university of groningen pharmacoeconomics of cardiovascular … · prediction models in...

University of Groningen

Pharmacoeconomics of cardiovascular disease preventionStevanovic, Jelena

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:2015

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):Stevanovic, J. (2015). Pharmacoeconomics of cardiovascular disease prevention. University of 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: 06-02-2021

https://www.rug.nl/research/portal/en/publications/pharmacoeconomics-of-cardiovascular-disease-prevention(2979abd7-00f8-43ea-baf7-e4e77f0794d8).html

https://www.rug.nl/research/portal/en/publications/pharmacoeconomics-of-cardiovascular-disease-prevention(2979abd7-00f8-43ea-baf7-e4e77f0794d8).html

Pharmacoeconomics of

cardiovascular disease prevention

Jelena Stevanović

Paranimfen: Milica Stanković-Brandl

Bianca Mulder

Cover design by Saša Dimitrijević

Layout by Henri van der Heiden

Printed by CPI Koninklijke Wöhrmann

ISBN (book) 978-90-367-8415-3

ISBN (electronic version) 978-90-367-8414-6

Printing of this thesis was financially supported by the University of Groningen, the Groningen

Graduate School of Science, and the Graduate School for Health Research SHARE.

© 2015 Jelena Stevanović. No part of this thesis may be reproduced or transmitted in any form or by

any means, electronical or mechanical, including photocopying, recording, or by any information

storage or retrieval system, without written permission of the author. The copyright of previously printed chapters of this thesis remains with the publisher or journal.

Pharmacoeconomics of cardiovascular disease

prevention

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

dinsdag 22 december 2015 om 11.00 uur

door

Jelena Stevanović

geboren op 13 juli 1983 te Niš, Servië

Promotor

Prof. dr. M.J. Postma

Copromotores

Dr. P. Pechlivanoglou

Dr. H.H. Le

Beoordelingscommissie

Prof. dr. B. Wilffert

Prof. dr. H.R. Büller

Prof. dr. G. Petrova

To my family

TABLE OF CONTENTS

Chapter 1: General Introduction 9

Part I: Simulating and synthesizing health and economic evidence in (primary)

cardiovascular disease prevention.

21

Chapter 2: A systematic review on the application of cardiovascular risk

prediction models in pharmacoeconomics, with a focus on primary

prevention.

23

Chapter 3: Economic evaluation of primary prevention of cardiovascular diseases

in mild hypertension. A scenario analysis for the Netherlands.

45

Annex: Incidence description and costs of acute heart failure in the

Netherlands.

73

Chapter 4: Multivariate meta-analysis of preference-based quality of life values in

coronary heart disease.

87

Part II: Pharmacoeconomic findings in secondary cardiovascular disease

prevention; Focus on oral anticoagulation.

119

Chapter 5: Budget impact of increasing market-share of patient self-testing and

patient self-management in anticoagulation.

121

Chapter 6: Economic evaluation of apixaban for the prevention of stroke in non-

valvular atrial fibrillation in the Netherlands.

141

Chapter 7: Economic evaluation of dabigatran for the treatment and secondary

prevention of venous thromboembolism.

165

Chapter 8: General Discussion 189

Summary 201

Samenvatting 203

List of publications 205

Acknowledgements 207

Research Institute of Science in Healthy Aging & health caRE (SHARE) 209

Curriculum Vitae 213

Chapter 1

General introduction

1


11

CARDIOVASCULAR DISEASES – HEALTH AND ECONOMIC BURDEN IN DEVELOPED

COUNTRIES

Cardiovascular diseases (CVD) comprise disorders that affect the heart, blood vessels or

both. Some of the most common CVD disorders are coronary heart disease ((CHD) i.e.

stable angina and acute coronary syndrome (ACS)), cerebrovascular disease (i.e. transient

ischemic attack and stroke), peripheral arterial disease, rheumatic heart disease,

congenital heart disease, heart failure (HF), venous thromboembolism (VTE) (1).

CVD had a leading position in contribution to global disease burden among the non-

communicable diseases in the last couple of decades. In 2008, 17 million persons died

from CVD worldwide. However, in some developed countries, a declining trend of CVD

mortality rates is observed in recent years. In the Netherlands, CVD are now the second

largest cause of death accounting for 29% of total deaths after cancer being the first with

a toll of 33% of total deaths (1). This significant drop in CVD mortality, reaching almost

50% in both men and women in the Netherlands, might be explained by the advances in

both treatment and CVD prevention strategies (2-4). Nevertheless, hospital admissions

due to CVD do not seem to follow the mortality trend and remain high (5). Moreover,

patients who experience non-fatal CVD events are severely affected with disability and

impairment of health-related quality of life (HRQoL). For example HRQol of patients after a

major stroke were on average half of perfect health (6). Given that this estimate is

positioned approximately only in the middle of the scale, the burden of major stroke on

patient’s HRQoL is undisputable. Although there is significant variation on the level of

HRQoL across CVD disorders and across severity levels it consistently indicates

impairment.

CVD do not only account for a major share of overall morbidity and mortality worldwide,

but also significantly contribute to global healthcare expenditures. Some of the major

drivers of resources used within the healthcare system are the costs of hospitalizations,

long-term management and rehabilitation care due to CVD. From a broader societal

viewpoint, next to healthcare costs accounting for approximately two thirds of total

economic burden due to CVD, one third of costs can be attributed to the productivity loss

costs associated with the CVD patients being absent from work and the costs of informal

care of these patients (7,8). In the Netherlands, economic burden of CVD led to healthcare

expenditures of €5.5 billion annually, while in the whole European Union it mounted to

€105 billion (7,8).

CARDIOVASCULAR DISEASES PREVENTION – RISK PREDICTION APPROACH

The most recent European guidelines on CVD prevention recommend the combination of

two different approaches to achieve the highest level of prevention (9). One approach

Chapter 1

12

focuses on promoting lifestyle and environmental changes in the population at large (i.e.

the public health/population-based approach) while an alternative approach aims to

modify risk factors only in individuals identified to be at a high risk (i.e. the high-risk

approach). Although the large-size benefits may be reached by applying the population

prevention approach, the benefits that can be achieved with individual risk assessment

should not be disregarded. Furthermore, implementing individual risk assessment to

identify high-risk individuals remains a recommended strategy in regular clinical practise

(9).

Nowadays risk assessment tools are derived using data from decades-long studies

designed to follow large cohorts for a sufficient length of time and for which both patient

characteristics (i.e. risk factor characteristics) and outcomes are known over time. For

example, studies based on the Framingham population were among the first ones to find

the association between a number of, often correlated, modifiable and/or non-modifiable

risk factors and the onset of CVD (10)(11). Elevated blood pressure (BP), cholesterol or

glucose levels and smoking have been identified as some of the major modifiable risk

factors, and age, gender and genetic predisposition are some of the non-modifiable risk

factors, whose conjoint influence can lead to CVD.

The aforementioned types of observations have set the grounds for the development of

various multivariable risk prediction models that attempt to translate this conjoint impact

of various risk factors on the onset of CVD into a mathematical relation. Numerous

differences exist between the various proposed CVD risk prediction models. These models

often differ in the specific risk factors they comprise, the underlying study populations

(different geographical regions, baseline CVD risks) as well as whether they reflect the risk

of overall CVD or one of its specific CVD forms (e.g. myocardial infarction, stroke etc.).

Moreover, they give predictions over different time horizons (e.g. 1-4 years (12), 10 years

(13) or 30 years (14)), for different clinical endpoints (e.g. fatal (13) or overall CVD (12))

and for primary or secondary CVD events (12).

Because of the aforementioned differences across risk models, the choice and consequent

utilization of a model for projecting CVD risk in a certain setting is not straightforward. The

choice of a model might need to be made with respect to specific characteristics of the

population of interest (e.g. middle-aged or elderly patients, or diabetics), or with respect

to the need to estimate CVD risk over a specific time-horizon. An example of a model for

CVD risk assessment in European populations is the SCORE model developed on data from

12 European cohorts (9,13). The high- and low-risk region-specific multivariable risk

functions of the SCORE model estimate the 10-year risk of a fatal CVD by accounting for

age, gender, level of SBP, level of cholesterol and prevalence of smoking. Thus, if the 10-

year risk of a fatal CVD needs to be estimated in individuals from a certain European

setting, the choice of the SCORE model risk function should be made carefully with respect

to the general population’s risk level. Notably, the robustness and validity of the risk

1


13

prediction in a specific setting needs to be assessed, and if needed, it can be enhanced by

various adjustments and validation methods.

PHARMACOLOGICAL INTERVENTIONS TO PREVENT CARDIOVASCULAR DISEASES

When it comes to implementing prevention strategies to reduce the overall CVD risk by

identifying individuals at risk with the use of a suitable CVD risk prediction model (9),

various national and international guidelines give different recommendations on the CVD

risk threshold above which pharmacotherapy should be introduced in addition to lifestyle

changes (15).

In the Netherlands, cardiology guidelines recommend pharmacotherapy to prevent CVD

for patients at a 10-year CVD mortality or morbidity risk of 20% or exceptionally at lower

levels of risk if accompanied with other comorbidities (e.g. diabetes or rheumatoid

arthritis) or previous CVD (16). Prevention strategies can include antihypertensive

treatment and/or statins, if high CVD risk is accompanied with systolic blood pressure

(SBP) greater than 140 mmHg and/or low-density lipoprotein (LDL) greater than 2.5mmol/l

(16). In patients with a medical history of previous CVD events and elevated SBP and/or

LDL levels, immediate pharmacotherapy is recommended due to the relatively high risk of

recurrent CVD events (16).

Moreover, in the presence of certain individual risk factors, such as atrial fibrillation (AF)

or diabetes, pharmacological interventions additional to the ones previously mentioned

may be indicated. For example, atrial fibrillation (AF) is considered to be an independent

risk factor leading to an even 5-fold increased risk of stroke (17). To prevent stroke in

patients with AF and minimum one stroke risk factor (i.e. congestive HF or left ventricular

dysfunction, Hypertension, Age≥75 (doubled risk), Diabetes, Stroke (doubled)-Vascular

disease, Age 65–74, Sex category (female) [CHA2DS2-VASc] ≥ 1), antithrombotic preventive

treatment is recommended (17). Yet, this preventive strategy does not apply to patients

younger than 65 years with an AF diagnosis and being female as the only risk factor.

Furthermore, once the criteria are satisfied for a patient to receive pharmacotherapy for

CVD prevention, a choice of the most suitable treatment option needs to be made. From

the healthcare perspective, this choice should consider both treatment effectiveness in

modifying the risk factors of interest and its safety profile. For example, to prevent stroke

in patients with AF, guidelines generally recommend the use of vitamin K-anticoagulants

(VKAs) or novel oral anticoagulants (NOACs)(17). However, these medicines differ in their

effectiveness to protect against stroke and other thromboembolic events as well as in the

safety profile reflected by the incidence of major bleeding events (18).

Finally, next to considering the health-related aspects of choosing a certain

pharmacological treatment, the economic consequences of that decision should not be

ignored due to limited healthcare resources. Together, the health and economic

Chapter 1

14

consequences of applying pharmacological interventions to prevent CVD should be

assessed in a formal way through pharmacoeconomic analyses.

CHALLENGES IN PHARMACOECONOMICS OF CARDIOVASCULAR DISEASES

PREVENTION

A number of governmental institutions worldwide request sound pharmacoeconomic

evidence to aid health policy and reimbursement decisions regarding newly introduced

interventions (19). Commonly, country-specific guidelines are available to assist analysts in

conducting coherent pharmacoeconomic analyses for reimbursement purposes. Yet, when

such analyses on pharmacological interventions to prevent CVD are needed, numerous

methodological challenges can be encountered whose solutions are not explicitly

discussed in the aforementioned guidelines.

To compare different pharmacological interventions in CVD prevention, numerous health

and economic evidence are needed. In ideal circumstances, evidence on long-term

treatment effectiveness of specific interventions (i.e. hard clinical endpoints, such as CVD

morbidity and mortality), HRQoL values collected alongside clinical trial and all the

relevant up-to-date country-specific costs would be available for such an analysis (Figure

1A). However, pragmatic pharmacoeconomic analysis needs to consider the use of less

perfect evidence, as often this evidence is not available at the time when the

pharmacoeconomic analysis is done (Figure 1B). Firstly, if the pharmacoeconomic analysis

investigates the use of a new pharmacological intervention that is directed to modifying a

certain risk factor, such an analysis would commonly have access only to short-term

(intermediate) treatment effectiveness evidence. This is particularly the case when

collecting pharmacological efficacy data from a relatively healthy population (commonly

indicated for primary CVD prevention) where conducting a clinical trial to provide hard-

clinical endpoints evidence would be too lengthy and costly. For example, the evidence on

effectiveness of different antihypertensive medicines would only indicate the change in

systolic and/or diastolic BP (SBP and/or DBP) when examined in patients with mild

hypertension and mild to moderate CVD risk. In case treatment evidence (short or long

term) is available from more than one study, analysts might consider the evidence

synthesis of all the relevant data (20). As long-term evidence is generally lacking, it is

necessary to apply a CVD risk prediction model to translate short-term treatment

effectiveness to long-term consequences. Notably, given the previously mentioned

differences across CVD risk prediction models, model choice and its application in

pharmacoeconomic analysis is not straightforward (21). Finally, in case health outcomes of

analysis will be expressed in terms of quality-adjusted life-years (QALYs) gained, all the

available HRQoL evidence reflecting preferences for living in a certain state of CVD might

need to be considered. Commonly, pharmacoeconomists tend to select a HRQoL value in a

1


15

less rigorous approach even if a couple of values are available in the literature. Yet, this

should not be a random process. One may consider a value that is close to analysis’s

setting, what seems reasonable, but still it is not robust. Therefore, further effort should

be made to also synthesize evidence on HRQoL, where appropriate.

Figure 1 Evidence availability in ideal (A) and pragmatic (B) pharmacoeconomic analysis of

pharmacological interventions in CVD prevention. HRQoL, health-related quality of life; CVD, cardiovascular diseases.

Embedding and integrating the concepts of evidence synthesis and prediction/simulation

in economic evaluation is one of the key challenges in providing reliable

pharmacoeconomic evidence on interventions for CVD prevention. Moreover, the

dynamic world of new pharmacological interventions for prevention and/or treatment of

CVD, and changes in country-specific health-related economic factors urge for new

pharmacoeconomic analysis accounting for those changes. Specifically, the introduction of

NOACs to the Dutch market for the prevention of stroke in non-valvular AF, and treatment

and prevention of VTE, requires sound pharmacoeconomic evidence to support

reimbursement decisions. This evidence should account for all the specificities of the

Dutch health system and indicate whether there is an additional value associated with the

use of NOACs compared to the standard treatment with VKAs. Notably, such analyses

should adequately consider core issues such as dealing with uncertainty and discounting in

secondary prevention within the context of changing Dutch health-care with increasing

roles for patient access schemes (such as price-volume arrangements), clinical guidelines-

driven cost-effectiveness analyses and cost consciousness.

Chapter 1

16

AIMS AND OUTLINE OF THIS THESIS

Part I of this thesis addresses some of the challenging issues on simulating and

synthesizing health and economic evidence in primary CVD prevention. Here, one of the

issues concerns the application of CVD risk prediction models in pharmacoeconomic

analysis to project changes in CVD incidence based on changes in risk factor levels

observed in short-term randomised clinical trials. In order to assess the adequacy of

incorporating risk prediction models in available pharmacoeconomic analysis for primary

CVD prevention interventions in high income countries, specify and propose methods for

enhancing the robustness of such studies, a systematic literature review is described in

Chapter 2. Based on the knowledge and caveats derived in this review, an attempt was

made to design a simplified model of primary CVD prevention with antihypertensives

using surrogate endpoints on lowering SBP projected to hard CVD endpoints. This

modelling approach is described in Chapter 3. The aforementioned model was informed

with various published Dutch cost estimates such as cost of HF, angina, stroke, etc. Annex

1 provides an up-to-date estimate of hospital costs associated with HF in the Netherlands.

Next, Chapter 4 describes an evidence synthesis of instrument-specific preference-based

HRQoL values in CHD and its underlying disease-forms (i.e. stable angina, ACS) while

accounting for study-level characteristics (i.e. covariates) and relevant correlations

between those values. This study aims to further enhance the robustness of HRQoL values

used to inform cost-utility analyses on interventions in CHD prevention and/or treatment.

Part II of this thesis describes a number of pharmacoeconomic findings in secondary CVD

prevention. In particular, this part of the thesis focuses on the health and economic

findings associated with the use of VKAs and NOACs such as apixaban and dabigatran, for

the prevention of CVDs in the Dutch setting. Chapter 5 describes a budget impact analysis

of increasing market-share scenarios of patient self-testing and patient self-management

in patients on long-term indication for anticoagulation with VKAs. In Chapter 6, the cost-

effectiveness of using apixaban compared to VKA was evaluated for the prevention of

stroke in non-valvular AF in the Netherlands. Furthermore, Chapter 7 focuses on the cost-

effectiveness of another NOAC, dabigatran, for treatment and secondary prevention of

VTE.

Finally, in Chapter 8 the main findings of this thesis are summarized and discussed in the

context of current knowledge and practice, and some recommendations for future

research are given.

1


17

REFERENCES

(1) Cardiovascular diseases. Available at: http://www.who.int/topics/cardiovascular_diseases/en/. Accessed

01/01, 2015.

(2) Peeters A, Nusselder WJ, Stevenson C, Boyko EJ, Moon L, Tonkin A. Age-specific trends in cardiovascular

mortality rates in the Netherlands between 1980 and 2009. Eur J Epidemiol 2011;26(5):369-373.

(3) Kelly BB, Fuster V. Promoting Cardiovascular Health in the Developing World:: A Critical Challenge to Achieve

Global Health. : National Academies Press; 2010.

(4) Vaartjes I, O'Flaherty M, Grobbee DE, Bots ML, Capewell S. Coronary heart disease mortality trends in the

Netherlands 1972-2007. Heart 2011 Apr;97(7):569-573.

(5) Reitsma JB, Dalstra JA, Bonsel GJ, van der Meulen JH, Koster RW, Gunning-Schepers LJ, Tijssen JG.

Cardiovascular disease in the Netherlands, 1975 to 1995: decline in mortality, but increasing numbers of patients

with chronic conditions. Heart 1999 Jul;82(1):52-56.

(6) Tengs TO, Lin TH. A meta-analysis of quality-of-life estimates for stroke. Pharmacoeconomics 2003;21(3):191-

200.

(7) Leal J, Luengo-Fernandez R, Gray A, Petersen S, Rayner M. Economic burden of cardiovascular diseases in the

enlarged European Union. Eur Heart J 2006 Jul;27(13):1610-1619.

(8) Poos M, Smit J, Groen J, Kommer G, Slobbe L. Kosten van ziekten in Nederland 2005. Bilthoven: Rijksinstituut

voor Volksgezondheid en Milieu 2008.

(9) Perk J, De Backer G, Gohlke H, Graham I, Reiner Z, Verschuren M, Albus C, Benlian P, Boysen G, Cifkova R,

Deaton C, Ebrahim S, Fisher M, Germano G, Hobbs R, Hoes A, Karadeniz S, Mezzani A, Prescott E, Ryden L, et al.

European Guidelines on cardiovascular disease prevention in clinical practice (version 2012). The Fifth Joint Task

Force of the European Society of Cardiology and Other Societies on Cardiovascular Disease Prevention in Clinical

Practice (constituted by representatives of nine societies and by invited experts). Eur Heart J 2012

Jul;33(13):1635-1701.

(10) Kannel W, Gordon T. The probability of developing certain cardiovascular diseases in eight years at specified

values of some characteristics. The Framingham Study: An epidemiological investigation of cardiovascular

disease.Springfield, MA: US Department of Commerce National Technical Information Service 1987.

(11) Anderson KM, Odell PM, Wilson PWF, Kannel WB. Cardiovascular disease risk profiles. Am Heart J

1991;121(1):293-298.

(12) D'Agostino RB, Russell MW, Huse DM, Ellison RC, Silbershatz H, Wilson PWF, Hartz SC. Primary and

subsequent coronary risk appraisal: new results from the Framingham study. Am Heart J 2000;139(2):272-281.

(13) Conroy R, Pyörälä K, Fitzgerald AP, Sans S, Menotti A, De Backer G, De Bacquer D, Ducimetiere P, Jousilahti P,

Keil U. Estimation of ten-year risk of fatal cardiovascular disease in Europe: the SCORE project. Eur Heart J

2003;24(11):987.

(14) Pencina MJ, D'Agostino RB, Larson MG, Massaro JM, Vasan RS. Predicting the 30-Year Risk of Cardiovascular

Disease. Circulation 2009;119(24):3078-3084.

(15) Ferket BS, Colkesen EB, Visser JJ, Spronk S, Kraaijenhagen RA, Steyerberg EW, Hunink MM. Systematic

review of guidelines on cardiovascular risk assessment: which recommendations should clinicians follow for a

cardiovascular health check? Arch Intern Med 2010;170(1):27-40.

(16) Nederlands Huisartsen Genootschap. Multidisciplinaire richtlijn cardiovasculair risicomanagement,

herziening 2011. Utrecht. Available at: https://www.nvvc.nl/media/richtlijn/106/2011_MDR_CVRM.pdf.

Accessed 02/04, 2013.

(17) Goldstein LB, Bushnell CD, Adams RJ, Appel LJ, Braun LT, Chaturvedi S, Creager MA, Culebras A, Eckel RH,

Hart RG, Hinchey JA, Howard VJ, Jauch EC, Levine SR, Meschia JF, Moore WS, Nixon JV, Pearson TA, American

Heart Association Stroke Council, Council on Cardiovascular Nursing, et al. Guidelines for the primary prevention

of stroke: a guideline for healthcare professionals from the American Heart Association/American Stroke

Association. Stroke 2011 Feb;42(2):517-584.

Chapter 1

18

(18) Lopes RD, Al-Khatib SM, Wallentin L, Yang H, Ansell J, Bahit MC, De Caterina R, Dorian P, Easton JD, Erol C.

Efficacy and safety of apixaban compared with warfarin according to patient risk of stroke and of bleeding in

atrial fibrillation: a secondary analysis of a randomised controlled trial. The Lancet 2012.

(19) Pharmacoeconomic Guidelines Around The World. Available at:

http://www.ispor.org/peguidelines/index.asp. Accessed 02/04, 2015.

(20) Evans D. Hierarchy of evidence: a framework for ranking evidence evaluating healthcare interventions. J Clin

Nurs 2003;12(1):77-84.

(21) Grieve R, Hutton J, Green C. Selecting methods for the prediction of future events in cost-effectiveness

models: a decision-framework and example from the cardiovascular field. Health Policy 2003;64(3):311-324.

PART I

Simulating and synthesizing health and

economic evidence in (primary)


Chapter 2

A systematic review on the application of

cardiovascular risk prediction models in

pharmacoeconomics, with a focus

on primary prevention

Jelena Stevanovic, Maarten J Postma, Petros Pechlivanoglou.

European Journal of Preventive Cardiology. 2012 Aug;19(2 Suppl):42-53.

Chapter 2

24

ABSTRACT

Background: Long-term trials on the effectiveness of pharmacological treatment for

primary cardiovascular disease prevention are scant. Risk prediction models are used as a

tool to project changes in cardiovascular disease incidence due to changes on risk factor

levels observed in short-term randomised clinical trials. In this article, we summarize the

literature on the application of these risk models in pharmacoeconomic studies for

primary cardiovascular disease prevention interventions in high income countries.

Methods and results: We systematically reviewed the available literature on the

application of cardiovascular disease risk models in pharmacoeconomic studies and

assessed the quality of incorporation of risk models in these studies. Quality assessment

indicated the distance between the characteristics of populations of the risk model and

the studies reviewed, the frequent disagreement between risk model- and study- time

horizons and the lack of proper consideration of the uncertainty surrounding risk

predictions.

Conclusion: Given that utilizing a risk model to project the effect of a pharmacological

intervention to cardiovascular events provides an estimate of the intervention's clinical

and economical impact, consideration should be paid to the agreement between the

study and risk model populations as well as the level of uncertainty that these predictions

add to the outcome of a decision-analytic model. In the absence of hard endpoint trials,

the value of risk models to model pharmacological efficacy in primary cardiovascular

disease prevention remains high, although their limitation should be acknowledged.

2

Cardiovascular risk prediction models in pharmacoeconomics: systematic review

25

INTRODUCTION

Cardiovascular disease (CVD) accounts for a major share of overall morbidity and mortality

in the Western world (1,2). Furthermore, due to the high inpatient costs of most common

CVDs, it is the most significant contributor of health care expenditures, e.g. treatment

costs of CVD in European Union comprise 12% of all health care costs (€105 billion)(3).

Given the health and economic burden of CVD, its primary prevention is of high

importance to society.

The onset of CVD has been associated with a number of, often correlated, risk factors.

Elevated blood pressure (BP), cholesterol or glucose levels, smoking and age have been

identified as the most important of them, whose conjoint influence can lead to CVD.

Based on these observations, during the last 25 years, different multivariable risk

prediction models have been developed, to assist identification of patients at risk of CVD

in clinical practice (4-23). These models have been generally based on large cohorts that

are being followed for a sufficient length of time and for which both risk factors and

outcomes are known over time. The models often differ in the risk factors they comprise

and the underlying study populations, as well as the definition of the CVD risk studied.

Furthermore, they make predictions for different time horizons and for different clinical

CVD-specific endpoints.

The utilization of CVD risk prediction models in the economic evaluation of

pharmacological primary prevention has already been established in various studies

(24,25). Their value stems from the fact that there is limited evidence on the long-term

benefits of the drugs used in CVD prevention. Therefore, these models provide a linkage

between the treatment effectiveness on intermediate endpoints, observed in short-term

randomised-clinical trials (RCTs) (e.g. change in cholesterol, BP), and the long-term

benefits of treatment (e.g. reduced number of CVD events). Furthermore, prediction of

the number of CVD events avoided due to treatment also offers a way to estimate the

future financial savings on these events.

In this article, we systematically summarize the available literature on the application of

CVD risk prediction models in pharmacoeconomic studies that explore interventions for

primary CVD prevention in high income countries.

METHODS

Pharmacoeconomic studies of different primary CVD prevention strategies were

systematically reviewed using the PubMed database and NHS Economic Evaluation

Database. We used the following keywords: "risk assessment" or "risk prediction model"

or "risk model" or "risk function" or "risk equation" or "risk factor" or "risk score" and

"cardiovascular diseases" or "cardiovascular" or "coronary heart disease" and "cost-

benefit" or "costs and cost analysis" or "health care costs" or "cost-benefit" or "cost-

Chapter 2

26

effectiveness" or "cost-utility" or "economic evaluation" or "decision model". The search

was limited to studies published until December 2011.

Studies were included in the reviewing process if they:

1) were strictly classifiable in one of the pharmacoeconomic categories of cost-

effectiveness analysis (CEA), cost-utility analysis (CUA) or cost-benefit analysis

(CBA);

2) were written in English and published in peer-reviewed journals;

3) presented a decision-analytic model that has utilized a risk prediction model;

and

4) explored pharmacological, primary prevention of overall CVD, or some of its

constituents, in populations with no prior CVD events or diabetes.

Hence, we excluded studies exploring either secondary, or primary combined with

secondary CVD prevention, or studies where CVD was considered as comorbidity to an

underlying disease (e.g. HIV/AIDS or diabetes).

Editorials, letters, clinical conference abstracts and reviews were excluded. Finally, we

focused only on pharmacoeconomic studies in high income countries (defined according

to their gross national income per capita (26). All abstracts retrieved from the electronic

databases were independently screened and reviewed by two team members (JS and PP).

Disagreements were resolved through discussions.

The appropriateness of the application of the CVD risk models in the pharmacoeconomic

analyses was evaluated using a set of the criteria similar to that proposed by Grieve et

al.(24). These criteria relate to:

1) the agreement between the time horizons of the risk model and the decision-

analytic model;

2) the comparability between the study population and the risk model

population;

3) whether the risk model comprises the key risk variable influenced through

pharmacological intervention;

4) the availability of evidence on prediction accuracy in the study population;

and

5) the approach followed in incorporating the uncertainty related to CVD risk

projection into the overall decision-analytic model uncertainty(25,27).

RESULTS

Through our search strategy, we retrieved 810 studies. The flow chart of the literature

search is presented in Figure 1. After implementing all the inclusion criteria, our search

highlighted 12 relevant studies. Of the 12 studies included, four referred to the United

States (US)(28-31), two to the United Kingdom (UK)(32,33), two to Sweden (34,35), and

2


27

one to each of Germany (36), Greece (37), Canada (38), and Japan (39). The main

characteristics of the pharmacoeconomic studies are summarized in Table 1, while

information regarding the utilization of CVD risk models and their incorporation in the

decision-analytic models is summarized in Table 2. A more detailed description of the

reviewed studies can be found in the appendix.

Figure 1 Search results and culling process of retrieving pharmacoeconomic studies of

pharmacological interventions for primary CVD prevention in high income countries that utilized published CVD risk prediction models. CVD, cardiovascular diseases

Summarizing the CVD risk models

The majority of the studies reviewed have utilized CVD risk prediction models based on a

Framingham population. We encountered five Framingham risk prediction models

projecting different endpoints for various time horizons (4,5,10,16,22). Such widespread

application of the Framingham risk models is not surprising. These risk models are widely

used in clinical guidelines (40,41) and are validated in not only Caucasian- and African-

American but also in some European and Asian populations, to accurately predict CVD risk

in asymptomatic patients (42). They comprise a variety of easily obtained and clinically

relevant risk factors (43). Furthermore, their advantage, which facilitates their

applicability, is that they are developed on a general population.

The risk prediction models developed by Anderson et al.(4) were used in five studies

(28,29,32-34) and their adjusted form in one (39), while another Framingham model

predicting coronary heart disease (CHD) risk(5) was used in one study (36). Johannesson

Chapter 2

28

(34) utilized a different risk model (16) that was also developed on data from the

Framingham population. Pignone et al. (28) used two Framingham based risk models for

stroke (10,22) next to the one from Anderson et al. (4). The CHD Policy Model (44) was

also designed using Framingham data (45).

Out of 12 studies, three utilized CVD risk models that were not developed on a

Framingham population. Specifically, Maniadakis et al. (37) utilized a risk model by Glynn

et al.(11), which was developed on two RCTs populations (46,47). These RCTs investigated

the effects of aspirin and beta-carotene in male, and the effects of aspirin and vitamin E in

female health professionals, on prevention of CVD and cancer. The Cardiovascular Disease

Life Expectancy (CVDLE) Model (48) used by Grover et al.(38) was based on a logistic

regression function developed on the Lipid Research Clinic (LRC) cohort (49,50). This

cohort consisted of patients older than 30 years, from 10 clinics in North America that

were followed for more than 12 years. Lofroth et al.(35) developed separate risk

equations for myocardial infarction (MI) and stroke that utilized data from two Swedish

hospital registers and from the Swedish Nora study (35,51).

Critical assessment of the risk models

Time horizon

Five of the reviewed studies (28,29,33,34,39) utilized the CVD risk prediction functions of

Anderson et al.(4) to estimate annual transition probabilities for Markov models with

lifetime horizons. However, these risk models are designed to predict a cumulative CVD

risk in a 4-12 years time horizon. This discrepancy evokes two concerns. The first relates to

the method the authors used to transform the cumulative risk to annual probabilities and

the second considers the extrapolation of these probabilities to periods beyond the risk

model's time horizon. Pignone et al. (28) and Ernshaw et al.(29) assumed an underlying

exponential distribution to obtain annual CVD probabilities for estimating the Markov

models. Saito et al.(39) neither reported their method of obtaining annual CHD

probabilities nor their assumptions related to extrapolating them to a lifetime horizon.

The risk models by Anderson et al.(5) also provide cumulative CHD risk predictions for 4-

12 years. Brennan et al. applied this risk model annually and for a time horizon of 5

years.(36) This application is also of concern given that this risk model in quite short

horizons cannot specify the exact time of onset of the events.(5)

The CHD Policy Model (44) used by Lazar et al.(31) and Pletcher et al.(30) is a Markov

model with a 30-year time horizon, where the age- and gender- specific CHD risk is based

on logistic regression models that are estimated using longitudinal Framingham data(45).

This approach, although reducing the problems related to the translation or the

extrapolation of the risk, suffers from the need of long follow-up periods and a large

cohort size.

2


29

Johannesson (34) applied a previously published Markov model based on a risk model by

Kannel et al.(16) This risk model predicts the 8-year CHD and stroke risk. However, his

decision-analytic model is run for a lifetime horizon. The 8-year cumulative risks are

transformed into annual probabilities but no exact information is given on how this

transformation takes place.

A risk model developed by Glynn et al.(11) that predicts a 5-year CVD risk, was used in the

model of Maniadakis et al. (37) assuming a lifetime horizon. Maniadakis et al. (37) noted

that they transformed 5-year probabilities into annual ones, but without reporting the

method for that or for extrapolating them through lifetime.

Finally, the CVDLE Model (46) was utilized in a study by Grover et al.(38,48) Even though

this model is run through a lifetime, it utilizes logistic regression functions developed on

the LRC cohort (47,51), which gives 10-year predictions. In the article where the original

model was published (48), the authors assumed equal annual risks. To obtain these annual

risks they have falsely divided the 10-year risk by 10. This generally results in an

underestimation of the annual risk (52).

Population characteristics

Predictions of CVD risk in decision-analytic modelling can be biased due to differences

between the risk model and the study populations. Such differences appear more

frequently when the risk model is estimated using a RCT population. These risk models

usually overestimate the CVD risk and therefore application of them in a general

population is not recommended, at least not without proper validation (24). Furthermore,

significant differences in the socio-economic and educational status between populations

could lead to considerable prediction discrepancies (43).

In order to compare these populations in the reviewed studies we evaluated the

differences in the risk factor levels between the study and the risk model populations. In

the CUA by Pignone et al. (28), the study population had lower levels of risk factors

compared with the Framingham one. Therefore, applying the risk model by Anderson et

al. in this study population might lead to a risk overestimation (4). Risk factors of the study

cohort in the analysis by Ernshaw et al. (29) were comparable with the Framingham

population, for the corresponding age and gender. In the study of Brennan et al. (36),

specific risk factor characteristics of the populations investigated were not provided,

which hindered us from making any comparisons. Population characteristics in the study

by Caro et al.(32) were comparable with the ones in Framingham. In the studies by

Montgomery et al. (33) and Lofroth et al.(35) different strata of risk profiles were

assumed; therefore, the appropriateness of the risk models for extreme risk factor strata

would be questionable. Saito et al. (39) did not fully report risk factor characteristics (e.g.

smoking, left ventricular hypertrophy), thus complicating the comparison. However, the

model they used has been adjusted to the studied population. Risk factor levels of the

study population in the analysis by Johannesson were not explicitly stated (34).

Chapter 2

30

Maniadakis et al.(37) studied a population with risk factor levels that were slightly

elevated compared with those in the risk model study. Another discrepancy was that the

risk model used was based on a population of health care professionals from two RCTs,

which was not representative of the general population. Grover et al.(38) populated the

CVDLE Model (46) with risk factor levels from a Canadian population that was comparable

with the risk model’s population. With respect to the CHD Policy Model (44), given that it

was calibrated to reproduce USA data on the distribution of the risk factors and CHD

events, its utilization in the studies by Lazar et al.(31) and Pletcher et al.(30) concerning

USA population older than 35 years, seems appropriate.

Key risk variable

The assessment of the long-term effectiveness of a change in a risk factor level can be

most clearly observed through a risk model that comprises the risk variable of concern

(24). The majority of the studies reviewed, investigated either the influence of

antihypertensives on systolic BP (SBP) and/or diastolic BP (DBP) levels (37,39) or the

influence of statins on cholesterol levels (30,31,38). In cases where there was no risk

factor that could be modifiable through a certain treatment, such as for aspirin, a relative

risk reduction of that treatment, observed through RCTs was modelled to lead to a risk

reduction (28,29). The same method was also implemented in studies investigating

treatments for obesity that in general lead to a change in body mass index (32,36). Finally,

in three studies, although the risk model used could accommodate the effect of statin or

antihypertensive treatment through cholesterol or BP reductions, the authors

incorporated it as a multiplicative effect on the CVD risk, drawn from published RCTs (33-

35). The drawback of such method is the potential bias of the risk prediction due to

differences between the RCT population and the population studied.

Validation of the risk prediction model

When making the decision to apply a risk model to a certain population, one should first

consider the existing evidence of its performance in this population. A necessity for using

the externally validated risk model is even more pronounced in the case of geographically

or racially diverse populations (53). Additionally, the time when validation took place is of

significant importance, given the recent decline in CVD mortality in the western world.(54)

The Framingham based risk models (4) were found to generally overestimate the CVD risk

in a variety of European populations (55-57). Given this evidence, the use of Framingham

in a German setting, as used in the study by Brennan et al.(36), seems questionable. A

similar overestimation of CHD risk based on the Framingham study (5) was observed for

British men (55), which questions its utilization in the studies by Montgomery et al.(33)

and Caro et al. (32). On the contrary, even though Framingham based risk models were

not validated for a Japanese setting, utilization after adjustment for the Japanese

population, as done by Saito et al.(39), seems appropriate (4).

2


31

Maniadakis et al.(37) used a risk model by Glynn et al.(11) that has not been validated for

a Greek setting. The risk model utilized by Grover et al.(38) has been recently shown to

accurately predict the 10-year CVD risk for a Canadian setting (58). Finally, the risk models

developed by Lofroth et al. were not validated for a Swedish setting, although the model

assumes the age- and sex- specific incidence of MI and stroke in Sweden (35).

Sensitivity analysis

Nowadays, it is almost a compulsory requirement from pharmacoeconomic analyses to

provide an insight into uncertainties surrounding the model inputs as well as the structure

and the assumptions related to the model (25,27). In the reviewed studies, however, the

risk prediction uncertainty was rarely incorporated. Specifically, only one study has

incorporated the uncertainty of the risk model parameters in the probabilistic sensitivity

analysis (PSA). We considered that a major drawback for the studies lacking incorporation

of the risk model uncertainty in the PSA, since this uncertainty can be responsible for a

large share of the overall model uncertainty.

Chapter2

32

Ta

ble

1 B

asic

ch

arac

teri

stic

s o

f th

e p

har

mac

oe

con

om

ic s

tud

ies

incl

ud

ed

in t

he

sys

tem

atic

re

vie

w.

Mai

n a

uth

or

(ye

ar),

co

un

try

Ye

ar o

f va

lue

s,

curr

en

cy

Stu

dy

ob

ject

ive

Eva

luat

ion

ty

pe

(o

utc

om

e

me

asu

re)

Dis

cou

nt

rate

(c

ost

s;

he

alth

b

en

efi

ts)

Sou

rce

of

eff

icac

y/

eff

ect

ive

ne

ss

dat

a

Co

st c

om

po

ne

nts

(s

tud

y p

ers

pe

ctiv

e)

So

urc

e o

f co

st

dat

a R

esu

lts

Pig

no

ne

(28

)(2

00

7),

U

S

20

05

, US

$

CU

A o

f p

rim

ary

CV

D

pre

ven

tio

n w

ith

asp

irin

in

65

-ye

ar-o

ld w

om

en

CU

A

(QA

LY)

3;

3

Pu

blis

he

d

sou

rce

s (m

ost

ly o

n

me

n)

Co

sts

of

dru

g,

adve

rse

dru

g e

ven

ts a

nd

CV

D

eve

nts

(h

ea

lth

care

)

Pu

blis

he

d

lite

ratu

re a

nd

n

atio

nal

d

atab

ase

s

13

30

0 $

/QA

LY

Ern

shaw

(2

9)(

20

11

),

US

20

09

, US

$

CU

A o

f as

pir

in a

lon

e o

r co

mb

ine

d w

ith

PP

I in

me

n a

t d

iffe

ren

t ri

sks

for

CH

D a

nd

GI

ble

ed

ing

CU

A

(QA

LY)

3;

3

Me

ta-

anal

ysis

an

d

RC

Ts

Co

sts

of

dru

g, C

VD

e

ven

ts, o

utp

atie

nt

ph

ysic

ian

vis

it

(he

alth

car

e)

Co

sts

retr

ieve

d

fro

m a

uth

ors

’

pre

vio

us

stu

dy

Stra

tegy

- an

d a

ge-

spe

cifi

c C

U

ou

tco

me

Bre

nn

an

(36

)(2

00

6),

Ge

rman

y

20

03

, €

CU

A o

f si

bu

tram

ine

co

mb

ine

d w

ith

no

n-

ph

arm

aco

logi

cal

pra

ctic

e (

die

t an

d

life

styl

e a

dvi

ce)

vers

us

no

n-p

har

mac

olo

gica

l p

ract

ice

alo

ne

CU

A

(QA

LY)

5;

5

RC

Ts

C

ost

s o

f d

rug,

d

ise

ase

-sp

eci

fic

cost

s (h

ea

lth

car

e)

Pu

blis

he

d

lite

ratu

re,

surv

eys

an

d

auth

ors

op

inio

ns

13

70

6 €

/QA

LY

Car

o

(32

)(2

00

7),

UK

20

05

, UK

£

CU

A o

f ad

din

g ri

mo

nab

ant

for

1 y

ea

r

to e

xerc

ise

an

d d

iet

in

ob

ese

pat

ien

ts w

ith

an

d w

ith

ou

t d

iab

ete

s,

com

par

ed

to

exe

rcis

e

and

die

t al

on

e

CU

A

(QA

LY)

3.5

; 3

.5

R

CT

s C

ost

s o

f d

rug,

ad

vers

e d

rug

eve

nts

, dis

eas

e-

spe

cifi

c co

sts

(he

alth

car

e)

Co

sts

of

ho

spit

aliz

atio

ns

fro

m in

-pat

ien

t U

K r

eco

rds,

oth

er

cost

s fr

om

NH

S

guid

elin

es

85

74

£/Q

ALY

Ple

tch

er

(30

)(2

00

9),

U

S

20

06

, US

$

CU

A o

f ad

he

ren

ce t

o

the

AT

P II

I gu

ide

line

s ag

ain

st o

the

r ri

sk-

and

ag

e-

bas

ed

lip

id

low

eri

ng

stra

tegi

es

CU

A

(QA

LY)

3;

3

Pu

blis

he

d

stu

die

s

Co

sts

of

dru

gs,

adve

rse

dru

g e

ven

ts a

nd

CH

D

eve

nts

(h

ea

lth

care

)

Nat

ion

al s

pe

cifi

c

cost

s

Stra

tegy

-sp

eci

fic

CE

ou

tco

me

2

33

Systematic review application of cardiovascular risk prediction models in pharmacoeconomics

M

ain

au

tho

r

(ye

ar),

co

un

try

Ye

ar o

f

valu

es,

cu

rre

ncy

Stu

dy

ob

ject

ive

Eva

luat

ion

typ

e

(ou

tco

me

m

eas

ure

)

Dis

cou

nt

rate

(c

ost

s;

he

alth

be

ne

fits

)

Sou

rce

of

eff

icac

y/

eff

ect

ive

ne

ss

dat

a

Co

st c

om

po

ne

nts

(stu

dy

pe

rsp

ect

ive

)

Sou

rce

of

cost

dat

a

Re

sult

s

Laza

r (3

1)(

20

11

),

US

20

08

, US

$

CU

A o

f lo

w-c

ost

ge

ne

ric

stat

ins

com

par

ed

to

no

inte

rve

nti

on

an

d

exa

min

e t

he

ci

rcu

mst

ance

s u

nd

er

wh

ich

th

eir

ap

plic

atio

n

wo

uld

be

co

st-e

ffe

ctiv

e

CU

A

(QA

LY)

3;

3

Pu

blis

he

d

stu

die

s C

ost

s o

f d

rug,

ad

vers

e d

rug

eve

nts

an

d C

HD

e

ven

ts (

he

alt

h

care

)

Ass

um

pti

on

of

dru

g co

sts,

oth

er

cost

s co

me

fro

m

nat

ion

al a

nd

C

alif

orn

ian

da

ta

Stra

tegy

-sp

eci

fic

CE

ou

tco

me

Joh

ann

ess

on

(3

4)(

20

01

),

Swe

de

n

19

99

, US

$

To

est

imat

e a

5-y

ear

co

ron

ary

risk

at

wh

ich

cho

lest

ero

l lo

we

rin

g w

ith

sta

tin

s is

co

st-

eff

ect

ive

CU

A

(QA

LY)

3;

3

Pu

blis

he

d

stu

die

s C

ost

s o

f d

rug,

d

ise

ase

-sp

eci

fic

cost

s, c

ost

s d

ue

to

p

rod

uct

ivit

y lo

ss

(so

cie

tal)

Pu

blis

he

d s

tud

ies

Fixe

d C

E o

utc

om

e

Mo

ntg

om

ery

(3

3)(

20

03

),

UK

20

02

, UK

£

CU

A o

f B

P lo

we

rin

g co

mp

are

d t

o n

o

inte

rve

nti

on

CU

A

(QA

LY)

6;

1.5

P

ub

lish

ed

st

ud

ies

Co

sts

of

tre

atm

en

t,

cost

s o

f

card

iova

scu

lar

eve

nts

(h

ea

lth

Pu

blis

he

d s

tud

ies

Stra

tegy

-sp

eci

fic

CE

ou

tco

me

Man

iad

akis

(3

7)(

20

11

),

Gre

ece

20

08

, €

CU

A o

f h

ype

rte

nsi

on

tr

eat

me

nt

wit

h

losa

rtan

, val

sart

an a

nd

ir

be

sart

an w

he

n

com

bin

ed

wit

h

hyd

roch

loro

thia

zid

e

CU

A

(QA

LY)

3;

3

RC

Ts

Co

sts

of

dru

gs,

cost

s o

f C

VD

eve

nts

, co

sts

of

mai

nte

nan

ce o

f p

atie

nts

po

sth

osp

ital

d

isch

arge

(h

eal

th

care

)

De

rive

d f

rom

n

atio

nal

dat

abas

es

ICE

Rs

no

t p

rese

nte

d

Gro

ver

(38

)(2

00

8),

Can

ada

20

02

, C

anad

ian

$

CE

A o

f an

tih

ype

rte

nsi

ve a

nd

lipid

-lo

we

rin

g tr

eat

me

nt

advo

cate

d

by

Can

adia

n g

uid

elin

es

CE

A (

YO

LS)

3;

3

RC

Ts

Co

sts

of

dru

gs,

cost

s o

f

card

iova

scu

lar

eve

nts

(h

ea

lth

care

)

Re

sou

rce

use

fr

om

nat

ion

al

surv

eys

16

70

0/Y

OLS

fo

r st

atin

s

37

10

0/Y

OLS

fo

r an

tih

ype

rte

nsi

ves

Sait

o

(39

)(2

00

5),

2

00

4(2

00

3),

Ja

pan

ese

C

EA

of

dif

fere

nt

anti

hyp

ert

en

sive

dru

g C

EA

(Y

OLS

) 3

; 3

R

evi

ew

an

d

pu

blis

he

d

Co

sts

of

dru

gs,

dis

eas

e-s

pe

cifi

c P

ub

lish

ed

stu

die

s IC

ER

no

t p

rese

nte

d

Chapter2

34

Mai

n a

uth

or

(ye

ar),

co

un

try

Ye

ar o

f

valu

es,

cu

rre

ncy

Stu

dy

ob

ject

ive

Eva

luat

ion

typ

e

(ou

tco

me

m

eas

ure

)

Dis

cou

nt

rate

(c

ost

s;

he

alth

be

ne

fits

)

Sou

rce

of

eff

icac

y/

eff

ect

ive

ne

ss

dat

a

Co

st c

om

po

ne

nts

(stu

dy

pe

rsp

ect

ive

)

Sou

rce

of

cost

dat

a

Re

sult

s

Jap

an

yen

re

gim

es

stu

die

s co

sts,

o

utp

atie

nt

man

agem

en

t (h

eal

th c

are

)

Lofr

oth

(3

5)(

20

06

),

Swe

de

n

20

00

, €

CU

A o

f an

tih

ype

rte

nsi

ve, l

ipid

lo

we

rin

g o

r sm

oki

ng

cess

atio

n in

terv

en

tio

ns

wit

h t

he

aim

of

op

tim

al

reso

urc

e a

lloca

tio

n

CU

A

(QA

LY)

3;

3

Re

vie

w a

nd

p

ub

lish

ed

st

ud

ies

Co

sts

of

dru

gs,

cost

s o

f an

ti-

smo

kin

g

cou

nse

ling,

d

ise

ase

-sp

eci

fic

cost

s, c

ost

s d

ue

to

p

rod

uct

ion

loss

(s

oci

eta

l)

Pu

blis

he

d

sou

rce

s St

rate

gy-s

pe

cifi

c C

E o

utc

om

e

US,

Un

ite

d S

tate

s; C

UA

, co

st-u

tilit

y an

alys

is;

CV

D,

card

iova

scu

lar

dis

eas

es;

QA

LY,

qu

alit

y-o

f-lif

e-y

ear

; P

PI,

pro

ton

-pu

mp

in

hib

ito

rs;

CH

D,

coro

nar

y h

ear

t d

ise

ase

; G

I,

gast

ro-i

nte

stin

al;

RC

T,

ran

do

miz

ed

clin

ical

tri

als;

CU

, co

st-u

tilit

y; E

UR

, e

uro

; U

K,

Un

ite

d K

ingd

om

; N

HS,

Nat

ion

al H

eal

th S

erv

ice

; A

TP

, ad

ult

tre

atm

en

t p

ane

l; C

E,

cost

-e

ffe

ctiv

en

ess

; B

P, b

loo

d p

ress

ure

, IC

ER, i

ncr

em

en

tal c

ost

-eff

ect

ive

ne

ss r

atio

; C

EA

, co

st-e

ffe

ctiv

en

ess

an

alys

is;

YO

LS, y

ear

s o

f lif

e s

ave

d.

2


35

Table 2 Study design of economic evaluations on cardiovascular diseases (CVD) primary prevention with pharmacological treatment. Studies utilizing risk prediction models.

CVD, cardiovascular diseases; PSA, probabilistic sensitivity analysis; CHD, coronary heart disease; LRC, Lipid Research Clinic follow-up cohort *Validated in White and African-American

Study Modeling

(time horizon)

CVD risk model Risk model time

span

Risk model

validated / recalibrated

PSA of risk

model parameters

Pignone et al.(28) Markov

model (Lifetime)

Framingham

based risk (4,10,22)

4-12 years (CHD);

10-years (stroke)

Yes* No

Ernshaw et al.(29)

Markov model (Lifetime)

Framingham based risk(4)

4-12 years Yes* No

Brennan et al.(36)

Decision tree (5-years)


4-12 years No No

Caro et al.(32) Markov

model (Lifetime)

Framingham

based risk(4)

4-12 years No No

Pletcher et al.(30) CHD Policy Model44 (Lifetime)


Not-stated Yes Yes

Lazar et al. (31) CHD Policy Model44

(Lifetime)


Not-stated Yes No

Johannesson (34) Markov

model (Lifetime)

Framingham

equations(16)

8-years No No

Montgomery et al.(33)

Markov model (Lifetime)


4-12 years No No

Maniadakis et al.(37)

Markov model

(Lifetime)

Glynn et al. equation(11)

5-years No No

Grover et al.(38) CVD Life

Expectancy Model(46) (Lifetime)

Logistic regression

model based on the LRC(47,51)

10-years Yes No

Saito et al.(39) Markov model

(Lifetime)

Adjusted Framingham

based risk(4)

4-12 years Yes No

Lofroth et al.(35) Markov

model (Lifetime)

Functions

developed on Swedish registers(48)

10-years No No

Chapter 2

36

DISCUSSION

This systematic review of pharmacoeconomic analyses of primary CVD prevention

strategies in high income countries identified 12 studies that have used published CVD risk

prediction models. The majority of these prediction models were based on a Framingham

population and focused on the CHD and stroke risk. The studies demonstrated the

usefulness of projecting intermediate effectiveness endpoints to long term, health and

cost related, benefits. However, after assessing the quality of incorporation of these risk

models in pharmacoeconomic analyses we identified a distance between the populations

of the risk model and the study population, a frequent disagreement between risk model-

and study- time horizons and a lack of proper consideration of the uncertainty

surrounding the prediction of the risk.

