor de geneesmiddelen welke vergeleken worden in een gevoeligheidsanalyse, zodat de intrinsieke waarde van de farmacogenetische test vastgesteld kan worden. Het gebrek aan robuust klinisch bewijs van de effecten van farmacogenetische testen werd geïdentificeerd als een belangrijke tekortkoming van de meeste economische evaluaties. Kosteneffectiviteitsanalyses naar CYP enzymen in combinatie met psychotrope middelen werden niet gevonden. Niet-temin is de relatie tussen een hoge exposure aan nortriptyline en genetische variaties in het CYP2D6 enzym een goed gedocumenteerde relatie. Er bestaan nationale en internationale richtlijnen voor het aanpassen van de dosering op basis van genetische variaties, maar niet voor wanneer er getest moet worden om deze informatie te verkrij-gen. Om te bestuderen of routinematig genotyperen voorafgaand aan de behandeling met nortriptyline kosteneffectief is, werd er een kosteneffectiviteitsmodel gemaakt. Dit model wordt beschreven in hoofdstuk 4.2. van dit proefschrift. In het model werd een screening voor CYP2D6, als aanvulling op TDM, in een patiëntenpopulatie van ouderen die gehospitaliseerd was vanwege depressie, gesimuleerd. De kosten voor het genotyperen werden gebaseerd op het testen voor vijf varianten van het CYP2D6 gen (~€200 per test). Van deze varianten waren klinische effecten bestudeerd in een

212 Chapter 5

farmacokinetisch model. Op basis van het economische model kon geconcludeerd worden dat genotyperen niet kosteneffectief is (€50 000 per gewonnen levensjaar in goede gezondheid), behalve wanneer de kosten voor het genotyperen verlaagd wer-den naar €40 per test. Het was opmerkelijk dat wanneer doseringsverlagingen voor IMs werden opgenomen in het model, de kosteneffectiviteit van genotyperen afnam. Dit werd veroorzaakt door het relatief hoge risico op het tot expressie brengen van een EM fenotype in de IM genotypegroep, waardoor een te lage dosering aan deze patiënten werd gegeven. Het werd aangenomen dat deze lage dosering ook tot een lagere effectiviteit van het antidepressivum zou leiden, waardoor deze patiënten een lagere kwaliteit van leven hadden, wat het verschil in kosteneffectiviteit veroorzaakte. Onze uitkomsten bevatten onzekerheden wat voornamelijk veroorzaakt wordt door het gebrek aan robuust klinisch bewijs wat aantoont dat genotyperen daadwerkelijk voor een betere dosering van nortriptyline kan zorgen. Het is daarom belangrijk dat de effecten van genotyperen worden onderzocht waarbij er weinig kans op vertekening van de resultaten is. Voor dergelijk onderzoek wordt gerandomiseerd gecontroleerd onderzoek nog steeds beschouwd als de gouden standaard en een dergelijk onderzoek werd ontworpen en wordt uitgevoerd. Het ontwerp van de ‘CYp Screening Elderly’ stu-die wordt beschreven in hoofstuk 4.3. De inclusie van patiënten ging van start in 2013. Op het moment dat dit proefschrift geschreven wordt, gaat de inclusie van patiënten in dit onderzoek nog steeds door. De resultaten van dit onderzoek zullen besluitvormers en clinici informeren over de toegevoegde waarde van genotyperen voor CYP2D6 onder ouderen welke starten met venlafaxine of nortriptyline.

Concluderend werden in dit proefschrift diverse strategieën voor het optimaliseren van farmacotherapie met psychotrope middelen beschreven. DBS analyse werd gebruikt voor TDM van nortriptyline en venlafaxine in de CYSCE studie en door deze DBS analyse was het mogelijk TDM van een nationale trial met meerdere centra in één laboratorium uit te voeren. Gebaseerd op een farmacoeconomisch model werd geconcludeerd dat genotyperen als aanvulling op TDM niet kosteneffectief (€50 000 per gewonnen levens-jaar in goede gezondheid) was voor de huidige kosten, maar wel wanneer de kosten verlaagd werden naar €40 per test. De uitkomsten bevatten echter substantiële onze-kerheid. De verwachte uitkomsten van de CYSCE trial, zullen in de nabije toekomst deze onzekerheid aanzienlijk kunnen verminderen. Met het oog op de toekomst zal door de ontwikkelingen in de technologie die beschikbaar is voor genetische analyses, de toenemende kennis van de relatie tussen genotype en fenotype en het betrekken van meerdere genen in het voorspellen van fenotypes, precisiegeneeskunde op basis van genetisch eigenschappen steeds dichter bij komen. Er is echter verder onderzoek nodig om de (logistieke) knelpunten die implementatie van farmacogenetica in de weg staan te identificeren en op te lossen. Daarbij is harmonisatie nodig omtrent indicatiestelling

Samenvatting 213

voor het uitvoeren van farmacogenetische testen, het aantal te bepalen genetische variaties en de manier van rapporteren en vastleggen van de genetische informatie.