In order for a risk model to be able to predict accurately the CVD events of a study

population, there must be a relative agreement between the study and the risk model

populations with respect to the main CVD risk factors (24). In the reviewed studies, this

criterion was generally met. However, characteristics that have been earlier identified as

influential on CVD risk and relate to temporal, regional, demographic and socio-economic

characteristics have not been adequately assessed in the reviewed studies. For example,

the Framingham population, which was widely used among the reviewed studies, has

been earlier found to overestimate CVD risk in the western world, especially in lower

socio-economic populations (55,59,60). Similar problems can be encountered when the

risk model is based on special populations, such as populations from RCTs. One such

example is that of Glynn et al. where the risk model population consisted of health care

providers, who cannot be considered as representative of the general population (11).

Pharmacoeconomic analyses generally require the estimation of both costs and effects for

a lifetime horizon. In this review, most studies were characterized by a disagreement

between the risk model and the decision-analytic model horizons. Specifically, risk models

provided a considerably smaller time horizon, forcing the authors to project the estimated

risks for time horizons for which the risk models were not validated. However, increasingly

more risk predicting models that are developed on cohorts with longer follow up times are

becoming available. The risk model by Pencina et al., which is based on the Framingham

population and can estimate a 30-year CVD risk, might be a valid tool for a more accurate

estimation of lifetime risk, which is necessary for pharmacoeconomic analyses (19). One

other such long-term risk model is the QRISK2 model (12), developed using general

practitioner collected data from the UK, which can be utilized for the estimation of

lifetime CHD and stroke risk in the general population.

The usage of a CVD risk model to project the effect of a pharmacological intervention from

risk factors to CVD events provides an estimate of the intervention's clinical and

2


37

economical impact. However, this projection introduces an additional uncertainty level

associated with the accuracy of the CVD prediction on the specific population. Although

pharmacoeconomic evaluations require this uncertainty to be studied together with the

overall model uncertainty, this has been rarely observed in our review. The means of

incorporation of this uncertainty has been described in detail in the literature and involve

the consideration of the level of uncertainty and correlation of the parameters of the risk

model (27).

Concluding, we provide some suggestions with respect to the appropriate use of a CVD

risk model in pharmacoeconomic analysis. First, incorporating a risk model into a

pharmacoeconomic analysis must be done after the compatibility of the risk model and

the study populations has been thoroughly investigated. Additionally, the study should

report the method used for the translation of cumulative risks to short-term probabilities

and should clearly describe the method behind extrapolating this risk to a lifetime horizon.

Finally, the uncertainty introduced due to CVD risk prediction should be incorporated in a

transparent way in the estimation of the overall uncertainty of the decision-analytic

model, as this prediction parameter is one of the main contributors of uncertainty.

Appendix

General description of selected studies

Pignone et al.(28) performed a cost-utility (CUA) of the use of aspirin for primary

prevention compared to no intervention in 65-year-old women in a USA setting. For this

purpose, they developed a Markov model that incorporated disease-specific health states

and health states related to adverse effects of aspirin. The cardiovascular diseases (CVD)

risk prediction models used were based on the Framingham Heart Study (4,10,22). The

cost-effectiveness (CE) of aspirin treatment was modelled in a lifetime horizon. The

authors concluded that preventive use of aspirin is beneficial in women with a higher risk

for ischemic stroke but potentially harmful for women with a relatively low stroke risk.

A CUA was conducted by Ernshaw et al.(29) to compare aspirin alone or in combination

with a proton-pump inhibitor (PPI) with no treatment, in primary CVD prevention. The

authors used a Markov model similar to the one of Pignone et al. (28) However, their

pharmacoeconomic study was focused on a population of 45-year-old men. The CVD risk

prediction was based on the Framingham Heart Study. Results were explored in time

horizons of 5, 10, 25 years and lifetime. Higher incremental cost-utility ratios were

observed for shorter time horizons. The authors concluded that CVD prevention with

aspirin alone in 45-year-old men was less costly and more effective than no treatment.

Brennan et al.(36) investigated the cost-utility of sibutramine combined with diet and

lifestyle advice in obese patients, in comparison with a non-pharmacological intervention.

They constructed a decision tree model where they modelled weight loss after one year of

drug treatment and four years of follow up. Two Framingham based risk models (5,61)

Chapter 2

38

were used to translate weight loss into coronary heart disease (CHD) risk reduction. Their

conclusion was that, for German, obese patients, sibutramine is more cost-effective

against non-pharmacological interventions.

Caro et al.(32) explored the cost-utility of adding rimonabant to exercise and diet,

compared to exercise and diet alone, in obese patients with and without diabetes. Since

our focus is on primary prevention in non-diabetic populations, we incorporated in our

review only the part that applies to non-diabetic patients. The authors designed a one-

month cycle Markov model with three CVD-related health states and a lifetime horizon.

Framingham risk models (4) were used for calculating the primary CVD risk, as a composite

of the risk for CHD and stroke. Given that those models do not incorporate body mass

index (BMI) as a risk factor, the authors incorporated in the estimation of the risk of CHD

an additional risk for patients with high BMI (62).The authors concluded that adding

rimonabant to exercise and diet seems to be cost-effective.

Pletcher et al.(30) explored the cost-utility of statin treatment adherent to the Adult

Treatment Panel III (ATP III) guidelines (63) against various other risk- and aged- based

statin treatment strategies. For this purpose, they utilized an updated version of the,

previously published, CHD Policy Model (44). This is a Markov model based on a USA

population older than 35 years of age. Risk functions for incident CHD and non-CHD

related death were based on longitudinal data from the Framingham Heart Study (45). The

cost-utility estimates were calculated for a 30-year time horizon. The main effect of statin

use was the reduction of low-density lipoprotein (LDL), with a subsequent estimated

effect on CHD risk reduction. Pletcher et al.(30) concluded that lipid-lowering strategies

adherent to the ATP III guidelines are relatively more cost-effective compared to other

risk- and age- guided alternatives.

A CUA by Lazar et al.(31) assessed primary prevention with low-cost generic statins

compared with no intervention and examined under which circumstances their

application is cost-effective. For this purpose, they utilized the CHD Policy Model (44).

Lazar et al. (31) incorporated statin-induced diabetes as an additional possible adverse

effect to statin treatment, next to myopathy and hepatitis. Statin induced diabetes was

directly incorporated in the model, given that it is considered as one of the CHD risk

factors. Lazar et al. (31) concluded that statin use in people with even modestly elevated

LDL or any CHD risk factor could be considered as a cost-effective option.

Johannesson conducted a study to estimate the 5-year coronary risk level at which statin

application is considered to be cost-effective in Sweden (34). He utilized a previously

published Markov model for cost-effectiveness analysis (CEA) in cardiovascular prevention

(64). In this model, transition probabilities to each of the CHD states were estimated

through logistic risk models from a study based on the Framingham population (16). In

order to assess the coronary risk level at which statin use was cost-effective, the author

2


39

varied the CHD risk level until the cost per QALY reached a certain willingness-to-pay

threshold.

Montgomery et al. explored the cost-utility of an antihypertensive intervention compared

with no intervention in cohorts of 20 different strata of age, gender and high or low CVD

risk (33). They used a Markov model in which CVD-related transition probabilities were

estimated through a Framingham Heart study (4). The authors modified the estimated

baseline CVD risk depending on the probability of successful systolic blood pressure (SBP)

control. According to Montgomery et al. (33), estimates of cost-utility ratios were higher

for low-risk women compared with men, while high-risk patients had more favourable

cost-utility ratios than low-risk patients.

A CUA of hypertension treatment among combinations of different angiotensin-receptor

blockers (ARBs) and hydrochlorothiazide (HCTZ) in a Greek setting was performed by

Maniadakis et al. (37). The authors designed a Markov model with eight CVD-specific

health states for a lifetime horizon. The risk models developed by Glynn et al., in

combination with gender- and disease-specific incidence rates were utilized to model

transitions between CVD states (11). The authors concluded that application of irbesartan

in combination with HCTZ was the most favourable compared with the other ARBs.

Grover et al. performed a CEA of antihypertensive and lipid-lowering treatment as

recommended by the 2003 Canadian Working Group guidelines, in comparison with no

treatment (38). They utilized the Cardiovascular Disease Life Expectancy (CVDLE) Model

(48). Probabilities of CVD events in this model were based on logistic regression functions

developed using data from the Lipid Research Clinic cohort (LRC)(49,50). The authors

concluded that lipid-lowering and antihypertensive therapies in accordance with the

Canadian guidelines are cost-effective in most of the age-gender groups of patients.

A CEA in a Japanese setting concerning different antihypertensive drug regimes was

conducted by Saito et al.(39). The authors utilized a Markov model in which a hypothetical

cohort of 55-year old patients with essential hypertension was followed lifelong and

incidence of CHD and stroke was monitored (65). Transition probabilities to CHD states

were calculated through adjusted Framingham risk models (4). These risk models were

adjusted to a Japanese setting under the assumption that the incidence of CHD in Japan is

20% of that in USA. Saito et al. did not observe any relative advantage in terms of costs

and expected survival among the applied drug regimes.

Lofroth et al. performed a CUA of primary CVD prevention through a combination of

antihypertensive, lipid-lowering or smoking cessation interventions, with the aim of

optimal resource allocation within the Swedish health care system (35). The authors

designed a Markov model with disease-specific health states reflecting myocardial

infarction (MI) and stroke. The model was run through a lifetime horizon. In order to

calculate the risk of MI and stroke, the authors developed risk functions that were based

on the Swedish Nora study and the Swedish Hospital Patients Discharge Register and

Chapter 2

40

Cause of Death Register (51). The authors based the 1-year risk for MI and stroke on a

combination of the annual age- and gender- specific incidence and a number of risk

factors (cholesterol, blood pressure, smoking). The study displayed that resources in the

studied regions are inefficiently allocated and should be primarily reallocated towards

smoking cessation interventions and lipid-lowering therapies.

2


41

REFERENCES

(1) Graham I, Atar D, Borch-Johnsen K, et al. European guidelines on cardiovascular disease prevention in clinical

practice: executive summary. Fourth Joint Task Force of the European Society of Cardiology and other societies

on cardiovascular disease prevention in clinical practice (constituted by representatives of nine societies and by

invited experts). Eur J Cardiovasc Prev Rehabil 2007 Sep;14 Suppl 2:E1-40.

(2) Roger VL, Go AS, Lloyd-Jones DM, et al. Executive summary: heart disease and stroke statistics--2012 update:

a report from the american heart association. Circulation 2012 Jan 3;125(1):188-197.

(3) Leal J, Luengo-Fernandez R, Gray A, et al. Economic burden of cardiovascular diseases in the enlarged

European Union. Eur Heart J 2006 Jul;27(13):1610-1619.

(4) Anderson KM, Odell PM, Wilson PWF, et al. Cardiovascular disease risk profiles. Am Heart J 1991;121(1):293-

298.

(5) Anderson KM, Wilson P, Odell PM, et al. An updated coronary risk profile. A statement for health

professionals. Circulation 1991;83(1):356.

(6) Assmann G, Cullen P, Schulte H. Simple scoring scheme for calculating the risk of acute coronary events based

on the 10-year follow-up of the prospective cardiovascular Münster (PROCAM) study. Circulation

2002;105(3):310-315.

(7) Conroy R, Pyörälä K, Fitzgerald AP, et al. Estimation of ten-year risk of fatal cardiovascular disease in Europe:

the SCORE project. Eur Heart J 2003;24(11):987.

(8) D’Agostino RB, Vasan RS, Pencina MJ, et al. General cardiovascular risk profile for use in primary care.

Circulation 2008;117(6):743-753.

(9) D'Agostino RB, Russell MW, Huse DM, et al. Primary and subsequent coronary risk appraisal: new results from

the Framingham study. Am Heart J 2000;139(2):272-281.

(10) D'Agostino RB, Wolf PA, Belanger AJ, et al. Stroke risk profile: adjustment for antihypertensive medication.

The Framingham Study. Stroke 1994;25(1):40-43.

(11) Glynn RJ, L’Italien GJ, Sesso HD, et al. Development of predictive models for long-term cardiovascular risk

associated with systolic and diastolic blood pressure. Hypertension 2002;39(1):105-110.

(12) Hippisley-Cox J, Coupland C, Robson J, et al. Derivation, validation, and evaluation of a new QRISK model to

estimate lifetime risk of cardiovascular disease: cohort study using QResearch database. BMJ: British Medical

Journal 2010;341.

(13) Hippisley-Cox J, Coupland C, Vinogradova Y, et al. Derivation and validation of QRISK, a new cardiovascular

disease risk score for the United Kingdom: prospective open cohort study. BMJ 2007;335(7611):136.

(14) Hippisley-Cox J, Coupland C, Vinogradova Y, et al. Predicting cardiovascular risk in England and Wales:

prospective derivation and validation of QRISK2. BMJ 2008;336(7659):1475-1482.

(15) Kannel WB, D'Agostino RB, Silbershatz H, et al. Profile for estimating risk of heart failure. Arch Intern Med

1999;159(11):1197.

(16) Kannel W, Gordon T. The probability of developing certain cardiovascular diseases in eight years at specified

values of some characteristics. The Framingham Study: An epidemiological investigation of cardiovascular

disease.Springfield, MA: US Department of Commerce National Technical Information Service 1987.

(17) Lloyd-Jones DM, Leip EP, Larson MG, et al. Prediction of lifetime risk for cardiovascular disease by risk factor

burden at 50 years of age. Circulation 2006;113(6):791-798.

(18) Merry AHH, Boer J, Schouten LJ, et al. Risk prediction of incident coronary heart disease in the Netherlands:

re-estimation and improvement of the SCORE risk function. European Journal of Cardiovascular Prevention &

Rehabilitation 2011.

(19) Pencina MJ, D'Agostino RB, Larson MG, et al. Predicting the 30-Year Risk of Cardiovascular Disease.

Circulation 2009;119(24):3078-3084.

(20) Ridker PM, Buring JE, Rifai N, et al. Development and validation of improved algorithms for the assessment

of global cardiovascular risk in women. JAMA: The Journal of the American Medical Association 2007;297(6):611.

Chapter 2

42

(21) Ridker PM, Paynter NP, Rifai N, et al. C-reactive protein and parental history improve global cardiovascular

risk prediction. Circulation 2008;118(22):2243-2251.

(22) Wolf PA, D'Agostino RB, Belanger AJ, et al. Probability of stroke: a risk profile from the Framingham Study.

Stroke 1991;22(3):312-318.

(23) Woodward M, Brindle P, Tunstall-Pedoe H. Adding social deprivation and family history to cardiovascular risk

assessment: the ASSIGN score from the Scottish Heart Health Extended Cohort (SHHEC). Heart 2007;93(2):172.



(25) Sculpher M, Claxton K. Establishing the Cost-Effectiveness of New Pharmaceuticals under Conditions of

Uncertainty—When Is There Sufficient Evidence? Value in Health 2005;8(4):433-446.

(26) World Bank. Data and statistics. Country group. Available at: http://data.worldbank.org/about/country-

classifications. Accessed 02/28, 2012.

(27) Briggs AH, Claxton K, Sculpher MJ. Decision modelling for health economic evaluation. Oxford: Oxford

University Press; 2006.

(28) Pignone M, Earnshaw S, Pletcher MJ, et al. Aspirin for the primary prevention of cardiovascular disease in

women: a cost-utility analysis. Arch Intern Med 2007 Feb 12;167(3):290-295.

(29) Earnshaw SR, Scheiman J, Fendrick AM, et al. Cost-utility of aspirin and proton pump inhibitors for primary

prevention. Arch Intern Med 2011 Feb 14;171(3):218-225.

(30) Pletcher MJ, Lazar L, Bibbins-Domingo K, et al. Comparing impact and cost-effectiveness of primary

prevention strategies for lipid-lowering. Ann Intern Med 2009;150(4):243-254.

(31) Lazar LD, Pletcher MJ, Coxson PG, et al. Cost-effectiveness of statin therapy for primary prevention in a low-

cost statin era. Circulation 2011 Jul 12;124(2):146-153.

(32) Jaime Caro J, Ozer Stillman I, Danel A, et al. Cost effectiveness of rimonabant use in patients at increased

cardiometabolic risk: estimates from a Markov model. Journal of Medical Economics 2007;10(3):239-254.

(33) Montgomery AA, Fahey T, Ben-Shlomo Y, et al. The influence of absolute cardiovascular risk, patient utilities,

and costs on the decision to treat hypertension: a Markov decision analysis. J Hypertens 2003 Sep;21(9):1753-

1759.

(34) Johannesson M. At what coronary risk level is it cost-effective to initiate cholesterol lowering drug treatment

in primary prevention? Eur Heart J 2001 Jun;22(11):919-925.

(35) Lofroth E, Lindholm L, Wilhelmsen L, et al. Optimising health care within given budgets: primary prevention

of cardiovascular disease in different regions of Sweden. Health Policy 2006;75(2):214-229.

(36) Brennan A, Ara R, Sterz R, et al. Assessment of clinical and economic benefits of weight management with

sibutramine in general practice in Germany. Eur J Health Econ 2006 Dec;7(4):276-284.

(37) Maniadakis N, Ekman M, Fragoulakis V, et al. Economic evaluation of irbesartan in combination with

hydrochlorothiazide in the treatment of hypertension in Greece. Eur J Health Econ 2011 Jun;12(3):253-261.

(38) Grover S, Coupal L, Lowensteyn I. Preventing cardiovascular disease among Canadians: Is the treatment of

hypertension or dyslipidemia cost-effective? Can J Cardiol 2008;24(12):891.

(39) Saito I, Kobayashi M, Matsushita Y, et al. Pharmacoeconomical evaluation of combination therapy for

lifetime hypertension treatment in Japan. JAPAN MEDICAL ASSOCIATION JOURNAL 2005;48(12):574.

(40) Genest J, McPherson R, Frohlich J, et al. 2009 Canadian Cardiovascular Society/Canadian guidelines for the

diagnosis and treatment of dyslipidemia and prevention of cardiovascular disease in the adult–2009

recommendations. Can J Cardiol 2009;25(10):567.

(41) New Zealand Guidelines Group. The assessment and management of Cardiovascular risk. 2003; Available at:

http://www.nzgg.org.nz/. Accessed 02/28, 2012.

(42) D'Agostino RB, Grundy S, Sullivan LM, et al. Validation of the Framingham coronary heart disease prediction

scores. JAMA: the journal of the American Medical Association 2001;286(2):180.

(43) Ward S, National Co-ordinating Centre for HTA (Great Britain), NHS R & D HTA Programme (Great Britain). A

systematic review and economic evaluation of statins for the prevention of coronary events. Tunbridge Wells:

Gray Pub. on behalf of NCCHTA; 2007.

2


43

(44) Weinstein MC, Coxson PG, Williams LW, et al. Forecasting coronary heart disease incidence, mortality, and

cost: the Coronary Heart Disease Policy Model. Am J Public Health 1987;77(11):1417.

(45) Framingham Heart Study [CD-ROM]. Washington, DC: Department of health and human services. 2005.

(46) Lee IM, Cook NR, Manson JAE, et al. β-Carotene supplementation and incidence of cancer and

cardiovascular disease: the Women's Health Study. J Natl Cancer Inst 1999;91(24):2102-2106.

(47) Steering Committee of the Physicians' Health Study Research Group. Final report on the aspirin component

of the ongoing Physicians' Health Study. N Engl J Med 1989;321:129-135.

(48) Grover SA, Paquet S, Levinton C, et al. Estimating the benefits of modifying risk factors of cardiovascular

disease: a comparison of primary vs secondary prevention. Arch Intern Med 1998;158(6):655.

(49) Heiss G, Tamir I, Davis CE, et al. Lipoprotein-cholesterol distributions in selected North American

populations: the lipid research clinics program prevalence study. Circulation 1980;61(2):302.

(50) Lipid Research Clinics Program Epidemiology Committee. Plasma lipid distributions in selected North

American populations: the Lipid Research Clinics Program prevalence study. Circulation 1979;60(2):427-439.

(51) Dobson AJ, Evans A, Ferrario M, et al. Changes in estimated coronary risk in the 1980s: data from 38

populations in the WHO MONICA Project. Ann Med 1998;30(2):199-205.

(52) Fleurence RL, Hollenbeak CS. Rates and probabilities in economic modelling: transformation, translation and

appropriate application. Pharmacoeconomics 2007;25(1):3-6.

(53) Brindle P, Beswick A, Fahey T, et al. Accuracy and impact of risk assessment in the primary prevention of

cardiovascular disease: a systematic review. Heart 2006;92(12):1752-1759.

(54) Cooney MT, Dudina A, D'Agostino R, et al. Cardiovascular risk-estimation systems in primary prevention.

Circulation 2010;122(3):300-310.

(55) Brindle P, Jonathan E, Lampe F, et al. Predictive accuracy of the Framingham coronary risk score in British

men: prospective cohort study. BMJ 2003;327(7426):1267.

(56) Hense HW, Schulte H, Löwel H, et al. Framingham risk function overestimates risk of coronary heart disease

in men and women from Germany—results from the MONICA Augsburg and the PROCAM cohorts. Eur Heart J

2003;24(10):937.

(57) Neuhauser HK, Ellert U, Kurth BM. A comparison of Framingham and SCORE-based cardiovascular risk

estimates in participants of the German National Health Interview and Examination Survey 1998. J Cardiovasc

Risk 2005;12(5):442.

(58) Grover SA, Hemmelgarn B, Joseph L, et al. The role of global risk assessment in hypertension therapy. Can J

Cardiol 2006;22(7):606.

(59) Quirke T, Gill P, Mant J, et al. The applicability of the Framingham coronary heart disease prediction function

to black and minority ethnic groups in the UK. Heart 2003;89(7):785.

(60) Brindle PM, McConnachie A, Upton MN, et al. The accuracy of the Framingham risk-score in different

socioeconomic groups: a prospective study. Br J Gen Pract 2005 Nov;55(520):838-845.

(61) Hubert HB, Feinleib M, McNamara PM, et al. Obesity as an independent risk factor for cardiovascular

disease: a 26-year follow-up of participants in the Framingham Heart Study. Circulation 1983;67(5):968-977.

(62) Mora S, Yanek LR, Moy TF, et al. Interaction of body mass index and Framingham risk score in predicting

incident coronary disease in families. Circulation 2005;111(15):1871-1876.

(63) Grundy SM, Cleeman JI, Bairey Merz CN, et al. Implications of recent clinical trials for the national

cholesterol education program adult treatment panel III guidelines. J Am Coll Cardiol 2004;44(3):720-732.

(64) Johannesson M, Hedbrant J, Jönsson B. A computer simulation model for cost-effectiveness analysis of

cardiovascular disease prevention. Informatics for Health and Social Care 1991;16(4):355-362.

(65) Saito I, Kobayashi K, Saruta T. Economic analysis of antihypertensive agents in treating patients with

essential hypertension. J Clin Ther Med 2003;19:777-788.

Chapter 3

Economic evaluation of primary prevention of

cardiovascular diseases in mild hypertension.

A scenario analysis for the Netherlands

Jelena Stevanović, Anouk C O’Prinsen, Bram G Verheggen, Nynke Schuiling-Veninga,

Maarten J Postma, Petros Pechlivanoglou

Clinical Therapeutics 2014;36(3);368-384.

Chapter 3

46

ABSTRACT

Background: In the Netherlands, antihypertensive treatment for patients with mild

hypertension is recommended if the 10-year cardiovascular disease (CVD) risk exceeds

20%. Recent evidence suggests that lifelong CVD risk estimates might be more informative

than 10-year ones. In addition, the cost of antihypertensive treatment in the Netherlands

has declined over the last decade. The aim of this study is to estimate the cost-

effectiveness of lowering systolic blood pressure (SBP) in patients ineligible for treatment,

both in a 10-year and in a lifetime horizon.

Methods: A Markov model was developed to assess the cost-effectiveness of SBP

reduction compared to no reduction in patients with mild hypertension and low CVD risk.

Modified SCORE risk estimates were used to predict fatal and non-fatal CVD events. We

analyzed scenarios for different age groups, genders and SBP reductions. Specifically, SBP

reductions due to hydrochlorothiazide (HCT) 25mg and hypothetical reductions with

HCT/Losartan combination were assumed. Parameter uncertainty was assessed through a

probabilistic sensitivity analysis.

Results: In a 10-year horizon, in scenarios of SBP reduction with HCT 25mg, the

incremental cost-effectiveness ratio (ICER) estimates for men varied across different ages

in the range of €6,032 to €58,217 per life year gained (LYG), while for women ICER

estimates were in the range of €12,345 to €361,064 per LYG. In a lifetime horizon, the

cost-effectiveness estimates were favorable for both genders. In scenarios of hypothetical

SBP reductions, more favorable ICER estimates compared to no reduction were found. A

large uncertainty around the cost-effectiveness estimates was observed among all

scenarios.

Conclusions: Larger SBP reductions were found to be cost-effective in both a 10-year and

lifetime horizon. These findings might call for more aggressive SBP reductions in patients

with mild hypertension. Yet, a high level of uncertainty surrounds these cost-effectiveness

estimates since they are based on CVD risk prediction modelling.

3

Economic evaluation of primary prevention of cardiovascular diseases in mild hypertension

47

INTRODUCTION

Cardiovascular disease (CVD) is the largest cause of morbidity and a major cause of

premature death and reduced quality of life in Europe. At the same time, CVD is the most

significant contributor to health care expenditures; treatment costs of CVD in EU comprise

12% of all health-care costs (€ 105 billion)(1), while in the Netherlands they reach 8% of

the total health-care budget (€ 5.5 billion)(2). Due to the fact that people with moderate

or high hypertension (>160mmHg), previous CVD events or diabetes mellitus (DM) are

generally considered to be at higher risk for CVD, preventive treatment in these groups is

rather straightforward and common (3,4). The question arises when considering

preventive treatment in patients with mild hypertension (140-160 mmHg), no prior CVD

events or DM, and overall low CVD risk (4), given the economic implications and ethical

considerations of such a decision. Currently, national guidelines recommend lifestyle

changes as a first step in lowering blood pressure in this patient population (5). Failure of

this intervention should be followed initially with diuretic treatment (e.g.

hydrochlorothiazide (HCT)) and if necessary combinations of diuretic and Angiotensin

converting enzyme inhibitor (ACEi) or Angiotensin receptor blocker (ARB) (5). In parallel,

the cost of these antihypertensive agents has declined considerably the last years in the

Netherlands, primarily due to generic introduction (6), causing a reduction on the cost of

treating patients with hypertension. Guideline recommendations on preventive treatment

of CVD in people with neither prior events nor DM necessitate the estimation of a patient-

specific risk for fatal or non-fatal CVD (3-5,7-9). Several studies have focused on identifying

and correcting for risk modifiers (e.g. reducing systolic blood pressure (SBP) or cholesterol

level) that can be beneficial for cardiovascular risk reduction (10-15). Considering the

overall prevalence of hypertension and the fact that elevated SBP is a major

cardiovascular risk factor, considerable health and economic benefits are expected from

its control (16-19).

The effective control of SBP through antihypertensive medication has been established

through numerous clinical trials (20). However, the effectiveness of these drugs often

varies across trials. Under these circumstances, the proper consideration of the exact long

term health and economic consequences of SBP reduction in patients with mild

hypertension and no prior CVD events requires evidence-synthesized estimates of SBP

reduction (e.g. through meta-analysis and mixed-treatment comparisons)(21,22). Such

estimates are currently available only for older antihypertensive agents, such as diuretics

(20) but not for newer antihypertensive agents such as ACEis and ARBs.

The aim of this study is to estimate the health effects and economic consequences of SBP

reduction through antihypertensive treatment in patients with mild hypertension but with

no history of CVD or diabetes. SBP reduction was used as a surrogate endpoint that was

expected to, consequently, result in a CVD risk reduction. As we are primarily interested in

Chapter 3

48

the impact of such an intervention in the Dutch setting, we utilized a CVD risk calculation

model recently validated for this population; i.e. the low-risk version of SCORE (23,24).

Furthermore, we applied cost estimates for cardiovascular complications as well as drug

costs reflecting the Dutch situation. We investigated the cost-effectiveness (CE) of

different scenarios of SBP reductions achieved by antihypertensive treatment in different

age groups of men and women, in comparison to no treatment, both in a 10-year and in a

lifetime horizon.

METHODS

Model structure

A Markov model was developed to compare the long-term costs and health benefits of

primary CVD prevention through SBP reduction. The model included five health states:

baseline (healthy with mild hypertension but no CVD history), acute non-fatal CVD, stable

non-fatal CVD, fatal CVD and non-CVD related death (Figure 1). Patients remained in the

baseline state until a fatal or non-fatal CVD event occurred or they died due to other, non-

CVD related, causes. From the acute CVD state at the end of one cycle, patients could

move to a stable non-fatal CVD state, experience a subsequent, fatal or non-fatal, CVD

event or die from other causes. From the stable CVD state, patients could experience a

subsequent fatal or non-fatal CVD event or die from other causes. The model allows for

multiple recurrent non-fatal CVD events to occur with a restriction to one event per cycle.

Transition probabilities between health states were modelled through reparametrizing the

10-year risks of CVD mortality and morbidity and non-CVD related mortality into

respective one-year probabilities as given in the Appendix.

The simulations were run using two time horizons: 10-year and lifetime, both applying

cycles of one year in the Markov model. All patients were assumed to be dead by the age

of 100 years. The model was developed using the statistical software R (version 3.0.2)(25).

Figure 1 Markov model on the progression of CVD in patients with mild hypertension. CVD, cardiovascular disease

3


49

Scenario analysis

The Markov model was applied in a set of different scenarios separately for men and

women, where the population analyzed varied with respect to its age (age groups: 40, 50,

60, 65), baseline SBP (hypertensive groups of SBP at 140 mmHg, 150 mmHg, 160 mmHg)

and level of SBP reduction achieved by the treatment. For the base-case analysis,

antihypertensive treatment with HCT 25 mg was assumed. According to a recent meta-

analysis, SBP reduction achieved through HCT 25 mg, in a similar population, was

estimated on average to be 11.3 mmHg (20). In the absence of evidence-synthesized

estimates on SBP reduction achieved by combinations of HCT and ACEis or ARBs,

hypothetical scenarios of SBP reductions (20 and 30 mmHg) achieved with a fixed-dose

combination of HCT 12.5 mg/Losartan 50 mg were assumed. All comparisons were made

against a no treatment alternative. Additionally, a scenario analysis focusing on male and

female smokers, examined the cost-effectiveness of a 10-year treatment with HCT 25mg

against no treatment in this patient population.

It was assumed that antihypertensive medication was the sole cause of SBP reduction.

Additionally, SBP reduction was assumed to be instant and remain constant throughout

the time horizon. Also, every patient with elevated SBP was immediately diagnosed and

treated (in the treatment scenario) or left untreated (in the no treatment scenario).

Finally, the base-case scenario was adjusted to incorporate patients’ adherence to HCT.

Adherence was defined as the percentage of prescribed dosage taken over a one year

period, and was assumed to be 78.6% (26). Assuming that poorly adherent patients would

have lower benefits from the treatment with HCT, the level of adherence was modelled

through a reduction in the antihypertensive effect of HCT (i.e. 11.3 × 0.786 = 8.8818) and

drug treatment related costs (i.e. drug costs and pharmacy fee)(27)(28,29). All the patients

in the treatment alternative were assumed to be fully persistent with the treatment.

Transition probabilities: CVD mortality risks

In order to calculate the probability of a fatal CVD event, the updated SCORE model for

low-risk regions was utilized (23) (personal communication: Fitzgerald T, Department of

Epidemiology & Public Health/Statistics, University College Cork). The low-risk SCORE

model has been shown to produce better prediction estimates for the Dutch population in

comparison to both the high-risk SCORE and the so-called SCORE-NL models (3,24). The

SCORE model only gives the 10-year cumulative probability of a fatal CVD event. However,

due to the use of 1-year-cycles in our model, annual transition probabilities were needed.

Annual transition probabilities were obtained for every time point and for each of the

scenarios by estimating the age- and SBP- specific CVD event probability conditional on

the patient’s survival up to this time point (30). In other words, one’s probability to have

an event in a certain year is dependent on whether the patient has survived until that year

without having any events. A detailed description of the process and assumptions on

Chapter 3

50

transforming the SCORE cumulative probabilities to annual probabilities is provided in the

Appendix.

The explanatory covariates used in the estimation of fatal CVD risk through the SCORE

model are age, gender, level of SBP, level of cholesterol and prevalence of smoking.

Therefore, age-specific prevalence of smoking and cholesterol levels that were

representative of the Dutch population were used (Table 1)(31,32). These baseline patient

characteristics were updated through the time horizon observed in order to account for

the time dependent change in the risk of CVD events. Finally, it was assumed that patients

with a non-fatal CVD event were at least as likely to experience a subsequent event as

patients with no CVD history.

Transition probabilities: CVD morbidity risks

Given that the SCORE model only estimates the probability of a fatal CVD event, a

functional form linking the annual risk of mortality to the annual risk of morbidity and

mortality was estimated using the risk estimates presented by the Dutch College of

General Practitioners (NHG)(9).

)0003055.0()0202491.0()2767607.0(

001.0867.1083.72

++−= xxy ,

where y represents the annual overall risk of morbidity and mortality and x the annual

mortality risk. In order to obtain annual probabilities for non-fatal CVD events, the annual

probability of a fatal CVD event was subtracted from the respective estimated overall CVD

probability.

Transition probabilities: non-CVD mortality

The transition probabilities representing the annual risk of death from non-CVD causes

were estimated using 2010 Dutch population data (33-35). In particular, the non-CVD

related probabilities of death were estimated by subtracting the number of CVD deaths

from the total number of deaths in 5-year age groups and dividing them by the total

number of people in the same age groups. Since 5-year interval probabilities were

calculated, it was further assumed that these probabilities are representative for a person

of average age within each group. However, in the Markov model one-year cycles were

applied, hence estimates of age-dependent annual probabilities were necessary.

Therefore, the estimates of non-CVD related annual probabilities of death from the age of

40 until 100 were extrapolated and interpolated using non-linear regression modelling

(36).

Finally, in order to account for the increased risk of death in patients experiencing non-

fatal CVD events, a two-fold increase in both CVD and non-CVD mortality was applied

(27,37-39).

3


51

Table 1 Age-stratified baseline patient characteristics in Dutch population.

Age

SBP (mmHg)a32 Cholesterol (mmol/l)32

Smoking31 (%) SBP (mmHg) a32 Cholesterol (mmol/l)32

Smoking31 (%)

Men Women

40 126.10 5.41 36.3 115.10 5.05 27.4

50 131.67 5.57 33.9 123.16 5.52 28.8

60 138.87 5.55 29.3 131.13 5.77 24.3

65 141.66 5.50 24.1 134.99 5.79 19.3

70 144.44 5.45 18.9 138.86 5.81 14.3

80 147.04 5.35 16.1 143.37 5.76 11.1

90 148.26 5.26 16.1 145.69 5.71 11.1

100

148.26 5.26 16.1 145.69 5.71 11.1

aAge-stratified patient levels of SBP are used only for the purpose of model validation. SBP, systolic blood pressure.

Costs

The definition of CVD in the SCORE risk model comprises a variety of CVD outcomes. That

wide range of outcomes within the CVD definition complicates assigning accurate cost

estimates to the overall CVD. To overcome this, the cost of CVD was assumed to be a

composite of the costs related to acute myocardial infarction (AMI), unstable angina,

heart failure (HF) and stroke, as these are cumulatively responsible for the majority of CVD

events. Therefore, Dutch estimates of incidence for each of the outcomes were used to

weight the overall CVD cost estimate (40,41). Unit cost estimates from previous costing

studies conducted in The Netherlands were updated to the year 2013 using the Dutch

inflation index (Table 2)(27,42-45). A distinction was made between the acute costs of the

first year after the event and the cumulative cost of health resources used in subsequent

years. Finally, an additional cost for a fatal CVD event was assumed (27).

The drug prices applied referred to the cheapest generic alternatives available on the

market and were defined as price per guideline-proposed dosing regimen (Table 2)(46). In

the first year of treatment, patients were assumed to visit the general practitioner (GP)

twice, and to subsequently refill their prescription by phone, every three months. In order

to establish and monitor the appropriateness of the antihypertensive treatment dosage

applied, four laboratory tests estimating the glomerular filtration rate and level of

potassium in the blood were added to the first year treatment costs (47)(5). In the

subsequent years, the monitoring of treatment dosage was assumed to consist of one GP

visit and laboratory test per year. It was assumed that the patient will visit a Dutch

pharmacy for every refill, hence a pharmacist fee would apply (48). The patients were

assumed to take the same antihypertensive drug according to the proposed dosing

regimen throughout the year until they experienced an event.

Cost-effectiveness

The costs of antihypertensive treatment and cardiovascular events were aggregated

separately for each scenario. Costs of treatment alternative included overall drug

Chapter 3

52

treatment costs, as well as costs related to non-fatal and fatal CVD events. Costs of a no

treatment alternative consisted only of costs related to CVD events.

As a measure of effectiveness, the life-years gained (LYG) that were obtained by the SBP

lowering strategy were utilized. Initially, the life years lost (LYL) for every SBP, age and

gender scenario due to the CVD mortality were estimated. The LYG were defined as the

difference in LYL between the no treatment alternative, where patients were assumed to

remain with a certain baseline SBP level, and the treatment alternative, where patients

were assumed to achieve a reduction of SBP.

The CE of SBP lowering was expressed through the incremental cost-effectiveness ratio

(ICER). The ICER thus reflects the additional cost, which, by choosing a treatment

alternative, is imposed over the no treatment alternative in order to gain one year of life.

Future costs and benefits were discounted by 4% and 1.5% annually according to the

Dutch guidelines for pharmacoeconomic research (49). The CE analysis was performed

from a healthcare perspective and only direct pharmacy and medical costs associated with

CVD events were considered.

Validation

We validated the performance of our Markov model by comparing the model’s estimate

of life expectancy (LE) for an average Dutch patient of age 40, 50, 60 and 65, to the

respective life tables from the Dutch Bureau of Statistics (Table 3)(50). A general

agreement between the observed and modeled LE was observed, although the model

slightly overestimated the LE for both genders. These differences can be attributed to a

number of factors such as the uncertainty on average SBP, cholesterol and smoking levels,

the inability of SCORE to predict life-long estimates and the differences between the

SCORE and the current Dutch population.

Probabilistic sensitivity analysis

Simultaneously incorporating the uncertainty around all parameters in the CE analysis was

done through a probabilistic sensitivity analysis (PSA) with 10,000 iterations for each

estimation. Key input parameters in the deterministic analysis that were assumed to be

random variables were: the antihypertensive effect of HCT on SBP, the level of population

adherence to HCT treatment, the covariate estimates of the SCORE model, the incidence

rates and cost estimates of CVD, the costs of pharmacy, GP fees and laboratory tests, and

the probability of non-CVD death. When cost estimates were available only as single point

estimates, they were assumed to follow a log-normal distribution with a coefficient of

variation equal to 0.25. In order to capture the uncertainty around the parameters of the

SCORE model the Cholesky decomposition on the variance-covariance matrix of the SCORE

regression model was used (personal communication: Fitzgerald T, Department of

Epidemiology & Public Health/Statistics, University College Cork). Through this technique,

correlated random draws can be generated from the parameters’ multivariate normal

3


53

distribution (30,51). A beta distribution was assigned to the CVD specific incidence rates, a

log-normal distribution to the cost estimates and a normal distribution to the rest of

varied parameters (Table 2).

Results from the PSA are presented through cost-effectiveness acceptability curves (CEAC)

and 95% confidence interval (CI) estimates around the incremental effects and the

incremental costs. The CEAC assesses the probability that the estimated ICER is under a

certain willingness to pay (WTP) threshold. Finally, in order to approximate the

contribution of the individual parameters to the overall uncertainty of the model the

Analysis of Covariance (ANCOVA) method was used. This method links the proportion of

the variance in the PSA-simulated incremental effects and incremental costs to the

variation of the model input parameters.(30) When the impact of each parameter on

overall model variation was examined, the impact of the antihypertensive effect of HCT

and the level of adherence with HCT was separately assessed, while all SCORE parameters,

CVD related costs, incidence rates of separate CVD events, costs of pharmacy fees, GP

related fees and laboratory tests, and the probability of death due to non-CVD causes

were assessed jointly.

Table 2 Key parameters and corresponding distributions for deterministic and probabilistic analyses.

Parameter Men Women Distribution

First year Subsequent years

First year Subsequent years

Cardiovascular complication (2013, €)

Stroke(44) 37,488 (10.50, 0.25)

7,448 (8.89, 0.25)

37,488 (10.50, 0.25)

11,405(9.31, 0.25)

Log-normal (μ,σ)

Myocardial infarction(27) 19,126 (9.86, 0.01)

1,162 (7.06, 0.03)

19,126 (9.86, 0.01)

1,162 (7.06, 0.03)

Log-normal (μ,σ)

Heart failure42 5,625 (8.60, 0.25)

N/A 5,625(8.60, 0.25)

N/A Log-normal (μ,σ)

Unstable angina43 5,358 (8.56, 0.25)

N/A 5,358(8.56, 0.25)

N/A Log-normal (μ,σ)

Estimated CVDa 19,252 2,533 17,128 2,772

Death27 2,976 (8.00, 0.02) 2,976 (8.00, 0.02) Log-normal

(μ,σ)

Incidence (per 100,000 person-years)

Stoke41 182 (12,324 , 6,727,241) 116 (7,818 , 6,731,747) Beta (p,q)

Myocardial infarction40 212.2 (481 ,226,219) 212.2 (481 ,226,219) Beta (p,q)

Heart failure40 66.4 (154 , 231,846) 66.4 (154 , 231,846) Beta (p,q)

Unstable angina 171.8 (390 , 226,610) 171.8 (390 , 226,610) Beta (p,q)

Meta-analyzed estimate of SBP reduction due to HCT 25 mg20

11.3 mmHg (11.3 , 0.86) Normal (μ,σ)

Adherence to HCT(26) 78.6% (78.6 , 0.38) Normal (μ,σ)

Chapter 3

54

Parameter Men Women Distribution

First year Subsequent

years

First year Subsequent

years

Annual drug costs (2013, €)

HCT 25 mg46 7.3 Fixed

Fixed-dose combination

HCT 12.5mg/Losartan 50mg46

10.95 Fixed

General practitioner visit48 31 (3.39, 0.25) Log-normal

(μ,σ)

Laboratory test47 2 (0.55, 0.25) Log-normal

(μ,σ)

Telephone contact/refill48 15 (2.69, 0.25) Log-normal

(μ,σ)

Pharmacy prescription

fee48

6 (1.84, 0.25) Log-normal

(μ,σ)

Additional pharmacy fee

for the first drug issuing48

4 (1.24, 0.25) Log-normal

(μ,σ)

Discount rate (costs)49 4% Fixed

Healthy life expectancy 50 Life-table Fixed

Probability of non-CVD death33-35

Derived from cited sources Normal (μ,σ)

Discount rate (health)49 1.5% Fixed

a Based on the assumption that the cost of CVD will be a composite of the costs related to stroke, MI, HF and

unstable angina whose contribution to the estimated CVD costs is calculated according to their annual incidence in The Netherlands.

CVD, cardiovascular diseases; HCT, hydrochlorothiazide; N/A, not available; SBP, systolic blood pressure.

Table 3 Predicted life expectancy for an average Dutch patient of different age groups.

Gender Age Dutch Bureau of Statistics50 (LE) Model prediction (LE)

Man 40 40.31 40.60

50 30.85 31.20

60 22.03 22.52

65 17.95 18.58

Woman 40 43.98 44.97

50 34.45 35.43

60 25.47 26.42

65 21.19 22.19

LE, life expectancy.

RESULTS

Base-case scenario

A 60 year-old man with baseline SBP equal to 160 mmHg was estimated to be in a 10-year

risk of 5.2% for a fatal and 5.1% for a non-fatal CVD event. The corresponding 10-year risk

of a fatal and a non-fatal CVD event after treatment with HCT dropped to 4.4% and 4.6%

respectively, resulting in 0.0847 LYG. For a 60 year-old woman with the same SBP level, a

3


55

10-year risk of 2.1% for a fatal and 2.9% for a non-fatal CVD event was estimated. After

HCT treatment the estimated 10-year fatal and non-fatal CVD risks were 1.8% and 2.6%

respectively. Treatment with HCT was estimated to result in 0.0413 LYG in a 10-year time

horizon.

In a lifetime horizon, the fatal CVD risk for a 60 year-old man with baseline SBP equal to

160 mmHg was estimated at 22.9% and the non-fatal CVD risk at 17.1%. The

corresponding fatal risk after HCT treatment was estimated at 20%, while the non-fatal

CVD risk was estimated at 15.7%. This resulted to an estimated 0.2874 LYG. A woman with

similar characteristics was estimated to have a lifelong fatal CVD risk of 20.2% while the

non-fatal CVD risk was estimated at 15.9%. After HCT treatment these estimated risks

dropped to 17.7% and 14.6% respectively. This reduction resulted to an estimated 0.3197

LYG.

Figure 2 presents the distribution of costs for a 10-year and a lifetime horizon for a 60-

year-old man and a 60-year-old woman with baseline SBP of 160 mmHg. The total costs

for a man with the aforementioned characteristics who was assumed not to receive

antihypertensive treatment, were €2,014 in a 10-year and €5,608 in a lifetime horizon.

When assuming that this person is treated with HCT 25mg, the total costs in a 10-year and

a lifetime horizon raised to €2,604 and €6,496, respectively. For a woman with similar

characteristics, assumed not to be receiving treatment, the costs in a 10-year and a

lifetime horizon were €1,210 and €5,443, respectively. Applying antihypertensive

treatment with HCT 25 mg in a woman with these characteristics resulted in 10-year and

lifetime costs of €1,938 and €6,553, respectively. It can be observed that the pharmacy fee

and the cost of GP visit bear most of the antihypertensive treatment related costs, while

non-fatal CVD events are responsible for the vast majority of medical costs.

The estimated ICERs after 10-year treatment with HCT in men were in a range from

€6,032/LYG (65-year old men with a baseline SBP of 160 mmHg) to €58,217/LYG (40-year

old men with a baseline SBP of 140 mmHg); the estimated ICERs for women were in a

range from €12,345/LYG (65-year old women with a baseline SBP of 160 mmHg) to

€361,064/LYG (40-year old women with a baseline SBP of 140 mmHg). In a lifetime

horizon the estimated ICERs for men were in a range from € 3,076/LYG (65-year old men

with a baseline SBP of 160 mmHg) to €4,221/LYG (65-year old men with a baseline SBP of

140 mmHg); while for women were in a range from €3,074/LYG (40-year old women with

a baseline SBP of 160 mmHg) to €4,645/LYG (65-year old women with a baseline SBP of

140 mmHg).

The estimated, per patient, 10-year and lifetime costs and LYL for the treatment scenario

which assumed treatment with a fixed-dose combination of HCT 12.5mg/Losartan 50mg

as well as for the no treatment alternative, for different gender, age and SBP reduction

levels, can be found in the Appendix (Table A.I and A.II).

Chapter 3

56

The ICERs for all the SBP reduction scenarios in a 10-year and a lifetime horizon, compared

to no reduction are presented in Figure 3. In general, the ICERs indicate that SBP reduction

would be more cost-effective and even cost-saving in older men and for greater SBP

reductions. In a 10-year horizon the estimated ICERs were significantly more favorable in

men, while this difference between genders in a lifetime horizon was less pronounced.

Treatment vs. no treatment in male and female smokers scenario

Table 4 presents the estimated 10-year fatal and non-fatal CVD risks in a 60 year-old man

and woman with baseline SBP equal to 160 mmHg and the corresponding risk levels after

treatment with HCT 25mg. Applying HCT in male smokers in a 10-year time horizon was

associated with ICERs in a range from €2,727/LYG (65-year old men with a baseline SBP of

160 mmHg) to €37,964/LYG (40-year old men with a baseline SBP of 140 mmHg); the

estimated ICERs for women were in a range from €6,161/LYG (65-year old women with a

baseline SBP of 160 mmHg) to €229,456/LYG (40-year old women with a baseline SBP of

140 mmHg).

Table 4 Estimated 10-year fatal and non-fatal CVD risks in a 60 year-old man and woman with

baseline SBP equal to 160 mmHg and the corresponding risk levels after treatment with HCT 25mg.

Risk (%) with no treatment at a baseline SBP=160mmHg

Risk (%) after treatment with HCT 25mg

Man

Fatal CVD 8.3 7.1

Non-fatal CVD 7.3 6.5

Woman

Fatal CVD 3.6 3.1

Non-fatal CVD 4.0 3.6

3


57

Figure 2 Allocation of overall medical costs in a treatment and no treatment alternative. A: Allocation of the total cost in the treatment (left) and no-treatment (right) alternative in the scenario of a 60-

year-old man in a 10-year time horizon with baseline SBP of 160 mmHg receiving HCT 25 mg as an antihypertensive treatment.

B: Allocation of the total cost in the treatment (left) and no-treatment (right) alternative in the scenario of a 60-year-old woman in a 10-year time horizon with baseline SBP of 160 mmHg receiving HCT 25 mg as an antihypertensive treatment.

C: Allocation of the total cost in the treatment (left) and no-treatment (right) alternative in the scenario of a 60-year-old man in a lifetime horizon with baseline SBP of 160 mmHg receiving HCT 25 mg as an antihypertensive treatment.

D: Allocation of the total cost in the treatment (left) and no-treatment (right) alternative in the scenario of a 60-year-old woman in a lifetime horizon with baseline SBP of 160 mmHg receiving HCT 25 mg as an

antihypertensive treatment. SBP, systolic blood pressure; HCT, hydrochlorothiazide; CVD, cardiovascular disease; GP, general practitioner

Probabilistic sensitivity analysis (PSA)

Application of multivariate PSA to the base-case scenario revealed a wide uncertainty

range around the incremental costs and incremental effects (Figure 4). This uncertainty

was found to be greater in scenarios with higher baseline SBP levels. In Figure 5, the

probabilities that the estimated ICERs would be below various WTP thresholds are

Chapter 3

58

presented for a 60-year-old man and a 60-year-old woman with baseline SBP levels of

160, 150 and 140 mmHg assuming to be treated with HCT 25 mg, in a 10-year and a

lifetime horizon. In a 10-year time horizon, in men, the probabilities that the SBP

reduction from baseline SBP of 160, 150 and 140 mmHg due to HCT 25mg would be below

the WTP of €20,000/LYG were 0.8012, 0.7469 and 0.6818, respectively (Figure 4 A). In a

lifetime horizon the corresponding probabilities were 0.9618, 0.9510 and 0.9324 (Figure 4

C). In a 10-year time horizon, in women, the corresponding probabilities were 0.5786,

0.4855 and 0.3972 (Figure 4 B), while in a lifetime horizon these probabilities were 0.9865,

0.9821 and 0.9719 (Figure 4 D).

The results of ANCOVA, applied for a 60 year-old man whose SBP was lowered due to

treatment with HCT 25mg from a baseline SBP of 160 mmHg in a 10-year time horizon,

clearly indicate that the parameters of the SCORE model have the most significant

contribution to the uncertainty around the incremental effects and incremental cost

estimates, being responsible for 75.7% and 46% of the variance of the respective PSA

simulations respectively (Figure 6).

DISCUSSION

In the present analysis, both the long-term health benefits and the economic

consequences of SBP reduction in a Dutch population with mild hypertension were

estimated. In the base-case scenario, where SBP reduction was assumed to be achieved

through antihypertensive treatment with HCT, significant health and economic benefits

were observed in all scenarios for both genders in a lifetime horizon. However, ICER

estimates for HCT treatment compared to no treatment in a 10-year horizon found SBP

reductions to be more favorable when targeted to older patients and yet more in men

than in women. In the hypothetical SBP reduction scenarios achieved through fixed-dose

combination HCT 12.5mg/Losartan 50mg, the long-term health and economic benefits

estimated were greater than in the base-case scenario for both a 10-year and a lifetime

horizon. Thus, the CE estimates across all scenarios were highly influenced by the model’s

time horizon.

In the scenario comparing a 10-year treatment with HCT 25mg vs. no treatment in male

and female smokers, the estimated ICERs were more favorable than the ICERs estimated

in the base-case scenario in patients with comparable age-, gender- characteristics but

with the prevalence of smoking representative for Dutch population.

The aforementioned results reflect the higher levels of CVD risk estimated for male and

female smokers (Table 4) compared to patients with similar characteristics but with

average smoking prevalence, as well as the greater benefits of risk reduction in smoking

patient population.

3


59

The impact of the model parameters’ uncertainty on the ICER estimates was assessed

through a PSA. The results of the PSA show that although the deterministic approach is

likely to favor the CE of SBP reduction, the leap from the reduction of a risk factor to that

of CVD risk is surrounded by a significant amount of prediction uncertainty.

Our findings, in terms of both health and economic benefits of SBP reduction are similar to

the results of other studies investigating the CE of SBP reduction in low CVD risk patients,

both in a 10-year and in a lifetime horizon (52-57).

Figure 3 Discounted ICER estimates for different scenarios of SBP reductions. A: Discounted ICER estimates for different scenarios of SBP reductions in a 10-year time horizon in men aged 40, 50, 60 and 65, compared to no treatment.

B: Discounted ICER estimates for different scenarios of SBP reductions in a 10-year time horizon in women aged 40, 50, 60 and 65, compared to no treatment. C: Discounted ICER estimates for different scenarios of SBP reductions in a lifetime horizon in men aged 40, 50,

60 and 65, compared to no treatment. D: Discounted ICER estimates for different scenarios of SBP reductions in a lifetime horizon in women aged 40,

50, 60 and 65, compared to no treatment. ICER, incremental cost-effectiveness ratio; SBP, systolic blood pressure; LYG, life year gained

Chapter 3

60

10-year time horizon Lifetime horizon

Figure 4 Uncertainty around the incremental costs and incremental effects. Bar plots show the incremental cost and the incremental effect estimates from the deterministic analysis. The lines above the bars represent the upper 95% confidence interval limit estimated through a probabilistic

sensitivity analysis. A: Uncertainty around the incremental costs in 60-year-old men for different SBP baseline levels, assumed to be

treated with hydrochlorothiazide 25 mg. B: Uncertainty around the incremental effects in 60-year-old men for different SBP baseline levels, assumed to be treated with hydrochlorothiazide 25 mg.

C: Uncertainty around the incremental costs in 60-year-old women for different SBP baseline levels, assumed to be treated with hydrochlorothiazide 25 mg. D: Uncertainty around the incremental effects in 60-year-old women for different SBP baseline levels, assumed

to be treated with hydrochlorothiazide 25 mg. SBP, systolic blood pressure

3


61

Figure 5 Cost-effectiveness acceptability curves. Cost-effectiveness acceptability curves show the probabilities that SBP reductions achieved with HCT 25 mg as antihypertensive treatment from baseline SBP levels of 160, 150 and 140 mmHg observed in a 10-year and in a lifetime horizon, in 60-year-old men A,C and 60-year-old women B,D, are cost-effective at different willingness to

pay threshold values. A: Cost-effectiveness acceptability curves for the treatment with HCT 25 mg assumed in the treatment scenarios

of 60-year-old men in a 10-year time horizon. B: Cost-effectiveness acceptability curves for the treatment with HCT 25 mg assumed in the treatment scenarios of 60-year-old women in a 10-year time horizon.

C: Cost-effectiveness acceptability curves for the treatment with HCT 25 mg assumed in the treatment scenarios of 60-year-old men in a lifetime horizon.

D: Cost-effectiveness acceptability curves for the treatment with HCT 25 mg assumed in the treatment scenarios of 60-year-old women in a lifetime horizon. SBP, systolic blood pressure; HCT, hydrochlorothiazide

Chapter 3

62

Figure 6 ANCOVA analysis of proportion of sum squares for incremental effects and incremental

costs. ANCOVA analysis of proportion of sum squares for incremental effects (left) and incremental costs (right) driven by the contribution of the individual model parameter uncertainty.

ANCOVA, analysis of covariance; CVD, cardiovascular disease; GP, general practitioner; HCT, hydrochlorothiazide.

However, some discrepancies were found in terms of the estimated ICERs. For example,

through our analysis, it was found that as age increases, lowering of SBP becomes more

cost-effective when looking in a 10-year time horizon; while the CE of SBP reduction does

not vary considerably across ages when looking in a lifetime horizon. This is in contrast to

two other simulation analysis, where middle-aged patients were identified as being those

mostly benefiting from SBP reductions (52,56). Yet, comparability between studies is

hampered due to differences regarding the choice of primary CE outcome (QALY)(52,54-

57), and specific assumptions in the model (e.g. differences in baseline patient

characteristics, modelling of different CVD-related health states, level of SBP reduction

and country specific cost estimates)(52-57).

To our knowledge, similar CE studies on the impact of SBP reduction in patients with mild

hypertension for various age, gender and specific baseline SBP levels and adapted to the

Dutch situation, have not yet been conducted. From the abundance of different CVD risk

prediction models (e.g. the Framingham, QRISK, ASSIGN etc.(58-63)), we decided to

implement the SCORE model (23) given that it has recently been validated for the Dutch

population.

We chose to conduct our simulations both in a 10-year and a lifetime horizon. There is a

variety of reasons for choosing a 10-year time horizon. Firstly, the SCORE model has only

been validated for 10-year horizons (23). Secondly, if a lifetime horizon is to be applied,

the CVD risk estimates in the elderly populations might deviate from the CVD risk

estimated from SCORE as it is validated only for patients under 65 (64,65). Finally, for

these reasons mentioned above, it can be anticipated that the application of a 10-year

3


63

horizon is more conservative than a lifetime. However, in order to make a comparison to

similar studies and to see the influence of its choice on final outcomes, our simulations

were also conducted in a lifetime horizon.

In order to examine the uncertainty in the input parameters, multivariate PSA was

applied. Interval estimates for the scenarios observed revealed a large uncertainty range

around the expected values from the deterministic analysis regardless of the applied time

horizon. When investigating the contribution of each of the included parameters to the

overall uncertainty (ANCOVA), the coefficient estimates of the SCORE model were found

to have the most impact. This simply suggests that even though a certain risk prediction

model has been shown to be applicable for a certain population, building decisions upon

the model should still be made cautiously and being aware of the inevitable uncertainty

surrounding estimates of effectiveness and CE that are based on long-term projections

(51).

As every analysis, ours is also confronted with some limitations. The risk of having a non-

fatal CVD event was obtained through a functional form linking the annual risk of mortality

to the annual risk of morbidity and mortality. This approach is based on the fact that the

SCORE model only predicts a CVD mortality risk. Other recent risk prediction models (e.g.

the Framingham, QRISK, ASSIGN etc.(58-63)) generally provide estimates of an overall CVD

risk (i.e. CVD morbidity and mortality) with no insight in the fraction of fatal cases. Given

the requirement of our model to distinguish between fatal and non-fatal CVD events as

well as the fact that the SCORE model was the only currently validated model in the Dutch

population, we incorporated it in our analysis for the prediction of CVD mortality together

with the functional form linking the annual risk of mortality to the annual risk of morbidity

and mortality. Furthermore, the probability of subsequent non-fatal CVD events was

assumed to be equal to the probability of a first non-fatal CVD event. This is supported by

the fact that the SCORE risk model is limited to prediction of first events only and that

there are no available estimates of risk of subsequent events in the population that the

SCORE was developed on. Moreover, the fact that the occurrence of a composite CVD

event was modelled in our analysis hampers the estimation of a subsequent CVD event

risk. A subsequent CVD event could occur in any CVD form (e.g. stroke, MI) and, therefore,

the risk of its re-occurrence would be dependent on the specific primary CVD event form.

Notably, we could expect more cost-effective outcomes if risks of subsequent non-fatal

events following a first event were naturally assumed to be higher. Therefore, in order to

alleviate the underestimation of the CVD risk, an increased mortality risk was assumed in

patients experiencing non-fatal CVD events. Furthermore, the impact of the joint

interaction of variables SBP and antihypertensive treatment on the risk prediction is not

accounted for in the SCORE model. Such a limitation of the SCORE model prevents

modelling the impact of SBP level on risk prediction with and without the use of

antihypertensive treatment differently. Furthermore, lowering of SBP was assumed to be

Chapter 3

64

exclusively achieved by specific antihypertensive treatment. In a real life setting,

differences in persistence to treatment or the presence of other risk factors could

obviously result in significant discrepancies between real life and the simulated results of

our analysis. In the base-case scenario, the real life effectiveness of HCT 25mg and

patients’ adherence to this treatment were incorporated. However, in the absence of

synthesized evidence on the effectiveness of the HCT 12.5mg/Losartan combination,

hypothetical SBP reductions of 20 and 30 mmHg were assumed. The latter assumption

was made in order to incorporate the treatment costs in the overall cost estimate of SBP

reduction and not to derive an estimate of CE for the combination treatment per se.

Additionally, an instant, incremental SBP reduction was assumed. In reality, however, this

is highly unlikely. Moreover, side effects resulting from antihypertensive treatment were

not incorporated. Societal costs were also not included in the estimations of CE which

makes our estimates more conservative. However, it could be expected that implementing

a societal perspective is likely to lead to more favourable ICERs especially at younger

populations that encounter greater loss of productivity and costs related to it. Finally, it

should be noted that CVD risk reduction can also be accomplished through dietary

measures or increased exercise, potentially leading to comparable (or perhaps even more

favourable) ICERs (55,66).

Recent evidence suggest that lifelong CVD risk estimates might be more informative than

10-year ones.(67) However, treatment recommendations by the latest Dutch guidelines

are based on the levels of patient’s 10-year risks. Specifically, patients can receive

treatment if their 10-year risk for CVD is more than 20%, or if they have accompanying risk

factors (5). The cumulative 10-year risk for our patient population was less than 20%, in all

cases. Hence, if a decision on whether to apply antihypertensive treatment would have

been made merely in accordance to the guidelines, none of our patients would have

received treatment. Notably, guideline recommendations are made with respect to CE of

treatment application. However, due to the recent changes in pharmaceutical policy in the

Netherlands (e.g. preference policy (6)), the cost of antihypertensive treatment is now

considerably lower. This might lead to the conclusion that the current recommendations

limit treatment to a population with a relatively high 10-year risk, whereas favourable CE

is also likely in groups with lower risks. Furthermore, younger patients, who are currently

at a relatively low 10-year risk, are likely to experience an increase in risk later in life, due

to their characteristics. Therefore, especially for this population, early preventive

treatment is expected to provide significant increases in LE, as illustrated in our model

when applying the lifetime horizon (67).

Conclusion

Even relatively small SBP reductions can lead to significant lifetime health and economic

benefits. Given that, nowadays, low-cost generic drugs account for the smallest share in

3


65

the overall medical costs in primary CVD prevention, there are more opportunities for a

wider and less strict application of CVD prevention strategies.

SBP reductions in low-risk elderly men were found to be more cost-effective when

compared to women, in a 10-year time horizon. When employing a lifetime horizon, the

CE estimates indicated a comparable eligibility of both genders for treatment. However,

our results clearly indicate that decision-making should highly consider the amount of

uncertainty surrounding the estimated values of the analysis.

Appendix

Detailed description of the process and assumptions on transforming the SCORE

cumulative risk to annual probabilities

The SCORE model applies a Weibull proportional hazards model to provide a 10-year

cumulative risk of CVD mortality. In our analysis we used the coefficients of an updated

SCORE model for low-risk regions (personal communication: Fitzgerald T, Department of

Epidemiology & Public Health/Statistics, University College Cork).

The SCORE risk model included covariate adjustments for a number of risk factors. We

used Dutch-specific values for these risk factors in order to produce predictions of 10-year

CVD mortality risk that would correspond to the Dutch population and we consequently

estimated annual probabilities of CVD mortality.

Specifically, the underlying baseline 10-year probability of survival without CVD �� was

estimated for different ages and SBP levels after being corrected for the risk factors total

cholesterol and the prevalence of smoking using Dutch-specific values.

�� = exp(− exp(α ∗ (age − 20��(��, ω = exp(β�� ∗ (sbp − 140/10 + β !"# ∗ (chol − 5 + β�)"*�+ ∗ smoker ,

where α, ρ, β�� , β !"#, β�)"*�+ are the coefficients of the updated SCORE model; sbp is

the level of systolic blood pressure investigated (mmHg); chol is the level of total

cholesterol (mmol/l); smoker stands for the prevalence of current smoking.

Annual probabilities were obtained for every Markov cycle and for each of the scenarios

by estimating the age- and SBP- specific CVD event probability conditional on the patient’s

survival up to this time point. For example, survival in the first year ��/0 where t=1 from

the start of the observation is conditional on the patient’s underlying baseline survival ��.

Survival probability ��/0 at the time point t is: ��/0 = exp1− exp(α ∗ (age − 20 + t��(��3, ω = exp(β�� ∗ (sbp − 140/10 + β !"# ∗ (chol − 5 + β�)"*�+ ∗ smoker,

where t stands for the year for which the probability of CVD mortality needs to be

estimated. t can take values from 1 to the end of time horizon observed.

Chapter 3

66

Process of calculating the annual probability of CVD mortality

Step 1

In order to estimate the risk of having a fatal CVD event in the first year of the model

cycle, calculate �� and ��/0, where t has a value 1. CVD7890:;<0= 9<>? =@:9 A = 1 − (��/A ��⁄

Step 2

Calculating the yearly risks of having a fatal CVD event in the second and the following

years continues in the following manner: CVD7890:;<0= 9<>? =@:9 C = 1 − (��/C ��/A⁄ CVD7890:;<0= 9<>? =@:9 D = 1 − (��/D ��/(D/A⁄

Table A.I Estimated 10-year costs of treatmenta assumed with a fixed-dose combination HCT

12.5mg/Losartan 50 mg and no treatmentb and life years lost (LYL) for different age groups and

different SBP scenariosc per patient.

Gender Age Baseline SBP

(mmHg)

10-year cost of no treatment

(€)

LYL SBP reduction (mmHg)

10-year cost of

treatment

(€)

LYL

Men

40 160 888 0.14 20 1,706 0.10

30 1,674 0.08

150 842 0.12 20 1,674 0.08

50 160 1,364 0.36 20 2,004 0.25

30 1,910 0.21

150 1,229 0.30 20 1,910 0.21

60 160 2,014 0.58 20 2,409 0.40

30 2,233 0.34

150 1,766 0.48 20 2,233 0.34

65 160 2,261 0.59 20 2,550 0.42

30 2,342 0.35

150 1,972 0.50 20 2,342 0.35

Women

40 160 704 0.02 20 1,595 0.02

30 1,590 0.01

150 697 0.02 20 1,590 0.01

50 160 823 0.11 20 1,646 0.08

30 1,618 0.06

150 783 0.09 20 1,618 0.06

60 160 1,209 0.28 20 1,868 0.20

30 1,785 0.16

150 1,090 0.23 20 1,785 0.16

3


67

Gender Age Baseline SBP

(mmHg)

10-year cost of no treatment

(€)

LYL SBP reduction (mmHg)

10-year cost of

treatment

(€)

LYL

65 160 1,482 0.37 20 2,027 0.26

30 1,907 0.21

150 1,312 0,31 20 1,907 0,21 aTreatment alternative assumed costs of antihypertensive treatment and costs due to cardiovascular morbidity and mortality that in a certain lower degree appears in a treated cohort, bNo treatment alternative assumed costs due to cardiovascular morbidity and mortality, c Different age-, gender- and SBP reduction related scenarios were assumed for a treatment and no treatment alternative, Treatment with a fixed-dose combination HCT 12,5mg/Losartan 50 mg was assumed to reduce the

SBP from a baseline SBP of 160 mmHg for either 20 or 30 mmHg; or to reduce the SBP from a baseline SBP of 150 mmHg for 20 mmHg, No treatment alternative assumed that a patient would remain at a baseline SBP level respective to the one assumed in the treatment scenario (160 or 150 mmHg),

LYL, life years lost; SBP, systolic blood pressure

Table A.II Estimated lifetime costs of treatmenta assumed with a fixed-dose combination HCT 12.5mg/Losartan 50 mg and no treatmentb and life years lost (LYL) for different age groups and

different SBP scenariosc per patient.

Gender Age Baseline

SBP

(mmHg)

Lifetime cost

of no

treatment

(€)

LYL SBP

reduction

(mmHg)

Lifetime cost

of treatment

(€)

LYL

Men

40 160 6,434 4.33 20 7,532 3.28

30 7,096 2.83

150 5,959 3.78 20 7,096 2.83

50 160 6,098 3.43 20 6,875 2.60

30 6,431 2.25

150 5,593 3.00 20 6,431 2.25

60 160 5,608 2.54 20 6,040 1.92

30 5,601 1.66

150 5,092 2.22 20 5,601 1.66

65 160 5,221 2.09 20 5,508 1.58

30 5,086 1.36

150 4,721 1.82 20 5,086 1.36

Women

40 160 6,524 4.43 20 7,741 3.30

30 7,303 2.83

150 6,053 3.83 20 7,303 2.83

50 160 5,974 3.57 20 6,977 2.67

30 6,556 2.29

Chapter 3

68

Gender Age Baseline

SBP

(mmHg)

Lifetime cost

of no

treatment

(€)

LYL SBP

reduction

(mmHg)

Lifetime cost

of treatment

(€)

LYL

150 5,511 3.10 20 6,556 2.29

60 160 5,443 2.73 20 6,146 2.05

30 5,737 1.76

150 4,978 2.37 20 5,737 1.76

65 160 5,142 2.31 20 5,680 1.73

30 5,279 1.49

150 4,679 2.01 20 5,279 1.49 a Treatment alternative assumed costs of antihypertensive treatment and costs due to cardiovascular morbidity and mortality that in a certain lower degree appears in a treated cohort, bNo treatment alternative assumed costs due to cardiovascular morbidity and mortality, c Different age-, gender- and SBP reduction related scenarios were assumed for a treatment and no treatment alternative, Treatment with a fixed-dose combination HCT 12,5mg/Losartan 50 mg was assumed to reduce the

SBP from a baseline SBP of 160 mmHg for either 20 or 30 mmHg; or to reduce the SBP from a baseline SBP of 150 mmHg for 20 mmHg, No treatment alternative assumed that a patient would remain at a baseline SBP level respective to the one assumed in the treatment scenario (160 or 150 mmHg),

LYL, life years lost; SBP, systolic blood pressure

Acknowledgements

We acknowledge Dr. Tony Fitzgerald (Department of Epidemiology & Public

Health/Statistics, University College Cork) for providing the updated SCORE risk prediction

model.

The study was supported by Sanofi (Gouda, Netherlands). The authors declare study

results were not influenced by Sanofi funding.

3


69

REFERENCES

(1) Leal J, Luengo-Fernandez R, Gray A, et al. Economic burden of cardiovascular diseases in the enlarged

European Union. Eur Heart J 2006 Jul;27(13):1610-1619.

(2) Poos M, Smit J, Groen J, et al. Kosten van ziekten in Nederland 2005. Bilthoven: Rijksinstituut voor

Volksgezondheid en Milieu 2008.

(3) Smulders Y, Burgers J, Scheltens T, et al. Clinical practice guideline for cardiovascular risk management in the

Netherlands. Neth J Med 2008;66(4):169-174.

(4) Mancia G, De Backer G, Dominiczak A, et al. 2007 ESH-ESC practice guidelines for the management of arterial

hypertension: ESH-ESC task force on the management of arterial hypertension. J Hypertens 2007;25(9):1751-

1762.

(5) Genootschap NH. Multidisciplinaire richtlijn Cardiovasculair risicomanagement, herziening 2011. Utrecht.

2011.

(6) Bos E. Door preferentiebeleid. Pharmaceutisch Weekblad 2010;May:13-15.

(7) De Backer G, Ambrosionie E, Borch-Johnsen K, et al. European guidelines on cardiovascular disease

prevention in clinical practice: third joint task force of European and other societies on cardiovascular disease

prevention in clinical practice (constituted by representatives of eight societies and by invited experts). European

Journal of Cardiovascular Prevention & Rehabilitation 2003;10(1 suppl):S1-S78.

(8) Graham I, Atar D, Borch-Johnsen K, et al. European guidelines on cardiovascular disease prevention in clinical

practice: executive summary Fourth Joint Task Force of the European Society of Cardiology and other societies

on cardiovascular disease prevention in clinical practice (constituted by representatives of nine societies and by

invited experts). Eur Heart J 2007;28(19):2375-2414.

(9) Dutch Institute for Healthcare Improvement CBO. Dutch Guideline on Cardiovascular Risk Management.

Dutch Institute for Healthcare Improvement CBO and Dutch College of General Practitioners. Utrecht:2006.

Available at: www.cbo.nl/Downloads/211/guide_cvrm_07.pdf. Accessed Jun 14, 2012.

(10) Kannel WB, D'Agostino RB, Sullivan L, et al. Concept and usefulness of cardiovascular risk profiles. Am Heart

J 2004;148(1):16.

(11) Staessen J, Fagard R, Thys L, et al. for the Syst-Eur trial investigators. Randomized double blind comparison

of placebo and active treatment for older patients with isolated systolic hypertension. Lancet 1997;350:757-764.

(12) SHEP Cooperative Research Group. Prevention of stroke by antihypertensive drug treatment in older

persons with isolated systolic hypertension: final results of the Systolic Hypertension in the Elderly Program

(SHEP). JAMA 1991;265(24):3255-3264.

(13) Shepherd J, Cobbe SM, Ford I, et al. Prevention of coronary heart disease with pravastatin in men with

hypercholesterolemia. N Engl J Med 1995;333(20):1301-1308.

(14) Rich-Edwards JW, Manson JAE, Hennekens CH, et al. The primary prevention of coronary heart disease in

women. N Engl J Med 1995;332(26):1758-1766.

(15) Downs JR, Clearfield M, Weis S, et al. Primary prevention of acute coronary events with lovastatin in men

and women with average cholesterol levels. JAMA: the journal of the American Medical Association

1998;279(20):1615-1622.

(16) Simons WR. Comparative cost effectiveness of angiotensin II receptor blockers in a US managed care setting:

olmesartan medoxomil compared with losartan, valsartan, and irbesartan. Pharmacoeconomics 2003;21(1):61-

74.

(17) Scheltens T, Bots M, Numans M, et al. Awareness, treatment and control of hypertension: the ‘rule of

halves’ in an era of risk-based treatment of hypertension. J Hum Hypertens 2006;21(2):99-106.

(18) Kannel WB. Blood pressure as a cardiovascular risk factor. JAMA: the journal of the American Medical

Association 1996;275(20):1571-1576.

(19) Ezzati M, Lopez AD, Rodgers A, et al. Selected major risk factors and global and regional burden of disease.

Lancet 2002;360(9343):1347.

Chapter 3

70

(20) Peterzan MA, Hardy R, Chaturvedi N, et al. Meta-analysis of dose-response relationships for

hydrochlorothiazide, chlorthalidone, and bendroflumethiazide on blood pressure, serum potassium, and urate.

Hypertension 2012;59(6):1104-1109.

(21) Cappelleri JC, ScD MD. conducting indirect-treatment-comparison and network-meta-analysis studies: report

of the ISPOR Task Force on indirect treatment comparisons good research practices—part 2. Value in Health

2011;14:429-437.

(22) Jansen JP, Fleurence R, Devine B, et al. Interpreting indirect treatment comparisons and network meta-

analysis for health-care decision making: report of the ISPOR Task Force on Indirect Treatment Comparisons

Good Research Practices, part 1. Value in Health 2011;14(4):417-428.

(23) Conroy R, Pyörälä K, Fitzgerald AP, et al. Estimation of ten-year risk of fatal cardiovascular disease in Europe:

the SCORE project. Eur Heart J 2003;24(11):987-1003.

(24) van Dis I, Kromhout D, Geleijnse JM, et al. Evaluation of cardiovascular risk predicted by different SCORE

equations: the Netherlands as an example. European Journal of Cardiovascular Prevention & Rehabilitation

2010;17(2):244-249.

(25) R Development Core Team (2008). R: A language and environment for statistical computing. R Foundation

for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org. .

(26) Elliott WJ, Plauschinat CA, Skrepnek GH, et al. Persistence, adherence, and risk of discontinuation associated

with commonly prescribed antihypertensive drug monotherapies. The Journal of the American Board of Family

Medicine 2007;20(1):72-80.

(27) Greving J, Visseren F, de Wit G, et al. Statin treatment for primary prevention of vascular disease: whom to

treat? Cost-effectiveness analysis. BMJ 2011;342.

(28) Hiligsmann M. The importance of integrating medication adherence in pharmacoeconomic analyses: the

example of osteoporosis. 2011.

(29) Ström O, Borgström F, Kanis J, et al. Incorporating adherence into health economic modelling of

osteoporosis. Osteoporosis Int 2009;20(1):23-34.

(30) Briggs AH, Claxton K, Sculpher MJ. Decision modelling for health economic evaluation. Oxford: Oxford

University Press; 2006.

(31) Den Haag/Heerlen. Central Institute of Statistics Web site. Available at:

statline.cbs.nl/StatWeb/publication/?DM=SLEN&PA=03799ENG&D1=202&D2=31-

35,4044&D3=0&D4=l&LA=EN&VW=T. Accessed April 14, 2012.

(32) Imperial College London. School of public health. Available at:

www1.imperial.ac.uk/publichealth/departments/ebs/projects/eresh/majidezzati/healthmetrics/metabolicriskfac

tors. Accessed April 24, 2012.


statline.cbs.nl/StatWeb/publication/?DM=SLEN&PA=7052ENG&D1=0&D2=1-2&D3=9-21&D4=l&LA=EN&VW=T.

Accessed April 24, 2012.


statline.cbs.nl/StatWeb/publication/?DM=SLEN&PA=7052ENG&D1=42&D2=1-2&D3=9-

21&D4=l&LA=EN&HDR=G3,G1,G2&STB=T&VW=T. Accessed April 24, 2012.


statline.cbs.nl/StatWeb/publication/?DM=SLEN&PA=7461ENG&D1=0&D2=1-2&D3=108-

120&D4=60&LA=EN&VW=T. Accessed April 24, 2012.

(36) O'Lorcain P, Deady S, Comber H. Mortality predictions for colon and anorectal cancer for Ireland, 2003–17.

Colorectal Disease 2006;8(5):393-401.

(37) Brønnum-Hansen H, Davidsen M, Thorvaldsen P. Long-term survival and causes of death after stroke. Stroke

2001;32(9):2131-2136.

(38) Brønnum-Hansen H, Jørgensen T, Davidsen M, et al. Survival and cause of death after myocardial infarction::

The Danish MONICA study. J Clin Epidemiol 2001;54(12):1244-1250.

3


71

(39) Lampe F, Whincup P, Wannamethee S, et al. The natural history of prevalent ischaemic heart disease in

middle-aged men. Eur Heart J 2000;21(13):1052-1062.

(40) Merry AHH, Boer JMA, Schouten LJ, et al. Validity of coronary heart diseases and heart failure based on

hospital discharge and mortality data in the Netherlands using the cardiovascular registry Maastricht cohort

study. Eur J Epidemiol 2009;24(5):237-247.

(41) Vaartjes I, Reitsma J, De Bruin A, et al. Nationwide incidence of first stroke and TIA in the Netherlands.

European Journal of Neurology 2008;15(12):1315-1323.

(42) Michel B, Al M, Remme W, et al. Heart failure Economic aspects of treatment with captopril for patients with

asymptomatic left ventricular dysfunction in The Netherlands. Eur Heart J 1996;17(5):731-740.

(43) Boersma C, Radeva JI, Koopmanschap MA, et al. Economic evaluation of valsartan in patients with chronic

heart failure: results from Val-HeFT adapted to the Netherlands. Journal of Medical Economics 2006;9(1-4):121-

131.

(44) Baeten SA, van Exel NJA, Dirks M, et al. Lifetime health effects and medical costs of integrated stroke

services-a non-randomized controlled cluster-trial based life table approach. Cost Effectiveness and Resource

Allocation 2010;8(1):21.


statline.cbs.nl/StatWeb/publication/?DM=SLEN&PA=71099ENG&D1=0&D2=12,25,38,51,64,77,90,103,116,129,1

42,155,168,181,l&LA=EN&HDR=T&STB=G1&VW=T. Accessed April 24, 2012.

(46) Dutch drug prices [in Dutch]. Available at: www.medicijnkosten.nl. Accessed April 24, 2012.

(47) van Gils PF, Over EA, Hamberg-van Reenen HH, et al. The polypill in the primary prevention of cardiovascular

disease: cost-effectiveness in the Dutch population. BMJ open 2011;1(2).

(48) Hakkaart-van Roijen L, Tan S, Bouwmans C. Handleiding voor kostenonderzoek: Methoden en standaard

kostprijzen voor economische evaluaties in de gezondheidszorg. : College voor zorgverzekeringen; 2010.

(49) College voor zorgverzekeringen. College voor zorgverzekeringen. Guidelines for pharmacoeconomic

research, updated version. 2006; Available at:

http://www.cvz.nl/binaries/content/documents/cvzinternet/en/documents/procedures/guidelines-

pharmacoeconomic-research.pdf. Accessed April 24, 2012.


statline.cbs.nl/StatWeb/publication/?DM=SLNL&PA=37360NED&D1=3&D2=a&D3=a&D4=l&VW=T. Accessed

April 24, 2012.

(51) Stevanovic J, Postma MJ, Pechlivanoglou P. A systematic review on the application of cardiovascular risk

prediction models in pharmacoeconomics, with a focus on primary prevention. European Journal of Preventive

Cardiology 2012;19(2 suppl):42-53.

(52) Montgomery AA, Fahey T, Ben-Shlomo Y, et al. The influence of absolute cardiovascular risk, patient utilities,

and costs on the decision to treat hypertension: a Markov decision analysis. J Hypertens 2003 Sep;21(9):1753-

1759.

(53) Grover S, Coupal L, Lowensteyn I. Preventing cardiovascular disease among Canadians: Is the treatment of

hypertension or dyslipidemia cost-effective? Can J Cardiol 2008;24(12):891.

(54) Löfroth E, Lindholm L, Wilhelmsen L, et al. Optimising health care within given budgets: Primary prevention

of cardiovascular disease in different regions of Sweden. Health Policy 2006;75(2):214-229.

(55) Barton P, Andronis L, Briggs A, et al. Effectiveness and cost effectiveness of cardiovascular disease

prevention in whole populations: modelling study. BMJ: British Medical Journal 2011;343.

(56) Gandjour A, Stock S. A national hypertension treatment program in Germany and its estimated impact on

costs, life expectancy, and cost-effectiveness. Health Policy 2007;83(2):257-267.

(57) Ekman M, Bienfait-Beuzon C, Jackson J. Cost-effectiveness of irbesartan/hydrochlorothiazide in patients with

hypertension: an economic evaluation for Sweden. J Hum Hypertens 2008;22(12):845-855.

(58) Pencina MJ, D'Agostino Sr RB, Larson MG, et al. Predicting the 30-Year Risk of Cardiovascular Disease The

Framingham Heart Study. Circulation 2009;119(24):3078-3084.

Chapter 3

72

(59) Anderson KM, Odell PM, Wilson PWF, et al. Cardiovascular disease risk profiles. Am Heart J 1991;121(1):293-

298.

(60) Ridker PM, Buring JE, Rifai N, et al. Development and validation of improved algorithms for the assessment

of global cardiovascular risk in women. JAMA: the journal of the American Medical Association 2007;297(6):611-

619.

(61) Hippisley-Cox J, Coupland C, Vinogradova Y, et al. Derivation and validation of QRISK, a new cardiovascular

disease risk score for the United Kingdom: prospective open cohort study. BMJ 2007;335(7611):136.

(62) Woodward M, Brindle P, Tunstall-Pedoe H. Adding social deprivation and family history to cardiovascular risk

assessment: the ASSIGN score from the Scottish Heart Health Extended Cohort (SHHEC). Heart 2007;93(2):172-

176.

(63) Merry AHH, Boer JMA, Schouten LJ, et al. Risk prediction of incident coronary heart disease in the

Netherlands: re-estimation and improvement of the SCORE risk function. European Journal of Preventive

Cardiology 2012;19(4):840-848.

(64) De Ruijter W, Westendorp RGJ, Assendelft WJJ, et al. Use of Framingham risk score and new biomarkers to

predict cardiovascular mortality in older people: population based observational cohort study. BMJ: British

Medical Journal 2009;338.

(65) Anand SS, Islam S, Rosengren A, et al. Risk factors for myocardial infarction in women and men: insights

from the INTERHEART study. Eur Heart J 2008;29(7):932-940.

(66) Lindgren P, Fahlstadius P, Hellenius ML, et al. Cost-effectiveness of primary prevention of coronary heart

disease through risk factor intervention in 60-year-old men from the county of stockholm—a stochastic model of

exercise and dietary advice. Prev Med 2003;36(4):403-409.

(67) Lloyd-Jones DM. Cardiovascular Risk Prediction Basic Concepts, Current Status, and Future Directions.

Circulation 2010;121(15):1768-1777.

Personal communication: Dr. Tony Fitzgerald (Department of Epidemiology & Public Health/Statistics, University

College Cork; Date:20-07-2012).

3


73

Annex

Incidence description and costs of

acute heart failure in the Netherlands

Jelena Stevanović, Maarten J. Postma

This annex is based on an oral presentation given at ISPOR 17th Annual European

Congress.

Chapter 3

74

ABSTRACT

Background: Acute heart failure (AHF) is frequent, severe and costly, however detailed

population-based epidemiological data are currently unavailable for the Netherlands. Our

aim was to characterize the incidence, patient characteristics and outcomes of AHF, and

estimate associated hospitalizations costs in the Netherlands.

Methods: Using the 2010 and 2011 National Medical Registration (LMR), we identified all

patients admitted to hospital with AHF as a primary diagnosis. We extracted hospital data

on resource use in 2011 associated with AHF, as it provided more contemporary

information, in order to calculate its cost. Major components of hospital resource use

included in this estimate were inpatient care, diagnostic procedures and major

interventions conducted during the hospital stay and autopsy associated with inhospital

mortality.

To quantify the cost of inpatient care, we applied four different scenarios. Scenario 1

assumed the aforementioned cost to be associated only with the cost of stay in general

wards. Other scenarios (2-4) assumed the cost of stay in emergency wards to contribute

with 10, 25 and 50% to the cost estimate associated with inpatient care

Results: Analysis of the data identified 33,973 hospitalizations in Netherlands in 2010 due

to main diagnosis of AHF. In 2011, the identified number of hospitalizations was 15,731 for

men and 15,517 for women. The average patient age in AHF in 2011 was 74.4 (±11.8) and

79 (±11.6) years for men and women, respectively. The most common comorbid

conditions were atrial fibrillation, old myocardial infarction, diabetes, unspecified

hypertension, COPD and other diseases of endocardium. The mean hospital length of stay

was 7.9 days and it was similar for men and women. Finally, the estimated hospital cost of

AHF in scenario 1 was €3,902 for women and €4,044 for men, while in scenarios 2, 3 and 4

the cost was even higher reaching up to €8,088 for women and €8,078 in men.

Conclusions: Our study provides important insights into the characteristics and costs of

AHF hospitalizations in the Netherlands. Further analysis directed to linking the available

LMR data with hospital claims data will indicate the cost of medication use and therefore

provide more comprehensive hospital AHF cost estimates.

3


75

INTRODUCTION

Heart failure (HF) is a syndrome presented with symptoms and signs caused by cardiac

dysfunction that commonly affects elderly people (1). In the United States, approximately

5.8 million people have HF, while in Europe this syndrome affects more than 15 million

people (2,3). In line with international findings, HF is a highly prevalent condition in the

Netherlands causing a major health and economic burden to the healthcare system.

According to the national RIVM-data, an incidence of 37,000 and prevalence of 127,000

patients was recorded in 2008 (4). In addition, patients with HF may suffer not only from

different levels of physical activity impairment, but also lower quality of life (QoL)

compared to the general population. Specifically, two studies assessing the QoL in Dutch

patients with HF used the EQ-5D instrument (measures an individual’s preference for

living in HF state compared to other health states) to summarize QoL between values of

0.47 and 0.77 (5,6). This range in the QoL observed in the two studies may be due to the

differences in HF severity, time of assessment relative to disease onset and other

underlying patients’ characteristics (e.g. age, comorbidities or type of treatment). In

agreement with the high health burden associated with HF, the economic burden to the

Dutch healthcare system was marked with the cost of €455 million or 0.6% of the total

healthcare budget in 2007 (4). The major share of expenditures (i.e. 60%) was recorded

for hospitalizations for HF. Notably, hospitalizations might occur due to the new onset of

HF (‘de novo’) or deterioration of chronic HF with symptoms that warrant hospitalization.

Both of the aforementioned presentations of HF can be described as acute heart failure

(AHF) according to ESC guidelines for the diagnosis and treatment of acute and chronic

heart failure 2012 (1). Notably, the number of HF hospitalizations in Dutch hospitals

increased in the period 1998-2010(4). Given such a trend in hospitalizations and the

indications of higher, both short- and long-term, mortality risk following index

hospitalization (7), it remains important to have up-to-date information on the incidence

of AHF.

In the last decade, there have been major changes in the clinical practice and treatment of

HF marked with the introduction of new medications and medical devices. These

treatment innovations are expected to lead to a better survival and shorter length of

hospital stay (1,8,9). Though clinical trials on new treatments might indicate favorable

health outcomes in HF patients, the decision on their application in clinical practice are

also made with respect to the economic consequences associated with their use.

Importantly, to conduct robust economic analyses on treatment strategies in HF patients,

analysts should be provided with the most contemporary evidence on both health

outcomes and costs associated with those strategies. Given that the costs of

hospitalization are a major contributor to the total health care expenditures, the up-to-

Chapter 3

76

date information on the use of hospital care resources is essential for such evidence, ergo,

ultimate economic evaluations.

In this study, we aim to characterize the incidence, patient characteristics and outcomes

of AHF, and estimate associated hospital costs in the Netherlands.

METHODS

Data source

The National Medical Registration (LMR) for 2010-2011 was used to identify patients who

were hospitalized for AHF as primary (discharge) diagnosis either due to the new-onset HF

or decompensation of chronic HF. Hospital data coverage was 87% and 82% in 2010 and

2011, respectively. LMR provide information on demographic characteristics (e.g. age,

gender), the main (discharge) diagnosis, secondary (other) diagnoses, date of admission

and discharge, medical procedures performed during the stay (i.e. surgery, imaging, etc.),

prior location of patient (own home, hospital, etc.), discharge destination (death, own

home, another hospital, rehabilitation center, etc.) and characteristics of the hospital (i.e.

general or academic hospital). In the LMR dataset, diagnoses are coded using the

International Classification of Diseases, Ninth Revision (ICD-9).

Identification of cases

Cases were selected for our analysis if their discharge diagnosis was coded with one of the

ICD-9 codes detailed in Table 1. Here the term ‘discharge’ refers to both live discharges

and deaths. Additionally, patients admitted to hospital in 2009 but discharged in 2010

were also selected for the analysis. The absolute number of cases (including day

admissions) was identified for both 2010 and 2011, however, the description of patient

characteristics and outcomes was provided only for cases observed in 2011 as it provided

more contemporary information. Specific outcomes described are the length of hospital

stay, medical procedures performed and patient flow in 2011. When patient

characteristics, such as the presence of comorbidities and age, were analyzed, recurrent

cases of HF hospitalizations were excluded. The presence of comorbidities was identified

by analyzing the secondary diagnosis if reported.

All the analysis was performed using SPSS version 22.

Cost estimation

In this study, a top-down approach was used to estimate the hospital costs associated

with AHF. In this approach, cost components were valued by identifying age- and gender-

specific resource use from annual AHF cases in 2011. This resulted in estimating hospital

resource use (e.g. average length of stay) for an average male and female patient.

Additionally, separate resource use estimates were estimated for different age groups (i.e.

<70, 70-85 and >86). Hospital resource use components were subgrouped into four

3


77

categories: inpatient care, diagnostics, interventions and autopsy. Unit costs associated

with those categories were based on national tariffs or collected from published Dutch

studies (Table 2)(10-12). All costs were inflated to 2014 levels using the Dutch inflation

index.

The LMR provides no information on whether patients were admitted to a general or

emergency hospital ward. Therefore, we applied four different scenarios to quantify the

cost of inpatient care. Scenario 1 assumed the cost of inpatient care to be associated only

with the cost of stay in general wards. Other scenarios (2-4) assumed the cost of stay in

emergency wards to contribute with 10, 25 and 50% to the cost of inpatient care.

Table 1 Identification of patients hospitalized for heart failure

Description ICD-9 code

Hypertensive heart disease malignant with HF 402.01

Hypertensive heart disease benign with HF 402.11

Hypertensive heart disease unspecified with HF 402.91

Hypertensive heart and renal disease malignant with HF 404.01

Hypertensive heart and renal disease malignant with HF and renal failure 404.03

Hypertensive heart and renal disease benign with HF 404.11

Hypertensive heart and renal disease benign with HF and renal failure 404.13

Hypertensive heart and renal disease unspecified with HF 404.91

Hypertensive heart and renal disease unspecified with HF and renal failure 404.93

Congestive HF, unspecified 428.00

Left HF 428.10

Systolic HF unspecified 428.20

Systolic HF acute 428.21

Systolic HF chronic 428.22

Systolic HF acute on chronic 428.23

Diastolic HF unspecified 428.30

Diastolic HF acute 428.31

Diastolic HF acute on chronic 428.33

Combined systolic and diastolic HF unspecified 428.40

Combined systolic and diastolic HF acute 428.41

Combined systolic and diastolic HF chronic 428.42

Combined systolic and diastolic HF acute on chronic 428.43

HF unspecified 428.90

ICD-9, International Classification of Diseases, 9th Revision; HF, heart failure.

Chapter 3

78

Table 2 Cost estimates applied in the analysis.

Key resource items Unit cost (€, 2014)

Length of stay in general ward (daily) 469

Length of stay in emergency ward (daily) 1,524

Electrocardiography 22

Echocardiography 53

Computer tomography scan 202

Doppler 41

Catheterisation 815

ICD implantation 35,893

PTCA 3,553

Pacemaker 10,152

Cost of autopsy 427

ICD, implantable cardioverter defibrillator; PTCA, Percutaneous transluminal coronary angioplasty

RESULTS

In 2010, a total of 33,973 hospitalizations were coded with a main diagnosis of HF at

discharge in Dutch hospitals. Throughout this period, there was a similar level of

hospitalization between men and women (i.e. 49.9%:50.1%). In 2011, the identified

number of hospitalizations was 15,731 for men and 15,517 for women.

The average patient age in AHF was 74.4 (±11.8) and 79 (±11.6) years for men and women

respectively. The distribution of patient ages for those hospitalized for AHF is reported in

Figure 1. The mean number of comorbidities per patient was 2.6 (±2.4). Some of the most

frequent comorbidities in AHF patients were atrial fibrillation (8.3%), old myocardial

infarction (4.1%), diabetes mellitus (3.9%), unspecified hypertension (3.8%), COPD (3.3%)

and other diseases of endocardium (3.2%).

Figure 2 depicts the patients flow for hospitalizations occurring in 2011. In 95% of the

cases, home was identified as a prior location of a patient. The majority of hospitalizations

took place in general hospitals (i.e. 92.6%). Day admissions accounted for 6% of all

admissions. After the hospitalization, 78.7% of patients returned to home, 11.5% were

transferred to nursing care or other hospitals and 9.8% died by the end of their hospital

stay. The mean length of hospital stay was 7.9 (±9) days and it similar for similar for men

and women (Figure 3).

3


79

Me

n

Wo

me

n

Figure 1 Distribution of hospitalized acute heart failure patients according to their age.

Hospital costs of AHF

Inpatient care cost was the largest contributor to the overall AHF cost estimate in both

women and men at 95% and 88%, respectively (Figure 4). Echocardiography and

electrocardiography were identified as some of the most frequently used diagnostic

procedures, while the most common interventions were various catheterizations,

percutaneous coronary angioplasty, implantable cardioverter defibrillator (ICD)

implantation. Based on the resource use of the aforementioned categories and their

respective national cost estimates (Table 2), the estimated hospital cost of AHF in scenario

1 was €3,902 for women and €4,044 for men (Figure 4). After adjusting for age- and

gender-specific level of resource use, the estimated cost of hospitalization due to AHF in

women was in range from €3,623 in women older than 86 years to €4,251 in women

younger than 70 years (Table 3). In men, the corresponding cost of hospitalization was in

range from €3,709 (>86 years old) to €4,453 (<70 years old) (Table 3). Scenarios 2, 3 and 4

assuming the cost of stay in emergency wards to contribute with 10, 25 and 50% to the

inpatient care cost, are presented in Table 4. At a cost of stay in emergency ward

contributing with 10% to the total inpatient care cost, the estimated AHF hospital cost was

€4,739 in women and €4,851 in men. Expanding the emergency ward’s contribution to the

Chapter 3

80

inpatient care cost to 50%, resulted in the total AHF hospital costs of €8,088 in women

and €8,078 in men.

Figure 2 Patient flow in 2011.

Figure 3 Length of hospital stay (days) related to acute heart failure.

3


81

Figure 4 The estimated cost of hospital admissions associated acute heart failure for women (left)

and men (right). Scenario 1.

Table 3 Age- and gender-specific cost of hospitalization (excl. medication use) for AHF. Scenario 1.

Age group Resource Women Men

<70 years Inpatient care €4,251 0.03% € 4,453 0.02%

Diagnostics 16.72% 19.51%

Interventions 83.20% 80.42%

Autopsy 0.05% 0.05%

71-85 years Inpatient care €4,065 94.62% €4,028 88.56%



Autopsy 0.05% 0.06%

>86 years Inpatient care €3,623 98.74% €3,709 97.77%



Autopsy 0.07% 0.06%

Table 4 The estimated cost of hospital admissions associated with AHF. Scenarios assuming the

emergency ward costs to contribute to the inpatient care cost.

Scenario Cost of hospitalization (excl. medication use) for AHF (€)

Women Men

2: emergency ward 10% 4,739 4,851

3: emergency ward 25% 5,995 6,061

4: emergency ward 50% 8,088 8,078

AHF, acute heart failure

Chapter 3

82

DISCUSSION

In the 2010 and 2011 LMR data and for AHF cases only, there were 33,973 and 31,248

hospitalizations, respectively. Based on the hospital resource use associated with AHF in

2011, we estimated the hospital cost of AHF in scenario 1 to be €3,902 for women and

€4,044 for men. However, this is under the assumption that the resource use of inpatient

care was associated only with the costs of stay in general hospital wards. In fact, the

overall hospital cost of AHF rose significantly when the cost of stay in emergency wards

was assumed to contribute to the cost of inpatient care. In particular, with the

contribution of emergency wards costs of 50%, the overall hospital cost of AHF would

double. Furthermore, our findings on the age- and gender-allocation of hospital resources

indicated lower AHF costs in older patients and in women compared to men. The

estimated decrease in costs in older age groups was due to lower use of intervention

resources.

The absolute number hospitalizations due to AHF decreased in 2011 compared to 2010,

however, this finding is based on the LMR data with incomplete hospital coverage.

Therefore, in order to properly assess the most recent trend in AHF hospitalizations, 2012-

2014 Dutch hospital data with complete hospital coverage should be analysed.

Interestingly, our analysis indicated a relatively low presence of certain comorbidities in

AHF patients in comparison to other international findings (13). For example,

hypertension as a secondary diagnosis was reported only in 3.8% of AHF patients in the

2011 LMR data, while in other studies on AHF patients such as the ADHERE, OPTIME and

VMAC, the prevalence of hypertension was approximately 70% (13). Failure of some of the

Dutch hospitals to report secondary diagnoses may be a reason for the aforementioned

findings.

Comparison to other studies

In their report on HF, Koopman et al. indicated a lower number of hospitalizations due to

HF in the Netherlands in 2010 (i.e. 14,808 men and 15,030 women) in comparison to our

observations (4). This discrepancy might be due to the possible differences in the data

coverage at the time of the analysis and cases selection criteria (e.g. the choice of the ICD-

9 codes, inclusion of patients admitted in 2009 but discharged in 2010, or inclusion of

cases with day admissions).

A Dutch study reported the cost for HF hospitalisation in the Netherlands in 1999 to be

€4,795 (14). However, a direct comparison of this cost estimate to our findings is hindered

given the disparity in reporting hospital resources between the two studies. Moreover,

changes in clinical practice and introduction of new innovations for the treatment of AHF

during the last decade may contribute to potential differences in cost estimates.

3


83

Limitations

Our analysis is confronted with several limitations that may restrict the interpretation of

results. One limitation is that the identification of AHF cases in our study relies on

discharge diagnosis. Specifically, problems with the interpretation of the data may arise if

incomplete and/or imprecise coding of diagnosis occurs. A couple of studies provided

evidence of the number of hospital events related to HF to be underestimated when the

analysis was solely based on hospital discharge codes (15-17). Furthermore, incomplete

coding might also be the reason for a relatively low prevalence of certain comorbidities

(e.g. hypertension, diabetes, atrial fibrillation, etc.) in AHF patients that our study

observed. Another limitation of our study concerns the missing information on medication

resource use in AHF. This inevitably underestimates the estimated hospital AHF cost.

Further investigation should be directed to linking the available LMR data with hospital

claims data. This will indicate the cost of medication use and therefore provide more

comprehensive hospital AHF cost estimates. Notably, such a complete analysis should

provide a valid, up-to-date hospital AHF cost estimate that could find its use in economic

evaluation on HF treatments.

Acknowledgements

The study was supported by Novartis. The authors declare study results were not

influenced by Novartis funding.

Chapter 3

84

REFERENCES

(1) McMurray JJ, Adamopoulos S, Anker SD, et al. ESC Guidelines for the diagnosis and treatment of acute and

chronic heart failure 2012. European journal of heart failure 2012;14(8):803-869.

(2) Dickstein K, Cohen-Solal A, Filippatos G, et al. ESC Guidelines for the diagnosis and treatment of acute and

chronic heart failure 2008‡. European journal of heart failure 2008;10(10):933-989.

(3) WRITING GROUP MEMBERS, Lloyd-Jones D, Adams RJ, et al. Heart disease and stroke statistics--2010 update:

a report from the American Heart Association. Circulation 2010 Feb 23;121(7):e46-e215.

(4) Koopman C, van Dis I, Bots ML, Vaartjes I. Feiten en cijfers Hartfalen. 2012; Available at:

https://www.hartstichting.nl/downloads/factsheet-hartfalen. Accessed 12/30.

(5) Janssen DJ, Franssen FM, Wouters EF, et al. Impaired health status and care dependency in patients with

advanced COPD or chronic heart failure. Quality of life research 2011;20(10):1679-1688.

(6) Kraai I, Luttik M, Johansson P, et al. Health-related quality of life and anemia in hospitalized patients with

heart failure. Int J Cardiol 2012;161(3):151.

(7) Vaartjes I, Hoes AW, Reitsma JB, et al. Age- and gender-specific risk of death after first hospitalization for

heart failure. BMC Public Health 2010 Oct 22;10:637-2458-10-637.

(8) Dickstein K, Vardas PE, Auricchio A, et al. 2010 Focused Update of ESC Guidelines on device therapy in heart

failure. European journal of heart failure 2010;12(11):1143-1153.

(9) Teerlink JR, Cotter G, Davison BA, et al. Serelaxin, recombinant human relaxin-2, for treatment of acute heart

failure (RELAX-AHF): a randomised, placebo-controlled trial. The Lancet 2012.

(10) Nederlandse Zorgautoriteit beleidsregel CU-2065. Tarieflijst Instellingen 2012. Available at:

www.nza.nl/Bijlage 2 bij Beleidsregel CU-2065 Tarieflijst Instellingen 2012. Accessed 09/01, 2014.

(11) Heeg BM, Peters RJ, Botteman M, et al. Long-term clopidogrel therapy in patients receiving percutaneous

coronary intervention. Pharmacoeconomics 2007;25(9):769-782.

(12) Ringborg A, Nieuwlaat R, Lindgren P, et al. Costs of atrial fibrillation in five European countries: results from

the Euro Heart Survey on atrial fibrillation. Europace 2008 Apr;10(4):403-411.