CV 215

CV

Elizabeth J.J. (Lisette) Berm (1986) was born in Nieuwegein, the Netherlands. She studied nursing for two years in Utrecht after which she worked one year, mainly in psychoge-riatric elderly care and then came to Groningen and started to study Pharmacy at the University of Groningen. She finished her master’s degree in 2012. During her master study she did her master research project at Lareb (The Netherlands Pharmacovigilance Centre) about patients motivation for participating in Lareb Intensive Monitoring. After graduating she started as a PhD candidate in September 2012. The PhD position was funded by a ZonMw grant, regarding Priority Medicines Elderly. The project aims to determine the added value of pharmacogenetic screening at the start of treatment with nortriptyline or venlafaxine in elderly. For this purpose, the CYp Screening Elderly (CYSCE) trial was started. During the project she studied and developed analytical dried blood spot methods, which resulted in three chapters of her PhD thesis. Due to slow en-rollment of patients in the CYSCE trial, enrollment time had to be extended and the proj-ect will be continued by a colleague PhD candidate. Nevertheless, a cost-effectiveness analysis was completed based on experience during the CYSCE trial and on literature to give an indication about the added value of pharmacogenetic testing among old aged patients starting nortriptyline pharmacotherapy.

She is now employed as consultant at Inventiv Health, performing services as a Clini-cal Research Associate for Janssen-Cilag B.V. Currently she is working on clinical studies in de field of cardiology.

Dankwoord 217

DAnkwOORD

Voor de totstandkoming van dit proefschrift gaat allereerst mijn dank uit naar mijn promotoren Prof. Wilffert en Prof. Hak en co-promotor Dr. Maring.Prof. Wilffert, beste Bob, bedankt voor het vertrouwen dat je me gegeven hebt door mij op dit project aan te nemen. Het was erg prettig om met jou als dagelijkse supervisor samen te werken. Niet alleen ben je zeer kundig in je vakgebied, ook daarbuiten werk je met een vriendelijke zorgvuldigheid die erg gewaardeerd wordt. Prof. Hak, beste Eelko, bedankt voor je scherpe beoordelingsvermogen, “helikopter view” en betrokkenheid bij het CYSCE onderzoek en mijn promotietraject. Dr. Maring, beste Jan Gerard. Zonder jou en het lab wat achter je staat, had de inhoud van mijn proefschrift er heel anders uitgezien. Ik heb erg veel van jullie geleerd. Bedankt dat je zoveel tijd en werk in dit onderzoek hebt gestoken. De heren van de leescommissie: Prof. E. Buskens, Prof. H.J. Guchelaar, en Prof. N.C. van der Merbel, dank voor het lezen en beoordelen van mijn proefschrift.

Inge van Doornik, bedankt voor je betrokkenheid en zorgvuldigheid tijdens het uitrol-len van de CYSCE trial en vervolgens voor het waarnemen van de dagelijkse gang van zaken in de CYSCE trial. Bedankt voor de gezellige autoritjes naar alle uithoeken van Nederland.

Prof. Postma, beste Maarten, bedankt voor het toevoegen van kosteneffectiviteit onder-zoek aan mijn promotietijd. Ik vind dit zeer waardevol. Prof Touw, beste Daan, bedankt voor je betrokkenheid en samenwerking in de wereld van ‘dried blood spots’. Ik heb veel van je kunnen leren en de samenwerking verliep altijd zeer vlot, wat in tijden van tijdsdruk erg gewaardeerd werd. Graag bedank ik alle verpleegkundigen, psychiaters, psychiaters in opleiding en arts-as-sistenten die betrokken zijn bij de CYSCE trial voor hun inzet. Ondanks dat de resultaten van dit onderzoek nog niet bekend zijn, weet ik zeker dat de uitkomst van dit onderzoek van grote waarde zal zijn voor de implementatie van farmacogenetica in de psychiatrie. Alle coauteurs, bedankt voor jullie inbreng tijdens het schrijven van de diverse artikelen!Ik wil alle patiënten die mee hebben gedaan en gaan doen aan het CYSCE onderzoek van harte danken voor hun vertrouwen en medewerking. De CYSCE trial was niet mogelijk geweest zonder de inzet van diverse ziekenhuisapo-thekers en analisten waaronder Dr. Hans Mulder, Dr. Arne Risselada, René Schutte, Ellen Brummel-Mulder en Jolanda Paardekoper, bedankt voor jullie betrokkenheid, nauwkeu-righeid en oplettendheid.

218 Chapter 5

De studenten farmacie die ik heb mogen begeleiden tijdens mijn promotietijd wil ik bedanken voor hun inzet. Bachelor studenten: Manon Bakker en Simone Jans, veel suc-ces met het afronden van jullie studie! Judith, samen met jou konden we een concrete kosten-effectiviteitanalyse maken omtrent een onderwerp waar weinig goede gegevens voor beschikbaar waren. Bedankt voor al je werk en succes met je promotieonderzoek. Blessing, jou inzet ging verder dan wat een promovendus van een student mag verwa-chten. Ik denk met plezier terug aan onze samenwerking.

Ik dank al mijn collega’s voor de fijne werkomgeving waarin ik heb mogen werken. Bert, Sipke en Jens, bedankt voor de paashazen, verhalen over Vietnamese avonturen en zeiltips. Anja, ik heb genoten van ons korte yoga avontuur en geklets. Riekje en Josta, bedankt voor het meedenken tijdens onze wandelingen over de Brug. De GIMMICS crew, ik zal het “acteren” missen! My roommates: Irene, Fenneke, Hao, Tran, Thao, Linda, Willem and Karin thank you for your pleasant company. Hao, if I ever notice my shoes are shrink-ing I will tell you! Irene bedankt voor het goede karma wat er rond jouw stoel hangt. Fenneke, bedankt voor je optimisme en relativeringsvermogen. Ik vind het erg leuk dat we nog samen in het NTvG terecht gekomen zijn! Linda, succes met je promotie! Hao, Doti and Aziati thank you for your shared interest in pharmacogenetics. Aizati &Tuan, thank you so much for taking care of me in Rome when my passport was stolen and for scouting me for your movie! All my colleagues from the no-longer -“other”-department, it was a pleasure to work and eat with you. I will never forget the black eggs at Pieters home! Job, bedankt voor je bijdrage aan onze review. Maarten Bijlsma, bedankt voor de statistische support. Jos bedankt voor je hulp met de PSA en succes met doceren en pro-moveren! Koen, na jou vertrek was de gym niet meer hetzelfde, maar we zijn er nog wel geweest! Christiaan en Eva, kijken waar je heen gaat! Jurjen, ik hoop dat de overname van mijn werkzaamheden voor de CYSCE trial je veel werkplezier zullen geven. Ik laat de studie in ieder geval in goede handen achter. Jannie, bedankt voor al je hulp met de meest uiteenlopende problemen en ideeën voor oplossingen. Ik ga de Nieuwjaarslunch missen!

Alle (surf ) vrienden en vriendinnen, bedankt voor jullie nabijheid en alle gezelligheid. Een dagje op zee, in de stad, of in het bos op het MTB-pad met jullie is het beste medicijn.

Lieve paranimfen Lisette en Eva, ik verheug me erop om tussen jullie in te staan tijdens mijn verdediging. Lisette, we delen niet alleen onze naam, maar ook lief en leed. Bedankt voor je vriendschap. Eva, jammer dat we niet lang samen op de basiseenheid hebben gewerkt, maar allicht is dit wel beter voor de productiviteit. Ik hoop dat de vaccin trial en je promotie voorspoedig zullen verlopen, als iemand de doorzettingskracht heeft die er voor nodig is, dan ben jij het.

Dankwoord 219

Hanny en Wim bedankt voor jullie interesse in mijn (paramedische) werkzaamheden en alle steun en betrokkenheid, zelfs tijdens de meest recente verhuizing. Mathijs en Yannick, al bijna 10 jaar mag ik genieten van jullie gezelschap en zorgzaamheid tijdens diverse vakanties, maar ook dicht bij huis, waarvoor dank. Gabrielle en Ronald-Jan, bedankt voor jullie mentale en fysieke nabijheid. Niet meer in Groningen wonen heeft zeker een groot nadeel.