(13) Adams Jr KF, Fonarow GC, Emerman CL, et al. Characteristics and outcomes of patients hospitalized for heart

failure in the United States: rationale, design, and preliminary observations from the first 100,000 cases in the

Acute Decompensated Heart Failure National Registry (ADHERE). Am Heart J 2005;149(2):209-216.



131.

(15) Merry AH, Boer JM, Schouten LJ, et al. Validity of coronary heart diseases and heart failure based on hospital

discharge and mortality data in the Netherlands using the cardiovascular registry Maastricht cohort study. Eur J

Epidemiol 2009;24(5):237-247.

(16) Khand AU, Shaw M, Gemmel I, et al. Do discharge codes underestimate hospitalisation due to heart failure?

Validation study of hospital discharge coding for heart failure. European journal of heart failure 2005;7(5):792-

797.

(17) Ingelsson E, Ärnlöv J, Sundström J, et al. The validity of a diagnosis of heart failure in a hospital discharge

register. European journal of heart failure 2005;7(5):787-791.

Chapter 4

Multivariate meta-analysis of preference-based

quality of life values in coronary heart disease

Jelena Stevanović, Petros Pechlivanoglou, Marthe A. Kampinga, Paul F.M. Krabbe, Maarten

J. Postma

This chapter is based on an oral presentation given at 36th Annual North American

Meeting Medical Decision Making. Submitted manuscript.

Chapter 4

88

ABSTRACT

Background: There are numerous health-related quality of life (HRQol) measurements

used in coronary heart disease (CHD) in the literature. However, only values assessed with

preference-based instruments can be directly applied in a cost-utility analysis (CUA).

Objective: To summarize and synthesize instrument-specific preference-based values in

CHD and the underlying forms of CHD, stable angina and post-acute coronary syndrome

(post-ACS), for developed countries, while accounting for study-level characteristics and

within-study correlation.

Methods: A systematic review was conducted to identify studies reporting preference-

based values in CHD. A multivariate meta-analysis was applied to synthesize the HRQoL

values. Meta-regression analyses examined the effect of study level covariates age,

publication year, prevalence of diabetes and gender.

Results: A total of 40 studies providing preference-based values were detected.

Synthesized estimates of HRQoL in post-ACS ranged from 0.64 (Quality of Well-Being) to

0.92 (EuroQol European ”tariff”), while in stable angina they ranged from 0.64 (Short form

6D) to 0.89 (Standard Gamble). Similar findings were observed in estimates applying to

general CHD. No significant improvement in model fit was found after adjusting for study-

level covariates. Large between-study heterogeneity was observed in all the models

investigated.

Conclusions: The main finding of our study is the presence of large heterogeneity both

within and between instrument-specific HRQoL values. This large between-study

heterogeneity among instrument-specific HRQoL estimates may be explained by both

observed and unobserved underlying methodological differences across instruments and

study-level characteristics. Current economic models in CHD ignore this between-study

heterogeneity. Multivariate meta-analysis can quantify this heterogeneity and provide

estimates for CUAs.

4

Quality of life values in coronary heart disease: multivariate meta-analysis

89

INTRODUCTION

A large number of health-related quality of life (HRQoL) measures for patients with

coronary heart disease (CHD) is available in the literature (1-4). Those measures either

describe the HRQoL of patients suffering from CHD overall or distinguish across patients

suffering from one of the underlying forms of CHD, specifically stable angina or post-acute

coronary syndrome (post-ACS). This interest in estimating the level of HRQoL in CHD is

mainly due to its increasing economic and clinical burden (5), the number of CHD

prevention and treatment strategies available, and the necessity to assess the impact of a

treatment on HRQoL for use as input parameter in cost-utility analyses (CUAs) (6,7).

From the abundance of HRQoL measurements in CHD, only the preference-based HRQoL

values can be directly applied in CUA (6,8). These values express the individual’s

preference for living with CHD compared to other health states on an interval scale where

the value of zero is assigned to death, and the value of one to full health (7). Their

application in CUA is essential for the computation of the quality-adjusted life years

(QALYs) health summary measure (7,9,10). Preference-based values can be generated

with direct elicitation techniques, such as the time trade-off (TTO), the standard gamble

(SG) and the rating scale (RS). Another approach is based on preference-based multi-

attribute instruments, such as, the EuroQol 5D (EQ-5D), Short form 6D (SF-6D)) health

utility index (HUI) and the quality of well-being (QWB) scale (7). Alternatively, in the

absence of HRQoL values obtained by one of the aforementioned instruments, other

HRQoL measurements can be mapped onto a preference-based estimate (11-13).

There is significant variation in the published literature with respect to the preference-

based HRQoL value of patients with CHD (1,14-16). Methodological differences between

direct preference-based techniques such as the exact specifications of the questions on

individuals’ preferences, have been shown to have considerable impact on HRQoL values

(7). Some of the differences in HRQoL values measured with preference-based multi-

attribute instruments reflect the variation across instruments on the sensitivity of

different health attributes across different severity levels as well as the use of different

direct valuation techniques (7). Finally, HRQoL values can largely vary due to differences in

study-level covariates such as patients’ characteristics (e.g. underlying CHD forms, age,

and comorbidities), types of treatment applied or time points of measuring HRQoL relative

to the disease onset or treatment initiation.

All the aforementioned differences in the underlying methodology indicate that the choice

and interpretation of the HRQoL value to be applied in CUA need to consider the impact of

both instrument-specific properties sometimes limited to country-specific

pharmacoeconomic guidelines’ requirements (e.g. the EQ-5D in the UK(8)), and study-level

covariates for specific HRQoL values. The choice of the instrument can have a significant

impact on the CUA outcome and the interpretation of CUA results. For example, Sach et

Chapter 4

90

al. compared the outcomes of a CUA on different treatment options for patients with knee

pain, where HRQoL was measured with both the EQ-5D and SF-6D instruments (17). The

authors demonstrated that the treatment option that was estimated to be the most

favourable differed depending on the HRQoL instrument used in the aforementioned

patient population. Notably, it is this sensitivity of CUA outcomes on the HRQoL values

(17) that urges for robust estimates of HRQoL. This issue becomes even more complex

when multiple candidate sources for the HRQoL value exist in the literature.

Evidence based decision making requires the use of all available relevant evidence for

decision making through a form of evidence synthesis as such exercise can provide better

estimates around the mean and variance of HRQoL. Examples of meta-analysis in HRQoL

can be found in various disease areas (18-20). Those meta-analyses accounted for the

underlying differences in the instruments by examining their impact on the level of HRQoL

in a meta-regression framework (19,21), and by conducting univariate meta-analyses for

the values measured with each instrument (22). The application of meta-analysis on

HRQoL values is straightforward when all values are measured with the same instrument.

However, a number of studies provide multiple and correlated HRQoL values measured in

the same population but using different instruments. Conducting separate univariate

meta-analyses for each HRQoL is inappropriate as ignoring the within-study correlation

might lead to biased mean and standard error (SE) estimates (23,24). Instead, multivariate

meta-regression analysis is recommended (23,24). Therefore, in this study we aim to

systematically summarize and synthesize the published preference-based HRQoL values in

CHD and its underlying disease-forms (i.e. stable angina and post-ACS) for developed

countries. To account for the underlying differences in the instruments used to measure

the HRQoL, and the correlation between instrument-specific values both within and

between studies, the synthesis of HRQoL values is conducted on an instrument-specific

level. Additionally, the impact of study-level covariates on HRQoL is explored in regression

analysis.

METHODS

Data collection

This study was conducted following the PRISMA guidelines. Relevant studies reporting

HRQoL in CHD were independently screened and systematically reviewed by two team

members [JS and PP]. Studies were searched using MEDLINE and EMBASE. The search

terms used were: (“coronary disease” or “coronary heart disease” or “myocardial

infarction” or “angina” or “acute coronary syndrome”) and (“utility” or “quality of life” or

“outcome assessment”) and ("Health Utilities Index" or "quality of well-being" or "rating

scale" or "standard gamble" or "time trade-off" or "15D" or "SF-6D" or "EQ-5D" or

"HALex"). The search was limited to studies applying to developed countries published

4


91

between 1990 and November 2014. By including only studies from developed countries,

we aimed to limit the variation on HRQoL associated with the relation between socio-

economic status and HRQoL (25). Additionally, the references of the identified articles and

other systematic reviews were searched for relevant studies not included in the above-

mentioned databases (snowballing). Studies were considered eligible for this analysis if

they: 1) applied a preference-based instrument of measuring HRQoL (TTO, SG, RS, EQ-5D,

SF-6D, SF-15D, HUI, QWB and HALex) 2) reported mean preference-based HRQoL values

measured three months or more after the initiation of CHD-treatment or after the onset

of CHD, 3) reported standard deviations (SDs) and sample sizes or confidence intervals

(CIs)/SEs of those measurements. Duplicate studies were excluded as well as editorials,

letters, clinical conference abstracts, reviews and studies that reported median but not

mean HRQoL values. Disagreements were resolved through discussion.

Data were extracted from the eligible studies regarding study origin (authors, publication

year, country), study design, participants (age, study sample, percentage of men,

percentage of diabetics, underlying form of CHD) as well as the HRQoL (mean value, SD or

CI/SE of the mean), type of instrument for measuring HRQoL and correlation coefficients

between instrument-specific values.

Assumptions and data adjustments

We distinguished between two underlying CHD forms: stable angina and post-ACS. The

stable angina and the post-ACS groups comprised of stable angina patients and patients

with unstable angina or myocardial infarction, respectively, in whom HRQoL was

measured at least three months after diagnosis or treatment initiation. Our analysis was

limited to HRQoL values measured three months after onset of CHD or treatment

initiation hypothesizing that the impact of the acute disease onset or a treatment effect

on HRQoL will be stabilized by then. When allocating recorded HRQoL values to underlying

CHD forms based on the patients’ characteristics reported in the studies, certain

assumptions were made. If information on patients’ characteristics was insufficient to

provide their allocation to either stable angina or post-ACS, these HRQoL values were

analysed as values for general CHD patients. Furthermore, no distinction was made

between the HRQoL values measured in patient subgroups other than the ones reflecting

underlying CHD forms (e.g. different treatment or socio-economic subgroups). In cases

where studies provided HRQoL values in a subgroup classification that was not of interest,

the HRQoL values across subgroups were synthesized to provide a weighted mean and

variance estimate for a specific CHD form per study.

Statistical model

A multivariate meta-analysis was used to estimate synthesized, instrument-specific HRQoL

estimates in post-ACS, stable angina and general CHD (23). This approach accounts for the

correlation (both within and between studies) between HRQoL values assessed with

Chapter 4

92

different instruments and was extended to a multivariate meta-regression analysis to

account for the impact of study-level covariates where appropriate. The general structure

of the model applied is presented below in matrix form. E< = F<G< + H< + I< i = 1,…,n (1)

Here E< stands for a vector of size p whose elements comprise the instrument-specific

HRQoL values for study i, F< is a matrix of p instruments and k covariates, G< is the vector

of regression coefficients of size k, H< is a vector of random-effects terms of size p and I< is

a vector of random sampling errors of size p. We assumed that H<~MVN(0, N, where

N = O PAC ⋯ RS(TTUPAPV⋮ ⋱ ⋮RS(TTUPAPV ⋯ PVCY. (2)

Here ∆ represents the between-study variance–covariance matrix and its elements are PVC,

the between-study instrument-specific variance, and RS(TTU, the between-study

correlation coefficient assessed when measuring the HRQoL values with Z and Z[ instruments. Additionally, it was assumed that I<~MVN(0, \<, where

\< = O ]<AC ⋯ RTTU]<A]<V⋮ ⋱ ⋮RTTU]<A]<V ⋯ ]<VCY (3)

The matrix \< is the within-study variance–covariance matrix with elements ]<VC , the

within-study instrument-specific variance and RTTU , the within-study correlation

coefficient.

A common problem in multivariate meta-regression is the presence of missing data for

variables of interest in eligible studies. In our analysis, missing data were anticipated for

some of the E< elements and their corresponding variances, as well as for some within-

study correlation coefficients. Missing values in E< may occur when HRQoL in a particular

study was not estimated with all the instruments included in the meta-regression. We

resolved this by setting the missing values in E<as equal to zero and their corresponding

variances equal to an arbitrary large number (1,000) (11). In this way the contribution of

these values to the summarized HRQoL estimate was insignificant. The problem related to

missing values of RTTU was resolved by retrieving correlation estimates from a more

general population without severe comorbidities. Elements of RTTU that still remained

missing were assumed to be equal to zero (24). In order to observe the impact of ignoring

the presence of correlation and to account for possibly different values of correlation

coefficients, we undertook a sensitivity analysis by assuming correlation coefficients to

take values of 0 and 0.5 (26).

In the regression analysis, the covariates incorporated in F< were examined for their

statistical significance and their impact on reducing some of the between-study

heterogeneity on those HRQoL values. There are no guidelines to suggest the minimal

number of studies per outcome of interest (i.e. an instrument-specific HRQoL value)

4


93

required for the regression analysis to be plausible in the multivariate setting. For this

reason, we adopted the guidelines for the univariate setting that suggest a data set

sample size of approximately 10 measurements may be sufficient for conducting a

regression analysis with one covariate at a time (26). The regression analysis was limited

to those outcomes where at least 10 measurements were available.

In order to indicate the extent of heterogeneity in the true population level of HRQoL that

is unexplained by study-level covariates and random sampling error in the multivariate

setting, we calculated the ^_C and ^̀C statistics, recently suggested by Jackson et al (27). We

used ^̀C to measure the impact heterogeneity for all HRQoL estimates jointly and the ^_C

statistic to measure the level of heterogeneity both jointly and separately for instrument-

specific HRQoL estimates (27). For the estimation of the ^_C, the aCstatistic was used as a

basis. In the multivariate setting, the aCstatistic can be interpreted as the ratio of the

volumes of CIs for summarized estimates under the random effect model and the volumes

of CIs for summarized estimates under the fixed effects models (27). Under such a

notation of the aCstatistic, Jackson et al. suggested that the ^_C can be estimated as ^_C = (aC − 1 aC⁄ (27). Furthermore, the ^̀C statistic, was estimated as ^̀C = (bC − 1 bC⁄ , where the bC statistic represents the ratio of a generalized version of

Cohran’s Q statistic and its associated degrees of freedom (27).

The multivariate meta-regression analysis was performed using the package mvmeta (28)

in the statistical software R (version 2.15.3) (29).

RESULTS

Study characteristics

A total of 40 eligible studies representing over 30,575 patients were identified after the

inclusion and exclusion criteria were applied (Figure 1). The main characteristics of these

studies are detailed in Table 1 (1-4,14-16,30-64). The studies were published between

1991 and 2014, with most of them between 2005 and 2010. Of the 40 studies included, 10

referred to the UK, 8 to the US, 6 to the multiple country settings, 4 to Germany, 3 to

Canada and Finland, 2 to Norway and Sweden, and 1 to each of Australia and Korea.

Various study designs were implemented, ranging from randomised clinical trial to

different observational study designs. Of the HRQoL values present in those 40 studies,

31% were observed in patients with stable angina and 29% in patients with post-ACS. The

remaining 40% of the HRQoL values were associated with patients suffering from any form

of CHD (including stable angina and post-ACS). The majority of patients were men (71%)

with an average age of 65.35 years. There was large variation in patients’ characteristics,

such as the presence and prevalence of various comorbidities (e.g. prevalence of diabetes

was in range 7-38%), and treatment under evaluation (e.g. surgical procedures,

pharmacological treatment or cardiovascular rehabilitation) across the studies. When

Chapter 4

94

reported, the time of HRQoL assessment from disease onset or treatment initiation was

between 4 months and 10 years.

The most commonly applied instrument for measuring HRQoL values was the EQ-5D

(63.5%), while 15D, QWB, SF-6D, HUI, SG, TTO, RS and HALex were less prevalent (7.7, 7.7,

5.8, 5.8, 3.8, 1.9, 1.9 and 1.9%, respectively). The values measured with the EQ-5D

instrument varied by the type of TTO “tariffs” utilized (i.e. UK, US, Europe and Korea). The

scoring of most of the EQ-5D values was based on the UK “tariff” (53%) while US,

European and Korean “tariff” were less present (19, 8, 8 and 3%, respectively). Three

studies provided no explicit information on the EQ-5D scoring “tariff” used, however, in

order to allow for the evidence synthesis these values were grouped together with the UK

“tariffs”. The values measured with the HUI instrument were presented in the studies as

both mark 2 (i.e. HUI2) and mark 3 (i.e. HUI3).

Finally, the correlation coefficients between the instrument-specific HRQoL values,

necessary for conducting a multivariate meta-analysis on the data set formed, were

reported in only one of the studies included in the data set (39). Therefore, some of the

correlation coefficients were retrieved from other studies on cardiovascular patients or

general populations without severe comorbidities (Table 2)(39,44,65-68). Nevertheless, a

great number of within the instrument-specific correlation coefficients remained missing.

Multivariate meta-analysis estimates in post-ACS, stable angina and general CHD

Table 3 summarizes the instrument-specific estimates synthesized through multivariate

meta-analysis in the post-ACS, stable angina subgroups and general CHD assessed on the

full data set. The values for HRQoL in estimates synthesized in post-ACS ranged from 0.64

(QWB) to 0.92 (EQ-5D European ”tariff”), while in stable angina it ranged from 0.64 (SF-

6D) to 0.89 (SG). In general CHD, the values ranged from 0.60 (HALex) to 0.89 (SG).

Between-study SDs and variance-covariance matrices for HRQoL in the post-ACS, stable

angina subgroups and general CHD, when these parameters could be estimated, are

reported in Tables 4-6.

In this evidence synthesis, some of the instrument-specific HRQoL values included in the

data set were present only as single inputs. Notably, the output of the multivariate meta-

analysis in the aforementioned cases reflected the initial instrument-specific HRQoL

inputs. Because the EQ-5D UK ”tariff” values were the only instrument-specific values

available as multiple inputs in both post-ACS and stable angina subgroups, a relatively fair

comparison of the level of summarized HRQoL between the two subgroups would only be

possible for this instrument-specific subgroup. This comparison indicated slightly lower

estimates in post-ACS (i.e. 0.76) than the ones in stable angina (i.e. 0.78).

4


95

Figure 1 PRISMA flow diagram summarizing the study selection process. HRQoL, health-related quality of life; SD, standard deviation; SE, standard error; CI, confidence interval.

Large variations are noticeable across instrument-specific estimates of HRQoL in CHD

presented in Table 3. The highest level of HRQoL was observed for the SG, TTO and 15D

estimates, which were followed by slightly lower EQ-5D and HUI estimates, while the RS,

SF-6D, QWB and HALex were the instruments with lowest HRQoL estimates.

Interestingly, substantial unexplained between-study but within-instrument heterogeneity

was observed in all the multivariate meta-analysis models with the most excessive levels

of heterogeneity observed in overall CHD (Table 3). There was a general agreement

between both ^̀C and ^_C statistics on the level of heterogeneity.

Regression analysis

The EQ-5D UK “tariff” values were the only instrument-specific values available with more

than 10 observations per instrument used and therefore in line with our requirement for

regression analysis. This led to reducing the analysis from a multivariate to a univariate

meta-regression where the impact of disease subgroup, age, publication year, and

prevalence of diabetes and of men was examined on the EQ-5D UK “tariff” values subset.

Here, increasing age was found to generally reduce the level of HRQoL while prevalence of

diabetes and higher proportion of men increased its level. The impact of publication year

was inconclusive. However, none of the covariates examined was found to provide a

Chapter 4

96

significant improvement in the model fit nor did they reduce the between-study

heterogeneity (Table 7).

Sensitivity analysis on the correlation coefficients between instrument-specific values was

undertaken. The result of this sensitivity analysis was that the model was overall robust to

different values of the correlation coefficients (Table 8). Moreover, the SEs of instrument-

specific estimates were generally insensitive to ignoring the correlation between the

instrument-specific values or setting the values of correlation coefficients to 0.5.

DISCUSSION

This is the first study that systematically summarized and synthesized instrument-specific

preference-based HRQoL values in CHD and its underlying disease-forms, post-ACS and

stable angina in developed countries. Pooled mean HRQoL values for patients in post-ACS,

in stable angina and overall CHD were estimated and a large variation was observed both

within and between the instrument-specific values. This variation could be explained by

the large underlying heterogeneity in the study populations, and the observed and

unobserved variation between the HRQoL instruments. Other factors possibly include the

impact of treatment applied, initial (acute) disease severity level, national and socio-

economic characteristics, various comorbidities present or the time of assessment.

Moreover, the fact that direct TTO “tariffs” assessed in general populations for the

preference-based scoring of the EQ-5D instrument vary across countries (69), suggests

that variations in the HRQoL values assessed across CHD populations from various national

or multinational settings could be larger than the ones observed across general

populations. The variation was larger in overall CHD, which was expected due to variation

in the patients’ form of underlying CHD across studies.

Additional arguments emphasize that unobservable differences in cultural or socio-

economic status may also be present in representatives of a general population selected

for the assessment of tariffs (70). The consequence of this would then be a greater

underlying variability in HRQoL values in CHD assessed with instruments utilizing those

tariffs. In essence, the aforementioned concerns may have a direct impact on the

generalizability of country-specific HRQoL values to various national or multinational

settings.

The regression analysis indicated no significant association between available study-level

covariates and HRQoL estimates. However, the reduction and increase in HRQoL observed

with advancing age and higher proportion of men, respectively is in line with other

published information (71). Surprisingly, studies with a higher proportion of patients with

diabetes had a higher average HRQoL estimate what contrasts the finding by Xie et al.

(72). This may be due to the missing information of the prevalence of diabetes across the

studies as well as an example of ecological fallacy.

4


97

The variation between the instruments in the HRQoL estimates observed in our study is in

agreement with the findings from other studies that demonstrated the differences across

instrument-specific HRQoL values (73-75). Similarly to the study by Johnson et al, we

observed higher levels of HRQoL when summarizing the EQ-5D US “tariffs” compared to

the UK “tariffs”(69). Additionally, SG values commonly exceed TTO values and RS values, a

finding that was also confirmed in our study (76,77). Furthermore, our synthesized HUI3

estimates in CHD were lower compared to the EQ-5D estimates, but similar to the findings

of O’Brien et al. in patients at increased risk of sudden cardiac death and receiving

implantable defibrillator therapy(74). In our study, both the summarized and study-level

SF-6D values were lower than EQ-5D values in CHD what contrasts the observations of

Brazier et al.(1,57). Differences such as the valuation technique, bounds of scale and

sensitivity to change after treatment are only some of potential reasons for the variation

between the instrument-specific HRQoL estimates (73-75).

For the evidence synthesis we applied multivariate meta-analysis given that when

information on various instrument-specific values is sparse, it allows for “borrowing of

strength” from the values available by accounting for the correlation between them

(23,24). Though such an approach may provide more precise summarized estimates

(23,24) than a meta-analysis where the correlation is ignored, this did not hold in our

study due to a high between-study variation.

Potentials for direct comparisons of our analysis to other synthesized HRQoL values are

limited due to the lack of studies meta-analysing preference-based values in CHD. A meta-

analysis of 84 studies identified to address HRQoL in cardiac patients by Kinney et al. may

be one potential comparator to our study (78). Kinney et al. investigated the effect of

pharmacological, surgical, nursing or other treatment on HRQoL and found a small positive

effect of treatment (i.e. standardised mean difference (d) = 0.31). Despite certain

similarity in the patient populations investigated between the two studies can be

acknowledged, numerous differences such as the study inclusion criteria (i.e. any

measurement of HRQoL including the ones of single health attributes), choice of study

effect size (i.e. standardised mean difference), the period of data collection (i.e. 1987-

1991) and the methodology used for conducting meta-analysis (i.e. fixed-effect model)

hamper adequate comparisons.

Another comparator to our study may be a review by Dyer et al. on the EQ-5D values in

cardiovascular disease (CVD). Dyer et al. summarized and stratified the EQ-5D values

across different CVD subgroups (e.g. ischemic heart disease (IHD) such as

angina/myocardial infarction/CHD, heart failure etc.) and, when feasible, across three

severity level categories defined by the percentage of patients in a given group in class

III/IV of NYHA or CCS class. Their stratification of the EQ-5D values with IHD collected at

baseline resulted in the range of values from 0.45 for moderate/severe angina to 0.80 for

mild angina. The authors did not synthesise these values across different severity levels

Chapter 4

98

due to the high heterogeneity observed (i.e. I2 of 82-96%), but suggested more rigorous

study inclusion criteria and the possibility to expand their data set with more recent

publications (i.e. studies published after 2008) as a method to reduce some of the

heterogeneity observed. However, the more rigorous inclusion criteria that our study

proposed such as the inclusion of only mean HRQoL values measured in patients in stable

or post-acute disease state, and incorporating HRQoL values from a wider publication

range (i.e. 1990-2014) in the data set, did not reduce the high heterogeneity observed

across studies. Notably, the heterogeneity indices observed in the study by Dyer et al. and

the ones observed between the EQ-5D values in our study cannot be directly compared.

Differences between the two data sets and the method for disease stratification (i.e. post-

ACS and stable angina subgroups vs. CCS class categories reported in ICH) are limiting such

a comparison.

Our analysis is confronted with certain limitations. The main limitation of our study is that

we analysed study-level and not patient-level data. Analysing patient-level data might

provide significant improvements in our analysis. If detailed information on patients

currently classified to suffer from CHD was available, this would allow for a more accurate

disease-specific allocation of HRQoL values. Conducting the multivariate meta-analysis on

such a data set could possibly lead to estimates with lower level of between-study

heterogeneity. Another limitation of our study was that we conducted the meta-

regression analysis only on the subset of the EQ-5D UK “tariff” values due to a relatively

small number of studies providing other instrument-specific values. This regression

analysis was also limited with the respect to the variety of covariates investigated.

Covariates such as patients’ socioeconomic status, presence of comorbidities other than

diabetes or the impact of treatment applied were not investigated in the regression

analysis due to their scarce information across the studies (79). Expectedly, this study was

confronted with the missing information on within-study correlation coefficients. This was

solved by retrieving the correlation coefficients reported in other studies on

cardiovascular patients or general populations. The sensitivity of the study results on the

correlation coefficients utilized was tested in the sensitivity analysis. Finally, information

on HRQoL in CHD measured with non-preference-based and disease-specific instruments

was not included in our analysis.

Importantly, our study did not aim to investigate what the most appropriate and reliable

instrument to measure HRQoL in CHD is, but rather to summarize and synthesize all the

available evidence of preference-based instrument-specific values. The decision on the

most robust instrument-specific estimate to be applied in a CUA depends not only on the

appropriateness and reliability of an instrument to measure HRQoL in CHD but also on its

agreement with instrument-specific values available for other health states modelled in

the CUA. Furthermore, some decision-makers might argue that considering the previously

discussed reasons for country- and centre-specific variability in HRQoL values, one should

4


99

simply choose a single country-specific and CHD-form specific HRQoL value. However, in

the case where multiple country-specific values exist or even a single such value is

unavailable, the decision on the most robust value becomes more complex and needs to

rely on an evidence-synthesis exercise Moreover, although distinguishing between

underlying CHD-forms seems as more clinically relevant, CUAs in both primary and

secondary prevention of CHD often model a general CHD health state (80-83). An evidence

synthesis of all available instrument-specific HRQoL values may again be considered to

select a robust HRQoL estimate in overall CHD. Such an estimate could then reflect more

appropriately the complex nature of CHD and its various manifestations. This motivated us

to consider an evidence-synthesis of HRQoL values in overall CHD in this study.

Finally, characterizing the between study heterogeneity not only provides a better mean

estimate for HRQoL values used in CUA but it also provides a better understanding on the

uncertainty around the HRQoL value and how this translates into uncertainty around the

CUA outcomes. This uncertainty, as we showed in our study, is considerable and as it is

mostly found between studies, it is ignored when a single value from an individual study is

selected. Dias et al proposed that in the presence of between-study heterogeneity, using

the predictive distribution is the appropriate way to characterize parameter uncertainty

when embedding synthesized evidence in CUA (84). Researchers using the findings from

this study for economic evaluation purposes will therefore have to rely in generating

values that incorporate both within- and between-study standard deviation (predictive

distribution) provided in the results section of this article.

Conclusions

This study represents the first evidence synthesis of instrument-specific preference-based

HRQoL values in post-ACS, stable angina and CHD in general. Considerable differences in

mean HRQoL estimates were observed both within and between the instruments. These

differences characterized by large between-study heterogeneity may be explained by both

the observed and unobserved methodological differences across instruments and

underlying study-level characteristics. Current CUAs in CHD ignore this between-study

study heterogeneity. Therefore, multivariate meta-analysis can facilitate quantifying this

heterogeneity for HRQoL estimates and offer the means for uncertainty around HRQoL

values to be translated to uncertainty in economic models.

100

Chapter 4

T

ab

le 1

Su

mm

ary

of

stu

die

s re

po

rtin

g p

refe

ren

ce-b

ase

d v

alu

es

in d

eve

lop

ed

co

un

trie

s in

co

ron

ary

he

art

dis

eas

e.

Au

tho

r (Y

ear

), C

ou

ntr

y St

ud

y d

esi

gn

CH

D

sub

gro

up

Stu

dy

sam

ple

(Sa

mp

le s

ize

)

Inst

rum

en

t A

ge

HR

Qo

L

valu

e

(SE

)

Tim

e

(mo

nth

s)*

Me

n

(%)

Dia

be

tics

(%)

Al R

uzz

eh

(1)(

20

08

), U

K

RC

T St

able

angi

na

Ran

do

mis

ed

to

OP

CA

B (

n=

66

) E

Q-5

D U

K

64

0

.65

80

(0.2

50

0)

6

84

2

1

R

and

om

ise

d t

o C

AB

G-C

PB

(n

=5

5)

6

4

0.6

54

0

(0.2

10

0)

8

3

24

Asc

ion

e (

20

04

), U

K(4

7)

RC

T A

CS

Pat

ien

ts w

ith

pre

vio

us

MI o

r

req

uir

em

en

t fo

r b

ypas

s su

rge

ry

ran

do

mis

ed

to

CA

BG

-CP

B (

n=

15

1)

EQ

-5D

UK

6

4

0.8

20

0

(0.0

23

0)

36

8

1

33

P

atie

nts

wit

h p

revi

ou

s M

I or

req

uir

em

en

t fo

r b

ypas

s su

rge

ry

ran

do

mis

ed

to

OP

CA

B (

n=

16

4)

EQ

-5D

UK

6

6

0.8

10

0

(0.0

18

7)

36

8

3

43

Bak

hai

(2)(

20

12

), F

ran

ce,

Spai

n, U

K

Pro

spe

ctiv

e

ob

serv

atio

nal

coh

ort

stu

dy

AC

S P

atie

nts

un

de

rgo

ing

PC

I (N

=1

14

0)

EQ

-5D

*

62

0

.81

00

(0.0

07

1)

12

7

8

NA

Bo

hm

er(

3)(

20

11

), N

orw

ay

Op

en

RC

T

AC

S R

and

om

ise

d t

o e

arly

inva

sive

:

angi

ogr

aph

y an

d P

CI (

n=

13

4)

15

D

61

0

.88

90

(0.0

13

8)

7

80

6

R

and

om

ise

d t

o la

te in

vasi

ve:

angi

ogr

aph

y an

d P

CI (

n=

13

2)

6

2

0.8

72

0

(0.0

15

8)

7

1

8

Bu

rstr

om

(4)(

20

01

),

Swe

de

n

Re

tro

spe

ctiv

e

cro

ss-s

ect

ion

al

surv

ey

Stab

le

angi

na

Re

spo

nd

en

ts w

ith

se

lf-r

ep

ort

ed

angi

na

(N=

19

1)

EQ

-5D

UK

N

A

0.7

00

0

(0.0

18

0)

NA

N

A

NA

Re

spo

nd

en

ts w

ith

se

lf-r

ep

ort

ed

angi

na

(N=

18

0)

RS

NA

0

.69

00

(0.0

15

0)

NA

N

A

NA

Ch

on

g(5

)(2

00

9),

Au

stra

lia

Cro

ss-s

ect

ion

al

surv

ey

(pri

son

er

po

pu

lati

on

)

Stab

le

angi

na

Re

spo

nd

en

ts w

ith

se

lf-r

ep

ort

ed

angi

na

(n=

81

)

SF-6

D

NA

0

.64

40

(0.0

14

6)

NA

N

A

NA

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

angi

na

& M

T (

n=

17

)

0.6

20

0

(0.0

41

0)

4

101

Multivariate meta-analysis of preference-based quality of life values in coronary heart disease

A

uth

or

(Ye

ar),

Co

un

try

Stu

dy

de

sign

C

HD

sub

gro

up

Stu

dy

sam

ple

(Sa

mp

le s

ize

)

Inst

rum

en

t A

ge

HR

Qo

L

valu

e

(SE

)

Tim

e

(mo

nth

s)*

Me

n

(%)

Dia

be

tics

(%)

Co

he

n(6

)(2

01

1),

Eu

rop

e

and

No

rth

Am

eri

ca

Pro

spe

ctiv

e

sub

stu

dy

as p

art

of

RC

T

chd

R

and

om

ise

d t

o C

AB

G (

n=

81

0)

EQ

-5D

US

66

0

.84

70

(0.0

05

4)

6

79

3

9

R

and

om

ise

d t

o P

CI (

n=

81

5)

6

6

0.8

61

0

(0.0

05

2)

7

6

38

De

nvi

r(7

)(2

00

6),

UK

P

rosp

ect

ive

ob

serv

atio

nal

stu

dy

chd

P

atie

nts

fro

m h

igh

SE

S gr

ou

p

un

de

rgo

ing

PC

I (n

=8

76

)

EQ

-5D

* 6

2

0.7

50

0

(0.0

08

8)

12

6

9

10

P

atie

nts

fro

m lo

w S

ES

gro

up

un

de

rgo

ing

PC

I (n

=4

62

)

6

2

0.6

30

0

(0.0

14

0)

6

3

13

Du

nn

ing

(20

08

), U

K(4

8)

Pro

spe

ctiv

e c

ross

-

sect

ion

al

chd

P

atie

nts

un

de

rgo

ing

CA

BG

(N

=6

21

) E

Q-5

D*

71

0

.70

00

(0.0

12

3)

12

0

NA

N

A

Elli

s(9

)(2

00

5),

US

Cro

ss-s

ect

ion

al

surv

ey

AC

S P

atie

nts

wit

h h

isto

ry o

f A

CS

(N=

49

0)

EQ

-5D

US

66

0

.81

00

(0.0

08

1)

6

71

N

A

Fryb

ack(

10

)(1

99

3),

US

Lon

gitu

din

al

coh

ort

stu

dy

Stab

le

angi

na

Re

spo

nd

en

ts w

ith

se

lf-r

ep

ort

ed

angi

na

(n=

68

)

QW

B

64

0

.66

00

(0.0

01

5)

12

N

A

NA

AC

S

Re

spo

nd

en

ts w

ith

se

lf-r

ep

ort

ed

AC

S (n

=2

0)

6

4

0.6

40

0

(0,0

17

5)

Gar

ste

r(1

1)(

20

09

), U

S C

ross

-se

ctio

nal

ran

do

m-d

igit

-

dia

lled

su

rve

y

chd

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D n

ot

taki

ng

che

st p

ain

MT

(n=

26

5)

EQ

-5D

US

70

0

.82

00

(0.0

09

2)

NA

5

7

30

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D c

urr

en

tly

taki

ng

che

st p

ain

MT

(n=

21

8)

6

9

0.7

40

0

(0.0

14

2)

4

9

47

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D n

ot

taki

ng

che

st p

ain

MT

(n=

26

5)

HU

I3

70

0

.75

00

(0.0

15

4)

5

7

30

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D c

urr

en

tly

taki

ng

che

st p

ain

MT

6

9

0.5

60

0

(0.0

23

7)

4

9

47

102

Chapter 4

A

uth

or

(Ye

ar),

Co

un

try

Stu

dy

de

sign

C

HD

sub

gro

up

Stu

dy

sam

ple

(Sa

mp

le s

ize

)

Inst

rum

en

t A

ge

HR

Qo

L

valu

e

(SE

)

Tim

e

(mo

nth

s)*

Me

n

(%)

Dia

be

tics

(%)

(n=

21

8)

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D n

ot

taki

ng

che

st p

ain

MT

(n=

26

5)

SF-6

D

70

0

,75

00

(0,0

08

0)

5

7

30

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D c

urr

en

tly

taki

ng

che

st p

ain

MT

(n=

21

8)

6

9

0.6

70

0

(0.0

10

2)

4

9

47

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D n

ot

taki

ng

che

st p

ain

MT

(n=

26

5)

QW

B

70

0

.58

00

(0.0

08

6)

5

7

30

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D c

urr

en

tly

taki

ng

che

st p

ain

MT

(n=

21

8)

6

9

0.5

20

0

(0.0

09

5)

4

9

47

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D n

ot

taki

ng

che

st p

ain

MT

(n=

26

5)

HU

I2

70

0

.80

00

(0.0

20

3)

5

7

30

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D c

urr

en

tly

taki

ng

che

st p

ain

MT

(n=

21

8)

6

9

0.6

90

0

(0.0

27

7)

4

9

47

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D n

ot

taki

ng

che

st p

ain

MT

(n=

26

5)

HA

Lex

70

0

.68

00

(0.0

27

5)

5

7

30

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D c

urr

en

tly

taki

ng

che

st p

ain

MT

(n=

21

8)

6

9

0.5

00

0

(0.0

28

9)

4

9

47

Gri

ffin

(12

)(2

00

7),

UK

P

rosp

ect

ive

ob

serv

atio

nal

stu

dy

chd

P

atie

nts

un

de

rgo

ing

CA

BG

(n

=1

00

) E

Q-5

D U

K

65

0

.66

00

(0.0

31

0)

72

7

8

13

P

atie

nts

un

de

rgo

ing

PC

I (n

=1

08

)

65

0

.65

00

(0,0

28

9)

4

103


A

uth

or

(Ye

ar),

Co

un

try

Stu

dy

de

sign

C

HD

sub

gro

up

Stu

dy

sam

ple

(Sa

mp

le s

ize

)

Inst

rum

en

t A

ge

HR

Qo

L

valu

e

(SE

)

Tim

e

(mo

nth

s)*

Me

n

(%)

Dia

be

tics

(%)

P

atie

nts

re

ceiv

ing

MT

(n

=1

31

)

65

0

.61

00

(0.0

26

2)

Kat

tain

en

(13

)(2

00

5),

Fin

lan

d

Lon

gitu

din

al

ob

serv

atio

nal

stu

dy

chd

P

atie

nts

un

de

rgo

ing

CA

BG

(n

=3

93

) 1

5D

6

3

0.8

58

0

(0.0

00

4)

6

73

2

0

P

atie

nts

un

de

rgo

ing

PC

I (n

=1

53

)

61

0

.82

40

(0,0

00

7)

6

7

Kie

sslin

g(1

4)(

20

05

),

Swe

de

n

Pro

spe

ctiv

e R

CT

ch

d

Pat

ien

ts w

ith

CA

D r

and

om

ise

d t

o

CM

L G

P a

tte

nd

ing

sem

inar

s –

sup

po

rte

d li

pid

-lo

we

rin

g st

rate

gy

(n=

45

)

EQ

-5D

UK

6

5

0.8

00

0

(0.0

04

2)

24

8

2

11

P

atie

nts

wit

h C

AD

ran

do

mis

ed

to

CM

L G

P f

ollo

win

g lo

cal g

uid

elin

es

–

sup

po

rte

d li

pid

-lo

we

rin

g st

rate

gy

(n=

43

)

6

4

0.7

60

0

(0.0

07

0)

8

8

14

P

atie

nts

wit

h C

AD

ran

do

mis

ed

to

CM

L sp

eci

alis

t gr

ou

p –

su

pp

ort

ed

lipid

-lo

we

rin

g st

rate

gy (

n=

16

7)

6

1

0.7

60

0

(0.0

01

4)

7

4

16

Kim

(15

)(2

00

5),

UK

R

CT

AC

S P

atie

nts

ran

do

mis

ed

to

max

imal

MT

plu

s e

arly

co

ron

ary

arte

rio

grap

hy

wit

h p

oss

ible

myo

card

ial r

eva

scu

lari

zati

on

(n=

80

6)

EQ

-5D

UK

6

3

0.7

52

0

(0.0

09

0)

12

6

1

15

P

atie

nts

ran

do

mis

ed

to

max

imal

MT

plu

s is

che

mia

- o

r sy

mp

tom

-

pro

voke

d a

ngi

ogr

aph

y an

d

reva

scu

lari

zati

on

(n

=8

20

)

6

2

0.7

36

0

(0.0

10

0)

6

4

12

Kra

me

r(2

4)(

20

12

),

Ge

rman

y

Qu

asi-

exp

eri

me

nta

l

de

sign

chd

C

HD

pat

ien

ts u

nd

erg

oin

g

de

velo

pe

rs t

rea

tme

nt

pat

hw

ay

(n=

12

8)

EQ

-5D

Eu

rop

e

69

0

.78

12

(0.0

15

3)

6

79

N

A

104

Chapter 4

A

uth

or

(Ye

ar),

Co

un

try

Stu

dy

de

sign

C

HD

sub

gro

up

Stu

dy

sam

ple

(Sa

mp

le s

ize

)

Inst

rum

en

t A

ge

HR

Qo

L

valu

e

(SE

)

Tim

e

(mo

nth

s)*

Me

n

(%)

Dia

be

tics

(%)

C

HD

pat

ien

ts u

nd

erg

oin

g u

sers

tre

atm

en

t p

ath

way

(n

=7

0)

6

9

0.6

93

6

(0.0

25

1)

6

1

NA

C

HD

pat

ien

ts u

nd

erg

oin

g co

ntr

ols

tre

atm

en

t p

ath

way

(n

=9

2)

7

1

0.6

64

5

(0.0

26

5)

5

9

NA

Lace

y(1

6)(

20

03

), U

K

Re

tro

spe

ctiv

e

lon

gitu

din

al

surv

ey

AC

S P

ost

-MI p

atie

nts

(N

=2

22

) E

Q-5

D U

K

63

0

.71

80

(0.0

16

3)

12

7

5

NA

Lee

(2

01

4),

Ko

rea(

51

) C

ross

-se

ctio

nal

surv

ey

chd

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D (

N=

70

8)

EQ

-5D

Ko

rea

64

0

.83

1

(0.0

09

0)

82

5

3

27

Lop

on

en

(17

)(2

00

9),

Fin

lan

d

Pro

spe

ctiv

e

ob

serv

atio

nal

stu

dy

Stab

le

angi

na

Pat

ien

ts u

nd

erg

oin

g C

AB

G (

n=

21

3)

15

D

67

0

.85

79

(0.0

07

5)

6

79

2

6

P

atie

nts

un

de

rgo

ing

PC

I (n

=2

08

65

0

.84

56

(0.0

07

3)

6

9

18

Nic

ho

l (1

99

6),

Can

ada(

52

) O

bse

rvat

ion

al

surv

ey-

bas

ed

stu

dy

Stab

le

angi

na

Re

spo

nd

en

ts u

nd

erg

oin

g e

lect

ive

card

iac

cath

ete

riza

tio

n (

n=

41

)

SG

58

0

.83

00

(0.0

42

2)

NA

8

7

NA

No

rris

(18

)(2

00

8),

Can

ada

P

rosp

ect

ive

lon

gitu

din

al

coh

ort

stu

dy

chd

W

om

en

un

de

rgo

ing

cath

ete

riza

tio

n (

n=

47

9)

EQ

-5D

* 6

7

0.8

00

0

(0.0

04

6)

12

0

2

1

M

en

un

de

rgo

ing

cath

ete

riza

tio

n

(n=

17

27

)

6

5

0.9

00

0

(0.0

02

4)

12

1

00

2

2

No

we

ls(1

9)(

20

05

), U

S C

ross

-se

ctio

nal

stu

dy

AC

S P

ost

-MI p

atie

nts

(C

CSG

cla

ss I)

(n=

67

)

EQ

-5D

UK

6

5

0.7

80

0

(0.0

24

4)

6

69

NA

P

ost

-MI p

atie

nts

(C

CSG

cla

ss II

)

(n=

17

)

6

5

0.7

20

0

(0.0

28

9)

Pe

tte

rse

n(2

0)(

20

08

),

No

rway

Co

ho

rt s

tud

y

surv

ey-

bas

ed

AC

S P

ost

-MI p

atie

nts

wit

h L

VE

F>5

0%

(n=

16

0)

EQ

-5D

UK

6

4

0.8

30

0

(0.0

14

2)

30

7

1

4

P

ost

-MI p

atie

nts

wit

h L

VE

F=4

0-

50

% (

n=

53

)

6

5

0.7

20

0

(0.0

37

1)

4

105


A

uth

or

(Ye

ar),

Co

un

try

Stu

dy

de

sign

C

HD

sub

gro

up

Stu

dy

sam

ple

(Sa

mp

le s

ize

)

Inst

rum

en

t A

ge

HR

Qo

L

valu

e

(SE

)

Tim

e

(mo

nth

s)*

Me

n

(%)

Dia

be

tics

(%)

P

ost

-MI p

atie

nts

wit

h L

VE

F<4

0%

)

(n=

30

)

6

6

0.7

60

0

(0.0

25

6)

Ose

(25

)(2

01

2),

Au

stri

a,

Be

lgiu

m, U

K, F

ran

ce,

Ge

rman

y, N

eth

erl

and

s,

Slo

ven

ia a

nd

Sw

itze

rlan

d

Cro

ss-s

ect

ion

al

ob

serv

atio

nal

stu

dy

chd

C

HD

pat

ien

ts (

n=

26

56

) E

Q-5

D

Eu

rop

e

68

0

.73

00

(0.0

04

3)

NA

7

0

NA

Pu

skas

(21

)(2

00

4),

US

RC

T St

able

angi

na

Ran

do

mis

ed

to

OP

CA

B (

n=

77

) E

Q-5

D U

K

63

0

.79

00

(0.0

28

5)

12

7

8

33

R

and

om

ise

d t

o C

AB

G (

n=

79

)

64

0

.80

40

(0.0

25

9)

7

7

33

Saar

ni(

22

)(2

00

6),

Fin

lan

d

Surv

ey-

bas

ed

,

stra

tifi

ed

clu

ste

r,

sam

plin

g d

esi

gn

chd

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D (

n=

55

5)

15

D

70

0

.82

10

(0.0

05

0)

NA

5

3

NA

R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d

CH

D (

n=

55

5)

EQ

-5D

UK

7

0

0.6

84

0

(0.0

12

0)

Sch

we

ike

rt(5

4)(

20

09

),

Ge

rman

y

Ob

serv

atio

nal

surv

ey-

bas

ed

stu

dy

AC

S P

atie

nts

wit

h h

isto

ry o

f M

I

(N=

29

50

)

EQ

-5D

UK

6

8

0.8

65

0

(0.0

02

8)

10

9

79

N

A

Sch

we

ike

rt(5

3)(

20

09

),

Ge

rman

y

Co

mp

reh

en

sive

coh

ort

de

sign

AC

S P

atie

nts

un

de

rgo

ing

CR

(in

pat

ien

t

sett

ing)

(n

=1

00

)

EQ

-5D

Eu

rop

e

58

0

.89

10

(0.0

18

3)

12

7

9

17

P

atie

nts

un

de

rgo

ing

CR

(o

utp

atie

nt

sett

ing)

(n

=4

7)

5

5

0.9

41

0

(0.0

25

7)

7

6

14

Serr

uys

(23

)(2

00

1),

th

e

AR

TS

(mu

ltic

en

tre

19

cou

ntr

ies)

RC

T St

able

angi

na

Ran

do

mis

ed

to

PC

I (n

=5

93

) E

Q-5

D U

K

62

0

.86

00

(0.0

06

6)

6

77

1

9

R

and

om

ise

d t

o C

AB

G (

n=

57

9)

6

2

0.8

60

0

(0.0

06

2)

7

6

16

Shah

(55

)(2

00

9),

US

Ob

serv

atio

nal

surv

ey-

bas

ed

stu

dy

AC

S P

atie

nts

(7

0%

) u

nd

erg

oin

g P

CI

(N=

32

)

EQ

-5D

US

89

0

.78

00

(0.0

07

1)

14

3

8

17

106

Chapter 4

A

uth

or

(Ye

ar),

Co

un

try

Stu

dy

de

sign

C

HD

sub

gro

up

Stu

dy

sam

ple

(Sa

mp

le s

ize

)

Inst

rum

en

t A

ge

HR

Qo

L

valu

e

(SE

)

Tim

e

(mo

nth

s)*

Me

n

(%)

Dia

be

tics

(%)

Shar

ple

s(5

7)(

20

07

), U

K

RC

T St

able

angi

na

Pat

ien

ts w

ith

su

spe

cte

d o

r kn

ow

n

CA

D u

nd

erg

oin

g an

gio

grap

hy

(N=

89

8)

EQ

-5D

UK

6

2

0.8

02

0

(0.0

03

5)

6

69

1

3

P

atie

nts

wit

h s

usp

ect

ed

or

kno

wn

CA

D u

nd

erg

oin

g an

gio

grap

hy

(N=

89

8)

SF-6

D

62

0

.64

25

(0,0

01

2)

Shri

ve(5

6)(

20

07

), C

anad

a P

rosp

ect

ive

lon

gitu

din

al

coh

ort

stu

dy

chd

P

atie

nts

(7

0%

) u

nd

erg

oin

g P

CI

(N=

19

54

)

EQ

-5D

US

NA

0

.87

00

(0.0

03

4)

12

7

7

15

P

atie

nts

(7

0%

) u

nd

erg

oin

g P

CI

(N=

19

54

)

EQ

-5D

UK

N

A

0.8

30

0

(0.0

04

5)

Staf

ford

(2

01

1),

UK

(58

) C

ross

-se

ctio

nal

surv

ey

Stab

le

angi

na

Re

spo

nd

en

ts w

ith

se

lf-r

ep

ort

ed

angi

na

(n=

71

7)

EQ

-5D

UK

N

A

0.7

11

0

(0.0

26

5)

NA

N

A

NA

AC

S R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d M

I

(n=

55

0)

EQ

-5D

UK

N

A

0.6

36

0

(0.0

17

1)

NA

N

A

NA

Sulli

van

(59

)(2

00

6),

US

Surv

ey-

bas

ed

,

stra

tifi

ed

clu

ste

r

de

sign

AC

S R

esp

on

de

nts

wit

h s

elf

-re

po

rte

d M

I

(n=

24

4)

EQ

-5D

US

62

0

.70

40

(0.0

16

8)

NA

N

A

NA

Stab

le

angi

na

Re

spo

nd

en

ts w

ith

se

lf-r

ep

ort

ed

angi

na

(n=

22

8)

6

9

0.6

95

0

(0.0

20

1)

Tse

vat

(19

91

), U

S(6

0)

Ob

serv

atio

nal

surv

ey-

bas

ed

stu

dy

AC

S Su

rviv

ors

of

MI (

N=

80

) T

TO

6

1

0.8

70

0

(0.0

02

6)

12

7

9

NA

Vis

ser(

61

)(1

99

4),

UK

C

om

par

ativ

e

stu

dy

Stab

le

angi

na

An

gin

a p

atie

nts

(N

YH

A I)

re

ceiv

ing

MT

aga

inst

ch

est

pai

n (

n=

10

)

QW

B

65

0

.68

00

(0.0

31

6)

NA

7

3

NA

A

ngi

na

pat

ien

ts (

NY

HA

II)

rece

ivin

g

MT

aga

inst

ch

est

pai

n (

n=

25

)

6

6

0.6

20

0

(0.0

18

0)

A

ngi

na

pat

ien

ts (

NY

HA

III)

re

ceiv

ing

MT

aga

inst

ch

est

pai

n (

n=

21

)

6

7

0.6

20

0

(0.0

26

2)

We

intr

aub

(2

00

8),

US

and

Can

ada(

62

)

RC

T St

able

angi

na

Ran

do

mis

ed

to

PC

I (n

=7

01

) SG

6

3

0.9

30

0

(0.0

06

4)

6

85

3

2

4

107


A

uth

or

(Ye

ar),

Co

un

try

Stu

dy

de

sign

C

HD

sub

gro

up

Stu

dy

sam

ple

(Sa

mp

le s

ize

)

Inst

rum

en

t A

ge

HR

Qo

L

valu

e

(SE

)

Tim

e

(mo

nth

s)*

Me

n

(%)

Dia

be

tics

(%)

R

and

om

ise

d t

o M

T (

n=

66

5)

SG

63

0

.93

00

(0.0

05

8)

6

85

3

5

We

rdan

(63

)(2

01

2),

Ge

rman

y

No

n-

inte

rve

nti

on

al,

mu

ltic

en

tre

op

en

-

lab

el p

rosp

ect

ive

stu

dy

Stab

le

angi

na

An

gin

a p

atie

nts

re

ceiv

ing

MT

(iva

bra

din

e)

(n=

23

30

)

EQ

-5D

UK

6

6

0.8

27

0

(0.0

04

1)

4

59

3

3

Win

kelm

aye

r(6

4)(

20

06

),

UK

, Ire

lan

d a

nd

Ne

the

rlan

ds

Pro

spe

ctiv

e a

nd

cro

ss-s

ect

ion

al

de

sign

as

a p

art

of

RC

T

AC

S M

I pat

ien

ts r

ece

ivin

g M

T

(pra

vast

atin

) (N

=5

46

)

HU

I3

75

0

.73

50

(0.0

11

1)

NA

4

8

11

* T

he

tim

e p

oin

t o

f m

eas

uri

ng

HR

Qo

L re

lati

ve t

o t

he

dis

eas

e o

nse

t o

r tr

eat

me

nt

app

licat

ion

H

RQ

oL,

qu

alit

y o

f lif

e;

SE,

stan

dar

d e

rro

r; A

CS,

acu

te c

oro

nar

y sy

nd

rom

e;

CH

D,

coro

nar

y h

ear

t d

ise

ase

; H

ALe

x, H

eal

th a

nd

Act

ivit

y Li

mit

atio

n I

nd

ex;

HU

I, h

eal

th u

tilit

y in

de

x; Q

WB

, q

ual

ity

of

we

ll-b

ein

g; R

S, r

atin

g sc

ale

; SG

, st

and

ard

gam

ble

; TT

O,

tim

e tr

ade

-off

; U

K,

Un

ite

d K

ingd

om

; U

S, U

nit

ed

Sta

tes;

RC

T, r

and

om

ize

d c

linic

al

tria

l; N

A,

no

t av

aila

ble

; M

I, m

yoca

rdia

l in

farc

tio

n;

MT

, m

ed

ical

tre

atm

en

t; C

AB

G,

coro

nar

y ar

tery

byp

ass

graf

t; C

PB

, ca

rdio

pu

lmo

nar

y b

ypas

s; O

PC

AB

, o

ff-p

um

p c

oro

nar

y ar

tery

b

ypas

s; S

ES,

so

cio

-eco

no

mic

sta

tus;

CA

D,

coro

nar

y ar

tery

dis

eas

e;

CM

L, c

ase

me

tho

d l

ea

rnin

g; C

R,

card

iac

reh

abili

tati

on

; G

P,

gen

era

l p

ract

itio

ne

r;

PC

I, p

erc

uta

ne

ou

s co

ron

ary

inte

rve

nti

on

; C

CSG

; C

anad

ian

Car

dio

vasc

ula

r So

cie

ty C

lass

ific

ati

on

fo

r A

ngi

na

Pe

cto

ris;

LV

EF;

left

ve

ntr

icu

lar

eje

ctio

n f

ract

ion

108

Chapter 4

T

ab

le 2

Co

rre

lati

on

co

eff

icie

nts

be

twe

en

th

e H

RQ

oL

valu

es

asse

sse

d w

ith

dif

fere

nt

inst

rum

en

ts.