Arjan, Mariska en Rieneke, jullie zijn een bont en zeer dierbaar gezelschap die naast een luisterend oor, ook een altijd welkome afleiding van de wetenschap zijn. Pap en mam, bedankt voor jullie betrokkenheid en oprechte interesse in mijn onder-zoek. Jullie hebben mij geleerd om door te zetten, me niet al te druk te maken om wat anderen denken en mij gesteund in wat ik leuk vond. Pap, ik hoop dat je, je nu geen zorgen meer hoeft te maken over het aantal (niet) geïncludeerde patiënten en mijn werkgelegenheid.

Vincent, dankbaar ben ik dat je mijn duinpad bent gekruist op Texel. Ik zou niet weten hoe ik zonder onze gezamenlijke uurtjes op het water inspiratie voor het schrijven van een proefschrift zou hebben gehad. Gelukkig wou je mee naar het zuidwesten van het land en kunnen we hopelijk, naast onze nieuwe werkzaamheden, de golven van Scheveningen gaan verkennen. Bedankt voor de kennis uit de medische praktijk, steun en de liefde die je me geeft. Ik ben dankbaar voor ons verleden, en ik zie uit naar onze toekomst!

Lisette

SHARE publications 221

sHARE PuBliCAtiOns

This thesis is published within the Research institute sHARE (Science in Healthy Ageing and healthcaRE) of the University Medical Center Groningen / University of Groningen.

Further information regarding the institute and its research can be obtained from our internetsite: http://www.share.umcg.nl/

More recent theses can be found in the list below.((co-) supervisors are between brackets)

2016

Otter tA

Monitoring endurance athletes; a multidisciplinary approach(prof KAPM Lemmink, prof RL Diercks, dr MS Brink)

Bielderman JH

Active ageing and quality of life; community-dwelling older adults in deprived neigh-bourhoods(prof CP van der Schans, dr MHG de Greef, dr GH Schout)

Bijlsma mJ

Age-period-cohort methodology; confounding by birth in cardiovascular pharmacoepi-demiology(prof E Hak, prof S Vansteelandt, dr F Janssen)

Dingemans EAA

Working after retirement; determinants and conzequences of bridge employment(prof CJIM Henkens, dr ir H van Solinge)

Jonge l de

Data quality and methodology in studies on maternal medication use in relation to congenital anomalies(prof IM van Langen, prof LTW de Jong-van den Berg, dr MK Bakker)

Vries Fm de

Statin treatment in type 2 diabetes patients(prof E Hak, prof P Denig, prof MJ Postma)

222 Chapter 5

Jager m

Unraveling the role of client-professional communcation in adolescent psychosocial care(prof SA Reijneveld, prof EJ Knorth, dr AF de Winter, dr J Metselaar)

mulder B

Medication use during pregnancy and atopic diseases in childhood(prof E Hak, prof SS Jick, dr CCM Schuling-Veninga, dr TW de Vries)

Romkema s

Intermanual transfer in prosthetic training(prof CK van der Sluis, dr RM Bongers)

Diest m van

Developing an exergame for unsupervised home-based balance training in older adults(prof GJ Verkerke, prof K Postema, dr CJC Lamoth, dr J Stegenga)

waterschoot FPC

Nice to have or need to have? Unraveling dosage of pain rehabilitation(prof MF Reneman, prof JHB Geertzen, prof PU Dijkstra)

zijlema wl

(Un)healthy in the city; adverse health effects of traffic-related noise and air pollution(Prof JGM Rosmalen, prof RP Stolk)

zetstra-van der woude AP

Data collection on risk factors in pregnancy(prof LTW de Jong-van den Berg, dr H Wang)

mohammadi s

The intersecting system of patients with chronic pain and their family caregivers; cogni-tions, behaviors, and well-being(prof M Hagedoorn, prof R Sanderman, dr M Deghani)

Verbeek t

Pregnancy and psychopathology(prof MY Berger, prof CLH Bockting, dr H Burger, dr MG van Pampus)

SHARE publications 223

2015

Broekhuijsen k

Timing of delivery for women with non-severe hypertensive disorders of pregnancy(prof PP van den Berg, prof BWJ Mol, dr MTM Faassen, dr H Groen)

tuuk k van der

Who’s at risk? Prediction in term pregnancies complicated by hypertensive disorders(prof PP van den Berg, prof BWJ Mol, dr MG van Pampus, dr H Groen))

Vitkova m

Poor sleep quality and other symptoms affecting quality of life in patients with multiple sclerosis(prof SA Reijneveld, prof Z Gdovinova, dr JP van Dijk, dr J Rosenberger)

sudzinova A

Roma ethnicity and outcomes of coronary artery disease(prof SA Reijneveld, dr JP van Dijk, dr J Rosenberger)

For more 2015 and earlier theses visit our website