Inst

rum

en

t 1

5D

E

Q-5

D U

K

EQ

-5D

Eu

rop

e

EQ

-5D

US

EQ

-5D

Ko

rea

SF-6

D

TT

O

SG

HU

I2

HU

I3

QW

B

RS

HA

Lex

15

D

1

EQ

-5D

UK

0

.39

(67

) 1

EQ

-5D

Eu

rop

e

NA

0

.99

*(6

5)

1

EQ

-5D

US

NA

0

.99

(65

) 0

.99

*(6

5)

1

EQ

-5D

Ko

rea

NA

N

A

NA

N

A

1

SF-6

D

0.5

1(6

7)

0.4

6(6

7)

NA

0

.72

(65

) N

A

1

TT

O

0.0

95

(66

) -0

.05

(68

) N

A

NA

N

A

NA

1

SG

NA

-0

.06

(68

) N

A

NA

N

A

NA

0

.52

(44

) 1

HU

I2

0.5

2(6

7)

0.5

(67

) N

A

NA

N

A

0.4

4(6

7)

NA

N

A

1

HU

I3

0.4

4(6

7)

0.6

1(6

7)

NA

N

A

NA

0

.4(6

7)

0.1

5(6

8)

0

.16

(68

)

0.7

7(6

7)

1

QW

B

0.5

3(6

7)

0.4

4(6

7)

NA

N

A

NA

0

.43

(67

) 0

.41

(39

) N

A

0.4

7(6

7)

0.4

7(6

7)

1

RS

NA

0

.39

(68

) N

A

NA

N

A

NA

0

.46

(44

) 0

.4(4

4)

NA

0

.41

(68

) N

A

1

HA

Lex

NA

N

A

NA

N

A

NA

N

A

NA

N

A

NA

N

A

NA

N

A

1

*C

orr

ela

tio

n c

oe

ffic

ien

ts a

ssu

me

d t

o b

e t

he

sam

e a

s th

e o

ne

s b

etw

ee

n t

he

val

ue

s as

sess

ed

wit

h E

Q-5

D U

K a

nd

oth

er

inst

rum

en

ts.

H

RQ

oL,

he

alth

re

late

d q

ua

lity

of

life

; U

K,

Un

ite

d K

ingd

om

; U

S, U

nit

ed

Sta

tes;

HU

I, h

eal

th u

tilit

y in

de

x; Q

WB

, q

ual

ity

of

we

ll-b

ein

g;

RS,

rat

ing

scal

e;

SG,

stan

dar

d g

amb

le;

TT

O, t

ime

tra

de

-off

; H

ALe

x, H

eal

th a

nd

Act

ivit

y Li

mit

atio

n In

de

x.

4

109


T

ab

le 3

Po

st-a

cute

co

ron

ary

syn

dro

me

, st

able

an

gin

a an

d g

en

era

l C

HD

par

ame

ter

est

imat

es

for

HR

Qo

L an

d m

ult

ivar

iate

he

tero

gen

eit

y st

atis

tics

usi

ng

mu

ltiv

aria

te m

eta

-an

alys

is.

Inst

rum

en

t

N

Est

imat

e p

ost

- A

CS

^̀C ^ _C

N

Stab

le a

ngi

na

^̀C ^ _C

N

CH

D

15

D

1

0.8

81

6 (

0.0

07

4)

1

0.8

51

5 (

0.0

03

7)

4

0.8

49

5 (

0.0

06

9)

EQ

-5D

Eu

rop

e

1

0.9

17

0 (

0.0

10

5)

3

0.7

91

5 (

0.0

62

5)

EQ

5D

Ko

rea

1

0.8

31

0 (

0.0

09

0)

EQ

-5D

UK

8

0

.76

38

(0

.02

46

)

99

.3%

7

0

.77

92

(0

.02

50

)

99

.4%

2

2

0.7

59

1 (

0.0

12

2)

EQ

-5D

US

3

0.7

66

2 (

0.0

30

8)

9

7.3

%

1

0.6

95

0 (

0.0

20

1)

7

0.8

01

2 (

0.0

12

8)

HA

Lex

1

0.5

96

7 (

0.0

07

5)

HU

I2

1

0.7

62

6 (

0.0

06

1)

HU

I3

1

0.7

35

0 (

0.0

11

1)

2

0.7

25

9 (

0.0

11

8)

QW

B

1

0.6

40

0 (

0.0

17

5)

2

0.6

51

7 (

0.0

09

7)

9

7.5

%

4

0.6

28

7 (

0.0

18

9)

RS

1

0.6

90

0 (

0.0

15

0)

SF-6

D

2

0.6

41

3 (

0.0

01

7)

5

1.9

%

3

0.6

85

9 (

0.0

13

1)

SG

2

0.8

88

9 (

0.0

49

2)

9

9.9

%

2

0.8

88

9 (

0.0

49

2)

TT

O

1

0.8

70

0 (

0.0

02

6)

1

0.8

70

0 (

0.0

02

6)

8

6.8

%

70

.7%

6

8.1

%

92

.7%

All

mo

de

l co

eff

icie

nts

wit

h t

he

leve

l of

sign

ific

ance

p <

0.0

01

are

pre

sen

ted

in b

old

.

Stan

dar

d e

rro

rs o

f p

aram

ete

r e

stim

ate

s ar

e s

ho

we

d in

pa

ren

the

ses.

AC

S, a

cute

co

ron

ary

syn

dro

me

; C

HD

, co

ron

ary

he

art

dis

eas

e;

N,

nu

mb

er

of

qu

alit

y o

f lif

e v

alu

es

use

d f

or

est

imat

ion

; U

K,

Un

ite

d K

ingd

om

; U

S, U

nit

ed

Sta

tes;

HA

Lex,

H

eal

th a

nd

Act

ivit

y Li

mit

atio

n In

de

x; H

UI,

he

alth

uti

lity

ind

ex;

QW

B, q

ual

ity

of

we

ll-b

ein

g; R

S, r

atin

g sc

ale

; SG

, sta

nd

ard

gam

ble

; T

TO

, tim

e t

rad

e-o

ff.

Chapter 4

110

Table 4 Between-study SDs and variance-covariance matrix in post-ACS model.

Instrument SD Variance-Covariance matrix

EQ-5D UK EQ-5D US

EQ-5D UK 0.07 -

EQ-5D US 0.05 -0.09 -

ACS, acute coronary syndrome; SD, standard deviation; UK, United Kingdom; US, United States.

Table 5 Between-study SDs and variance-covariance matrix in stable angina model

Instrument SD Variance-Covariance matrix

EQ-5D UK QWB SF-6D SG

EQ-5D UK 0.06 -

QWB 0.01 0.00 -

SF-6D 0.00 1.00 0.01 -

SG 0.06 -0.01 0.02 -0.01 -

SD, standard deviation; UK, United Kingdom; QWB, quality of well-being; SG, standard gamble.

4

111


T

ab

le 6

Be

twe

en

-stu

dy

SDs

and

var

ian

ce-c

ova

rian

ce m

atri

x in

CH

D m

od

el.

Inst

rum

en

t

SD

Var

ian

ce-C

ova

rian

ce m

atri

x

15

D

EQ

-5D

Eu

rop

e

EQ

-5D

UK

E

Q-5

D U

S H

UI3

Q

WB

SF

-6D

SG

15

D

0.0

2

-

EQ

-5D

Eu

rop

e

0.1

1

-0.1

8

-

EQ

-5D

UK

0

.07

1

,00

0

-0.0

4

-

EQ

-5D

US

0.0

6

99

6.2

0

1.7

9

99

6.4

0

-

HU

I3

0.0

2

20

5.1

0

-19

.87

2

03

.60

1

19

.50

-

QW

B

0.0

4

20

6.3

0

-19

.83

2

04

.70

1

20

.60

1

,00

0

-

SF-6

D

0.0

7

-99

6.5

0

2.0

4

-99

6.4

0

-98

5.5

0

-28

6.0

0

-28

7.2

0

-

SG

0.0

6

-0.3

7

-0.6

1

-0.0

4

0.4

8

-4.3

4

-4.3

2

0.3

1

-

CH

D,

coro

nar

y h

ear

t d

ise

ase

; SD

, st

an

dar

d d

evi

atio

n;

UK

, U

nit

ed

Kin

gdo

m;

US,

Un

ite

d S

tate

s; H

eal

th a

nd

Act

ivit

y Li

mit

atio

n I

nd

ex;

HU

I, h

eal

th u

tilit

y in

de

x; Q

WB

, q

ual

ity

of

we

ll-b

ein

g; S

G, s

tan

dar

d g

amb

le.

Ta

ble

7 P

aram

ete

r e

stim

ate

s an

d h

ete

roge

ne

ity

stat

isti

cs f

or

the

EQ

-5D

UK

“ta

riff

” e

stim

ate

s u

sin

g u

niv

aria

te m

eta

-re

gre

ssio

n.

Mo

de

l N

P

ost

-AC

S ^C

N

An

gin

a ^C

N

Ge

ne

ral C

HD

^C

b0

18

0

.75

87

(0

.02

15

) 9

9.6

%

18

0

.75

42

(0

.01

78

) 9

9.5

%

18

0

.76

52

(0.0

16

5)

99

.6%

bD

ise

ase

typ

e

0

.01

65

(0

.03

39

)

0

.05

12

(0

.03

96

)

bA

ge

-0

.00

51

(0

.00

59

)

-0

.00

35

(0

.00

57

)

-0

.00

56

(0

.00

57

)

b0

22

0

.75

88

(0

.03

47

) 9

9.7

%

22

0

.74

97

(0

.03

55

) 9

9.6

%

22

0

.76

04

(0

.03

26

) 9

9.7

%

bD

ise

ase

typ

e

0

.00

59

(0

.03

33

)

0

.02

70

(0

.03

48

)

bP

ub

lica

tio

n y

ea

r

-0.0

00

2 (

0.0

05

2)

0.0

00

5 (

0.0

05

1)

-0.0

00

1 (

0.0

05

0)

7

b0

13

0

.72

75

(0

.05

10

) 9

9.7

%

13

0

.74

05

(0

.04

43

) 9

9.7

%

13

0

.73

40

(0

.04

26

) 9

9.7

%

bD

ise

ase

typ

e

0

.01

25

(0

.04

57

)

0

.03

06

(0

.04

50

)

bD

iab

ete

s

0.0

02

9 (

0.0

02

4)

0.0

01

7 (

0.0

02

6)

0.0

02

7 (

0.0

02

2)

b0

18

0

.59

82

(0

.14

56

) 9

9.7

%

18

0

.55

72

(0

.13

25

) 9

9.7

%

18

0

.59

64

(0

.14

07

) 9

9.7

%

112

Chapter 4

M

od

el

N

Po

st-A

CS

^C N

A

ngi

na

^C N

G

en

era

l CH

D

^C

bD

ise

ase

typ

e

0

.00

28

(0

.03

25

)

0

.05

40

(0

.03

31

)

bM

en

0

.00

24

(0

.00

20

)

0

.00

28

(0

.00

18

)

0

.00

25

(0

.00

19

)

All

mo

de

l co

eff

icie

nts

wit

h t

he

leve

l of

sign

ific

ance

p <

0.0

01

are

pre

sen

ted

in b

old

.

Stan

dar

d e

rro

rs o

f p

aram

ete

r e

stim

ate

s ar

e s

ho

we

d in

pa

ren

the

ses.

A

CS,

acu

te c

oro

nar

y sy

nd

rom

e;

CH

D, c

oro

nar

y h

ea

rt d

ise

ase

; N

, nu

mb

er

of

qu

alit

y o

f lif

e v

alu

es

use

d f

or

est

imat

ion

; U

K, U

nit

ed

Kin

gdo

m;

b0

, in

terc

ep

t.

Ta

ble

8 P

aram

ete

r e

stim

ate

s an

d m

ult

ivar

iate

he

tero

gen

eit

y st

atis

tics

in

th

e s

en

siti

vity

an

alys

is o

n d

iffe

ren

t co

rre

lati

on

co

eff

icie

nts

be

twe

en

HR

Qo

L

inst

rum

en

ts in

CH

D.

Inst

rum

en

t

No

co

rre

lati

on

C

orr

ela

tio

n c

oe

ffic

ien

t =

0.5

15

D

0.8

49

5 (

0.0

06

9)

0.8

49

2 (

0.0

05

9)

EQ

-5D

Eu

rop

e

0.7

91

5 (

0.0

62

6)

0

.79

15

(0

.06

27

)

EQ

5D

Ko

rea

0.8

31

0 (

0.0

09

0)

0.8

31

0 (

0.0

09

0)

EQ

-5D

UK

0

.75

91

(0.0

12

1)

0.7

59

9 (

0.0

12

1)

EQ

-5D

US

0.8

01

1 (

0.0

12

8)

0.7

98

8 (

0.0

14

8)

HA

Lex

0.5

92

6 (

0.0

07

6)

0.5

95

5 (

0.0

07

2)

HU

I2

0.7

59

6 (

0.0

06

2)

0.7

61

5 (

0.0

05

9)

HU

I3

0.7

25

8(0

.01

18

) 0

.72

65

(0

.01

14

)

QW

B

0.6

28

7 (

0.0

18

7)

0.6

30

5 (

0.0

16

7)

RS

0.6

90

0 (

0.0

15

0)

0.6

90

0 (

0.0

15

0)

SF-6

D

0.6

85

9 (

0.0

13

1)

0.6

73

9 (

0.0

11

5)

SG

0.8

88

9 (

0.0

49

0)

0.8

89

0 (

0.0

48

8)

TT

O

0.8

70

0 (

0.0

02

6)

0.8

70

0 (

0.0

02

6)

^̀C

90

.4%

9

1.2

%

All

mo

de

l co

eff

icie

nts

wit

h t

he

leve

l of

sign

ific

ance

p <

0.0

01

are

pre

sen

ted

in b

old

.

Stan

dar

d e

rro

rs o

f p

aram

ete

r e

stim

ate

s ar

e s

ho

we

d in

pa

ren

the

ses.

HR

Qo

L, h

eal

th-r

ela

ted

qu

alit

y o

f lif

e;

CH

D,

coro

nar

y h

ear

t d

ise

ase

; U

K,

Un

ite

d K

ingd

om

; U

S, U

nit

ed

Sta

tes;

HA

Lex,

He

alth

an

d A

ctiv

ity

Lim

itat

ion

In

de

x; H

UI,

he

alth

uti

lity

ind

ex;

QW

B, q

ual

ity

of

we

ll-b

ein

g; R

S, r

atin

g sc

ale

; SG

, sta

nd

ard

gam

ble

; T

TO, t

ime

tra

de

-off

.

4


113

REFERENCES

(1) Garster NC, Palta M, Sweitzer NK, et al. Measuring health-related quality of life in population-based studies of

coronary heart disease: comparing six generic indexes and a disease-specific proxy score. Quality of Life Research

2009;18(9):1239-1247.

(2) Kim J, Henderson RA, Pocock SJ, et al. Health-related quality of life after interventional or conservative

strategy in patients with unstable angina or non–ST-segment elevation myocardial infarction: one-year results of

the third Randomized Intervention Trial of unstable Angina (RITA-3). J Am Coll Cardiol 2005;45(2):221-228.

(3) Norris CM, Spertus JA, Jensen L, et al. Sex and gender discrepancies in health-related quality of life outcomes

among patients with established coronary artery disease. Circulation: Cardiovascular Quality and Outcomes

2008;1(2):123-130.

(4) Nowels D, McGloin J, Westfall JM, et al. Validation of the EQ-5D quality of life instrument in patients after

myocardial infarction. Quality of life research 2005;14(1):95-105.

(5) World Health Organization. World Health Organization. Fact sheet No 310 'Top ten causes of death'. 2008;

Available at: http://www.who.int/mediacentre/factsheets/fs310/en/index.html. Accessed 11/16, 2012.

(6) Drummond M. Introducing economic and quality of life measurements into clinical studies. Ann Med

2001;33(5):344-349.

(7) Drummond MF, Sculpher MJ, Torrance GW. Methods for the economic evaluation of health care

programmes. : Oxford University Press, USA; 2005.

(8) National Institute for Health and Care Excellence. Guide to the methods of technology appraisal 2013. 2013;

Available at: http://publications.nice.org.uk/guide-to-the-methods-of-technology-appraisal-2013-pmg9.

Accessed 10/08, 2013.

(9) Rawlins MD, Culyer AJ. National Institute for Clinical Excellence and its value judgments. BMJ

2004;329(7459):224-227.

(10) Ramsey S, Willke R, Briggs A, et al. Good research practices for cost-effectiveness analysis alongside clinical

trials: the ISPOR RCT-CEA Task Force report. Value in health 2005;8(5):521-533.

(11) Ara R, Brazier J. Predicting the short form-6D preference-based index using the eight mean short form-36

health dimension scores: estimating preference-based health-related utilities when patient level data are not

available. Value in Health 2009;12(2):346-353.

(12) Ara R, Brazier J. Deriving an algorithm to convert the eight mean SF-36 dimension scores into a mean EQ-5D

preference-based score from published studies (where patient level data are not available). Value in Health

2008;11(7):1131-1143.

(13) Hanmer J. Predicting an SF-6D preference-based score using MCS and PCS scores from the SF-12 or SF-36.

Value in Health 2009;12(6):958-966.

(14) Kiessling A, Zethraeus N, Henriksson P. Cost of lipid lowering in patients with coronary artery disease by case

method learning. Int J Technol Assess Health Care 2005;21(02):180-186.

(15) Kramer L, Hirsch O, Schlößler K, et al. Associations between demographic, disease related, and treatment

pathway related variables and health related quality of life in primary care patients with coronary heart disease.

Health and quality of life outcomes 2012;10(1):1-10.

(16) Ose D, Rochon J, Campbell SM, et al. Secondary prevention in patients with coronary heart diseases: what

factors are associated with health status in usual primary care? PloS one 2012;7(12):e51726.

(17) Sach TH, Barton GR, Jenkinson C, et al. Comparing cost-utility estimates: does the choice of EQ-5D or SF-6D

matter? Med Care 2009;47(8):889-894.

(18) van Tol BAF, Huijsmans RJ, Kroon DW, et al. Effects of exercise training on cardiac performance, exercise

capacity and quality of life in patients with heart failure: a meta-analysis. European journal of heart failure

2006;8(8):841-850.

(19) Tengs TO, Lin TH. A meta-analysis of quality-of-life estimates for stroke. Pharmacoeconomics

2003;21(3):191-200.

Chapter 4

114

(20) Rehse B, Pukrop R. Effects of psychosocial interventions on quality of life in adult cancer patients: meta

analysis of 37 published controlled outcome studies. Patient Educ Couns 2003;50(2):179-186.

(21) Tengs TO, Lin TH. A meta-analysis of utility estimates for HIV/AIDS. Medical Decision Making 2002;22(6):475-

481.

(22) Mohiuddin S, Payne K. Utility Values for Adults with Unipolar Depression: Systematic Review and Meta-

Analysis. Med Decis Making 2014 Apr 2;34(5):666-685.

(23) Jackson D, Riley R, White IR. Multivariate meta-analysis: Potential and promise. Stat Med 2011;30(20):2481-

2498.

(24) Mavridis D, Salanti G. A practical introduction to multivariate meta-analysis. Stat Methods Med Res 2012.

(25) Niedzwiedz CL, Katikireddi SV, Pell JP, et al. Socioeconomic inequalities in the quality of life of older

Europeans in different welfare regimes. Eur J Public Health 2014 Jun;24(3):364-370.

(26) Borenstein M, Hedges LV, Higgins JPT, Rothstein HR. Introduction to meta-analysis. : Wiley; 2011.

(27) Jackson D, White IR, Riley RD. Quantifying the impact of between-study heterogeneity in multivariate

meta-analyses. Stat Med 2012;31(29):3805-3820.

(28) CRAN project. Multivariate meta-analysis and meta-regression 0.3.1. Package ‘mvmeta’. 2012; Available at:

http://cran.r-project.org/web/packages/mvmeta/mvmeta.pdf. Accessed 11/16, 2012.

(29) R. The R foundation for statistical computing, Rversion 2.14.0. 2011.

(30) Al-Ruzzeh S, Epstein D, George S, et al. Economic Evaluation of Coronary Artery Bypass Grafting Surgery With

and Without Cardiopulmonary Bypass: Cost-Effectiveness and Quality-Adjusted Life Years in a Randomized

Controlled Trial. Artif Organs 2008;32(11):891-897.

(31) Ascione R, Reeves BC, Taylor FC, et al. Beating heart against cardioplegic arrest studies (BHACAS 1 and 2):

quality of life at mid-term follow-up in two randomised controlled trials. Eur Heart J 2004 May;25(9):765-770.

(32) BAKHAI A, FERRIERES J, IÑIGUEZ A, et al. Clinical Outcomes, Resource Use, and Costs at 1 Year in Patients

with Acute Coronary Syndrome Undergoing PCI. J Interv Cardiol 2012.

(33) Bøhmer E, Kristiansen IS, Arnesen H, et al. Health and cost consequences of early versus late invasive

strategy after thrombolysis for acute myocardial infarction. European Journal of Cardiovascular Prevention &

Rehabilitation 2011;18(5):717-723.

(34) Burström K, Johannesson M, Diderichsen F. Swedish population health-related quality of life results using

the EQ-5D. Quality of Life Research 2001;10(7):621-635.

(35) Chong CAKY, Li S, Nguyen GC, et al. Health-state utilities in a prisoner population: a cross-sectional survey.

Health and quality of life outcomes 2009;7(Articl).

(36) Cohen DJ, Van Hout B, Serruys PW, et al. Quality of life after PCI with drug-eluting stents or coronary-artery

bypass surgery. N Engl J Med 2011;364(11):1016-1026.

(37) Denvir M, Lee A, Rysdale J, et al. Influence of socioeconomic status on clinical outcomes and quality of life

after percutaneous coronary intervention. J Epidemiol Community Health 2006;60(12):1085-1088.

(38) Dunning J, Waller JR, Smith B, et al. Coronary artery bypass grafting is associated with excellent long-term

survival and quality of life: a prospective cohort study. Ann Thorac Surg 2008;85(6):1988-1993.

(39) Fryback DG, Dasbach EJ, Klein R, et al. The Beaver Dam Health Outcomes study Initial Catalog of Health-state

Quality Factors. Medical Decision Making 1993;13(2):89-102.

(40) Griffin S, Barber J, Manca A, et al. Cost effectiveness of clinically appropriate decisions on alternative

treatments for angina pectoris: prospective observational study. BMJ 2007;334(7594):624.

(41) Hatoum HT, Brazier JE, Akhras KS. Comparison of the HUI3 with the SF-36 preference based SF-6D in a

clinical trial setting. Value in Health 2004;7(5):602-609.

(42) Kattainen E, Sintonen H, Kettunen R, et al. Health-related quality of life of coronary artery bypass grafting

and percutaneous transluminal coronary artery angioplasty patients: 1-year follow-up. Int J Technol Assess

Health Care 2005;21(2):172-179.

(43) Lacey E, Walters S. Continuing inequality: gender and social class influences on self perceived health after a

heart attack. J Epidemiol Community Health 2003;57(8):622-627.

4


115

(44) Lalonde L, Clarke AE, Joseph L, et al. Comparing the psychometric properties of preference-based and

nonpreference-based health-related quality of life in coronary heart disease. Quality of Life Research

1999;8(5):399-409.

(45) Lee HT, Shin J, Lim YH, et al. Health-Related Quality of Life in Coronary Heart Disease in Korea: The Korea

National Health and Nutrition Examination Survey 2007 to 2011. Angiology 2014 May 2.

(46) Loponen P, Luther M, Korpilahti K, et al. HRQoL after coronary artery bypass grafting and percutaneous

coronary intervention for stable angina. Scandinavian Cardiovascular Journal 2009;43(2):94-99.

(47) Nichol G, Llewellyn-Thomas HA, Thiel EC, et al. The relationship between cardiac functional capacity and

patients' symptom-specific utilities for angina: some findings and methodologic lessons. Med Decis Making 1996

Jan-Mar;16(1):78-85.

(48) Ellis JJ, Eagle KA, Kline-Rogers EM, et al. Validation of the EQ-5D in patients with a history of acute coronary

syndrome*. Current Medical Research and Opinion® 2005;21(8):1209-1216.

(49) Pettersen KI, Kvan E, Rollag A, et al. Health-related quality of life after myocardial infarction is associated

with level of left ventricular ejection fraction. BMC cardiovascular disorders 2008;8(1):28.

(50) Puskas JD, Williams WH, Mahoney EM, et al. Off-pump vs conventional coronary artery bypass grafting: early

and 1-year graft patency, cost, and quality-of-life outcomes. JAMA: the journal of the American Medical

Association 2004;291(15):1841-1849.

(51) Saarni SI, Härkänen T, Sintonen H, et al. The impact of 29 chronic conditions on health-related quality of life:

a general population survey in Finland using 15D and EQ-5D. Quality of Life Research 2006;15(8):1403-1414.

(52) Schweikert B, Hahmann H, Steinacker JM, et al. Intervention study shows outpatient cardiac rehabilitation to

be economically at least as attractive as inpatient rehabilitation. Clinical research in cardiology 2009;98(12):787-

795.

(53) Schweikert B, Hunger M, Meisinger C, et al. Quality of life several years after myocardial infarction:

comparing the MONICA/KORA registry to the general population. Eur Heart J 2009 Feb;30(4):436-443.

(54) Shah P, Najafi AH, Panza JA, et al. Outcomes and quality of life in patients≥ 85 years of age with ST-elevation

myocardial infarction. Am J Cardiol 2009;103(2):170-174.

(55) Shrive FM, Ghali WA, Johnson JA, et al. Use of the US and UK Scoring Algorithm for the EuroQol-5D in an

Economic Evaluation of Cardiac Care. Med Care 2007;45(3):269-273.

(56) Serruys PW, Unger F, Sousa JE, et al. Comparison of coronary-artery bypass surgery and stenting for the

treatment of multivessel disease. N Engl J Med 2001;344(15):1117-1124.

(57) Sharples L, Hughes V, Crean A, Dyer M, Buxton M, Goldsmith K, et al. Cost-effectiveness of functional cardiac

testing in the diagnosis and management of coronary artery disease: a randomised controlled trial. The CECaT

trial. : Gray Pub.; 2007.

(58) Stafford M, Soljak M, Pledge V, et al. Socio-economic differences in the health-related quality of life impact

of cardiovascular conditions. Eur J Public Health 2012 Jun;22(3):301-305.

(59) Sullivan PW, Ghushchyan V. Preference-based EQ-5D index scores for chronic conditions in the United

States. Medical Decision Making 2006;26(4):410-420.

(60) Tsevat J, Goldman L, Lamas GA, et al. Functional status versus utilities in survivors of myocardial infarction.

Med Care 1991;29(11):1153-1159.

(61) Visser M, Fletcher A, Parr G, et al. A comparison of three quality of life instruments in subjects with angina

pectoris: the Sickness Impact Profile, the Nottingham Health Profile, and the Quality of Well Being Scale. J Clin

Epidemiol 1994;47(2):157-163.

(62) Weintraub WS, Boden WE, Zhang Z, et al. Cost-Effectiveness of Percutaneous Coronary Intervention in

Optimally Treated Stable Coronary PatientsCLINICAL PERSPECTIVE. Circulation: Cardiovascular Quality and

Outcomes 2008;1(1):12-20.

(63) Werdan K, Ebelt H, Nuding S, et al. Ivabradine in combination with beta-blocker improves symptoms and

quality of life in patients with stable angina pectoris: results from the ADDITIONS study. Clinical Research in

Cardiology 2012:1-9.

Chapter 4

116

(64) Winkelmayer WC, Benner JS, Glynn RJ, et al. Assessing health state utilities in elderly patients at

cardiovascular risk. Medical decision making 2006;26(3):247-254.

(65) Franks P, Hanmer J, Fryback DG. Relative disutilities of 47 risk factors and conditions assessed with seven

preference-based health status measures in a national US sample: toward consistency in cost-effectiveness

analyses. Med Care 2006;44(5):478-485.

(66) Honkalampi T, Sintonen H. Do the 15D scores and time trade-off (TTO) values of hospital patients' own

health agree? Int J Technol Assess Health Care 2010;26(01):117-123.

(67) Moock J, Kohlmann T. Comparing preference-based quality-of-life measures: results from rehabilitation

patients with musculoskeletal, cardiovascular, or psychosomatic disorders. Quality of Life Research

2008;17(3):485-495.

(68) Bosch JL, Hunink MM. Comparison of the Health Utilities Index Mark 3 (HUI3) and the EuroQol EQ-5D in

patients treated for intermittent claudication. Quality of Life Research 2000;9(6):591-601.

(69) Johnson JA, Luo N, Shaw JW, et al. Valuations of EQ-5D health states: are the United States and United

Kingdom different? Med Care 2005 Mar;43(3):221-228.

(70) Havranek EP, Steiner JF. Valuation of health states in the US versus the UK: two measures divided by a

common language? Med Care 2005 Mar;43(3):201-202.

(71) Emery CF, Frid DJ, Engebretson TO, et al. Gender differences in quality of life among cardiac patients.

Psychosom Med 2004;66(2):190-197.

(72) Xie J, Wu EQ, Zheng ZJ, et al. Patient-reported health status in coronary heart disease in the United States:

age, sex, racial, and ethnic differences. Circulation 2008 Jul 29;118(5):491-497.

(73) Van Stel HF, Buskens E. Comparison of the SF-6D and the EQ-5D in patients with coronary heart disease.

Health Qual Life Outcomes 2006;4:20.

(74) O'Brien BJ, Spath M, Blackhouse G, et al. A view from the bridge: agreement between the SF-6D utility

algorithm and the Health Utilities Index. Health Econ 2003;12(11):975-981.

(75) Brazier J, Roberts J, Tsuchiya A, et al. A comparison of the EQ-5D and SF-6D across seven patient groups.

Health Econ 2004;13(9):873-884.

(76) Bleichrodt H. A new explanation for the difference between time trade-off utilities and standard gamble

utilities. Health Econ 2002;11(5):447-456.

(77) Bleichrodt H, Johannesson M. Standard gamble, time trade-off and rating scale: experimental results on the

ranking properties of QALYs. J Health Econ 1997;16(2):155-175.

(78) Kinney MR, Burfitt SN, Stullenbarger E, et al. Quality of life in cardiac patient research: a meta-analysis. Nurs

Res 1996;45(3):173-180.

(79) Pinquart M, Sörensen S. Influences of socioeconomic status, social network, and competence on subjective

well-being in later life: a meta-analysis. Psychol Aging 2000;15(2):187.

(80) Gillespie P, O'Shea E, Murphy AW, et al. The cost-effectiveness of the SPHERE intervention for the secondary

prevention of coronary heart disease. Int J Technol Assess Health Care 2010;26(03):263-271.

(81) Pignone M, Earnshaw S, Tice JA, et al. Aspirin, statins, or both drugs for the primary prevention of coronary

heart disease events in men: a cost–utility analysis. Ann Intern Med 2006;144(5):326-336.

(82) Pletcher MJ, Lazar L, Bibbins-Domingo K, et al. Comparing impact and cost-effectiveness of primary

prevention strategies for lipid-lowering. Ann Intern Med 2009;150(4):243-254.



(84) Dias S, Welton NJ, Sutton AJ, et al. Evidence synthesis for decision making 5: the baseline natural history

model. Med Decis Making 2013 Jul;33(5):657-670.

PART II

Pharmacoeconomic findings in secondary

cardiovascular disease prevention;

Focus on oral anticoagulation

Chapter 5

Budget impact of increasing market-share

of patient self-testing and patient

self-management in anticoagulation

Jelena Stevanović, Maarten J Postma, Hoa H Le

Submitted manuscript.

Chapter 5

122

ABSTRACT

Background: Patient self-testing and/or patient self-management (PST and/or PSM) might

provide better coagulation care than monitoring at specialised anticoagulation centres.

Yet, it remains an underused strategy in the Netherlands.

Methods: Budget impact analyses of current and new market-share scenarios of PST

and/or PSM compared to monitoring at specialised centres were performed for a national

cohort of 260,338 patients requiring long-term anticoagulation testing. A healthcare payer

perspective and one to five year time horizons were applied. The occurrence of

thromboembolic and haemorrhagic complications in the aforementioned patient

population were assessed in a Markov model. Dutch specific costs were applied, next to

effectiveness data derived from a meta-analysis on self-monitoring of oral anticoagulation.

Sensitivity and scenario analyses were performed to assess uncertainty on budget impact

model results.

Results: Increasing PST and/or PSM usage on a national level from the current 15.4% to

50% was found to be associated with savings ranging from €8 million following the first

year to €184 million after 5 years. Further increases in the use of PST and/or PSM

produced greater savings up to €450 million by 5 years after a complete replacement of

anticoagulation testing with PST and/or PSM. Sensitivity analyses showed robust cost-

savings, with the extreme of thromboembolic risk being maximally unfavourable for the

budget impact of PST and/or PSM. Results of a scenario analysis exploring a linear increase

in the uptake of PST and/or PSM from the current 15.4% to the expected 50% indicated

savings from €2 million after the first year to €184 million cumulatively after the fifth year.

Finally, potential unfavourable budget impact was found in scenarios exploring an

increase in the use of PST alone as well as increase in market shares of PST and/or PSM in

patients with atrial fibrillation on long-term oral anticoagulation.

Conclusion: PST and/or PSM was found to be a more favourable alternative to monitoring

at specialised centres. However, using PoC devices solely for PST resulted in greater

expenditures compared to testing in anticoagulation clinics. Additionally, our study

indicated less favourable findings of using PST and/or PSM in patients with atrial

fibrillation.

5

Patient self-testing and -management in anticoagulation: budget impact analysis

123

INTRODUCTION

Oral anticoagulation therapy (OAT) with vitamin K antagonists (VKA) has been shown to

reduce the risks of thromboembolic events in a number of clinical situations (1). In the

Netherlands, indications for OAT include atrial fibrillation (AF), arterial diseases (e.g.,

cardiomyopathy, coronary syndromes and surgery, vascular surgery, cerebral embolism),

heart valve replacement and venous thromboembolism (2). Patients with AF represent the

majority of patients requiring OAT (i.e. 62%). Given the increase in numbers of AF

patients, it is not surprising that the number of patients requiring OAT has increased as

well over the past decades in Western countries (2). Furthermore, the population of

patients in need of OAT is projected to increase further in the coming decades (3,4). This is

partly due to the aging population in Western countries and the positive association

between age and the incidence and prevalence of AF (5).

While warfarin is commonly used worldwide, acenocoumarol and phenprocoumon (to a

lesser extent) are the VKAs of first choice in the Netherlands. Prophylaxis with VKA is

considered an effective strategy but it has some shortcomings, including multiple

interactions with food and other drugs as well as inter-individual and intra-individual

variability in pharmacodynamics (6,7). As a result, regular monitoring is required to

maintain the international normalized ratio (INR) within the therapeutic range. INR-testing

is typically performed at specialised anticoagulation testing centres, adding to the

cumbersomeness of VKA use for the patients. Notably, Point-of-Care (PoC) devices allow

for patient self-testing (PST), where trained patients can perform the INR-test but still

inform his/her healthcare provider for subsequent advice on anticoagulant dosing, or even

patient self-management (PSM), with trained patients performing the INR testing,

interpreting the results, and adjusting dosing accordingly.

In agreement with the increasing number of patients with indications for OAT, the number

of patients using INR-testing in the Netherlands has increased from approximately

320,000 in 2002 to 430,000 in 2012 (2). Again, this trend is expected to continue in the

coming years with the aging population. Also, between 2007 and 2011 annual incidence

for AF has steadily increased from under 40,000 in 2008 to 56,000 in 2012 (2).

These figures, however, may still underrepresent the actual number of patients in need of

OAT as many eligible patients do not receive anticoagulation because of concerns of the

patients or concerns of their physicians of being outside the INR range (8-12). PoC testing

may also address this issue. As supported by international findings (13,14), PST and/or

PSM can lead to better coagulation care in the Netherlands compared to regular

monitoring in specialized anticoagulation centres (2). This may be due in part to the

convenience of use, resulting in more frequent testing, which is associated with greater

time in therapeutic range (TTR) (15). Also, findings from meta-analytical studies suggest

that PST and/or PSM compared to regular monitoring has similar risks of bleedings but

Chapter 5

124

reduced risks of thromboembolic events and all-cause mortality (13). Finally, it has been

reported that the patient empowerment inherent in PoC-strategies in itself already

directly reduces the risks of complications and death even in the absence of any

measurable increase in the quality of anticoagulation control (16). Despite these positive

results, PST and/or PSM remains an underused strategy in the Netherlands.

Eligible patients for PST and/or PSM include all those on long-term OAT (regardless of

indication), who have passed the required training. To date, the estimated number of

patients on long-term OAT in the Netherlands is approaching 260,338. In the current

situation, 15.4% of this population utilizes PST and/or PSM (2). In this study, we will assess

the budget impact of the current situation and new scenarios where PST and/or PSM

represents 50%, 75% and 100% of INR monitoring in the Netherlands.

METHODS

A budget impact analysis (BIA) was performed, using a patient cohort approach. Patients

in the cohort exit the model after death, but no new patients enter the model (closed

model). The perspective of the study is that of a healthcare payer. In the present analysis,

the patient cohort includes all patients, who require anticoagulation monitoring for OAT

for any clinical indication.

The design and reporting of study outcomes followed the recommendations of

International Society for Pharmacoeconomics and Outcomes Research for BIAs (ISPOR)

Task Force (17)(18).

Model structure

Patients indicated for OAT are at risk of haemorrhagic and thromboembolic events. To

incorporate the course of disease in the BIA, a Markov model was developed. This model

includes the following health states: no complications, thromboembolic complications,

haemorrhagic complications, and death (Figure 1). A cycle length of one year was used.

The cumulative budget impact of the cohort was assessed each year up to 5 years.

Patients enter the model in the “no complications” health state.

Figure 1 Markov model structure. Patients enter the model in the “No complication” health state. Possible transitions are defined by the arrows.

5


125

In the base-case analysis, transition probabilities for thromboembolic and haemorrhagic

complications and death were based on a systematic review and meta-analysis of

individual patient data on self-monitoring of oral anticoagulation by Heneghan et al. (14).

This study estimated the relative risks (RR) of thromboembolic and haemorrhagic

complications and death between testing at anticoagulation centres and PST and/or PSM

for all indications of OAT (Table 1). To assign transition probabilities for each testing

strategy, baseline risks for patients visiting anticoagulation centres/clinics were also

needed. Data on the number of thromboembolic and haemorrhagic complications and

death, and duration of follow-up for this group were taken from studies by Menendez-

Jandula et al., Fitzmaurice et al., Matchar et al., and Siebenfofer et al. (Table 2 (19-22).

These studies were selected because they were included in the study by Heneghan et al.

and hence provide internal validity to our study (14). To estimate baseline risks of

complications and death, we conducted a meta-analysis on the aforementioned studies in

the statistical program R 3.0.2, using the “metafor” package (Table 1)(19-24). Because the

study populations are not homogeneous across the studies and do not fit the assumption

of a fixed-effect meta-analysis, a random-effect meta-analysis was applied.

Annual risks of complications and death for the PST and/or PSM group were calculated by

taking the product of the RRs reported by Heneghan et al. and the baseline risks estimated

through a random-effect meta-analysis (Table 1)(14). In addition, age-specific background

mortality rates for the Netherlands for 2012 were used to estimate transition from no

complications to death (25).

Cost parameters

Costs associated with thromboembolic and haemorrhagic complications were collected

from published Dutch studies. All costs were inflated to 2013 levels using the harmonised

index for consumer product (HICP) for the health sector for the Netherlands (26). Costs for

thromboembolic events were derived from costs of ischemic stroke (27), myocardial

infarction (28), and pulmonary embolism (29) with contributions of 71.43%, 24.32%, and

4.25%, respectively, to the overall estimate (30). Cerebral haemorrhage (31),

gastrointestinal bleeding (32), and other bleedings (32) were assumed to represent

10.66%, 30.46%, and 58.88% of the costs associated with haemorrhagic complications

(30). For each complication, costs were differentiated between first and subsequent years

in the analysis, given the differences in the nature of complications between these years.

No subsequent year cost was assumed for pulmonary embolism and bleeding events

though. Furthermore, death was not associated with any additional cost. The weighted

costs of thromboembolic complications were €41,866 for the first year and €8,750 for

subsequent years. For haemorrhagic complications, the weighted averages were €9,748

and €1,263 for first and subsequent years, respectively (Table 3).

Chapter 5

126

Table 1 Annual risks of clinical events for use in the Markov model – base-case analysis.

Specialized Centre PST and/or PSM

Annual baseline riska, % Relative riskb Annual riskc, %

Thromboembolic Event 3.22 (1.50-4.94) 0.44 (0.17-1.14) 1.42 (0.26-5.63)

Haemorrhagic Event 2.84 (1.16-4.52) 0.91 (0.74-1.12) 2.58 (0.86-5.06)

Death 2.87 (1.01-4.74) 0.82 (0.52-1.28) 2.35 (0.53-6.07)

PST, patient self-testing; PSM, patient self-management. aEstimated through a random-effects meta-analysis of annual risks for patients using anticoagulation testing

centre. bAdapted from the meta-analysis of individual patient data on self-monitoring of oral anticoagulation by Heneghan et al.(14). cEstimated by taking the product of the relative risks reported by Heneghan et al. (14) and the baseline annual risks estimated by the authors’ random-effects meta-analysis of annual risks for patients using anticoagulation

testing centre. 95% confidence intervals of risk estimates are showed in parentheses.

Table 2 Studies used to estimate annual risks of complications for patients using testing centres.

Study Country Open RCT

design

Events Patient-

years

Proportion SE

Menendez-Jandula

(2005)(19)

Spain Single-centre

Thromboembolic

events

20 363 0.055 0.012

Haemorrhagic events

7 363 0.019 0.007

Death 15 363 0.041 0.010

Fitzmaurice (2005)(20) UK Multi-centre

Thromboembolic

events

4 264 0.015 0.008

Haemorrhagic

events

3 264 0.011 0.007

Death 1 264 0.004 0.004

Matchar (2010)(21) USA Multi-centre

Thromboembolic events

101 4235 0.024 0.002

Haemorrhagic

events

199 4235 0.047 0.003

Death 157 4235 0.037 0.003

Siebenhofer (2008)(22) Germany Multi-centre

Thromboembolic events

13 290 0.045 0.012

Haemorrhagic events

10 290 0.034 0.011

Death 11 290 0.038 0.011

RCT, randomised clinical trial; SE, standard error.

5


127

Table 3 Costs parameters applied in the model (€2013 per patient).

First Year Subsequent Years

Stroke(27) 50,828 11,800

Myocardial infarction(28) 22.015 1.338

Pulmonary embolism(29)a 5.244 0

Weighted Average Thromboembolic Event b 41,866 8,750

Cerebral haemorrhage(27)(31)c 38,417 11,800

Gastrointestinal bleeding(32) 7,120 0

Other major bleeding(32) 5,884 0

Weighted Average Haemorrhagic eventd 9,748 1,263

Testing strategy

VKA(33) 16 16

Specialized Centres 248 248

Blood sampling & INR-measurements at centres(34)e 181 181

Additional tariff for blood sampling at home(34)f 67 67

PST and/or PSM 958 749

Initial training & instruction(34) 396 0

Monitoring & supervision (34)g 562 749

Additional tariff for phone consultation for PST only (32)h 210 210

VKA, vitamin K antagonists; PST, patient self-testing; PSM, patient self-management. aCost estimate of pulmonary embolism from the original source was corrected to exclude the cost of INR testing and coumarines. bOn the basis of the assumption that the cost of a thromboembolic event will be a composite of the costs related

to stroke, myocardial infarction and pulmonary embolism with contributions of 71.4%, 24.3%, and 4.2%, respectively, to the overall estimate. cCost estimate of cerebral haemorrhage from the original source was corrected to exclude the costs of home help and private transportation costs. dOn the basis of the assumption that the cost of a haemorrhagic event will be a composite of the costs related to

cerebral haemorrhage, gastrointestinal bleeding bleeding and other major bleeding with contributions of 10.7%, 30.5%, and 58.9%, respectively, to the overall estimate. eAssuming 21.1 tests per year(2). fAssuming 8.6 home blood samplings per year(2). gAssuming 3 supervision sessions in the first year and 4 supervision sessions in the subsequent years. hAssuming 12 phone consultations on dosing per year for patients who only self-test.

Acquisition cost of VKA and anticoagulation monitoring are presented in Table 3. Cost of

VKA was estimated as a weighted average cost of acenocoumarol and phenprocoumon

based on their usage in the Netherlands (i.e. 80%:20%, respectively). The annual cost of

VKA was estimated at €16.06 (33). Anticoagulation testing at centres may involve blood

sampling at the centres or at home. For blood sampling and measurement at testing

centres, 21.1 INR-testings per patient per year was assumed (2). In addition, there were

Chapter 5

128

8.6 home blood samplings per patient per year for the same patient group as those tested

in the centre (2). The annual cost of monitoring at testing centres is the same for each

year and estimated to be €248 (34).

The first year cost of PST and/or PSM consisted of the costs of the device and one initial

training session and 3 supervision sessions. Subsequent years of use required quarterly

supervision sessions. The costs of the first and subsequent years were estimated to be

€958 and €749, respectively (34). While these patients receive information for possible

adjustments in their VKA dose by e-mail or specific software, no additional costs were

added for dosing adjustments.

Budget impact analysis

In the analysis, a cohort population size was evaluated at 260,338. This cohort size

represents an estimate of all patients requiring long-term anticoagulation testing for OAT

for any indication on a national level for the Netherlands. Using estimates from the

Federation of Dutch Thrombosis Services (“Federatie Nederlandse Trombosediensten”;

FNT) Report 2012, the current share of PST and/or PSM among patients on long-term

monitoring was assumed to be 15.4% (2). This current situation was evaluated against

potential new market penetration scenarios of 50%, 75%, and 100% for PST and/or PSM.

The BIA was evaluated for each year up to 5 years. Costs were not discounted as

recommended by the ISPOR Task Force for BIAs (17).


To examine the impact of uncertainty in key model parameters (i.e. baseline and relative

risks of complications and death and cost parameters) univariate sensitivity analyses were

performed on the 5-year BIA results considering a market penetration scenario of 50% in

Dutch patients on long-term OAT. Here each parameter was varied over the 95%

confidence interval (CI) while holding all other parameters constant. Where CI or standard

deviation (SD) was unavailable, the SD was assumed to be 25% of the mean.

Scenario analyses

Three scenario analyses were conducted to investigate the impact of increasing the

market share of PST and/or PSM up to 50% under different decision analytic settings. The

first scenario explored a linear increase in the uptake of PST and/or PSM from the current

15.4% to the expected 50% in 5 years. In the second scenario, an increase in the market

share of PST alone from the current 6.16% (i.e. 40% of all patients with PoC devices) to

50% was explored. Here, transition probabilities for thromboembolic and haemorrhagic

complications in the Markov model were based on the RRs of utilizing PST alone compared

to testing at specialized centres as reported by Heneghan et al. while the baseline risks

were estimated through a random-effect model (Table 4). The costs of PST alone were

assumed to be associated with additional €210 per year, reflecting consultations for

dosing regime adjustments. In the third scenario, the BIA of increasing the market share of

5


129

PST and/or PSM from the assumed 15.4% to 50% in patients with AF was explored in

scenario 3. In this analysis, it was assumed that 62% of patients on long-term OAT are

affected with AF, thus a cohort population size was evaluated at 161,410 patients.

Transition probabilities in the Markov model were based on the RRs assessed in AF

patients (Table 4).

Table 4 Annual risks of clinical events for use in the Markov model – scenario analyses

Specialized

centre

PST AF

Annual baseline

riska

Relative riskb Annual riskc Relative riskb Annual riskc

Thromboembolic

Event

Mean 3.22% 0.74 2.38% 0.67 2.16%

Low 95% CI 1.50% 0.3 0.45% 0.28 0.42%

High 95% CI 4.94% 1.82 8.99% 1.57 7.76%

Haemorrhagic

Events

Mean 2.84% 0.84 2.39% 1.04 2.95%

Low 95% CI 1.16% 0.64 0.74% 0.81 0.94%

High 95% CI 4.52% 1.12 5.06% 1.34 6.06%

Death

Mean 2.87% 0.91 2.61% 0.72 2.07%

Low 95% CI 1.01% 0.75 0.76% 0.43 0.43%

High 95% CI 4.74% 1.11 5.26% 1.2 5.69%

PST, patient self-testing; AF, atrial fibrilation. aEstimated through a random-effects meta-analysis of annual risks for patients using anticoagulation testing

centre. bAdapted from the meta-analysis of individual patient data on self-monitoring of oral anticoagulation by Heneghan et al.(14). cEstimated by taking the product of the relative risks reported by Heneghan et al. (14) and the baseline annual risks estimated by the authors’ random-effects meta-analysis of annual risks for patients using anticoagulation

testing centre. 95% confidence intervals of risk estimates are showed in parentheses.

RESULTS

Base-case results

The estimation of total costs per patient associated with INR monitoring in specialized

anticoagulation centres and with PoC-devices for a time horizon of one to five years is

detailed in Table 5. Monitoring related costs were higher than event related costs only in

the first year in patients conducting INR-testing with PoC-devices. Costs associated with

thromboembolic and haemorrhagic events were responsible for the vast majority part of

total costs with PoC-devices in the longer time horizons and in all time horizons in patients

conducting testing in specialized centres. Expanding the aforementioned findings to the

Chapter 5

130

Dutch national cohort of 260,338 patients who are using long-term anticoagulation

testing, a current situation of 15.4% using PST and/or PSM with PoC-devices resulted in

cumulative costs of €486 million, €1.00 billion, €1.54 billion, €2.11 billion, and €2.70 billion

in the first, second, third, fourth, and fifth year, respectively (Figure 2). Increasing the use

of PST and/or PSM to 50% in the first year resulted in cost-saving of €8 million from the

healthcare budget. The savings increased exponentially each year, reaching estimated

savings of €184 million after the fifth year. Similarly, increasing PST and/or PSM market

penetration to 75% and 100% produced correspondingly greater 5-year cumulative

savings of €317 and €450 million, respectively. While it is not likely that PST and/or PSM

will completely replace INR-testing at specialised anticoagulation centres, this latter

scenario illustrates the potential maximum savings in long-term utilization.

Figure 2 Budget impact analysis from year 1 to year 5 of current and new market share scenarios for PST and/or PSM (in millions). PST, patient self-testing; PSM, patient self-management.

5


131

Table 5 Cost allocation of overall medical costs associated with anticoagulation monitoring with PST and/or PSM and within specialized anticoagulation centres (€, per patient). Current situation.

PST/PSM Specialised centre

Year Total Monitoring Events Total Monitoring Events

1 1,796 951 845 1,881 256 1,628

2 3,444 1,651 1,793 3,912 490 3,422

3 5,140 2,306 2,834 6,071 703 5,368

4 6,877 2,919 3,958 8,336 897 7,439

5 8,647 3,492 5,155 10,690 1,073 9,617

PST, Patient self-testing; PSM, patient self-management


The results of the univariate sensitivity analyses show the impact of uncertainty

surrounding the key model parameters, illustrating that the relative and baseline risk of

thromboembolic complications had the highest impact on the BIA results (Figure 3).

Specifically, when the relative risk of thromboembolic complications would drop to the

lower limit of the 95% CI, BIA results would indicate total health care expenditures of €302

million. At risks increasing to the upper limits of 95% CIs, a cost-saving of €382 million

would be observed. The univariate sensitivity analyses also showed the BIA results were

sensitive to the uncertainty around the cost parameters. Yet, these results generally

favored an increasing market penetration of PST and/or PSM. Notably, cost savings were

robust over the whole range of variations except at the extreme for thromboembolic risk,

reflecting a situation with this risk at its lower limit that is maximally unfavourable for our

BIA.

Figure 3 Tornado diagram illustrating results from sensitivity analyses for budget impact analysis in a

5 year horizon considering a market penetration scenario of 50% for PST and/or PSM in a Dutch

cohort of 260,338 patients (in millions). Grey bars denote influence of the low value of the 95% confidence interval range and black bars denote

influence of the high value for parameters investigated. PST, patient self-testing; PSM, patient self-management.

Chapter 5

132

Scenario analyses

The results of scenario analyses are presented in Table 6. A linear increase in the uptake of

PST and/or PSM from the current 15.4% to the expected 50% indicated savings from € 2

million after the first year to €184 million cumulatively after the fifth year (scenario 1).

Increasing the market share of PST alone from the current 6.16% to 50% resulted in

expenditures from €57 million after the first year to €123 million cumulatively after the

fifth year (scenario 2). Finally, increasing the market share of PST and/or PSM in a cohort

of 161,410 AF patients indicated an expenditure of €15 million after the first year but

resulted in cumulative savings of €2 million after 5 years (scenario 3).

Table 6 Results of scenario analyses on uptake of PST and/or PSM (€, in millions)

New scenario Current scenario Difference

Scenario 1: 260,338 patients and 15.4% PST and/or PSM

1 485 486 -2

2 983 1,000 -17

3 1,493 1,543 -50

4 2,007 2,112 -105

5 2,517 2,701 -184

Scenario 2: 260,338 patients and 6.16% PST

1 555 497 57

2 1,113 1,030 83

3 1,697 1,595 102

4 2,301 2,186 115

5 2,923 2,800 123

Scenario 3: 161,410 AF patients and 15.4% PST and/or PSM

1 325 310 15

2 652 638 15

3 996 985 11

4 1,354 1,348 6

5 1,723 1,725 -2

*values are rounded Scenario 1 explores a linear increase in the uptake of PST and/or PSM from the current 15.4% to the expected 50% in 5 years.

Scenario 2 explores an increase of market share of PST alone from 6.16% to 50%. Scenario 3 explores an increase of market share of PST and/or PSM in AF patients from 15.4% to 50%.

DISCUSSION

Our study presents a BIA of the current practice as well as new varying market penetration

scenarios of anticoagulation monitoring with PST and/or PSM compared to monitoring at

specialised anticoagulation centres in the Netherlands. Our findings in the base-case

analysis indicated that increasing PST and/or PSM usage for anticoagulation testing from

the current 15.4% to 50%, 75% and 100%, would lead to significant savings in all analysed

scenarios. Even though INR testing is 3.9 times and 3.0 times more costly for PST and/or

5


133

PSM compared to specialised anticoagulation centres during the first year and subsequent

years, cost-saving was still observed when considering total direct medical costs due to

considerably higher event-related costs. This is due to the greater risk reductions of

thromboembolic and haemorrhagic complications, associated with high medical costs. In

fact, increasing the number of patients switching from conventional testing to PST and/or

PSM by increasing market penetration, produced even greater savings in the time horizon

of five years. For example, considering a national-level cohort population, potential

maximum savings over the current situation of €450 million may be observed in 5 years.

However, this is under the unlikely scenario of 100% adoption of PST and/or PSM. Yet,

even if PST and/or PSM would be adopted by 50% of all patients requiring long-term INR

testing – a figure that is quite attainable in the coming years – would result in a savings

range from €8 million following the first year to €184 million after 5 years. These analyses

clearly demonstrated the value of PST and/or PSM strategy with PoC-devices in the

Netherlands. Univariate sensitivity analyses revealed the major impact of uncertainty in

baseline thromboembolic risk on the BIA results. The impact of the uncertainty in the

baseline thromboembolic risk can be directly attributed to its impact on the occurrence of

stroke, myocardial infarction and pulmonary embolism events and their related costs of

treatment reaching in the first year and follow-up years a weighted average cost of

€41,866 and €8,750 per patient respectively. Overall, univariate sensitivity analysis

showed robust cost savings, with the extreme of thromboembolic relative and absolute

risk being maximally unfavourable for the BIA of PST and/or PSM at the relatively unlikely

exception.

Finally, this study observed potential unfavourable budget impact of increasing market

shares of PST alone as well as increasing market shares of PST and/or PSM in AF patients

on long-term OAT (scenarios 2 and 3). Greater expenditures associated with increasing

market shares of PST alone in scenario 2 are due to not only the higher cost of PST

strategy compared to PST and PSM combined but also costs associated with lower number

of prevented complications in comparison to the base-case scenario. The findings in

scenario 3 may be attributed to lower number of thromboembolic complications

prevented in comparison to the base-case scenario with a corresponding greater number

of haemorrhagic complications with PST and/or PSM compared to monitoring in

specialized centers, which are associated with high costs.

Comparison with other studies

To our knowledge, published economic evaluations of PST and/or PSM compared to

monitoring at specialised anticoagulation centres or to routine clinic-based care are all

cost-effectiveness analyses (CEA). This hampers a direct comparison of our study findings

with the ones from these other studies. Yet, there is a general agreement in the

conclusions of the available CEAs with our study results regarding the preference for PST

and/or PSM for long-term use. Specifically, Regier et al. found the self-managed

Chapter 5

134

anticoagulation to be a more cost-effective alternative compared to physician-managed

anticoagulation from the Canadian healthcare payer perspective in a 5-year time horizon

with an incremental cost-effectiveness ratio (ICER) of CAD14,129 per quality-adjusted life

year (QALY) (35). In the same study, when use of self-management was limited to a 1-year

time horizon, an ICER of CAD236,667 per QALY was estimated (35). Furthermore, in the

study by Lafata et al., self-testing in a US setting was found to be a cost-effective

alternative to testing in anticoagulation clinics with an ICER of $24,818 per event avoided

in a 5-year time horizon (36). Finally, Jowett et al. found PSM compared to routine clinic-

based monitoring unlikely to be cost-effective from the UK healthcare system perspective

in a 1-year time horizon (i.e. ICER of £32,716 per QALY) (37). These findings may be mainly

attributed to greater local costs of PST/PSM compared to alternative testing strategy and

sources of effectiveness data. Across all the aforementioned studies, the cost of testing

with PST and/or PSM outweighed the cost of alternative strategy. The costs associated

with thromboembolic and haemorrhagic complications were greater with PST and/or PSM

strategy compared to alternative testing. This was mainly driven by the effectiveness data

used in those studies. In particular, Jowett et al. utilized patient-level data from a

randomised clinical trial (RCT) by Fitzmaurice et al. which indicate greater number of

thromboembolic and haemorrhagic events with PST and/or PSM compared to alternative

testing strategy (20). In the studies by Regier et al. and Lafata et al., the number of

thromboembolic and haemorrhagic events was estimated based on the TTR achieved

while using investigated testing strategies (i.e. 71.8% vs. 63.2% and 89% vs. 65%,

respectively) and risk of those events conditional on the TTR. This estimation resulted in a

relatively low number of events avoided with PST and/or PSM compared to alternative

testing strategy. Regier et al. found only 0.72 thrombotic and 0.17 haemorrhagic events

avoided per 100 patients with PST and/or PSM in the first year, and after five years this

summed up to 3.5 and 0.79 events avoided, respectively. Similarly, Lafata et al. observed

in total 4.9 events avoided per 100 patients with PST and/or PSM over a five year time

horizon.

Strengths and limitations

Inferences drawn from BIAs are related to the quality of the evidence that goes into the

model. One point of strength of the current analysis is that effectiveness inputs are based

on a synthesis of evidence (14). In the hierarchy of evidence pyramid, evidence synthesis

of multiple trials resides above evidence from a single RCT (38,39). Yet, it must be pointed

out that no studies investigating the effectiveness of PST and/or PSM have been

conducted in the Netherlands. In the present analysis, effectiveness measures were

derived from studies investigating PST and/or PSM versus specialized testing centres for

all OAT indications (19-22). Because of the heterogeneity between the studies, a random-

effect model was used to establish baseline risks for thromboembolic and haemorrhagic

complications for patients using anticoagulation testing centres. To estimate risks of

5


135

complications for PST and/or PSM patients, relative risk reductions were applied as

reported by Heneghan et al. (14). In addition to possibly providing some external validity,

an advantage of this approach over relying on data from a single RCT is that all indications

for OAT were considered. This reflects a more complete assessment of the impact of

different strategies on costs of anticoagulation testing. Finally, we examined the impact of

uncertainty surrounding the key model parameters on BIA results.

Our study has several limitations. Firstly, only direct medical costs were considered in our

analyses and no costs related to productivity loss were included. Costs related to

productivity loss may be reduced for PST and/or PSM patients because they may spend

less time away from work as a result of greater effectiveness in prevention of

complications (Table 1). Also, testing at home with a PoC-device avoids work time lost

because it eliminates the need for travel and waiting time at testing centres. Yet, one

caveat in considering productivity loss among patients indicated for OAT is that many are

elderly patients, such as those with AF, may already be retired. Notably, although the

current estimates on cost-savings in the 5-year time horizon applied are substantial, they

may still reflect an underestimation of the true savings in the patient groups where

accounting for productivity loss is considered to be appropriate (i.e. patients <65 years of

age). Secondly, our model design may also be a factor for underestimation. In this analysis,

clinical events and associated costs were followed for a cohort of patients for up to 5

years. An alternative approach is to dynamically add new patients each year as estimated

by annual incidence rates (i.e. assume an open cohort). Such an approach would include

more patients in the analysis because the incidence rates are expected to continue to rise

in the coming decades (3). Thirdly, the recent introduction of novel oral anticoagulants

(NOACs), such as dabigatran, rivaroxaban and apixaban, on the Dutch market for use in

patients with some of the indications for OAT, was not accounted for in our study (40,41).

Currently, such a comparison between NOACs and the VKAs managed with PST and/or

PSM is hampered given the lack of RCTs between the two comparators as well as data on

the current use of NOACs in clinical practice in the Netherlands. Fourthly, future uptake of

PST and/or PSM strategy (i.e. 50, 75 and 100%) in this study was based on an assumption.

Yet, it may be more informative to estimate this in relation to the factors influencing its

current low market share, for example due to patients’ preferences. In particular, there

are indications by some Dutch experts that some patients prefer to have more regular

contact with hospitals/anticoagulation centres rather than to self-manage. Additionally,

they indicate that an increase in the uptake of PoC strategies could be achieved if these

strategies would be actively offered to patients as an alternative to management in the

clinics what currently is not the case. Finally, estimates of the baseline risks used in this

study were not supported by local real-life data. Such information unavailable and is duly

needed given that the baseline information used from the RCTs is commonly based on

Chapter 5

136

highly selected patient populations with characteristics that may deviate from the usual

practice.

In conclusion, compared to regular anticoagulation testing at specialised centres, PST

and/or PSM with PoC-devices resulted in cost-savings. However, using PoC devices solely

for PST resulted in greater expenditures compared to testing in anticoagulation clinics.

Hence, this strategy may need to be disregarded. Additionally, our study indicated less

favourable findings of using PST and/or PSM in patients with AF. Further research is

needed to explore this strategy in other indications and confirm the aforementioned

findings with local real-life data.

Given the increasing number of patients with indications for OAT and high treatment costs

of thromboembolic events, the choice of the optimal monitoring and managing strategy is

of high importance, both regarding the costs considered here and the health effects as

well. Further research should be directed to perform formal CEAs comparing the two

strategies. This would provide the additional insights of both societal costs and long-term

effects of those strategies on health, such as expressed in terms of QALYs.

Acknowledgements

The study was supported by Roche Diagnostics. The authors declare study results were

not influenced by Roche Diagnostics funding.

5


137

REFERENCES

(1) Hirsh J. Oral anticoagulant drugs. N Engl J Med 1991 Jun 27;324(26):1865-1875.

(2) Federatie van Nederlandse Trombosediensten. Samenvatting medische jaarverslagen. 2012. . Available at: http://www.fnt.nl/media/docs/jaarverslagen/FNT_Medisch_jaarverslag_2012_WEB.pdf. Accessed 09/04, 2014. (3) Go AS, Hylek EM, Phillips KA, et al. Prevalence of diagnosed atrial fibrillation in adults: national implications

for rhythm management and stroke prevention: the AnTicoagulation and Risk Factors in Atrial Fibrillation (ATRIA)

Study. JAMA 2001;285(18):2370-2375.

(4) Colilla S, Crow A, Petkun W, et al. Estimates of current and future incidence and prevalence of atrial

fibrillation in the US Adult population. Am J Cardiol 2013;112(8):1142-1147. (5) Heeringa J, van der Kuip, Deirdre AM, Hofman A, et al. Prevalence, incidence and lifetime risk of atrial fibrillation: the Rotterdam study. Eur Heart J 2006;27(8):949-953.

(6) Greenblatt DJ, Moltke LL. Interaction of warfarin with drugs, natural substances, and foods. The Journal of Clinical Pharmacology 2005;45(2):127-132. (7) Hirsh J, Dalen JE, Anderson DR, et al. Oral anticoagulants: mechanism of action, clinical effectiveness, and

optimal therapeutic range. Chest Journal 2001;119(1_suppl):8S-21S. (8) Fang MC, Stafford RS, Ruskin JN, et al. National trends in antiarrhythmic and antithrombotic medication use in

atrial fibrillation. Arch Intern Med 2004;164(1):55-60. (9) De Schryver E, Van Gijn J, Kappelle L, et al. Non–adherence to aspirin or oral anticoagulants in secondary prevention after ischaemic stroke. J Neurol 2005;252(11):1316-1321.

(10) Hylek EM, Evans-Molina C, Shea C, et al. Major hemorrhage and tolerability of warfarin in the first year of therapy among elderly patients with atrial fibrillation. Circulation 2007 May 29;115(21):2689-2696. (11) Samsa GP, Matchar DB, Goldstein LB, et al. Quality of anticoagulation management among patients with

atrial fibrillation: results of a review of medical records from 2 communities. Arch Intern Med 2000;160(7):967-973.

(12) Boulanger L, Kim J, Friedman M, et al. Patterns of use of antithrombotic therapy and quality of anticoagulation among patients with non-valvular atrial fibrillation in clinical practice. Int J Clin Pract 2006;60(3):258-264.

(13) Garcia-Alamino JM, Ward AM, Alonso-Coello P, et al. Self-monitoring and self-management of oral anticoagulation. Cochrane Database Syst Rev 2010;4(CD003839). (14) Heneghan C, Ward A, Perera R, et al. Self-monitoring of oral anticoagulation: systematic review and meta-

analysis of individual patient data. The Lancet 2012;379(9813):322-334. (15) Horstkotte D, Piper C, Wiemer M. Optimal frequency of patient monitoring and intensity of oral

anticoagulation therapy in valvular heart disease. J Thromb Thrombolysis 1998;5(1):19-24. (16) Connock M, Stevens C, Fry-Smith A, et al. Clinical effectiveness and cost-effectiveness of different models of managing long-term oral anticoagulation therapy: a systematic review and economic modelling. 2007.

(17) Mauskopf JA, Sullivan SD, Annemans L, et al. Principles of good practice for budget impact analysis: report of the ISPOR Task Force on good research practices—budget impact analysis. Value in health 2007;10(5):336-347.

(18) Sullivan SD, Mauskopf JA, Augustovski F, et al. Budget impact analysis—principles of good practice: report of the ISPOR 2012 Budget Impact Analysis Good Practice II Task Force. Value in Health 2014;17(1):5-14. (19) Menéndez-Jándula B, Souto JC, Oliver A, et al. Comparing Self-Management of Oral Anticoagulant Therapy

with Clinic ManagementA Randomized Trial. Ann Intern Med 2005;142(1):1-10. (20) Fitzmaurice DA, Murray ET, McCahon D, et al. Self management of oral anticoagulation: randomised trial. BMJ 2005 Nov 5;331(7524):1057.

(21) Matchar DB, Jacobson A, Dolor R, et al. Effect of home testing of international normalized ratio on clinical events. N Engl J Med 2010;363(17):1608-1620.

(22) Siebenhofer A, Rakovac I, Kleespies C, et al. Self-management of oral anticoagulation reduces major outcomes in the elderly. A randomized controlled trial.Thromb Haemost 2008;100(6):1089-1098. (23) R Development Core Team (2008). R: A language and environment for statistical computing. R Foundation

for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org. . (24) Viechtbauer W. Conducting meta-analyses in R with the metafor package. J Stat Softw 2010;36(3):1-48. (25) Den Haag/Heerlen. Central Institute of Statistics Web site. Available at:

http://statline.cbs.nl/StatWeb/publication/?DM=SLNL&PA=37530NED&D1=0-1&D2=1-100&D3=0&D4=l&HDR=G1,G2&STB=G3,T&VW=T. Accessed 09/01, 2013.

Chapter 5

138

(26) DataMarket. HICP for Health for the Netherlands. Available at: http://datamarket.com/data/set/18qu/hicp-2005-100-annual-data-average-index-and-rate-of-change#!ds=18qu!mh6=2:mh7=m:6ggm=s&display=table. Accessed 11/01, 2013.

(27) Baeten SA, van Exel, N Job A, Dirks M, et al. Lifetime health effects and medical costs of integrated stroke services-a non-randomized controlled cluster-trial based life table approach. Cost Effectiveness and Resource

Allocation 2010;8(1):21. (28) Greving J, Visseren F, de Wit G, et al. Statin treatment for primary prevention of vascular disease: whom to treat? Cost-effectiveness analysis. BMJ 2011;342.

(29) Ten Cate-Hoek A, Toll D, Buller H, et al. Cost-effectiveness of ruling out deep venous thrombosis in primary care versus care as usual. Journal of thrombosis and haemostasis 2009;7(12):2042-2049. (30) Connolly SJ, Ezekowitz MD, Yusuf S, et al. Dabigatran versus warfarin in patients with atrial fibrillation. N Engl

J Med 2009;361(12):1139-1151. (31) Roos YB, Dijkgraaf MG, Albrecht KW, et al. Direct costs of modern treatment of aneurysmal subarachnoid

hemorrhage in the first year after diagnosis. Stroke 2002 Jun;33(6):1595-1599. (32) Hakkaart-van Roijen L, Tan S, Bouwmans C. Handleiding voor kostenonderzoek. Methoden en standaard kostprijzen voor economische evaluaties in de gezondheidszorg.Geactualiseerde versie 2010.

(33) Z-index WMG-geneesmiddelen. , 2014. (34) Nederlandse Zorgautoriteit (NZa). Bijlage 1 bij tariefbeschikking TB/CU-7078-01. Available at:

http://www.nza.nl/. Accessed 18/11, 2014. (35) Regier DA, Sunderji R, Lynd LD, et al. Cost-effectiveness of self-managed versus physician-managed oral anticoagulation therapy. CMAJ 2006 Jun 20;174(13):1847-1852.

(36) Lafata JE, Martin SA, Kaatz S, et al. Anticoagulation clinics and patient self-testing for patients on chronic warfarin therapy: A cost-effectiveness analysis. J Thromb Thrombolysis 2000;9(1):13-19. (37) Jowett S, Bryan S, Murray E, et al. Patient self-management of anticoagulation therapy: a trial-based

cost-effectiveness analysis. Br J Haematol 2006;134(6):632-639. (38) Evans D. Hierarchy of evidence: a framework for ranking evidence evaluating healthcare interventions. J Clin

Nurs 2003;12(1):77-84. (39) Ho PM, Peterson PN, Masoudi FA. Evaluating the evidence: is there a rigid hierarchy? Circulation 2008 Oct 14;118(16):1675-1684.

(40) Nederlandse Vereniging voor Cardiologie. Leidraad begeleide introductie nieuwe orale antistollingsmiddelen. Available at: https://www.nvvc.nl/richtlijnen/bestaande-richtlijnen. Accessed 09/13, 2013. (41) Camm AJ, Lip GY, De Caterina R, et al. 2012 focused update of the ESC Guidelines for the management of

atrial fibrillation An update of the 2010 ESC Guidelines for the management of atrial fibrillationDeveloped with the special contribution of the European Heart Rhythm Association. Europace 2012;14(10):1385-1413.

Chapter 6

Economic evaluation of apixaban for

the prevention of stroke in non-valvular

atrial fibrillation in the Netherlands

Jelena Stevanović, Marjolein Pompen, Hoa H. Le, Mark H. Rozenbaum, Robert G.

Tieleman, Maarten J. Postma

This chapter is based on the published manuscript:

PLOS ONE 2014

Chapter 6

142

ABSTRACT

Background: Stroke prevention is the main goal of treating patients with atrial fibrillation

(AF). Vitamin-K antagonists (VKAs) present an effective treatment in stroke prevention,

however, the risk of bleeding and the requirement for regular coagulation monitoring are

limiting their use. Apixaban is a novel oral anticoagulant associated with significantly lower

hazard rates for stroke, major bleedings and treatment discontinuations, compared to

VKAs.

Objective: To estimate the cost-effectiveness of apixaban compared to VKAs in non-

valvular AF patients in the Netherlands.

Methods: The previously published lifetime Markov model using efficacy data from the

ARISTOTLE and the AVERROES trial was modified to reflect the use of oral anticoagulants

in the Netherlands. Dutch specific costs, baseline population stroke risk and coagulation

monitoring levels were incorporated. Univariate, probabilistic sensitivity and scenario

analyses on the impact of different coagulation monitoring levels were performed on the

incremental cost-effectiveness ratio (ICER).

Results: Treatment with apixaban compared to VKAs resulted in an ICER of €10,576 per

quality adjusted life year (QALY). Those findings correspond with lower number of strokes

and bleedings associated with the use of apixaban compared to VKAs. Univariate

sensitivity analyses revealed model sensitivity to the absolute stroke risk with apixaban

and treatment discontinuations risks with apixaban and VKAs. The probability that

apixaban is cost-effective at a willingness-to-pay threshold of €20,000/QALY was 68%.

Results of the scenario analyses on the impact of different coagulation monitoring levels

were quite robust.

Conclusions: In patients with non-valvular AF, apixaban is likely to be a cost-effective

alternative to VKAs in the Netherlands.

6

Economic evaluation of apixaban for the prevention of stroke in non-valvular atrial fibrillation

143

INTRODUCTION

Atrial fibrillation (AF) is a heart disease common among elderly people. In the Netherlands

incidence rates increase with advancing age from approximately 1% among 55-year olds

to 18% among 85-year olds and related relevant risks of ischemic stroke (IS) and other

systemic thromboembolic events (1,2). In addition, patients with AF suffer not only from a

greater activity impairment and lower quality of life (QoL) compared to the general

population but also have a 50 – 90% increased risk of mortality (3,4). The majority of AF

patients suffer from non-valvular AF. Strokes related to AF are often characterized by

more severe disability and impairment of QoL in comparison to strokes due to other

causes (5). As a result, stroke related morbidity, which is driven by high hospitalization and

long-term maintenance costs, causes a high economic burden to the Dutch health care

system. Specifically, the 6-month cost of usual care for stroke patients range from €16,000

to €54,000 depending on severity (6). In parallel, the annual costs of treating patients with

AF in the Netherlands were estimated to mount up to €2,328 with 70.1% of the resources

allocated to the inpatient care and interventional procedures (7). Given the humanistic

implications of both AF and stroke and economic considerations of their management,

stroke prevention is the main focus of treatment strategies for patients with AF and could

be expected to lead to both health and economic benefits.

Until recently patients with AF and an estimated moderate to high risk of stroke (i.e.

cardiac failure, hypertension, age, diabetes, stroke (doubled) [CHADS2] score ≥ 2) were

recommended to receive vitamin-K antagonists (VKAs; e.g. warfarin, acenocoumarol or

phenprocoumon) for stroke prevention (8). However, although VKAs present a highly

effective treatment strategy in reducing the incidence of stroke, their optimal

effectiveness and safety is crucially safeguarded with regular coagulation monitoring due

to VKAs’ narrow therapeutic range (international normalized ratio [INR] limits of 2.0 and

3.0)(9). Failure to achieve the anticoagulant effect inside the required INR therapeutic

range increases the risk of IS and bleeding including haemorrhagic stroke (HS). The

complexity of regular monitoring, which in the Dutch healthcare system is handled by

thrombotic services, possibly followed by failure to achieve the safety range inside INR

limits, accompanied with multiple drug and food interactions, might lead to underuse of

VKAs or even result in an increase in medication-related hospital admissions as observed

in the HARM study (8,10).

Recently, a new class of anticoagulants became available (novel oral anticoagulant

(NOAC)) that are at least as effective or superior in reducing the risk of stroke or systemic

embolism (SE), have a better efficacy/safety profile and exclude the need for constant INR

monitoring, compared to VKAs (11,12),(13). Accordingly, NOACs have been included in

both international and national guidelines (8,14). One of them is apixaban, a NOAC of

which the efficacy and safety was tested in clinical trials on VKA suitable (ARISTOTLE trial

Chapter 6

144

[ClinicalTrials.gov Identifier, NCT00412984]) or unsuitable (AVERROES trial

[ClinicalTrials.gov Identifier, NCT00496769]) non-valvular AF patients with a high risk of

stroke (11,15). In the AVERROES trial, apixaban was shown to prevent more stroke or SE

events with no significant difference in the incidence of major bleedings (MBs) or

intracranial haemorrhages (ICHs) compared to acetylsalicylic acid (ASA)(11,15). Similarly,

in the ARISTOTLE trial, less stroke or SE events, less MBs and less fatal events related to

any cause were observed in the treatment with apixaban when compared to the

treatment with VKA (11,15). Despite obvious advantages of the NOACs, the choice of the

optimal treatment strategy for AF-patients always needs to be made with respect to both

health and economic consequences of the approach chosen, including a formal

comparison of apixaban and VKAs as one element (16).

The aim of this study is to evaluate the health and economic consequences of applying

apixaban compared to VKAs for stroke prevention in non-valvular AF patients in the

Netherlands. The health consequences associated with the use of apixaban and VKAs

reflecting the likelihoods of having stroke, other thromboembolic or bleeding events, are

mainly based on the data from the ARISTOTLE trial (11). The cost estimates of stroke and

other AF-related complications as well as drug costs, reflect the Dutch situation from the

healthcare payers’ perspective.

METHODS

Decision model

The previously published lifetime Markov model was modified and updated to reflect the

use of apixaban per defined daily dose and adjusted-dose warfarin in patients with non-

valvular AF in the Netherlands (17,18). The following health states were included in the

model: baseline (non-valvular AF), IS, HS, SE, myocardial infarction (MI), other ICH, other

MB and clinically-relevant non-major (CRNM) bleeding, other treatment discontinuation

and death (Figure 1). Notably, other treatment discontinuations reflect discontinuations

that are not directly related to having had a thrombotic or bleeding event. For the

purposes of this study, warfarin, studied versus apixaban in the ARISTOTLE trial (11), was

used as a comparator, as the Dutch reimbursement authorities presume the efficacy and

safety profile of warfarin and acenocoumarol/phenprocoumon (also VKAs) to be

interchangeable(19).

Base-case analysis followed a hypothetical cohort of 1,000 patients with non-valvular AF

whose characteristics were comparable to those in the ARISTOTLE trial. Specifically,

patients were predominantly male, aged 70 years, with an average CHADS2 score of 2.3

and a history of previous VKA use (Table 1) (11,20,21). The progression of patients with

non-valvular AF through the Markov model is detailed elsewhere (17,18). Briefly, patients

remained in the baseline state until a fatal or non-fatal event or treatment discontinuation

6


145

occurred, or they died due to other, non-cardiovascular related, causes. In order to reflect

daily life more closely, a distinction between different levels of IS and HS severity was

made in the model, i.e. mild, moderate, severe and fatal. The model allows one recurrent

stroke event to occur. The annual risk of recurrent stroke event was based on the 10-year

cumulative risk of recurrence derived from a population based study using the South

London stroke registry (22). Health states for thromboembolic events other than stroke

(i.e. SE and MI) were considered to be absorbing (i.e. patients remain there until death).

The probability of patient being in a particular health state was assessed every 6-weeks

which was the cycle length of the model.

Certain assumptions on the treatment following thromboembolic or bleeding events were

made. Firstly, upon the occurrence of IS or SE, patients surviving were assumed to stay on

the initially assigned anticoagulant treatment while those surviving HS and MI were

assumed only to receive long-term disease-specific maintenance treatment. Secondly,

patients experiencing other ICH, MB and CRNM bleeding were allocated between an

option to stay on the initially assigned treatment and an option to switch to ASA. Details

on the allocation of patients between the two treatment options are provided in

previously published studies (17,18). Patients staying on the initially assigned

anticoagulant treatment after an ICH that was not a HS, were additionally assumed to

have a six-week drug holiday. Finally, patients discontinuing the initial treatment for

reasons unrelated to stroke, SE and bleeding were assumed to switch to ASA.

The final outcome of the decision model is the incremental cost-effectiveness ratio (ICER)

of apixaban compared to VKA. As a measure of effectiveness, quality-adjusted life-years

(QALYs) and life years (LYs) gained were estimated. All relevant costs incorporated in the

model reflect the health care payer’s perspective and were inflated to price year 2013

using the Dutch consumer price index (23). Future costs and health effects were

discounted by 4% and 1.5% annually after the first year, according to the Dutch guidelines

for pharmacoeconomic research (24).

Table 1 Baseline characteristics of the patients included in the model. Characteristic Value Range Reference

Age 70 63-77 (11)

Gender (female, %) 35.3 34.1-36.5 (11)

CHADS2 (% of patients)

1 7 5.4-8.8 (20)

2 27 21.3-33.1 (20)

3 25 19.7-30.7 (20)

4 20 15.7-24.7 (20)

5 12 9.3-15 (20)

6 7 5.4-8.8 (20)

7 2 1.5-2.5 (20)

Average TTR in the Netherlands (%) 72.48 Fixed (21)

CHADS2, cardiac failure, hypertension, age, diabetes, stroke (doubled); TTR, time in therapeutic range.

Chapter 6

146

Figure 1 Model for the non-valvular AF population. Depicted in the diagram are the chance nodes (circles) and terminal nodes (triangles). Branches for apixaban and

VKA are identical except numerical risks. Patients that discontinue the initial anticoagulant treatment re-enter the model with identical Markov branches but under the assumption of switching their treatment to

acetylsalicylic acid. NVAF, non-valvular atrial fibrillation; ASA, acetylsalicylic acid; IS, ischemic stroke; HS, haemorrhagic stroke; SE, systemic embolism; MI, myocardial infarction; ICH, intracranial haemorrhage; CRNM, clinically-relevant non-

major; MB, major bleeding; Tmt, treatment.

Transition probabilities

Data from the ARISTOTLE and the AVERROES trial were the main sources used to estimate

the transition probabilities between the health states in the model for patients receiving

apixaban, VKA and ASA (11,15,17,18). Specifically, the rates of IS, MI, SE, ICH, other MB

and CRNM bleeding and other treatment discontinuations from the aforementioned trials,

were applied for deriving the transition probabilities between the health states similarly to

the previously published Markov models (Table 2)(17,18). Additionally, trial rates of IS,

ICH, other MB and CRNM bleeding were adjusted for the average level of risk dependent

on the level of INR control in the Netherlands represented by mean time in therapeutic

range (TTR) (i.e. 72.48%(21),(25)).

ICHs were further differentiated into HSs and other ICHs; other MBs were differentiated to

those that were or were not gastrointestinal (GI) by location. Details on the number of

patients experiencing one of the two types of ICHs, specific fatality rates after stroke, MI,

SE, ICH and other MB and the factors of age-related increasing risk of stroke, bleeding (i.e.

ICH, other MB and CRNM bleedings) and MI are provided elsewhere (17,18)(29).

6


147

Table 2 Rates of events while on apixaban, VKA and ASA used in estimating the transition probabilities in the model. Parameter Apixaban VKA ASA Reference

IS rate by CHADS2 score

1 0.52 0.46 (11,15,17,19)

2 0.52 0.46 (11,15,17,19)

3 0.95 0.93 (11,15,17,19)

4 1.53 1.94 (11,15,17,19)

5 1.53 1.94 (11,15,17,19)

6 1.53 1.94 (11,15,17,19)

7 1.53 1.94 (11,15,17,19)

IS HR by cTTR

cTTR < 52.38% 0.92 1.54 (11,15,17,19)

52.38% ≤ cTTR < 66.02% 1.00 1.00 (11,15,17,19)

66.02% ≤ cTTR < 76.51% 0.69 0.84 (11,15,17,19)

cTTR ≥ 76.51% 0.56 0.72 (11,15,17,19)

Rate of IS (per 100 patient years) 3.45 (11,15,17,19)

Rate of ICH (per 100 patient years) 0.33 0.80 0.32 (11,15,17,19)

ICH HR by cTTR

cTTR < 52.38% 0.58 1.05 (11,15,17,19)

52.38% ≤ cTTR < 66.02% 1.00 1.00 (11,15,17,19)

66.02% ≤ cTTR < 76.51% 0.69 0.68 (11,15,17,19)

cTTR ≥ 76.51% 0.36 0.78 (11,15,17,19)

Rate of other MBs (per 100 patient years) 1.79 2.27 0.89 (11,15,17,19)

Other MBs HR by cTTR

cTTR < 52.38% 0.72 0.84 (11,15,17,19)

52.38% ≤ cTTR < 66.02% 1.00 1.00 (11,15,17,19)

66.02% ≤ cTTR < 76.51% 1.69 1.13 (11,15,17,19)

cTTR ≥ 76.51% 1.77 1.37 (11,15,17,19)

Rate of CRNMBs (per 100 patient years) 2.08 2.99 2.94 (11,15,17,19)

CRNMBs HR by cTTR

cTTR < 52.38% 0.71 0.99 (11,15,17,19)

52.38% ≤ cTTR < 66.02% 1.00 1.00 (11,15,17,19)

66.02% ≤ cTTR < 76.51% 1.25 1.26 (11,15,17,19)

cTTR ≥ 76.51% 1.70 1.27 (11,15,17,19)

Rate of myocardial infarction (per 100 patient years) 0.53 0.61 1.11 (11,15,17,19)

Rate of other treatment discontinuations (unrelated to

stroke and bleedings) (per 100 patient years)

13.42 14.5

4

19.65 (11,15,17,19)

Rate of systemic embolism (per 100 patient years) 0.09 0.10 0.4 (11,15,17,19)

Death rate during trial period (per 100 patient years) 3.08 3.34 3.59 (11,15,17,19)

Background mortality after trial period Age- and gender-adjusted

non-CVD mortality

(26-28)

VKA, vitamin K-antagonist; ASA, acetylsalicylic acid; IS, ischemic stroke; HR, hazard ratio; cTTR, clinic time in

therapeutic range; ICH, intracranial haemorrhage; MB, major bleeding; CRNM, clinically relevant non-major

bleeding.

Chapter 6

148

The average risk of IS in patients receiving apixaban and VKA, was estimated as the joint

probability of having an event associated with a specific baseline population stroke risk

represented by CHADS2 score corrected for the average level of risk dependent on INR

control, and the probability of having an event associated with the level of INR control in

the Netherlands (Table 1)(21),(25). Baseline population stroke risk represented by CHADS2

score was determined by weighting the risk for each categorization of CHADS2 score by

the proportion of patients within each group of CHADS2 score in the Netherlands (20).

Published population based registries were used to estimate the transition probabilities

for recurrence of events (22). The annual risks for recurrent stroke events of 2.97 and 2.17

were assigned to patients surviving first IS and HS, respectively (22). The distribution of

stroke severity for recurrent stroke events was assumed to be the same as that of the first

stroke events in patients treated with apixaban.

Mortality due to causes other than cardiovascular while on apixaban and VKA, for the trial

period, were based on data from the ARISTOTLE trial (11). Beyond the duration of the trial

period (1.8 years), age- and gender-adjusted mortality due to causes other than

cardiovascular, was obtained from Statistics Netherlands (26-28). In addition to the

mortality due to causes other than cardiovascular, an increase in mortality rates

associated with AF, strokes by severity level, MI and SE, was incorporated as in the

previously published Markov models (17,18).

Utilities

A utility score specific for patients with AF was applied to all patients in the baseline

health state of the model (Table 3)(30). Upon the occurrence of stroke, MI or SE, utility

scores were adjusted to account for the level of utility for AF and comorbid

thromboembolic event jointly (30),(31). Utility decrements following the occurrence of a

certain bleeding event were applied additively for a specific time interval (32). Finally,

utility decrements reflecting the use of VKA (warfarin), apixaban and ASA were applied

(32)(33).

Costs

Prices of apixaban, defined as price per defined daily dose (2x 5 mg), VKAs and ASA (100

mg) were taken from the official Dutch price list (Z-index) (Table 4)(35). Cost of VKA was

estimated as a weighted average cost of acenocumarol and fenprocoumon based on their

usage in the Netherlands (80%:20%, respectively)(36). In addition to anticoagulants’ costs,

routine care cost representing medical specialist fee was added to all treatment

alternatives and cost due to INR testing was added to treatment with VKAs (24).

Acute care costs associated with clinical events (IS and HS with different levels of severity,

other ICH, other MB and CRNM bleedings, SE, MI) were adopted from previous costing

studies conducted in the Netherlands and updated to the year 2013 using the Dutch

6


149

inflation index (Table 4)(6,37,38)(23)(23). Patients surviving acute stroke and MI were

assigned with long-term maintenance costs (39).

Sensitivity analyses

Univariate sensitivity analyses were conducted in order to inspect the effects of the

uncertainty in key input parameters and assumptions on the uncertainty in the final cost-

effectiveness outcome. Furthermore, a probabilistic sensitivity analysis (PSA) was

performed in order to simultaneously incorporate the uncertainty around all parameters

in the CE analysis. Key input parameters in the deterministic analysis that were assumed

random variables in the PSA were: event rates, utilities and costs. A gamma distribution

was assigned to event rates, a beta distribution to utilities and a log-normal distribution to

cost estimates. Results from the PSA were plotted on a CE plane and transformed into CE-

acceptability curves (CEACs).

Finally, in order to investigate the impact of different levels of INR control on the

estimated ICER, as is the case in the various Dutch thrombotic centres, scenario analyses

were conducted. Four different scenarios were investigated. Specifically, one scenario

assumed patients were equally distributed across centers with different cTTR, similarly to

the patient allocation in the ARISTOTLE trial. Other scenarios assumed the allocation of all

patients to the one of specific cTTR range (i.e. cTTR< 52.38%, 52.38% ≤ cTTR<66.02% and

cTTR ≥ 76.51%) that was different from the range in the base-case analysis (i.e.

cTTR=72.48%).

Table 3 Utility parameters applied in the model.

Parameter Mean Range* Reference

Baseline Utility for AF 0.6980 0.5532-0.8250 (30)

Stroke mild 0.6704 0.5330-0.7944 (31)

Stroke moderate 0.6165 0.4925-0.7333 (31)

Stroke severe 0.4416 0.3561-0.5289 (31)

SE 0.5769 0.4622-0.6876 (30)

MI 0.5328 0.4279-0.6363 (30)

Disutility of other ICH (6 weeks) 0.1385 0.1125-0.1667 (32)

Disutility of other MBs (14 days) 0.1385 0.1125-0.1667 (32)

Disutility of CRNM bleedings (2 days) 0.06 0.0488-0.0723 (32)

Disutility of anticoagulation with VKA 0.013 0.0106-0.0157 (33)

Disutility of anticoagulation with ASA 0.002 0.0016-0.0024 (32)

Disutility of anticoagulation with apixaban 0.002 0.0016-0.0024 (32)

AF, atrial fibrillation; SE, systemic embolism; MI, myocardial infarction; ICH, intracranial haemorrhage; MB, major bleeding; CRNM, clinically relevant non-major; VKA, vitamin K-antagonist; ASA, acetylsalicylic acid.

*Utility estimates that were available only as single point estimates, were assumed to follow a beta distribution with a 10% standard deviation of the mean. §Utilities were calculated based on the method for predicting utility for joint health states by Bo Hu (34).

Chapter 6

150

Table 4 Cost parameters applied in the model.

Parameter Mean Range‡ Reference

Apixaban (daily) € 2.28 Fixed (35)

VKA (daily)§ € 0.03 Fixed (35)

ASA (daily) € 0.15 Fixed (35)

Monitoring visit (per year) € 224 163-308 (35)

Routine care (per visit) € 78.97 62-117 (24)

Stroke¶

Mild 0-6 months € 16,097 11,712-22,124 (6)

Mild 7-12 months € 4,470 3,252-6,144 (6)

Mild 13 -19 months men € 1,174 854-1,614 (6)

Mild 13 -19 months women € 1,174 854-1,614 (6)

Moderate 0-6 months € 44,640 32,479-61,354 (6)

Moderate 7-12 months € 21,146 15,385-29,063 (6)

Moderate 13 -19 months men € 7,115 5,177-9,779 (6)

Moderate 13 -19 months women € 11,745 8,545-16,142 (6)

Severe 0-6 months € 54,678 39,783-75,150 (6)

Severe 7-12 months € 26,711 19,43-36,712 (6)

Severe 13 -19 months men € 9,055 6,588-12,445 (6)

Severe 13 -19 months women € 15,069 10,964-20,711 (6)

Fatal stroke € 2,988 2,876-3,102 (39)

SE acute care (per episode)* € 4,995 3,634-6,865 (38)

Other ICH# € 20,326 14,789-27,937 Assumption

GI bleeds € 4,995 3,635-6,866 (38)

Non ICH and non-GI bleeds¥ € 4,995 3,635-6,866 Assumption

CRNMB (assume a visit to GP) € 30.71 22-42 (24)

MI acute care (per episode) € 5,021 4,936-5,106 (37)

MI maintenance (per month) € 196 183-206 (39)

VKA, vitamin K-antagonist; ASA, acetylsalicylic acid; SE, systemic embolism; ICH, intracranial haemorrhage; GI,

gastrointestinal; CRNMB, clinically relevant non-major bleeding; GP, general practitioner; MI, myocardial infarction. ‡ Cost estimates that were available only as single point estimates, were assumed to follow a log-normal distribution with a coefficient of variation equal to 0.25. § Cost of VKA was estimated as a weighted average cost of acenocumarol and fenprocoumon based on their

usage in the Netherlands (35) ¶ Stroke related costs were adjusted to fit the design of a decision model. Specifically, acute and long-term one-month maintenance costs were estimated.

*Assumed to be equal to the cost of pulmonary embolism. # Assumed to be the same as cost of acute mild stroke ¥ Assumed to be the same as cost of GI bleeds

RESULTS

The number of events associated with the use of VKAs and apixaban in a cohort of 1,000

patients with non-valvular AF, as well as the costs related to those events and the

anticoagulant treatment, are presented in Table 5. Specifically, the incremental difference

in the number of events observed over a lifetime horizon in the apixaban treatment

6


151

scenario compared to the VKA treatment scenario was: ten less stroke or SE events

(including first and recurrent IS and HS), nine less other ICHs, 12 more other MBs, three

less MIs and 58 less CRNM bleedings. A comparable number of other treatment

discontinuations was observed in both apixaban and VKA treatment scenarios (648 and

652 respectively). Finally, treatment with apixaban was estimated to provide an additional

0.18 QALYs or 0.18 LYs compared to treatment with VKA over a lifetime horizon (Table 6).

Costs associated with handling stroke and thromboembolic events were lower in the

apixaban treatment scenario compared to the VKA treatment scenario (€14,113 vs.

€14,904)(Table 5). However, the overall anticoagulant treatment costs including the drug

acquisition costs, costs of routine care and INR monitoring were higher with apixaban

compared VKA (€6,092 vs. €3,449)(Table 5). Accounting for all the aforementioned costs

resulted in an additional cost of €1,852, associated with the use of apixaban compared to

VKA.

Finally, the summarized lifetime health and economic consequences of applying apixaban

compared to VKA in 70-year old patients in the Netherlands yielded a base-case ICER of

€10,576 per QALY gained or €10,529 per LY gained (Table 6).


Figure 2 presents a tornado diagram illustrating the impact of varying each of key input

parameters on the ICER while holding all the other model parameters fixed. The

uncertainty around the absolute stroke risk under apixaban, the risks of treatment

discontinuations under both apixaban and VKA and the risk of ICH under VKA, showed the

highest impact on uncertainty in the estimated ICERs. In particular, when the absolute

stroke risk or treatment discontinuations risk under apixaban would reach the upper limit

of the 95% confidence interval (CI), ICERs would be €33,426 and €27,103 per QALY gained,

respectively. At risks dropping to lower limits of 95% CIs, ICERs would fall to €4,268 or

€5,086 per QALY gained, respectively. The uncertainty around the risks of treatment

discontinuations under VKA led to comparable variation in estimated ICERs ranging from

€4,236 to €25,811 per QALY gained.

The results of 2,000 iterations in PSA are presented through an incremental CE plane in

Figure 3. The ellipsoid shape of this incremental CE plane indicated a negative correlation

between incremental costs and incremental effects. Transforming the results of a CE plane

to CEACs shows that apixaban was cost-effective at alternative willingness to pay (WTP)

thresholds of €20,000/QALY and €30,000/QALY in 68% and 72% of simulations

respectively (Figure 4). Accordingly, VKA was estimated to be the preferred alternative

over apixaban at the aforementioned WTP thresholds in 32% and 28% of simulations

respectively.

The impact of different levels of INR control on the estimated ICER was explored through

scenario analyses. Specifically, the level of INR control is applied for the estimation of the

Chapter 6

152

rates of IS, ICH, other MB and CRNM bleeding and therefore can have an indirect impact

on the estimated ICER. Across the scenarios investigated, the majority of the

aforementioned rates was estimated to be lower with apixaban compared to VKA. IS rate

with apixaban was estimated to be higher than with VKA (i.e. 1.316 and 1.159,

respectively) only in the scenario assuming allocation of patients in the range 52.38% ≤

cTTR< 66.02%. Finally, the estimated ICER was in range from €27 to €12,662/QALY or from

€22 to €12,905/LY across the scenarios explored (Table 7).

Table 5 Stroke and other thromboembolic and bleeding complications and related costs within a hypothetical patient population of 1,000 subjects receiving apixaban and VKA over a lifetime

horizon.

Apixaban VKA

Number

of events

Acute

event

related

lifetime

costs p.p.

Long-

term

costs

Number

of events

Acute

event

related

lifetime

costs p.p.

Long-term costs

Ischemic stroke

Mild, non-fatal 96.32 € 1,429 € 638 93.01 € 1,379 € 614

Moderate, non-fatal 83.62 € 3,285 € 2,612 89.30 € 3,563 € 2,921

Severe, non-fatal 32.86 € 1,586 € 643 34.31 € 1,675 € 699

Fatal 30.36 € 67 29.05 € 64

Sum 243.15 € 10,259 245.67 € 10,915

Recurrent ischemic

stroke

Mild, non-fatal 10.63 € 149 € 38 10.81 € 152 € 37

Moderate, non-fatal 4.21 € 161 € 225 4.28 € 165 € 244

Severe, non-fatal 1.61 € 75 € 49 1.63 € 77 € 52

Fatal 3.61 € 7 3.67 € 8

Sum 20.06 € 705 20.40 € 735

Haemorrhagic stroke

Mild, non-fatal 4.86 € 79 € 41 5.64 € 93 € 49


Severe, non-fatal 4.47 € 226 € 105 5.67 € 295 € 147

Fatal 10.88 € 25 17.68 € 42

Sum 27.99 € 1,147 34.69 € 1,115

Recurrent haemorrhagic

stroke

Mild, non-fatal 0.29 € 4 € 1 0.29 € 4 € 1


Severe, non-fatal 0.13 € 6 € 5 0.13 € 6 € 6

Fatal 0.44 € 1 0.44 € 1

Sum 1.25 € 51 1.25 € 51

Systemic embolism

Non-fatal 24.10 € 86 24.60 € 88

6


153

Apixaban VKA

Number

of events

Acute

event

related

lifetime

costs p.p.

Long-

term

costs

Number

of events

Acute

event

related

lifetime

costs p.p.

Long-term costs

Fatal 2.50 € 0 2.55 € 0

Sum 26.60 € 86 27.14 € 88

Other ICH

Non-fatal 11.72 € 176 19.20 € 303

Fatal 1.75 € 0 2.87 € 0

Sum 13.47 € 176 22.07 € 303

Other major bleedings

Non-fatal GI

bleedings

78.14 € 306 69.49 € 274

Non-fatal Non ICH or

Non GI related major

bleedings

126.23 € 496 123.45 € 490

Fatal 4.17 € 0 3.94 € 0

Sum 208.54 € 802 196.88 € 764

Clinically relevant non-

major bleeding

314.94 € 7 372.69 € 9

Myocardial infarction

Non-fatal 76.39 € 283 € 596 78.85 € 295 € 630

Fatal 14.34 14.79

Sum 90.73 € 879 93.64 € 925

Other treatment

discontinuation

647.58 652.08

Cost of anticoagulants € 3,870 € 365

Cost of routine care € 2,120 € 2,067

Cost of INR monitoring € 102 € 1,018

Total costs € 20,205 € 18,353

VKA, vitamin K-antagonist; p.p., per patient; ICH, intracranial hemorrhage; GI, gastrointestinal; INR, international

normalized ratio.

Table 6 Incremental costs, QALYs and ICER for patients with non-valvular atrial fibrillation receiving

anticoagulation therapy.

Treatment Costs (€) QALYs LYs Δ Cost (€) Δ QALY Δ LYs ICER

(€/QALY)

ICER (€/ LYs)

VKA 18,353 7.00 10.26 1,852 0.18 0.18 10,576 10,529

Apixaban 20,205 7.18 10.44

QALY, quality adjusted life year; LY, life-year; ICER, incremental cost-effectiveness ratio; VKA, vitamin-K antagonist

Chapter 6

154

Figure 2 Tornado diagram illustrating results from sensitivity analyses for apixaban vs. vitamin-K

antagonists. Black bars denote influence of the high value of the 95% confidence interval range and grey bars denote

influence of the low value for parameters investigated. ICER, incremental cost-effectiveness ratio; QALY, quality adjusted life year; ICH, intracranial hemorrhage; AF, atrial fibrillation; MI, myocardial infarction; MB, major bleeding.

Figure 3 Incremental cost-effectiveness plane. Incremental cost-effectiveness plane presents the incremental cost-effectiveness ratios of apixaban compared to vitamin-K antagonists in patients with non-valvular atrial fibrillation, obtained through a probabilistic sensitivity analysis. Points below the diagonal dotted and the full line represent simulations in which apixaban was a cost-

effective alternative at a threshold of €30,000/QALY and €20,000/QALY, respectively. QALY, quality adjusted life year; VKA, vitamin-K antagonists

6


155

Figure 4 Cost-effectiveness acceptability curves for the treatment with apixaban and VKA in non-

valvular atrial fibrillation.

The cost-effectiveness acceptability curve assesses the probability that the estimated incremental cost-effectiveness ratio is under a certain willingness to pay threshold.

VKA, vitamin-K antagonist; QALY, quality adjusted life year.

156

Chapter 6

Ta

ble

7 S

cen

ario

an

aly

ses

on

th

e i

mp

act

of

dif

fere

nt

leve

ls o

f IN

R m

on

ito

rin

g. E

ven

t ra

tes

adju

ste

d f

or

the

le

vel

of

INR

mo

nit

ori

ng,

in

cre

me

nta

l co

sts,

QA

LYs

and

ICE

Rs.

Sce

nar

io

Tre

atm

en

t IS

rate

ICH

rate

Oth

er

MB

s ra

te

CR

NM

B

rate

Co

sts

(€)

QA

LYs

LYs

Δ

Co

st(€

)

Δ

QA

LY

Δ

LYs

ICE

R

(€/Q

ALY

)

ICE

R

(€/L

Y)

cTT

R<

52

.38

%

VK

A

1.7

87

0

.95

9

1.7

65

2

.62

2

20

,32

8

6.8

1

10

.00

9

0

.34

0

.40

2

7

22

Ap

ixab

an

1.2

13

0

.29

2

0.9

99

1

.27

1

20

,33

7

7.1

5

10

.40

52

.38

%

≤

cTT

R<

66

.02

%

VK

A

1.1

59

0

.91

2

2.0

93

2

.65

7

19

,03

4

6.9

2

10

.15

2

,03

9

0.1

6

0.1

5

12

,66

2

12

,90

5

Ap

ixab

an

1.3

16

0

.50

2

1.3

84

1

.78

8

21

,07

4

7.0

8

10

.30

cTT

R ≥

76

.51

%

VK

A

0.8

31

0

.70

8

2.8

57

3

.36

4

18

,28

5

7.0

0

10

.27

1

,26

1

0.2

4

0.2

6

5,2

32

4

,69

8

Ap

ixab

an

0.7

38

0

.18

1

2.4

42

3

.04

3

19

,54

6

7.2

4

10

.53

Eq

ual

d

istr

ibu

tio

n

of

pat

ien

ts a

cro

ss c

TT

Rs

VK

A

1.1

86

0

.80

0

2.2

70

2

.99

5

19

,01

3

6.9

3

10

.17

1

,28

3

0.2

3

0.2

5

5,5

99

5

,08

5

Ap

ixab

an

1.0

44

0

.33

0

1.7

90

2

.08

3

20

,29

6

7.1

6

10

.42

INR

, in

tern

atio

na

l n

orm

aliz

ed

rat

io;

QA

LY,

qu

alit

y ad

just

ed

lif

e y

ear

; IC

ER,

incr

em

en

tal

cost

-eff

ect

ive

ne

ss r

atio

; IS

, is

che

mic

str

oke

; IC

H,

intr

acra

nia

l h

aem

orr

hag

e;

MB

, m

ajo

r b

lee

din

g; C

RN

MB

, clin

ical

ly r

ele

van

t n

on

-maj

or

ble

ed

ing;

LY

, lif

e-y

ear

; cT

TR

, clin

ic t

ime

in t

he

rap

eu

tic

ran

ge;

VK

A, v

itam

in-K

an

tago

nis

t.

6


157

DISCUSSION

This economic evaluation estimated the CE of apixaban compared to VKA for prevention

of stroke and other thromboembolic events in patients with non-valvular AF in the

Netherlands. Apixaban was shown to be a cost-effective alternative to treatment with VKA

with an ICER of €10,576/QALY or €10,529/LY. Notwithstanding that both the Dutch-

specific level of INR monitoring (TTR=72.48%) and a weighted level of baseline CHADS2

stroke risk were incorporated in this analysis, the specific long-term health and economic

benefits of treatment with apixaban are evident. Those benefits correspond with a lower

number of stroke and thromboembolic events as well as a generally better safety profile

(i.e. less ICH and CRNM bleeding events) that is associated with the use of apixaban when

compared to VKA. However, the number of other MBs was higher in the apixaban

treatment scenario compared to VKA scenario. This finding can be explained by a

relatively small difference in risks of MBs between the two comparators and a higher

number of survivors in each model cycle that would be exposed to those risks in the

apixaban treatment scenario.

Yet, the base-case ICER was found to be below the Dutch informal WTP threshold of

€20,000/QALY in 68% of PSA simulations mainly reflecting the uncertainty in the apixaban

absolute stroke risk and the risks of treatment discontinuations under both comparators.

The major impact of uncertainty in those risks on both incremental effects and

incremental costs was visualized in univariate sensitivity analyses’ tornado diagram (Figure

2) and PSA’s incremental CE plane (Figure 3). Specifically, the estimated ICERs in univariate

and probabilistic sensitivity analyses, the specific shape of incremental CE plane and the

overall number of simulated ICERs below a certain WTP threshold are mainly driven by the

uncertainty in absolute stroke risk and treatment discontinuations risk. Notably, the

relevance of the uncertainty in the apixaban absolute stroke risk can be directly attributed

to its impact on the occurrence of stroke events and their related costs of treatment and

reduced quality of life. In particular, the influence of the risk of treatment discontinuations

can be indirectly explained by the choice of a second-line treatment that would follow

after those discontinuations. In this analysis ASA was chosen to be a second-line

treatment even though it provides less protection from various thromboembolic and

bleeding events. Therefore, uncertainty in the risk of treatment discontinuations would in

the case of a higher level of risk, lead to more patients being treated with ASA and

consequently to a higher number of stroke and thromboembolic events.

Finally, the results of the scenario analyses examining the impact of different levels of INR

control were quite robust, resulting in ICER estimates bellow €20,000/QALY in all the

scenarios.

Chapter 6

158

Comparison with other studies

Findings of this analysis regarding the long-term health effects and economic

consequences of using apixaban compared to VKA are similar to the results of other

analyses studying apixaban in the US setting (40-42). Harrington et al. evaluated the use of

NOACs compared to warfarin and estimated an ICER of $15,026/QALY for apixaban (40).

Furthermore, the use of apixaban was reported to be a cost-effective alternative with

$11,400/QALY compared to warfarin in the study of Kamel et al (41). Finally, Lee et al.

found apixaban to be a cost-saving option compared to warfarin (42). Differences in the

observed ICERs might be explained by different modelling assumptions and inputs that

were used. Also differences in economic consequences associated with the use of

anticoagulants could be attributed to the variability in country-specific cost estimates and

the choice of study perspective (e.g. societal (41,42)). Finally, differences in the underlying

patients’ characteristics, modes of INR control and various modelling assumptions such as

inclusion of multiple recurrent events (42) or different comorbid health states (e.g.

transient ischemic attack (41)) could additionally hinder comparability between the study

results.

Strengths and limitations

Our study examines the potential use of apixaban for the prevention of stroke and other

thromboembolic events in patients with non-valvular AF in the Dutch setting. Country-

specific cost estimates, nation-specific background mortality and conjoint influence of the

level of INR control and the allocation of baseline CHADS2 stroke risk in Dutch population

on the events rates were implemented in this analysis. Furthermore, the impact of

different levels of INR control was examined in this study. Comparable to similar CE

studies, the health states of IS, SE, ICH, MI and CRNM bleedings were incorporated in this

analysis. Finally, unlike other aforementioned analysis, utility estimates of stroke, MI and

SE were estimated to reflect the level of utility for joint health states, using specific

methodologies (34).

Our analysis has several limitations that may restrict the interpretation of results in a

broader context. One limitation might be that the event rates incorporated in the decision

model were assumed to be constant through life even though they were based on

ARISTOTLE trial with an average follow-up of 1.8 years. This assumption, however, was

partly corrected by applying event-specific rates that account for age-related increase in

risk. In addition, multiple thromboembolic events were not incorporated in the decision

model due to the lack of epidemiological evidence. Incorporating multiple

thromboembolic events could nevertheless lead to a more favourable ICER which makes

our analysis more conservative. A further potential limitation in our analysis is the

assumption that a certain number of patients who experience ICH or other MB as well as

all patients that discontinue treatment for reasons other than thromboembolic events,

switch to a treatment with ASA. Noticeably, second-line treatment with ASA will provide a

6


159

different level of protection from stroke and thromboembolic events when compared to

initially assigned anticoagulants. In a real-life setting, patients might alternatively be

switched to other NOACs or VKAs. However, current guidelines have no clear

recommendations for the treatment following bleeding events or treatment

discontinuations. Finally, our analysis does not reflect a societal perspective. In patients

with non-valvular AF younger than 65 years, incorporating a societal perspective is of

major importance given that this patient population encounters a significant productivity

loss. Accounting for the productivity loss in the aforementioned patient population, would

lead to even more favourable ICERs.

Implications for practice and future research

Our findings showed apixaban to be a cost-effective alternative to VKAs in the Dutch

setting with 68% of chances if a willingness-to-pay threshold is set to €20,000/QALY.

Having in mind the high cost of illness due to AF and associated comorbid events, applying

an effective and safe treatment in patients is of high importance. In the ARISTOTLE and

AVERROES trials apixaban was shown to be a valuable alternative to VKAs and ASA

regarding the effectiveness and safety issues. However “real life” long-term benefits of the

use of apixaban as well as other NOACs still need to be proven, primarily regarding the

patients’ adherence to them (43). Additionally, the regulations regarding the choice of a

second-line treatment in the case of adverse drug reactions need to be more clearly

specified.

Apixaban is a NOAC which was recently approved for use in Europe, just after two other

NOACs, rivaroxaban and dabigatran, were introduced. Further investigation should be

directed to estimating comparative effectiveness and CE among the individual NOACs in

the Dutch setting.

Acknowledgment

This study was funded with a grant by Bristol-Myers Squibb and Pfizer. The authors

declare study results were not influenced by this funding.

Chapter 6

160

REFERENCES

(1) Heeringa J, van der Kuip, Deirdre AM, Hofman A, et al. Prevalence, incidence and lifetime risk of atrial

fibrillation: the Rotterdam study. Eur Heart J 2006;27(8):949-953.

(2) Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: the Framingham

Study. Stroke 1991;22(8):983-988.

(3) Goren A, Liu X, Gupta S, et al. Quality of Life, Activity Impairment, and Healthcare Resource Utilization

Associated with Atrial Fibrillation in the US National Health and Wellness Survey. PloS one 2013;8(8):e71264.

(4) Benjamin EJ, Wolf PA, D’Agostino RB, et al. Impact of atrial fibrillation on the risk of death the Framingham

Heart Study. Circulation 1998;98(10):946-952.

(5) Steger C, Pratter A, Martinek-Bregel M, et al. Stroke patients with atrial fibrillation have a worse prognosis

than patients without: data from the Austrian Stroke registry. Eur Heart J 2004;25(19):1734-1740.

(6) Baeten SA, van Exel, N Job A, Dirks M, et al. Lifetime health effects and medical costs of integrated stroke



(7) Ringborg A, Nieuwlaat R, Lindgren P, et al. Costs of atrial fibrillation in five European countries: results from

the Euro Heart Survey on atrial fibrillation. Europace 2008;10(4):403-411.

(8) Camm AJ, Lip GY, De Caterina R, et al. 2012 focused update of the ESC Guidelines for the management of

atrial fibrillation An update of the 2010 ESC Guidelines for the management of atrial fibrillationDeveloped with

the special contribution of the European Heart Rhythm Association. Europace 2012;14(10):1385-1413.

(9) Hylek EM, Evans-Molina C, Shea C, et al. Major hemorrhage and tolerability of warfarin in the first year of

therapy among elderly patients with atrial fibrillation. Circulation 2007;115(21):2689-2696.

(10) Leendertse AJ, Egberts AC, Stoker LJ, et al. Frequency of and risk factors for preventable medication-related

hospital admissions in the Netherlands. Arch Intern Med 2008;168(17):1890.

(11) Granger CB, Alexander JH, McMurray JJ, et al. Apixaban versus warfarin in patients with atrial fibrillation. N

Engl J Med 2011;365(11):981-992.

(12) Connolly SJ, Ezekowitz MD, Yusuf S, et al. Dabigatran versus warfarin in patients with atrial fibrillation. N Engl

J Med 2009;361(12):1139-1151.

(13) Patel MR, Mahaffey KW, Garg J, et al. Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N Engl J

Med 2011;365(10):883-891.

(14) Nederlandse Vereniging voor Cardiologie. Leidraad begeleide introductie nieuwe orale

antistollingsmiddelen. Available at: https://www.nvvc.nl/richtlijnen/bestaande-richtlijnen. Accessed 09/13, 2013.

(15) Connolly SJ, Eikelboom J, Joyner C, et al. Apixaban in patients with atrial fibrillation. N Engl J Med

2011;364(9):806-817.

(16) Health Council of the Netherlands. New Anticoagulants: A well-dosed introduction. The Hague: Health

Council of the Netherlands, 2012; publication no. 2012/07. Available at:

http://www.gezondheidsraad.nl/sites/default/files/summary%20antistollingmiddelen%202012.pdf. Accessed

09/13, 2013.

(17) Dorian P, Kongnakorn T, Phatak H, et al. Cost-effectiveness of Apixaban against Current Standard of Care

(SoC) for Stroke Prevention in Atrial Fibrillation Patients. Accepted for publication. Eur Heart J 2014.

(18) Lip G, Kongnakorn T, Phatak H, et al. Cost-effectiveness of Apixaban against Other Novel Oral Anticoagulants

(NOACs) for Stroke Prevention in Atrial Fibrillation Patients. Accepted for publication. Clin Ther 2014.

(19) Zorginstituut Nederland website. Available at: http://www.zorginstituutnederland.nl/bin

aries/content/documents/zinl-www/documenten/publicaties/publications-in-english/2012/1206-

pharmacotherapeutic-report-on-dabigatran-etexilate-pradaxa-for-the-prevention-of-stroke-and-systemic-

embolism-in-adult-patients-with-nonvalvular-atrial-fibrillation-with-one-or-more-risk-factors/1206-

pharmacotherapeutic-report-on-dabigatran-etexilate-pradaxa-for-the-prevention-of-stroke-and-systemic-

embolism-in-adult-patients-with-nonvalvular-atrial-fibrillation-with-one-or-more-risk-fac

6


161

tors/Pharmacotherapeutic+report+on+dabigatran+etexilate+%28Pradaxa%29+for+the+pr

evention+of+stroke+and+systemic+embolism+in+adult+patients+with+nonvalvular+atri

al+fibrillation+with+one+or+more+risk+factors.pdf. Accessed 09/01, 2013.

(20) Pisters R, van Oostenbrugge RJ, Knottnerus IL, et al. The likelihood of decreasing strokes in atrial fibrillation

patients by strict application of guidelines. Europace 2010;12(6):779-784.

(21) Wallentin L, Lopes RD, Hanna M, et al. Efficacy and Safety of Apixaban Compared With Warfarin at Different

Levels of Predicted International Normalized Ratio Control for Stroke Prevention in Atrial Fibrillation. Circulation

2013;127(22):2166-2176.

(22) Mohan KM, Crichton SL, Grieve AP, et al. Frequency and predictors for the risk of stroke recurrence up to 10

years after stroke: the South London Stroke Register. Journal of Neurology, Neurosurgery & Psychiatry

2009;80(9):1012-1018.

(23) Statistics Netherlands, Den Haag/Heerlen. Consumer prices. Available at:

http://statline.cbs.nl/StatWeb/publication/?DM=SLEN&PA=71099ENG&D1=0&D2=64,77,90,103,116,129,142,15

5,168,181,194,207,220,l&LA=EN&VW=T. Accessed July, 2013.

(24) Hakkaart-van Roijen L, Tan S, Bouwmans C. Handleiding voor kostenonderzoek. Methoden en standaard

kostprijzen voor economische evaluaties in de gezondheidszorg.Geactualiseerde versie 2010.

(25) Lopes RD, Al-Khatib SM, Wallentin L, et al. Efficacy and safety of apixaban compared with warfarin according

to patient risk of stroke and of bleeding in atrial fibrillation: a secondary analysis of a randomised controlled trial.

The Lancet 2012.

(26) Statistics Netherlands website. Causes of death; main primary causes of death, sex, age. Available at:

http://statline.cbs.nl/StatWeb/publication/?DM=SLEN&PA=7052ENG&D1=0&D2=1-2&D3=9-

21&D4=l&LA=EN&VW=T. Accessed 04/01, 2013.

(27) Statistics Netherlands website. Causes of death; main primary causes of death, sex, age. Available at:


21&D4=l&LA=EN&HDR=G3,G1,G2&STB=T&VW=T. Accessed 04/01, 2013.

(28) Statistics Netherlands website. Population; sex, age and marital status. . Available at:


120&D4=60&LA=EN&VW=T.

(29) Vaartjes I, van Dis I, Grobbee D, et al. The dynamics of mortality in follow-up time after an acute myocardial

infarction, lower extremity arterial disease and ischemic stroke. BMC cardiovascular disorders 2010;10(1):57.

(30) Sullivan PW, Slejko JF, Sculpher MJ, et al. Catalogue of EQ-5D scores for the United Kingdom. Medical

Decision Making 2011;31(6):800-804.


2003;21(3):191-200.

(32) Pink J, Lane S, Pirmohamed M, et al. Dabigatran etexilate versus warfarin in management of non-valvular

atrial fibrillation in UK context: quantitative benefit-harm and economic analyses. BMJ: British Medical Journal

2011;343.

(33) Gage BF, Cardinalli AB, Owens DK. The effect of stroke and stroke prophylaxis with aspirin or warfarin on

quality of life. Arch Intern Med 1996;156(16):1829.

(34) Hu B, Fu AZ. Predicting utility for joint health states: a general framework and a new nonparametric

estimator. Medical Decision Making 2010;30(5):E29-E39.

(35) Z-index WMG-geneesmiddelen. , 2014.

(36) De Federatie van Nederlandse Trombosediensten. De waarde van trombosediensten in Nederland. 2011;

Available at: http://www.fnt.nl/media/docs/Positionpaper_FNT.pdf. Accessed 09/13, 13.

(37) Soekhlal R, Burgers L, Redekop W, et al. Treatment costs of acute myocardial infarction in the Netherlands.

Netherlands Heart Journal 2013:1-6.

(38) Ten Cate-Hoek A, Toll D, Buller H, et al. Cost-effectiveness of ruling out deep venous thrombosis in primary

care versus care as usual. Journal of thrombosis and haemostasis 2009;7(12):2042-2049.

Chapter 6

162



(40) Harrington AR. Cost-Effectiveness of Apixaban, Dabigatran, Rivaroxaban, and Warfarin for the Prevention of

Stroke Prophylaxis in Atrial Fibrillation. 2012.

(41) Kamel H, Easton JD, Johnston SC, et al. Cost-effectiveness of apixaban vs warfarin for secondary stroke

prevention in atrial fibrillation. Neurology 2012;79(14):1428-1434.

(42) Lee S, Mullin R, Blazawski J, et al. Cost-effectiveness of apixaban compared with warfarin for stroke

prevention in atrial fibrillation. PloS one 2012;7(10):e47473.

(43) Ansell J, Granger CB, Armaganijan LV. New Oral Anticoagulants Should Not Be Used as First-Line Agents to

Prevent Thromboembolism in Patients With Atrial FibrillationResponse to Ansell. Circulation 2012;125(1):165-

170.

Chapter 7

Economic evaluation of dabigatran

for the treatment and secondary

prevention of venous thromboembolism

Jelena Stevanović, Dvortsin Evgeni, Bregt Kappelhoff, Maarten Voorhaar, Maarten J.

Postma

Submitted manuscript.

Chapter 7

166

ABSTRACT

Background: Dabigatran was proven to have similar effect on the recurrence of venous

thromboembolism (VTE) and a lower risk of bleeding compared to vitamin K antagonists

(VKA). The aim of this study is to assess the cost-effectiveness (CE) of dabigatran for the

treatment and secondary prevention in high risk patients of VTE compared to VKAs in the

Dutch setting.

Methods: The previously published Markov model was modified and updated to assess

the CE of dabigatran and VKAs for the treatment and secondary prevention in high risk

patients of VTE from a societal perspective in the base-case analysis. The model was

populated with efficacy and safety data from major dabigatran trials (i.e. RE-COVER,

RECOVER II, RE-MEDY and RE-SONATE), Dutch specific costs, and utilities derived from

dabigatran trials or other published literature. Univariate, probabilistic sensitivity and a

number of scenario analyses on the impact of various decision-analytic settings (e.g. the

perspective of analysis, use of anticoagulants only for treatment or only for secondary

prevention) were tested on the incremental cost-effectiveness ratio (ICER).

Results: In the base-case analysis, patients on dabigatran gained an additional 0.583

discounted quality adjusted life years (QALYs) over a lifetime and savings of €1,996.

Results of univariate sensitivity analysis were quite robust. The probability that dabigatran

is cost-effective at a willingness-to-pay threshold of €20,000/QALY was 100%.

Except for the scenario comparing the use of dabigatran and VKAs from the healthcare

provider perspective and the one comparing dabigatran to placebo for the prevention of

recurrent VTE in patients who are at equipoise for anticoagulation treatment where the

ICERs for dabigatran compared to VKAs of €1,005 and €33,305 per QALY gained,

respectively were estimated, other scenarios showed dabigatran was cost-saving.

Conclusion: From a societal perspective, dabigatran is likely to be a cost-effective or even

cost-saving strategy for treatment and secondary prevention of VTE compared to VKAs in

the Netherlands.

7

Economic evaluation of dabigatran for the treatment and prevention of venous thromboembolism

167

INTRODUCTION

Venous thromboembolism (VTE) can manifest as deep vein thrombosis (DVT) and/or

pulmonary embolism (PE)(1). The health burden associated with VTE is mostly determined

with the risk of a fatal PE and risk of considerable late morbidity associated with the

development of post thrombotic syndrome (PTS) or chronic thromboembolic pulmonary

hypertension (CTEPH). Moreover, recurrent DVT (rDVT) occurs in approximately 7% of

patients per year and reaches to about one quarter to one third of patients within 8 years

(2)(3). The health related quality of life (HRQoL) in VTE patients is also affected. For

example, a Dutch study found distinctly lower HRQoL scores measured with SF-36

questionnaire in patients with PE compared to the general population on the subscales:

social functioning, role emotional, general health, role physical and vitality (4).

In the Netherlands, the DVT incidence was estimated to approximately 16,000 to 20,000

cases per year (5). Though, the overall incidence of PE in the Netherlands is unknown, a

survey among Dutch pulmonologists/internists indicated the incidence of suspected PE

was 2.6 per 1,000 patients per year (2), while in general practice, 0.2 PEs per 1,000

patients were reported (6).

Both national and international guidelines recommend anticoagulation therapy as an

effective measure to prevent thrombus propagation and recurrence in VTE patients (1,2).

For the initial treatment phase of VTE, low-molecular-weight heparins (LMWHs) for at

least 5 days combined with subsequent administration of vitamin K antagonists (VKAs; e.g.

warfarin, acenocoumarol or phenprocoumon), or rivaroxaban are recommend. For the

maintenance phase, the use of VKAs or rivaroxaban is recommended for at least three

months (1,7). The need to continue anticoagulation should be re-assessed in patients

based on individual patient risk-benefit balance every three months as there is no strong

distinction between treatment and prevention phase (1,7).

VKAs present a highly effective anticoagulation treatment which use is limited by a narrow

therapeutic range (international normalised ratio (INR) limits of 2.0 and 3.0) and several

interactions with other drugs and food. To achieve the anticoagulant effect inside the

required INR range, regular monitoring and dose-adjustment is required for treatment

with VKAs. In the Dutch healthcare system, INR-monitoring is handled by thrombotic

services or patient self-management. Though it is considered highly effective in the

Netherlands, INR-monitoring also affects expenditures from both healthcare provider and

societal perspective. In particular, next to the costs of material, labour, nurse visits,

training and material for self-management that affect healthcare providers, there are

various out-of-the pocket expenses (e.g. travel costs of patients) and productivity loss

costs associated with monitoring visits that have an impact on a societal economic burden.

Recently, in the RE-COVER and RE-COVER II trials, dabigatran, a novel oral anticoagulant

(NOAC) was shown to have similar effect on VTE recurrence and a lower risk for clinically

Chapter 7

168

relevant non-major bleeding (CRNMB) and for any bleeding compared to VKA (8,9). When

administered for the extended treatment in patients with VTE who had completed at least

three months of initial therapy, dabigatran was non-inferior in preventing rVTE events and

showed a better safety profile than VKA (the RE-MEDY trial), but a significantly better

efficacy in preventing rVTE and higher risk of bleedings than placebo (the RE-SONATE trial)

(10).

Importantly, both health and economic consequences associated with the use of

dabigatran compared to VKAs need to be considered when choosing the optimal

treatment strategy. A formal pharmacoeconomic comparison of the two anticoagulant

treatments should be conducted to account for all the relevant health consequences such

as likelihood of rVTE, bleedings, PTS, CTEPH, death and other adverse events, as well as all

relevant cost parameters including the costs of drugs, administration, INR-monitoring,

event-related costs and various indirect costs.

The aim of this study is to assess the cost-effectiveness (CE) of dabigatran (300mg per day)

for the treatment and secondary prevention in high risk patients of DVT and PE compared

to VKAs for the Dutch situation.

METHODS

Decision model

The previously published Markov model was modified and updated to assess the CE of

dabigatran and VKAs for the treatment and secondary prevention of DVT and PE in the

Dutch setting. The health states included in the model were: index VTE, rVTE, major or

clinically relevant bleeding (MCRB), CTEPH, PTS, other adverse events (i.e. myocardial

infarction (MI), unstable angina (UA) and dyspepsia), off-treatment and death from other

causes (Figure 1).

In the base-case, the use of dabigatran was compared to VKAs for up to 6 months of

treatment followed by up to 18 months of secondary prevention in high risk patients. The

flow of patients with an index VTE event through the Markov model is detailed elsewhere

(11). Shortly, at the start of the simulation, a hypothetical cohort of 10,000 adult patients

(mean age 54.7 years) for whom at least 6 months of anticoagulant therapy was

considered appropriate entered the model following an index VTE (i.e. index DVT or index

PE) event and received initial treatment with LMWHs followed by either dabigatran or

VKAs. The duration of treatment with LMWHs was assumed to be 5 days in dabigatran

treatment arm to follow the summaries of product characteristics (SCPs) for dabigatran,

and 9 days in VKAs arm to simulate the treatment in RE-COVER trials. Patients in the index

VTE state were exposed to the risk of rVTE, MCRB, CTEPH, PTS, other adverse events,

treatment discontinuation and death from other causes. After the initial 6 months of

treatment, patients who remained in the index health state were simulated to receive up

7


169

to 18 months of anticoagulants for the secondary prevention, reflecting patient profiles

from the RE-MEDY trial (10).

rVTE could occur in any model cycle, however, the model was restricted to a maximum of

two rVTEs (12). Furthermore, a distinction was made between different forms of rVTE (i.e.

a fatal VTE, non-fatal PE, proximal DVT and distal DVT). After a first rVTE, patients from

both treatment arms were assumed to stop the initial treatment and initiate or reinitiate a

6 months standard treatment course of LMWHs followed by VKAs.

For patients experiencing a MCRB, a distinction was made between an intracranial

haemorrhage (ICH), other major bleed (MB), and a CRNMB. If ICH occurred, it could lead

to permanent disability, death, or recovery. MBs were modelled to lead to death or

recovery. Furthermore, it was assumed that patients can experience up to two major

bleeds (ICH or MB) during the entire time they may spend on anticoagulation; one event

could be experienced during treatment phase with study medication and one event during

LMWHs/VKAs re-treatment (12). CRNMB could occur at every model cycle while on

anticoagulation (12). After a MB or ICH, all patients were assumed to discontinue

treatment altogether having no further risk of bleeding, but continuing to be exposed to a

risk of rVTE.

In the model, all patients who experienced a first or recurrent PE (rPE) were at risk to

develop CTEPH, while those with an index or rDVT were at risk of PTS.

Other adverse events of anticoagulant therapy captured by the model are UA, MI and

dyspepsia. Patients with a non-fatal MI or UA could suffer from chronic ischemic heart

disease (IHD), or recover.

During treatment or secondary prevention phases, all patients could discontinue

treatment prior to reaching the maximum planned duration of treatment due to reasons

other than rVTE or ICH/MB. If discontinuation occurs, patients move to off-treatment state

where they continue to experience a risk of rVTE, but no further risk of bleeds. Finally,

patients in any of the health states were at risk to die from other causes. Patient

movement between health states was modeled using 1-month cycles for 60 years or until

death.

The final outcome of the decision model is the incremental cost-effectiveness ratio (ICER)

of dabigatran compared to VKAs. Quality-adjusted life-years (QALYs) and life-years (LYs)

gained were estimated as a measure of effectiveness. All relevant costs reflect a societal

perspective in the base-case analysis and are inflated, if necessary, to price year 2013

using the Dutch consumer price index (13). Future costs and health effects were

discounted by 4% and 1.5% annually after the first year, according to the Dutch guidelines

for pharmacoeconomic research (14).

Transition probabilities

In the base-case, to estimate the transition probabilities between the health states in the

model during the treatment and prevention phase, data from a published meta-analysis of

Chapter 7

170

the RE-COVER and RE-COVER II trials (8)(9) and the RE-MEDY trial (10), respectively were

used, similarly to the previously published Markov model (11).

In particular, the baseline probabilities of rVTE and MCRB were calculated from the

observed incidence in VKA arm of the aforementioned trials. For the treatment phase, the

incidences of rVTE and MCRB were log transformed with respect to time, to better reflect

the occurrence pattern of these events in the trials. For the secondary prevention phase,

the incidences were not varied with time. To calculate the probabilities of events while on

dabigatran, the estimated treatment effect (hazard ratio (HR)) for each trial endpoint was

applied to the risk in the VKA arm (Table 1).

Furthermore, the probabilities of having a fatal VTE, non-fatal PE, proximal DVT, or distal

DVT, were based on the incidences of these events in the aforementioned trials, and they

were modelled to be conditional on having a VTE event. Similarly, the probabilities of

having an ICH, other MB, fatal MB (including ICH) or CRNMB were conditional on having a

MCRB. The proportion of ICH leading to permanent disability was assumed to be 65.3%

(15).

Beyond the duration of the anticoagulant treatment, the lifetime probability of rVTE was

calculated from the assumed 10-year cumulative incidence of 39.9% (16), assuming a

constant hazard. The risk of bleeding after treatment discontinuation was assumed at

zero.

Probabilities of MI, UA and dyspepsia were estimated from the dabigatran trials (8)(9)(10).

For the treatment followed by secondary prevention, probabilities of MI, fatal MI and UA

were calculated as the sum of probabilities in the treatment and secondary prevention

trials. Events were assumed to occur at a constant rate during the trial follow-up. For

simplicity, events were assigned to occur at the midpoint (i.e., three months). Additionally,

we assumed 14% of MIs and UAs would lead to IHD (17).

CTEPH rate for all patients in index PE was estimated to be 3.8% for two years (18). For

patients experiencing non-fatal rPE events, the risk of CTEPH was applied monthly up to 2

years (18).

Published evidence suggests that mild PTS has little detrimental effect on HRQoL(6),

therefore, the model included only severe PTS. For all patients in index DVT, the 5-year

rate of PTS was estimated to be 8.1% at model start (19). A monthly probability of PTS

subsequent to non-fatal rDVT events was applied up to 5 years (19). Finally, the

probability of death due to other causes was obtained from Statistics Netherlands.

7


171

Table 1 Distribution and parameter limits for the transition probabilities in the model as used in the probabilistic sensitivity analysis.

Clinical variable Value CI (95%) Distribution Reference

Incidence of rVTE (baseline risk),

treatment

2.43% - Beta (α=62, β=2492) (8,9)

Incidence of MCRB (baseline risk),

treatment

7.68% - Beta (α=189, β=2273) (8,9)

Treatment effects

Treatment phase

rVTE, dabigatran vs VKA (HR) 1.09 0.77 – 1.54 Normal (log scale) (8,9)

MCRB, dabigatran vs VKA (HR) 0.56 0.45 – 0.71 Normal (log scale) (8,9)

Secondary prevention

rVTE, dabigatran vs VKA (HR) 1.44 0.78 – 2.64 Normal (log scale) (10)

rVTE, dabigatran vs placebo (HR) 0.08 0.02 – 0.25 Normal (log scale) (10)

MCRB, dabigatran vs VKA (HR) 0.55 0.41 – 0.72 Normal (log scale) (10)

MCRB, dabigatran vs placebo (HR) 2.69 1.43 – 5.07 Normal (log scale) (10)

Type of recurrent VTE events

Treatment phase

Dabigatran

Non-fatal PE 33.8% Beta (α=23, β=45) (8,9)

Proximal DVT 63.2% Beta (α=43, β=25) (8,9)

VTE-related death 2.9% Beta (α=2, β=66) (8,9)

Distal DVT 0.0% n/a

VKA

Non-fatal PE 33.9% Beta (α=21, β= 41) (8,9)

Proximal DVT 61.3% Beta (α=38, β=24) (8,9)

VTE-related death 4.8% Beta (α=3, β=59) (8,9)

Distal DVT 0.0% n/a


Dabigatran (RE-MEDY trial)

Non-fatal PE 34.6% Beta (α=9, β=17) (10)

Proximal DVT 61.5% Beta (α=16, β=10) (10)

VTE-related death 3.8% Beta (α=1, β=25) (10)

Distal DVT 0.0% n/a

Dabigatran (RE-SONATE trial)



VTE-related death 0.0% Beta (α=0.5, β=4) (10)

Distal DVT 0.0% Fixed

VKA



VTE-related death 5.6% n/a

Distal DVT 0.0% Beta (α=1, β=17) (10)

After therapy discontinuation


Chapter 7

172



VTE-related death 11.5% Beta (α=43, β=330) (16)

Distal DVT 0.0% n/a

Type of bleeding events

Treatment phase

Dabigatran

ICH 1.8% Beta (α=2, β=107) (8,9)

Other MB 20.2% Beta (α=22, β=87) (8,9)

Fatal MB 4.2% Beta (α=1, β=23) (8,9)

CRNMB 78.0% Beta (α=85, β=24) (8,9)

VKA

ICH 2.1% Beta (α=4, β=185) (8,9)

Other MB 19.0% Beta (α=36, β=153) (8,9)

Fatal MB 5.0% Beta (α=2, β=38) (8,9)

CRNMB 78.8% Beta (α=149, β=40) (8,9)


Dabigatran (RE-MEDY)

ICH 2.5% Beta (α=2, β=78) (10)

Other MB 13.8% Beta (α=11, β=69) (10)

Fatal MB 0.0% Fixed (10)

CRNMB 83.8% Beta (α=67, β=13) (10)

Dabigatran (RE-SONATE)

ICH 0.0% Fixed (10)

Other MBE 5.6% Beta (α=2, β=34) (10)

Fatal MBE 0.0% Fixed (10)

CRNMB 94.4% n/a (10)

VKA

ICH 2.8% Beta (α=4, β=141) (10)

Other MB 14.5% Beta (α=21, β=124) (10)

Fatal MB 4.0% Beta (α=1, β=24) (10)

CRNMB 82.8% Beta (α=120, β=25) (10)

Other probabilities

Disabled from ICH 65.3% Beta (α=90.8, β=48.2) (15)

Probability of IHD after MI and UA 14% Beta (α=19, β=116) (17)

Cumulative incidence of CTEPH at 2

years in PE patients

3.8% Beta (α=7, β=184) (18)

Probability of CTEPH (per cycle) 0.16% (18)

5 years cumulative incidence of

severe PTS

8.1% Beta (α=43, β=485) (19)

Probability of severe PTS (per

cycle)

0.14% (19)

rVTE after therapy discontinuation 39.90% 35.40% - 44.40% Normal (SE=0.02) (16)

Discontinuation probabilities (per

cycle)

Treatment phase

Dabigatran 2.09% Fixed (8,9)

7


173


VKA 1.91% Fixed (8,9)

Secondary Prevention

Dabigatran 1.00% Fixed (10)

VKA 0.97% Fixed (10)

CI, confidence interval; r VTE, recurrent venous thromboembolism; MCRB, major or clinically relevant bleeding;

VKA, vitamin K antagonists; HR, hazard ratio; DVT, deep vein thrombosis; PE, pulmonary embolism;

CRNMB = clinically relevant non-major bleed event; ICH = intracranial haemorrhage; MB = major bleed;

MI = myocardial infarction; UA, unstable angina; CTEPH = chronic thromboembolic pulmonary hypertension;

PTS = post thrombotic syndrome.

Table 2 Utility parameters applied in the model.

Parameter Value Distribution Reference

Baseline utilities

Age 18-24 years (weight for males, females) 0.976, 0.925 Fixed (20)






Age ≥ 75 years (weight for males, females) 0.910, 0.787 Fixed (20)

Disutility during active VKA treatment e 0.012 Gamma (α=28.66, β=0.0004) (25)

Disutility of index and recurrent DVT c 0.250 Normal (SE=0.0054)a (11)

Disutility of index and recurrent PE c 0.250 Normal (SE=0.0152)a (11)

Disutility of ICH or other MB d 0.130 Gamma (α=100, β=0.001) (11)

Disutility of disabled from ICH f 0.380 Gamma (α=16, β=0.024)b (21)

Disutility of CRNMB d 0.040 Gamma (α=100, β=0.0004) (11)

Disutility of MI d 0.063 Gamma (α=22.57, β=0.003) (22)

Disutility of Angina d 0.085 Gamma (α=40.40, β=0.002) (22)

Disutility of Dyspepsia e 0.040 Gamma (α=16, β=0.003) b (23)

Disutility of CTEPH d 0.440 Gamma (α=16, β=0.028) b (24)

Disutility of severe PTS f 0.070 Gamma (α=39.22, β=0.002) (6)

CRNMB = clinically relevant non-major bleed event; DVT = deep vein thrombosis; ICH = intracranial haemorrhage;

MB = major bleed; MI = myocardial infarction; PE = pulmonary embolism; CTEPH = chronic thromboembolic

pulmonary hypertension; PTS = post thrombotic syndrome. a Change in mean from baseline to 3 months. In the probabilistic analysis, the mean baseline and 3-month value

were individually sampled from normal distributions defined by the mean and standard error (standard error was

calculated from the standard deviation and N) and the difference calculated for each simulation. b Variance was not reported; the standard error is assumed to be 25% of the mean. c The duration of disutility was assumed to be 6 weeks similarly to the previously published study. d A disutility is applied in the month of the event. Specifically, the duration of the impact of UA and MI on HRQoL

was assumed to be 3 months. e The disutility applied is assumed to last for the duration of treatment. f A disutility is applied for the remaining lifetime.

Chapter 7

174

Utilities

Table 2 summarises the utilities used in the model. Patients were assigned baseline age-

and gender-specific utilities derived from the general Dutch population (20). These

estimates formed the baseline from which the utility decrements associated with VTE,

bleeding and other adverse events were subtracted.

Utility decrements associated index and rVTE, MB and CRNMB were based on a meta-

analysis of EQ-5D data collected in the RE-COVER and RE-COVER II trials and applied in the

model similarly to the previously published study (11).

Utility decrements following the occurrence of other adverse events (i.e. MI, UA,

dyspepsia, disabled from ICH, CTEPH, and severe PTS) were derived from the published

studies and applied additively for a specific time interval in the model (21)(22)(23)(6)(24).

Finally, utility decrements reflecting the use of VKA were applied (25).

Costs

In the base-case analysis, all costs were collected from a societal perspective, therefore,

both direct (inside and outside healthcare) and indirect costs were included (Table 3).

Direct costs inside healthcare included the costs related to: drugs, visits to general

practitioner (GP), administration, INR-monitoring, event-related resources. Costs of

dabigatran, defined as price per defined daily dose (2x 150mg), VKAs and LMWHs were

taken from the official Dutch price list (Z-index) (26). Importantly, the price of dabigatran

extracted from Z-index is established for other registered indications of dabigatran in the

Netherlands (i.e. prevention of VTE in patients who have undergone elective total hip

replacement surgery or total knee replacement surgery, and prevention of stroke and

systemic embolism in non-valvular atrial fibrillation). Acenocumarol and phenprocoumon

are the only VKAs registered in the Netherlands, therefore, the cost of VKAs was

estimated as a weighted average cost of those drugs based on their usage in the

Netherlands (80%:20%, respectively)(27). The cost of LMWHs was assumed as a weighted

average cost of enoxaparin, dalteparin, tinzaparin and nadroparin (26). All treatment

alternatives were assumed to have a cost of one initial GP visit in the first month and one

follow up visit in the 4th month of a treatment.

The cost of administration of LMWHs was estimated to reflect the costs of administration

in hospital and at home. For patients receiving LMWHs in hospital, the costs of

administration was adjusted for the percentage of patients and time they spent being

hospitalized for DVT and for PE (28). The costs of administration at home accounted for

the costs of self-injection and costs for patients requiring a nurse visit for injection (Table

3)(28,29).

The costs of INR-monitoring reflected the costs of monitoring handled by thrombotic

services and costs for patient self-management (Table 3). In the Netherlands, self-

management is applied by 14.9% of patients on treatment with VKAs. Therefore, the costs

of initial training for self-management and monthly follow up costs associated with the

7


175

rental of equipment were applied for this patient population (Table 3). Resource use

associated with INR-monitoring at thrombotic services (i.e. number of visits) was based on

the data from the RECOVER trials (8)(9). In particular, in the first month of a treatment,

the cost of INR-monitoring by thrombotic services reflected the costs of 4.3 visits to

thrombotic services. In follow up months (2nd until 6th month), the cost of 1.9 visits per

month was assumed (8)(9). For the application of VKAs longer than 6 months (i.e.

secondary prevention phase), the costs of INR-monitoring by thrombotic services reflected

the cost of 1.5 visits per month (8)(9). Moreover, direct costs outside healthcare (i.e.

travel costs) were attributed to the nurse visits for injection of LMWHs. Acute care costs

associated with clinical events (e.g. DVT, PE, ICH, other MB, CRNMB, PTS, CTEPH, MI, UA,

and dyspepsia) were adopted from previous costing studies conducted in the Netherlands.

Patients surviving acute ICH, MI, PTS and CTEPH were assigned with long-term

maintenance costs.

Indirect costs outside healthcare included: productivity loss costs, caregiver time costs for

patients experiencing ICH and travel costs for the visits of patients to thrombotic services.

A 2-hour productivity loss costs were assumed for all INR-monitoring visits to thrombotic

services and all GP-related visits. Additionally, productivity loss costs associated with

hospitalizations due to DVT (0.63 days), PE (7 days) and MI (5.6 days) were included. The

number of productivity loss hours for each of the aforementioned hospitalizations was

estimated in order to account for regular working hours (8 hours per day) and was

corrected for the weekends (14,30).

Table 3 Cost parameters applied in the model.

Cost parameters Average cost

(2013, €)

Rangea Reference

Medication, administration and monitoring costs

VKA (daily) 0.04 0.03-0.05 (26)

Dabigatran (daily) 2.30 Fixed (26)

LMWH (daily) 10.65 7.99-13.31 (26)

LMWH at home, self-injection (one-off training) 16.77 9.59-25.93 (29)

LMWH at home, nurse injection (per day after

discharge)

17.50 10.00-27.05 (29)

LMWH, administration in clinic (per day after

discharge) incl. travel costs

16.54 9.45-25.57 (29)

LMWH at home, self-injection (domiciliary care) 6.74 3.85-10.43 (29)

GP visit 30.54 17.46-47.22 (14)

INR-control self-management initial monthly

cost

90.46 51.71-139.88 (14,31)

INR-control cost incl. travel costs (per visit) b 12.54 7.17-19.38 (14,31)

INR-control self-management (monthly) 12.29 7.03-19.01 (14,31)

Events costs

Chapter 7

176

Cost parameters Average cost

(2013, €)

Rangea Reference

DVT 1,187.23 679-1,836 (30)

PE 4,221.01 2,413-6,527 (30)

ER visit 167.28 96-259 (30)

Chest x-ray 156.15 89-241 (30)

Electrocardiogram 30 17-46 (30)

Acute ICH 32,754 18,722-50,646 (21)

ICH direct mild (annually) 2,367.97 1,354-3,662 (21)

ICH direct moderate (annually) 18,268 10,442-28,247 (21)

ICH direct severe (annually) 23,353 13,348-36,110 (21)

MB 4,969 2,840-7,683 (30)

CRNMB c 31 17-47 (14)

PTS (year 1) 25,073 14,331-38,769 (30)

PTS (year 2) d 61 35-94 (14)

MI acute 5,021 4,936- 5,106 (32)

MI follow up (monthly) 97 55-150 (33)

UA 5,351 5,236-5,467 (34)

Dyspepsia e 0.69 0.39-1.07 (26)

CTEPH acute f 7,121 4,070-11,011 (35)

CTEPH follow up (monthly) 84 48-130 (35)

Indirect costs

Productivity loss age group 55-60 (per hour) g 31 17-47 (14)

Productivity loss age group 60-65 (per hour) g 23 13-36 (14)

ICH informal care mild (annually) 12,369 7,070-15,462 (36)

ICH informal care moderate (annually) 16,345 9,343-25,274 (36)

ICH informal care severe (annually) 20,322 11,616-31,422 (36)

VKA, vitamin K antagonists; LMWH, low molecular weight heparins; INR, international normalised ratio; MCRB, major or clinically relevant bleeding; DVT, deep vein thrombosis; PE, pulmonary embolism; CRNMB = clinically relevant non-major bleed event; ICH = intracranial haemorrhage; MB = major bleed; MI = myocardial infarction;

UA, unstable angina; CTEPH = chronic thromboembolic pulmonary hypertension; PTS = post thrombotic syndrome GP, general practitioner. a Cost estimates that were available only as single point estimates, were assumed to follow a log-normal distribution with a coefficient of variation equal to 0.25. b Travel costs of patients included only in the base-case. c Assumed to be equal to the cost of a GP visit. d Assumed to be equal to the cost of two GP visit. e Assumed the cost of Omeprazol 20mg. f Based on the study by Mayer et al, pulmonary endarterectomy is applied to 56.8% of cases. g One hour of productivity loss costs was estimated as a weighted average cost for employed and non-employed

population in the Netherlands in the specific age group.


Univariate sensitivity analyses were performed to identify the key determinants of CE by

varying parameters individually over the ranges derived from their 95% confidence

intervals. Where confidence intervals and standard deviations of parameters were

unavailable, the standard error was assumed to be 25% of the mean. The exceptions were

7


177

made when varying discount rates which were varied between 0 and 5%, and the number

of days on treatment with LMWHs which were varied between 5 and 9 days. The results

were defined in terms of incremental cost per QALY and are presented diagrammatically

in the form of a tornado diagram.

Additionally, a probabilistic sensitivity analysis (PSA) was performed to assess the

robustness of the findings by performing 1,000 simulations to generate ICERs in which

event risks and HRs, costs and utilities were simultaneously varied randomly within their

ranges. HRs were sampled from a normal distribution on the log scale, probabilities were

sampled from a beta distribution, and costs were sampled from a gamma distribution. For

utilities, a gamma distribution was used, except for utilities assigned to DVT and PE, for

which a normal distribution was used. Results from the PSA were plotted on a CE plane.

Scenario analyses

To investigate the impact of applying dabigatran under different decision making settings

four scenario analyses were conducted. First scenario compared the use of dabigatran and

VKAs for treatment and secondary prevention in high risk patients from the healthcare

provider perspective. Second scenario compared the use of dabigatran to VKAs for up to 6

months of treatment only. Third scenario assessed the use of anticoagulants for up to 18

months of secondary prevention only (not considering the preceding treatment duration).

In the fourth scenario, the use of dabigatran for up to 6 months of secondary prevention

(not considering the preceding treatment duration) was compared to placebo. Here, study

population simulated the profile of the patients in the RE-SONATE trial (i.e. patients for

whom the need for secondary prevention is at equipoise (10)). Data from the RE-SONATE

trial were the main sources used to estimate the transition probabilities between the

health states in the model in this scenario.

Chapter 7

178

Figure 1 Markov model VTE, venous thromboembolism; DVT, deep vein thrombosis; PE, pulmonary embolism; r, recurrent; LMWH, low

molecular weight heparin; CRNMB = clinically relevant non-major bleed event; ICH = intracranial haemorrhage; MB = major bleed; MI = myocardial infarction; CTEPH = chronic thromboembolic pulmonary hypertension; PTS = post thrombotic syndrome; UA, unstable angina; IHD, ischemic heart disease.

RESULTS

Under base-case conditions, in a hypothetical cohort of 10,000 patients with VTE event

followed over their lifetime starting at age 54.7 years, dabigatran averted 720 MCRBs

compared with VKAs but resulted in an additional 86 rVTEs, and 65 MIs (Table 4). A

comparable number of PTS, CTEPH and UA was observed in both dabigatran and VKAs

treatment arms. Dabigatran was associated with a projected discounted quality-adjusted

life expectancy of 19.187 QALYs compared with 19.129 QALYs for patients receiving VKAs.

Costs allocation across different categories indicated that costs associated with handling

rVTE, bleeding and other adverse events were the greatest contributors to the total

expenditures. In VKAs treatment arm, these costs were higher compared to dabigatran

arm (€12,409 vs €11,072). Expenditures for event-related costs were followed by

monitoring and administration costs which were higher with VKAs compared to

dabigatran (€3,730 vs. €1,242), and the total drug costs that were higher with dabigatran

than VKAs (€1,549 vs. €298). Finally, accounting for all the aforementioned cost categories

resulted in the total lifetime costs varied from €12,254 per person for VKAs to €10,258 per

7


179

person for dabigatran at a discount rate of 4%. In total, the savings of €1,996 and an

additional 0.0583 discounted QALYs per patient where observed when applying

dabigatran compared to VKAs (Table 5).


The results of univariate sensitivity analyses for the top 15 parameters by the order of

influence they have to the ICERs are presented in the form of a tornado diagram (Figure

2). Specifically, the ICER was mostly influenced by variations in the probability of VTE-

related death, productivity loss costs in the age group 55-60, probability of MCRBs-related

and probability of ICH.

The results of 1,000 iterations in PSA are presented through an incremental CE plane in

Figure 3. The probability that dabigatran is cost-effective at a willingness-to-pay (WTP)

threshold of €20,000/QALY was 100%.

Scenario analyses

The results of the scenario analyses are presented in Table 5. In the scenario comparing

dabigatran to VKAs for the treatment, and in the one comparing them for the secondary

prevention in high-risk patients, dabigatran remained cost-saving. In the scenario

comparing dabigatran to VKAs for treatment and secondary prevention from the

healthcare provider perspective dabigatran was cost-effective with an ICER of €1,005 per

QALY gained. Finally, the scenario examining the prevention of recurrent VTE in patients

who are at equipoise for anticoagulation treatment with dabigatran compared to placebo,

yielded an ICER of €33,305 per QALY gained.

Chapter 7

180

Table 4 Recurrent VTE, bleeding complications and other adverse events and related costs within a hypothetical patient population of 10,000 subjects receiving dabigatran and VKA over a lifetime

horizon. Dabigatran VKA

Number of

events

Costs p.p.

(undiscounted)

Number

of events

Costs p.p.

(undiscounte

d)

Index VTE 10,000 €2,142 10,000 €2,142

All recurrent VTE 13,471 13,384

Recurrent non-fatal VTE 11,959 11,871

Non-fatal DVT 8,761 €1,071 8,713 €1,065

Non-fatal PE 3,198 €1,592 3,158 €1,570

VTE-related death 1,512 €0 1,513 €0

All MCRBs 1,351 2,071

Non-fatal MCRBs 1,342 2,052

ICH 28 €1,876 47 €3,262

Other MBs 230 €118 339 €177

CRNMBs 1,084 €9 1,665 €14

Deaths from bleeding 9 €0 19 €0

MI 86 €93 21 €23

UA 23 €23 23 €23

Dyspepsia 682 €0.05 112 €0.01

PTS 1,294 €3,482 1,290 €3,471

CTEPH 243 €667 242 €662

Medication

Investigational treatment €1,315 €24

LMWHs, index event €71 €110

Re-treatment recurrent event, VKA €12 €12

Re-treatment recurrent event, LMWHs €152 €152

Monitoring and administration

INR-monitoring, GP visits, administration

and productivity loss

€167 €2,622

Index event: Administration of LMWHs €43 €81

Re-treatment with VKA for recurrent

event: INR-monitoring, GP visits,

administration and productivity loss

€901 €896

Re-treatment recurrent event:

Administration of LMWHs

€131 €130

VKA, vitamin K antagonists; VTE, venous thromboembolism; DVT, deep vein thrombosis; PE, pulmonary

embolism; r, recurrent; LMWH, low molecular weight heparin; CRNMB = clinically relevant non-major bleed

event; ICH = intracranial haemorrhage; MB = major bleed; MI = myocardial infarction; CTEPH = chronic

thromboembolic pulmonary hypertension; PTS = post thrombotic syndrome; UA, unstable angina; INR,

international normalised ratio.

7


181

Table 5 Results of the base-case and scenario analyses. Base-case: 6 months treatment + 18 months secondary prevention (societal perspective)

Dabigatran VKA Difference

Discounted LYs 22.053 22.025 0.0282

Discounted QALYs 19.187 19.129 0.0583

Costs (€) undiscounted 13,865 16,437 -2,571

Costs (€) discounted 10,258 12,254 -1,996

ICER (€/ LYs) Cost-saving

ICER (€/ QALYs) Cost-saving

Scenario 1: 6 months treatment + 18 months secondary prevention (healthcare provider perspective)

Discounted LYs 22.053 22.025 0.0282


Costs (€) undiscounted 12,140 12,346 -206

Costs (€) discounted 9,067 9,008 59

ICER (€/ LYs) 2,078

ICER (€/ QALYs) 1,005

Scenario 2: 6-months treatment (societal perspective)

Discounted LYs 21.924 21.907 0.0170






Scenario 3: 18-months secondary prevention in high-risk patients (societal perspective)

Discounted LYs 22.044 22.030 0.0144






Scenario 4: 6-months secondary prevention with placebo in patients for whom the need for secondary prevention is at equipoise (societal perspective)

Discounted LYs 21.950 21.950 0.0003


Costs (€) undiscounted 7,971 7,873 98

Costs (€) discounted 4,940 4,823 117

ICER (€/ LYs) 428,158

ICER (€/ QALYs) 33,305 VKA, vitamin K antagonists; ICER, incremental cost-effectiveness ratio; QALY, quality adjusted life year; LY, life

year.

Chapter 7

182

Figure 2 Tornado diagram illustrating ICERs from sensitivity analyses for dabigatran vs. vitamin-K antagonists. Figure 2 presents a tornado diagram illustrating the impact of varying each of input parameters on the ICER while holding all the other model parameters fixed. Light grey bars denote influence of the high limit and dark grey bars denote influence of the low limit of the input parameters investigated on the ICER. The solid vertical line

represents the base case incremental costs per QALY for dabigatran compared to VKA. Horizontal bars indicate the range of incremental costs per QALY obtained by setting each variable to the values shown while holding all

other values constant. ICER, incremental cost-effectiveness ratio; QALY, quality adjusted life year; VKA, vitamin K antagonists; VTE, venous thromboembolism; r, recurrent; ICH = intracranial haemorrhage; MBE = major bleeding event; HR, hazard

ratio.

Figure 3 Incremental cost-effectiveness plane.

7


183

DISCUSSION

Our base-case decision analysis demonstrated dabigatran may be a cost–saving

alternative to VKAs for the treatment and secondary prevention of VTE from societal

perspective. Patients on dabigatran gained an additional 0.583 discounted QALYs over a

lifetime and savings of €1,996. The key drivers of the CE of dabigatran relative to trial-

based VKA are based on its ability to reduce MCRBs as found in RECOVER trials.

Particularly, the use of dabigatran resulted in 710 less MCRB events (i.e. 19 ICHs, 109

other MBs and 582 CRNMBs) in a cohort of 10,000 patients.

Results were sensitive to the probability of VTE-related death, productivity loss costs,

probability of MCRBs-related death and probability of ICH, yet, they all indicated

dabigatran to be cost-saving compared to VKAs. Moreover, the PSA showed that the

likelihood of dabigatran being cost-effective at WTP threshold of €20,000 per QALY was

100%.

The results of the scenario analyses comparing dabigatran to VKAs for the treatment,

secondary prevention in high-risk patients, were quite robust, all indicating dabigatran

may be cost-saving alternative to VKAs. However, in the scenario examining the CE of

anticoagulants for the treatment and secondary prevention from the healthcare provider

perspective, dabigatran was shown to be a cost-effective alternative to VKAs with an ICER

of €1,005 per QALY gained. Interestingly, although the variability in productivity loss costs

showed an impact on the estimated ICER in the univariate sensitivity analyses, excluding

these costs together with other indirect costs still led to highly cost-effective findings in

the aforementioned scenario. Finally, in the scenario examining the prevention of

recurrent VTE in patients who are at equipoise for anticoagulation treatment, unlikely to

be treated with anticoagulants in clinical practice, dabigatran was found to be cost-

effective compared to placebo with an ICER of €33,305 per QALY gained. This finding

reflects the higher total costs associated with greater number of MCRBs and drug costs in

dabigatran treatment arm compared to placebo arm.

To our knowledge, this is the first study that examined the use of dabigatran compared to

VKAs for the treatment and prevention of VTE in the Dutch setting. In terms of the

economic consequences of using dabigatran compared to VKAs, our findings are similar to

the ones by Braidy et al (37). A cost-minimisation analysis by Braidy et al. investigated the

use of NOACs and VKAs for the prevention of VTE and stroke in patients with atrial

fibrillation from the third-party payer perspective in Australian setting (37). They found

dabigatran to be dominant over VKA (∆c≈$AUS40) in terms of cost of drug administration

and therapeutic monitoring. Notably, a direct comparability between the two studies is

hampered due to differences in the underlying patients’ characteristics, safety and

effectiveness data used, country-specific cost estimates and study perspective.

Chapter 7

184

Our study is confronted with several potential limitations. One limitation might be that the

duration of initial treatment with LMWHs was assumed to be different in dabigatran and

VKAs treatment arms in the base-case analysis (i.e. 5 and 9 days respectively). However,

the duration of treatment with LMWHs was assumed to have no impact on the

effectiveness of the follow up use of dabigatran and VKAs. To cope with this limitation, we

varied the duration of LMWHs treatment between 5 and 9 days for both treatment

alternatives in univariate sensitivity analyses. The results remained robust to variability in

the duration of LMWHs use. Furthermore, this study simulated the occurrences of all

MCRBs further subdivided into ICHs, other MBs and CRNMBs, however, the meta-analysis

of the RE-COVER trials indicated that there was only a marginally significant reduction of

MBs observed in the double-dummy period in dabigatran arm compared to VKA.

Therefore, simulating the occurrences of MBs might overestimate the benefits in the

dabigatran arm compared to VKA. Similarly, acute coronary syndromes (i.e. MIs and UAs)

were modelled in this study although their incidence was only numerically higher with

dabigatran compared to VKAs. A further potential limitation in our study concerns the

assumption that patients in both treatment arms who experience a first recurrent VTE

event would switch to a 6-months standard treatment course of LMWH followed by VKAs.

This may not always be the case and patients might alternatively be switched to other

NOACs. However, there are currently no available efficacy and safety data that could

characterize such a switch. Furthermore, a maximum of two rVTEs over the lifetime of the

patients and two MBs during the anticoagulation treatment were modelled. This

assumption may be considered conservative given that a better safety profile of

dabigatran treatment would be associated with a lower number of MBs and consequent

lower costs compared to treatment with VKAs. Another limitation concerns treatment

discontinuation that was assumed for patients who experience a MB or ICH. In a real-life

setting such a decision would likely be based on individual patient characteristics. Finally,

given the lack of specific treatment recommendations for patients experiencing CRNMBs,

the discontinuation of treatment due to CRNMB was not modelled. Notably, in daily

practice, patients may discontinue with the treatment after a certain number of

consequent CRNMBs.

In conclusion, from a societal perspective, this modelling study suggests that the use of

dabigatran for treatment and secondary prevention of VTE is maximally likely to be a cost-

effective alternative to VKAs at a WTP threshold €20,000 per QALY in the Dutch setting.

Importantly, even when the comparison between dabigatran and VKAs was assessed from

the healthcare provider perspective, dabigatran remained highly cost-effective with an

ICER of €1,005 per QALY gained.

In addition to some established advantages of dabigatran (e.g. better safety profile than

VKAs; excludes the need for INR-monitoring), our study estimated the long-term economic

benefits associated with its use. Yet, it must be acknowledged that such benefits in a “real

7


185

life” setting are still to be proven. Notably, given that dabigatran is the second NOAC

registered in Europe for the treatment and secondary prevention of VTE, further

investigation is needed to estimate comparative effectiveness and CE among the

individual NOACs.

Acknowledgment

This study was funded with a grant by Boehringer Ingelheim. The authors declare study

results were not influenced by this funding.

Chapter 7

186

REFERENCES

(1) Guyatt GH, Akl EA, Crowther M, et al. Introduction to the ninth edition: antithrombotic therapy and

prevention of thrombosis: American College of Chest Physicians evidence-based clinical practice guidelines.

CHEST Journal 2012;141(2_suppl):48S-52S.

(2) CBO. Richtlijn Diagnostiek, Preventie en Behandeling van Veneuze Trombo-embolie en Secundaire Preventie

Arteriële Trombose. Hoofdstuk 1: Diagnostiek bij diep veneuze trombose; Hoofdstuk 2: Diagnostiek van

longembolie; Hoofdstuk 4: Medicamenteuze behandeling van veneuze trombo-embolie. . 2008; Available at:

https://www.nvkc.nl/kwaliteitsborging/documents/Richtlijnveneuzetrombodefinitief--130109.pdf. Accessed

06/08, 2014.

(3) Cushman M, Tsai AW, White RH, et al. Deep vein thrombosis and pulmonary embolism in two cohorts: the

longitudinal investigation of thromboembolism etiology. Am J Med 2004;117(1):19-25.

(4) van Es J, den Exter PL, Kaptein AA, et al. Quality of life after pulmonary embolism as assessed with SF-36 and

PEmb-QoL. Thromb Res 2013;132(5):500-505.

(5) Rijksinstituut voor Volksgezondheid en Milieu. RIVM rapport 260242001/2010. Risico's door gebrekkige

afstemming in de trombosezorg. Bilthoven: RIVM, 2010. Available at:

http://www.rivm.nl/dsresource?objectid=rivmp:11843&type=org&disposition=inline&ns_nc=1. Accessed 01/20,

2015.

(6) Lenert LA, Soetikno RM. Automated computer interviews to elicit utilities: potential applications in the

treatment of deep venous thrombosis. J Am Med Inform Assoc 1997 Jan-Feb;4(1):49-56.

(7) Goldhaber SZ, Bounameaux H. Pulmonary embolism and deep vein thrombosis. The Lancet

2012;379(9828):1835-1846.

(8) Schulman S, Kearon C, Kakkar AK, et al. Dabigatran versus warfarin in the treatment of acute venous

thromboembolism. N Engl J Med 2009;361(24):2342-2352.

(9) Schulman S, Kakkar AK, Goldhaber SZ, et al. Treatment of Acute Venous Thromboembolism with Dabigatran

or Warfarin and Pooled Analysis. Circulation 2013:CIRCULATIONAHA. 113.004450.

(10) Schulman S, Kearon C, Kakkar AK, et al. Extended use of dabigatran, warfarin, or placebo in venous

thromboembolism. N Engl J Med 2013;368(8):709-718.

(11) Jugrin A, Ustyugova A, Urbich M, et al. The Cost-utility of Dabigatran Etexilate Compared with Warfarin in

Treatment and Extended Anticoagulation of Acute Venous Thromboembolism in the UK. [Accepted for

publication]. Thromb Haemost 2015.

(12) Samenvatting consultatiebijeenkomst 16 sept 2014. .

(13) Statistics Netherlands, Den Haag/Heerlen. Consumer prices. Available at:

http://statline.cbs.nl/StatWeb/publication/?DM=SLEN&PA=71099ENG&D1=0&D2=64,77,90,103,116,129,142,15

5,168,181,194,207,220,l&LA=EN&VW=T. Accessed July, 2013.

(14) Hakkaart-van Roijen L, Tan S, Bouwmans C. Handleiding voor kostenonderzoek. Methoden en standaard

kostprijzen voor economische evaluaties in de gezondheidszorg.Geactualiseerde versie 2010.

(15) Rosand J, Eckman MH, Knudsen KA, et al. The effect of warfarin and intensity of anticoagulation on outcome

of intracerebral hemorrhage. Arch Intern Med 2004;164(8):880-884.

(16) Prandoni P, Noventa F, Ghirarduzzi A, et al. The risk of recurrent venous thromboembolism after

discontinuing anticoagulation in patients with acute proximal deep vein thrombosis or pulmonary embolism. A

prospective cohort study in 1,626 patients. Haematologica 2007 Feb;92(2):199-205.

(17) Brouwer MA, van den Bergh PJ, Aengevaeren WR, et al. Aspirin plus coumarin versus aspirin alone in the

prevention of reocclusion after fibrinolysis for acute myocardial infarction: results of the Antithrombotics in the

Prevention of Reocclusion In Coronary Thrombolysis (APRICOT)-2 Trial. Circulation 2002 Aug 6;106(6):659-665.

(18) Pengo V, Lensing AW, Prins MH, et al. Incidence of chronic thromboembolic pulmonary hypertension after

pulmonary embolism. N Engl J Med 2004;350(22):2257-2264.

(19) Prandoni P, Villalta S, Bagatella P, et al. The clinical course of deep-vein thrombosis. Prospective long-term

follow-up of 528 symptomatic patients. Haematologica 1997 Jul-Aug;82(4):423-428.

7


187

(20) Szende A, Williams A. Measuring self-reported population health: An international perspective based onEQ-

5D. EuroQol Group 2014.

(21) Baeten SA, van Exel, N Job A, Dirks M, et al. Lifetime health effects and medical costs of integrated stroke



(22) Sullivan PW, Slejko JF, Sculpher MJ, et al. Catalogue of EQ-5D scores for the United Kingdom. Med Decis

Making 2011 Nov-Dec;31(6):800-804.

(23) Gerson LB, Ullah N, Hastie T, et al. Patient-derived health state utilities for gastroesophageal reflux disease.

Am J Gastroenterol 2005;100(3):524-533.

(24) McKenna SP, Ratcliffe J, Meads DM, et al. Development and validation of a preference based measure

derived from the Cambridge Pulmonary Hypertension Outcome Review (CAMPHOR) for use in cost utility

analyses. Health Qual Life Outcomes 2008;6:65.

(25) Marchetti M, Pistorio A, Barone M, et al. Low-molecular-weight heparin versus warfarin for secondary

prophylaxis of venous thromboembolism: a cost-effectiveness analysis. Am J Med 2001;111(2):130-139.

(26) Z-index WMG-geneesmiddelen. , 2014.

(27) De Federatie van Nederlandse Trombosediensten. De waarde van trombosediensten in Nederland. 2011;

Available at: http://www.fnt.nl/media/docs/Positionpaper_FNT.pdf. Accessed 09/13, 13.

(28) van Bellen B, Bamber L, Correa de Carvalho F, et al. Reduction in the length of stay with rivaroxaban as a

single-drug regimen for the treatment of deep vein thrombosis and pulmonary embolism. Current Medical

Research & Opinion 2014;30(5):829-837.

(29) Postma MJ, Kappelhoff BS, van Hulst M, et al. Economic evaluation of dabigatran etexilate for the primary

prevention of venous tromboembolic events following major orthopedic surgery in the Netherlands. Journal of

medical economics 2012;15(5):878-886.

(30) Ten Cate-Hoek A, Toll D, Buller H, et al. Cost-effectiveness of ruling out deep venous thrombosis in primary

care versus care as usual. Journal of thrombosis and haemostasis 2009;7(12):2042-2049.

(31) Nederlandse Zorgautoriteit (NZa). Beleidsregel voor trombosediensten. Available at:

http://www.nza.nl/137706/145406/623721/BR-CU-2083-Trombosediensten.pdf. Accessed 06/06, 2014.

(32) Soekhlal R, Burgers L, Redekop W, et al. Treatment costs of acute myocardial infarction in the Netherlands.

Netherlands Heart Journal 2013:1-6.





131.

(35) CFH-rapport rivaroxaban. Rapportnr 12/13. Diemen. College voor zorgverzekeringen 2012.

(36) Van den Berg B, Brouwer W, van Exel J, et al. Economic valuation of informal care: lessons from the

application of the opportunity costs and proxy good methods. Soc Sci Med 2006;62(4):835-845.

(37) Braidy N, Bui K, Bajorek B. Evaluating the impact of new anticoagulants in the hospital setting. Pharmacy

practice 2011;9(1):1-10.

Chapter 8

General discussion

8

General discussion

191

In the last decades, a decline in cardiovascular diseases (CVD) mortality was observed

across most of the developed countries (1). This can likely be attributed to a broad

application of prevention strategies (2). In clinical practice, CVD prevention strategies aim

at identifying individuals at high CVD risk followed by enhancing risk factor(s) modification

(e.g. lifestyle changes and/or pharmacological interventions). This approach was shown to

successfully reduce the risk of CVD (3). However, the rate of CVD-related hospital

admission and mortality, as well as CVD attributable costs of acute treatment, long-term

management and rehabilitation care, all indicate that further spread of CVD prevention

approaches and interventions have the potential to be beneficial and further reduce the

disease burden (1)(4).

Pharmacological innovations open new opportunities to prevent or delay the onset of

CVD. Yet, innovative interventions often come at a higher cost what may be seen as a

burden to the healthcare budget and possibly a limitation to patient access. Because of

healthcare budget constraints, the decisions to implement innovative interventions in

clinical practice have to be made with respect to both health and economic consequences

associated with their use.

This thesis deals with pharmacoeconomic issues that relate to pharmacological preventive

interventions in CVD. In particular, it addresses challenges in simulating and synthesizing

pharmacoeconomic evidence in (primary) CVD prevention, and proposes potential

solutions to those challenges. Moreover, it derives pharmacoeconomic evidence on the

use of oral anticoagulants (i.e. vitamin K antagonists (VKAs) and novel oral anticoagulants

(NOACs)) for secondary CVD prevention. This chapter summarizes and discusses the main

results, findings, and ideas described in the previous chapters, including some future

perspectives. Lastly, to assess the main implications of this thesis, this chapter discusses

methods for providing robust pharmacoeconomic evidence in CVD prevention, and direct

implications of pharmacoeconomic evidence on the use of VKAs and NOACs for current

practice.

PRINCIPLE FINDINGS AND FUTURE PERSPECTIVES

Part I: Simulating and synthesizing health and economic evidence in (primary)


Part I of this thesis examined some of the challenging issues on simulating and

synthesizing health and economic evidence in (primary) CVD prevention. In Chapter 2, a

systematic literature review on the application of CVD risk prediction models in

pharmacoeconomic studies for primary CVD prevention in high income countries was

described. This review aimed at understanding the extent that methods for incorporating

CVD risk models in pharmacoeconomic studies are being used and assesses their quality.

Through a systematic search, 12 eligible pharmacoeconomic studies were identified. The

quality assessment indicated disagreement between the characteristics of populations of

Chapter 8

192

the intended population(s) of the risk model, frequent disagreement between risk model-

and study time horizons and a general lack of proper consideration of the uncertainty

surrounding risk predictions. Based on the findings in Chapter 2, it was concluded that

projecting intermediate effectiveness/efficacy evidence to long-term health and economic

benefits is possible and useful but careful characterization of the uncertainty should be

pursued as well as the limitations of this approach should be acknowledged. To assist in

the design of future pharmacoeconomic studies, Chapter 2 proposed a set of

recommendations for the appropriate use of a CVD risk model in pharmacoeconomic

analysis.

The knowledge and recommendations from Chapter 2 were utilized for designing a

pharmacoeconomic model in Chapter 3. This model explored the cost-effectiveness of

primary CVD prevention with antihypertensive treatment compared to no treatment in

patients with mild hypertension, ineligible for treatment according to the Dutch

guidelines. Surrogate endpoints on lowering systolic blood pressure (SBP) were projected

to hard CVD endpoints with the use of the modified SCORE CVD risk model. It was shown

that, the cost-effectiveness of antihypertensive treatment was highly influenced by the

model’s time horizon. In a lifetime horizon, significant health and economic benefits were

observed in all scenarios for both genders and ages when SBP reduction was assumed to

be achieved with the use of hydrochlorothiazide (HCT) 25 mg. However, in a 10-year

horizon, SBP reductions were more favourable when targeted to older patients and more

in men than in women. In scenarios assuming fixed-dose combination HCT 12.5

mg/Losartan 50 mg, long-term health and economic benefits were even more favourable

in both time horizons investigated. In the same chapter, a 10-year treatment with HCT 25

mg vs. no treatment was examined in male and female smokers. This analysis indicated

more favourable cost-effectiveness for applying antihypertensive treatment compared to

no treatment in smokers than the ones estimated for patients with comparable age-,

gender- characteristics but with the average prevalence of smoking as representative the

for Dutch population. This was due to the higher levels of CVD risk estimated for male and

female smokers as well as the greater benefits of risk reduction in this patient population.

The aforementioned findings suggest more aggressive SBP reductions in patients with mild

hypertension and in smokers, however, uncertainty in the results should be noted.

Notably, a relatively large overall uncertainty was observed surrounding the estimated

values in the analysis regardless of the applied time horizon. As expected from our

previous work, the largest contribution to the overall uncertainty came from the

uncertainty surrounding the risk prediction.

The model in Chapter 3 was informed with various published Dutch cost estimates such as

cost of heart failure (HF), angina, stroke, etc. Given the major changes in the clinical

practice and treatment of HF in the last decade, and the fact that hospitalization costs

substantially contribute to the total health care expenditures in HF patients, Annex 1

8

General discussion

193

aimed to provide up-to-date information on the characteristics and use of hospital care

resources associated with HF in the Netherlands. Notably, given that HF hospitalizations

might occur due to the new onset of HF (‘de novo’) or deterioration of chronic HF with

symptoms that warrant hospitalization, these are labelled acute heart failure (AHF)

hospitalizations. This study identified 31,248 HF hospitalizations in the 2011 Dutch

National Medical Registration (LMR) data. Based on the hospital resource use associated

with the hospitalisations in 2011 and under the assumption that the resource use of

inpatient care was associated only with the costs of stay in general hospital wards, the

minimum total hospital costs were estimated to be €3,902 and €4,044 for women and

men, respectively. When the cost of stay in emergency wards was assumed to plausibly

significantly contribute as well to the cost of inpatient care, the aforementioned total cost

estimates almost doubled. Due to the limitations of the LMR data to provide information

on medication use, further analysis should be directed to linking the available LMR data

with hospital claims data which could indicate the cost of medication use and, therefore,

provide more comprehensive AHF cost estimates.

Chapter 4 describes an evidence synthesis (a multivariate meta-analysis) of instrument-

specific preference-based health-related quality of life (HRQoL) values in coronary heart

disease (CHD) and its underlying disease-forms (i.e. stable angina, post-acute coronary

syndrome (post-ACS)) in developed countries. The study explicitly accounted for study-

level characteristics and relevant correlations between those values. Using a systematic

literature search, 40 studies were identified. In post-ACS, HRQoL estimates ranged from

0.64 (Quality-of-Well-Being) to 0.92 (EQ-5D European ”tariff”); in stable angina, HRQoL

ranged from 0.64 (SF-6D) to 0.89 (standard gamble) while in overall CHD the range of

values was comparable to the aforementioned (i.e. from 0.60 (HALex) to 0.89 (standard

gamble)). Chapter 4 indicated inequality in HRQoL values, both within and between the

instrument-specific values, potentially explained by both observed and unobserved

methodological differences across instruments and underlying study-level characteristics.

Current economic models in CHD generally ignore the between-study study heterogeneity

in HRQoL. Therefore, Chapter 4 suggests that multivariate meta-analysis should be applied

for quantifying this uncertainty to improve estimates used in cost-utility analyses.

Part II: Pharmacoeconomic findings in secondary cardiovascular disease prevention;

Focus on oral anticoagulation

Part II of this thesis described a number of pharmacoeconomic findings in secondary CVD

prevention. Specifically, the health and economic findings associated with the use of oral

anticoagulants - VKAs and NOACs such as apixaban and dabigatran - for the prevention of

CVD in the Dutch setting were presented.

In Dutch clinical practice, VKAs are still the most commonly applied anticoagulants in

indications such as atrial fibrillation (AF), arterial diseases or venous thromboembolism

Chapter 8

194

(VTE)(5). Yet, their use comes with a narrow therapeutic range of international normalised

ratio (INR) limits of 2.0 and 3.0, calling for regular monitoring. In the Netherlands, INR-

monitoring is mostly performed at specialised anticoagulation testing centres. An

alternative is patient self-testing and patient self-management (PST and/or PSM) that was

internationally shown to lead to better coagulation care compared to monitoring in

specialized anticoagulation centres (6,7). Chapter 5 explored the budget impact of

increasing market-share scenarios of PST and/or PSM in patients on long-term

anticoagulation with VKAs. The occurrence of thromboembolic and haemorrhagic

complications in the aforementioned patient population was assessed in a Markov model

that incorporated Dutch specific costs and effectiveness data derived from a meta-analysis

on self-monitoring of oral anticoagulation. Chapter 5 indicated that increasing PST and/or

PSM in anticoagulation monitoring from the current 15.4% to 50%, 75% and 100%, would

lead to significant savings in one to five year time horizons. This is driven mainly by

considerably lowered complications-related costs associated with PST and/or PSM

compared to conventional monitoring centres. Due to a lack of RCTs to compare the VKAs

managed with PST and/or PSM with NOACs, the use of NOACs was not accounted for in

this study. Further research should account for the potential increasing market share of

NOACs as well as provide a formal cost-effectiveness analysis comparing the two VKA

monitoring strategies.

Chapter 6 presents the cost-effectiveness analysis of using apixaban compared to VKAs for

the prevention of stroke in non-valvular AF in the Netherlands. Using a previously

developed Markov model (8,9), updated with the most recent Dutch costs and

epidemiological data, the incremental cost-effectiveness ratio (ICER) of using apixaban

compared to VKAs was €10,576 per quality adjusted life year (QALY). These results from

the healthcare provider’s perspective require confirmation from the studies on “real life”

long-term benefits of the use of apixaban, primarily regarding the patients’ adherence,

and extension to aspects relevant within broader societal perspective inclusive production

losses (10).

For the VTE indication, NOACs are not yet included in the Dutch reimbursement system.

To support any further future decision-making on reimbursement, a formal

pharmacoeconomic comparison of NOACs and standard treatment (i.e. VKAs) was needed.

Thus, Chapter 7 explored the cost-effectiveness of dabigatran, another NOAC, for

treatment and secondary prevention of VTE. In contrast to Chapters 3, 5 and 6, in which

the economic analyses were performed from a healthcare provider’s perspective, the

base-case analysis in this study considered the societal perspective. This perspective is

preferred by the Dutch authorities. The base-case findings indicated that patients on

dabigatran would gain an additional 0.585 discounted QALYs over a lifetime time horizon

and savings of €1,996 per patient. Chapter 7 explored various scenarios comparing

dabigatran to VKAs, for treatment as well as for secondary prevention in high-risk

8

General discussion

195

patients. All scenarios consistently showed dabigatran to be a cost-saving alternative to

VKAs. Finally, scenarios comparing the use of dabigatran and VKAs from the healthcare

provider perspective and comparing dabigatran to placebo for the prevention of recurrent

VTE in patients who are at equipoise for anticoagulation treatment, indicated the ICERs for

dabigatran compared to VKAs of €1,005 and €33,305 per QALY gained, respectively.

PROVIDING ROBUST PHARMACOECONOMIC EVIDENCE IN CARDIOVASCULAR

DISEASES (PRIMARY) PREVENTION

Healthcare decision making bodies recognize and/or require pharmacoeconomic evidence

to guide or assist in allocating healthcare resources in many developed countries (11). In

particular, decisions to reimburse pharmacological innovations for prevention of CVD

often need to be firmly grounded on both the evidence of added therapeutic value in

comparison to standard or usual intervention, and pharmacoeconomic evidence.

Moreover, if local health and economic parameters change, pharmacoeconomic evidence

is needed to guide changes in policy or reimbursement. Therefore, due to the health and

economic implications of health policy and reimbursement decisions, the most robust

evidence is needed nowadays. In that respect health technology assessment (HTA) has

rapidly evolved in the recent decades.

Various country-specific and international guidelines for conducting pharmacoeconomic

analyses assist analysts in providing sound and valid evidence. Some of the matters

addressed by guidelines concern: the quality of health and economic inputs, modelling,

sensitivity analyses, discounting future health effects and costs, etc. In assessing

pharmacoeconomic evidence on interventions in CVD prevention, analysts can apply

general methodological standards and criteria addressed in aforementioned guidelines.

Yet, in Chapter 1 of this thesis, it was suggested that the standards or solutions to some

potential challenges in analyses on interventions in (primary) CVD prevention are still to

be faced.

One challenge relates to the common preference of the healthcare decision making

bodies to assess the effectiveness of interventions with respect to their impact on life

expectancy and quality of life rather than intermediate (surrogate) outcomes. Specifically,

this can be a major issue in analyses on interventions in primary CVD prevention. Here,

preventive interventions are indicated in patients with no medical history of hard clinical

CVD events. Commonly, only short-term treatment effectiveness evidence (e.g.

modification of CVD risk factor levels) is measured in clinical trials on pharmacological

interventions for primary CVD prevention. For short-term effectiveness evidence to be

incorporated into a pharmacoeconomic analysis, surrogates need to be projected to long-

term hard clinical endpoints (e.g. fatal or non-fatal CVD events). In various studies, the use

of CVD risk prediction models has already been established to provide the aforementioned

projection adequately and validly (12,13). However, our review on pharmacoeconomic

Chapter 8

196

analyses incorporating CVD risk prediction models (Chapter 2) indicated a lack of

standardisation and quality consideration as well as failure to exactly report the

implications of incorporating these models with their specific methodologies. To provide

robust pharmacoeconomic evidence in analyses incorporating CVD risk prediction models,

it may be beneficial to specify and expand methodological criteria and considerations (12).

For example, a risk model can actually only be applied after the compatibility of the risk

model and the study populations has been thoroughly investigated and assessed.

Furthermore, to enhance the transparency of methods applied, the study should report

the method used for the translation of cumulative risks to short-term probabilities and

should clearly describe the method behind extrapolating this risk to longer time horizons,

if applicable. Finally, the uncertainty surrounding risk predictions should be incorporated

in a transparent way in the estimation of the overall uncertainty of the

pharmacoeconomic model. Thus, by examining current practices and identifying areas for

improvement in selecting and utilizing CVD risk prediction models in pharmacoeconomic

analyses, we specified and proposed methods for enhancing the robustness of such

studies. Moreover, we used these caveats and recommendations when conducting and

interpreting the findings of the pharmacoeconomic analysis in Chapter 3. In this analysis,

we used the modified SCORE CVD risk model to project surrogate endpoints on lowering

SBP to hard CVD endpoints. This model was selected due to the similarity between the risk

model’s and study’s populations characteristics, and its established validity in the Dutch

population. Furthermore, the methods used for translating cumulative 10-year risk to

short-term probabilities and extrapolating those risks to a lifetime horizon were explicitly

reported. Finally, the uncertainty surrounding risk predictions was provided in such a form

that could be incorporated in the overall probabilistic analysis.

Another challenging issue in conducting pharmacoeconomic analyses, in particular cost-

utility analyses, on interventions in CVD prevention and/or treatment relates to the

robustness of evidence on the HRQoL values used. When multiple HRQoL values are

available in literature, analysts often select a value assessed in a setting and using a

patient population that is comparable to their study/country. This seems highly

reasonable, but still may not be fully robust. Solid methods for evidence-synthesis can be

considered here as a potential tool for optimizing strategies confronting this challenge

(14-16). Importantly, not only may evidence-synthesis provide summary data that is

considered as the evidence of highest level in the hierarchy of evidence, but it can also

indicate the level of heterogeneity in the HRQoL values assessed across the studies (17). In

fact, the main finding of our evidence synthesis of HRQoL values in stable angina, post-ACS

and, in general CHD (Chapter 4) was the indication of large heterogeneity within and

between the instrument-specific values and inherent uncertainty if applied within cost-

utility analysis. This finding was despite the fact that this evidence-synthesis was

conducted on the instrument-specific level, accounting for underlying CHD disorder,

8

General discussion

197

correlation between the instrument-specific values and a number of study-level

covariates. Heterogeneity detected between the instrument-specific values may be partly

explained by underlying methodological differences across the HRQoL instruments used.

Still, the unobserved patients’ characteristics (e.g. socio-economic, ethnical, religious

characteristics, etc.) are potentially of even greater concern as crucially contributing to the

overall heterogeneity. Thus, evidence-synthesis may be considered as a relevant tool to,

where appropriate, synthesize HRQoL values in specific CVD disorders as well as to

indicate the level of heterogeneity and possibly its sources across the studies. In this

manner, robustness of HRQoL values used to inform cost-utility analyses on interventions

in CVD prevention and/or treatment can be further enhanced.

In general, we conclude that conducting pharmacoeconomic analyses on interventions in

CVD prevention can be confronted with challenges, with solutions not yet specified in

pharmacoeconomic guidelines. Addressing these challenges could benefit from further

standardization ideally with a consensus on methodological requirements and

recommendations across pharmacoeconomic guidelines (18). In this thesis, we tackled

some of the challenging issues and provided recommendations that could assist in

providing more robust pharmacoeconomic evidence on interventions in CVD prevention.

In fact, the expected benefit of standardizing methodological requirements and

recommendations for conducting and reporting pharmacoeconomic analyses including the

ones proposed in this thesis is to enhance the quality and validity of pharmacoeconomic

analyses as well as reduce possible bias (19).

DIRECT IMPLICATIONS FOR CURRENT PRACTICE: VITAMIN K ANTAGONISTS OR

NOVEL ORAL ANTICOAGULANTS IN THE DUTCH SETTING?

The work in the thesis may have some direct societal relevance and impact. During the last

60 years, VKAs have been established as highly effective anticoagulant treatment, shown

to reduce the risks of thromboembolic events in a number of clinical situations (20). Yet,

their position has recently been challenged when the European Medicines Agency granted

market authorizations to NOACs (e.g. dabigatran, rivaroxaban, apixaban) for prevention of

stroke in AF, and treatment of VTE. Notably, for NOACs to compete for the market share

of VKAs, reimbursement approval is essential. In the Netherlands, health authorities

require pharmacoeconomic evidence in reimbursement submissions. This evidence should

indicate whether there is an economic rationale next to added value of innovative

treatment (i.e. NOACs) compared to standard treatment (i.e. VKAs). Moreover, the health

and economic consequences associated with possible improvements in the quality of

standard care should be assessed.

In the Dutch healthcare system, the quality of anticoagulation care performed by

specialised anticoagulation testing centres is measured with the percentage of patients on

treatment with VKAs which INR measurements are within the limits of 2.0 and 3.5.

Chapter 8

198

Notably, the estimate of approximately 80% for quality of anticoagulation care in the

Dutch setting may be considered high (5). This suggests that further investments into

optimizing the standard anticoagulant care with VKAs may be considered as an alternative

to switching to NOACs. INR monitoring with PST and/or PSM may be considered as an

alternative approach to optimizing the standard anticoagulant care with VKAs. This

monitoring alternative was shown to be associated with lower occurrence of

thromboembolic and haemorrhagic complications in patients on long-term indication for

anticoagulation with VKAs, when compared to monitoring in specialized anticoagulation

centres (21-24). Although PST and/or PSM is more costly than monitoring in Dutch

specialized anticoagulation centres, Chapter 5 showed that increasing its market share

would lead to overall cost-savings due to lower total direct medical costs driven by lower

complication-related costs. However, using point-of-care devices solely for PST resulted in

greater expenditures compared to testing in anticoagulation centres. Additionally, this

study indicated less favourable findings of using PST and/or PSM in patients with atrial

fibrillation.

Furthermore, anticoagulant treatment with NOACs was proven to be superior or at least

non-inferior to VKAs for thromboprophylaxis, but also free from the need for regular INR-

monitoring (25,26)(27,28). Yet, some concerns of clinicians and decision makers regarding

the use of NOACs are still unresolved. These reflect the scarcity of antidotes for treatment

with NOACs (29), real-world adherence to NOACs and finally the economic consequences

in different treatment indications. In this thesis, to tackle some of the concerning issues

regarding the use of NOACs, we examined the health and economic consequences

associated with the use of apixaban for the prevention of stroke in non-valvular AF

(Chapter 6) and the use of dabigatran for treatment and prevention of VTE in the Dutch

setting (Chapter 7). Both apixaban and dabigatran were found to be favourable

alternatives to treatment with VKAs. Both studies importantly contributed to discussions

on reimbursement of NOACs in the Netherlands. Notably, the one on dabigatran was part

of the formal submission to the Dutch authorities by the manufacturer. Importantly, given

that these findings were based on RCTs evidence where drug adherence is usually up to

100%, further analyses are needed to examine the pharmacoeconomic findings that

reflect real-world adherence. Moreover, future research should also allow for comparison

between the VKAs managed with PST and/or PSM and NOACs. This was currently

hampered due to the lack of RCTs comparing VKAs managed with PST and/or PSM and

NOACs.

We can conclude that in the Dutch setting, NOACs may present a valuable alternative to

VKAs for thromboprophylaxis, supported by various other studies (30-32)(33). Still, fine-

tuning of recommendations for the most optimal anticoagulant treatments in the

Netherlands will require continuous confirmation and adaptation based on real-world

data as well as full consideration of all treatment alternatives.

8

General discussion

199

REFERENCES

(1) Cardiovascular diseases. Available at: http://www.who.int/topics/cardiovascular_diseases/en/. Accessed

01/01, 2015.

(2) Kelly BB, Fuster V. Promoting Cardiovascular Health in the Developing World:: A Critical Challenge to Achieve

Global Health. : National Academies Press; 2010.

(3) Manson JE, Tosteson H, Ridker PM, et al. The primary prevention of myocardial infarction. N Engl J Med

1992;326(21):1406-1416.

(4) Reitsma JB, Dalstra JA, Bonsel GJ, et al. Cardiovascular disease in the Netherlands, 1975 to 1995: decline in

mortality, but increasing numbers of patients with chronic conditions. Heart 1999 Jul;82(1):52-56.

(5) Federatie van Nederlandse Trombosediensten. Samenvatting medische jaarverslagen. 2012. . Available at:

http://www.fnt.nl/media/docs/jaarverslagen/FNT_Medisch_jaarverslag_2012_WEB.pdf. Accessed 09/04, 2014.

(6) Garcia-Alamino JM, Ward AM, Alonso-Coello P, et al. Self-monitoring and self-management of oral

anticoagulation. Cochrane Database Syst Rev 2010;4(CD003839).

(7) Heneghan C, Ward A, Perera R, et al. Self-monitoring of oral anticoagulation: systematic review and meta-

analysis of individual patient data. The Lancet 2012;379(9813):322-334.

(8) Dorian P, Kongnakorn T, Phatak H, et al. Cost-effectiveness of Apixaban against Current Standard of Care

(SoC) for Stroke Prevention in Atrial Fibrillation Patients. Accepted for publication. Eur Heart J 2014.

(9) Lip G, Kongnakorn T, Phatak H, et al. Cost-effectiveness of Apixaban against Other Novel Oral Anticoagulants

(NOACs) for Stroke Prevention in Atrial Fibrillation Patients. Accepted for publication. Clin Ther 2014.

(10) Ansell J, Granger CB, Armaganijan LV. New Oral Anticoagulants Should Not Be Used as First-Line Agents to

Prevent Thromboembolism in Patients With Atrial FibrillationResponse to Ansell. Circulation 2012;125(1):165-

170.

(11) Pharmacoeconomic Guidelines Around The World. Available at:

http://www.ispor.org/peguidelines/index.asp. Accessed 02/04, 2015.



(13) Sculpher M, Claxton K. Establishing the Cost-Effectiveness of New Pharmaceuticals under Conditions of

Uncertainty—When Is There Sufficient Evidence? Value in Health 2005;8(4):433-446.

(14) van Tol BAF, Huijsmans RJ, Kroon DW, et al. Effects of exercise training on cardiac performance, exercise

capacity and quality of life in patients with heart failure: a meta-analysis. European journal of heart failure

2006;8(8):841-850.


2003;21(3):191-200.

(16) Rehse B, Pukrop R. Effects of psychosocial interventions on quality of life in adult cancer patients: meta

analysis of 37 published controlled outcome studies. Patient Educ Couns 2003;50(2):179-186.

(17) Borenstein M, Hedges LV, Higgins JPT, Rothstein HR. Introduction to meta-analysis. : Wiley; 2011.

(18) Mullins CD, Ogilvie S. Emerging standardization in pharmacoeconomics. Clin Ther 1998;20(6):1194-1202.

(19) Ofman JJ, Sullivan SD, Neumann PJ, et al. Examining the value and quality of health economic analyses:

implications of utilizing the QHES. Journal of Managed Care Pharmacy 2003;9(1):53-61.

(20) Hirsh J. Oral anticoagulant drugs. N Engl J Med 1991 Jun 27;324(26):1865-1875.

(21) Menéndez-Jándula B, Souto JC, Oliver A, et al. Comparing Self-Management of Oral Anticoagulant Therapy

with Clinic ManagementA Randomized Trial. Ann Intern Med 2005;142(1):1-10.

(22) Fitzmaurice DA, Murray ET, McCahon D, et al. Self management of oral anticoagulation: randomised trial.

BMJ 2005 Nov 5;331(7524):1057.

(23) Matchar DB, Jacobson A, Dolor R, et al. Effect of home testing of international normalized ratio on clinical

events. N Engl J Med 2010;363(17):1608-1620.

(24) Siebenhofer A, Rakovac I, Kleespies C, et al. Self-management of oral anticoagulation reduces major

outcomes in the elderly. A randomized controlled trial.Thromb Haemost 2008;100(6):1089-1098.

Chapter 8

200

(25) De Commissie Farmaceutische Hulp (CFH). CFH rapport Rivaroxaban (Xarelto). 2012;2012059711.

(26) Health Council of the Netherlands. New Anticoagulants: A well-dosed introduction. 2012;2012/07.

(27) Schulman S, Kakkar AK, Goldhaber SZ, et al. Treatment of Acute Venous Thromboembolism with Dabigatran

or Warfarin and Pooled Analysis. Circulation 2013:CIRCULATIONAHA. 113.004450.

(28) Schulman S, Kearon C, Kakkar AK, et al. Dabigatran versus warfarin in the treatment of acute venous

thromboembolism. N Engl J Med 2009 Dec 10;361(24):2342-2352.

(29) Kaatz S, Kouides PA, Garcia DA, et al. Guidance on the emergent reversal of oral thrombin and factor Xa

inhibitors. Am J Hematol 2012;87(S1):S141-S145.

(30) Harrington AR. Cost-Effectiveness of Apixaban, Dabigatran, Rivaroxaban, and Warfarin for the Prevention of

Stroke Prophylaxis in Atrial Fibrillation. 2012.

(31) Kamel H, Easton JD, Johnston SC, et al. Cost-effectiveness of apixaban vs warfarin for secondary stroke

prevention in atrial fibrillation. Neurology 2012;79(14):1428-1434.

(32) Lee S, Mullin R, Blazawski J, et al. Cost-effectiveness of apixaban compared with warfarin for stroke

prevention in atrial fibrillation. PloS one 2012;7(10):e47473.

(33) Braidy N, Bui K, Bajorek B. Evaluating the impact of new anticoagulants in the hospital setting. Pharmacy

practice 2011;9(1):1-10.

Summary

201

Summary Cardiovascular diseases (CVD) account for a major share of overall morbidity and

mortality worldwide and significantly strain healthcare budgets. In reducing some of

this burden, pharmacological interventions for preventing CVD can be considered.

Notably, the decisions to implement pharmacological interventions in clinical practice

have to be made with respect to both health and economic consequences associated

with their use.

In this thesis, some of the methodological challenges in simulating and synthesizing

pharmacoeconomic evidence in CVD prevention were assessed and potential

solutions to those challenges were proposed. This includes a set of recommendations

for enhancing the robustness of pharmacoeconomic analyses that apply CVD risk

prediction models. The application of these recommendations was also explored in a

simplified model of primary CVD prevention with antihypertensives. Another

challenging issue relates to the robustness of evidence on the health-related quality

of life (HRQoL) values used in pharmacoeconomic analyses. Here, evidence-synthesis

was suggested as a relevant approach to, where appropriate, synthesize HRQoL

values in specific CVD disorders as well as to indicate the level of heterogeneity

across those values and possibly its sources. As an example, evidence-synthesis of

instrument-specific HRQoL values in coronary heart disease and its underlying

disease-forms was explored. In this study large heterogeneity within and between

the instrument-specific values and inherent uncertainty if applied within

pharmacoeconomic analysis, were found.

This thesis also derives pharmacoeconomic evidence on the use of oral

anticoagulants (i.e. vitamin K antagonists (VKAs) and novel oral anticoagulants

(NOACs) e.g. apixaban, dabigatran) for secondary CVD prevention that may have

direct societal relevance and implications in the Dutch setting. Firstly, optimizing the

standard anticoagulant care with VKAs was explored. Secondly, the health and

economic consequences associated with the use of apixaban for the prevention of

stroke in non-valvular atrial fibrillation, and dabigatran for treatment and prevention

of venous thromboembolism indicated that both apixaban and dabigatran were

favourable alternatives to treatment with VKAs.

In conclusion, several studies in this thesis emphasize that standardizing

methodological requirements and recommendations for conducting and reporting

pharmacoeconomic studies in CVD prevention including the ones proposed in this

thesis may enhance the quality and validity of pharmacoeconomic evidence and

reduce possible bias. Furthermore, this thesis provides pharmacoeconomic evidence

that NOACs may present a valuable alternative to VKAs for thromboprophylaxis in the

Dutch setting.

Samenvatting

203

Samenvatting

Cardiovasculaire ziekten (CVZ) zijn verantwoordelijk voor een belangrijk deel van de

algehele morbiditeit en mortaliteit wereldwijd en hebben daarmee een significante

invloed op de gezondheidszorgbudgetten. Om deze last enigszins te reduceren, kan

farmacologische interventie voor preventie van CVZ worden overwogen. In het

bijzonder moeten de beslissingen om farmacologische interventies te implementeren

in de klinische praktijk worden gemaakt tegen de achtergrond van zowel de

consequenties voor de gezondheid als de economische consequenties van het

gebruik ervan.

In dit proefschrift zijn een aantal methodologische uitdagingen in het simuleren en

synthetiseren van farmaco-economisch evidence in CVZ-preventie vastgesteld en zijn

potentiële oplossingen aangedragen. Deze bevatten een aantal aanbevelingen voor

het verbeteren van de robuustheid van farmaco-economische analyse waarin CVZ-

risicoschattingsmodellen worden toegepast. Het toepassen van de aanbevelingen is

ook verkend in een vereenvoudigd model voor CVZ-preventie door middel van het

gebruik van antihypertensiva. Een andere uitdaging wordt gevormd door de

robuustheid van de in farmaco-economische analyses gebruikte waarden voor health

related quality of life (HRQoL). Hier wordt evidence-synthese voorgesteld als een

relevante aanpak om, waar toepasbaar, HRQoL-waarden in specifieke CVZ-

stoornissen te synthetiseren alsmede om het niveau van heterogeniteit tussen deze

waarden en mogelijk haar oorsprong aan te duiden. Bij wijze van voorbeeld zijn

evidence-synthese van instrumentspecifieke HRQoL-waarden in coronaire hartziekte

en haar onderliggende ziektevormen verkend. In deze studie werden een hoge

heterogeniteit binnen en tussen de instrumentspecifieke waarden en inherente

onzekerheid wanneer toegepast in farmaco-economische analyse gevonden.

In dit proefschrift wordt ook farmaco-economische evidence voor het gebruik van

orale anticoagulantia afgeleid voor secundaire CVZ-preventie die directe

maatschappelijke relevantie en implicaties kunnen hebben in de Nederlandse

situatie. De beschouwde anticoagulantia zijn vitamine-K antagonisten (VKAs) en

nieuwe orale anticoagulantia (NOACs), zoals bijvoorbeeld apixaban en dabigatran.

In de eerste plaats is de optimalisatie van de standaard trombosedienst met VKAs

onderzocht. In de tweede plaats zijn het de economische en

gezondheidsconsequenties gekoppeld aan het gebruik van apixaban voor de

preventie van beroerten in niet-valvulaire atriumfibrillatie en dabigatran voor de

behandeling en preventie van veneuze trombo-embolie die laten zien dat zowel

apixaban als dabigatran te prefereren alternatieven zijn voor behandeling met VKAs.

Concluderend kan worden gesteld dat verscheidene studies in dit proefschrift

benadrukken dat het standaardiseren van methodologische voorwaarden en

Samenvatting

204

aanbevelingen voor het uitvoeren en rapporteren van farmaco-economische studies

in CVZ-preventie, waaronder de in dit proefschrift voorgestelde, de kwaliteit en

geldigheid van farmaco-economisch bewijs verbeteren alsmede de mogelijke bias

verkleinen. Voorts verschaft dit proefschrift farmaco-economische evidence dat

NOACs een waardevol alternatief kunnen zijn voor VKAs voor tromboprofylaxe in de

Nederlandse situatie.

List of publications

205


Peer reviewed international publications supporting this thesis

Stevanovic J, Postma MJ, Pechlivanoglou P. A systematic review on the application of

cardiovascular risk prediction models in pharmacoeconomics, with a focus on primary

prevention. European Journal of Preventive Cardiology. 2012 Aug;19(2 Suppl):42-53.

Stevanović J, O’Prinsen AC, Verheggen BG, Schuiling-Veninga N, Postma MJ,

Pechlivanoglou P. Economic Evaluation of Primary Prevention of Cardiovascular

Diseases in Mild Hypertension: A Scenario Analysis for the Netherlands. Clinical

Therapeutics 2014;36(3);368-384.

Stevanović J, Pompen M, Le HH, Rozenbaum MH, Tieleman RG, Postma MJ. Economic

evaluation of apixaban for the prevention of stroke in non-valvular atrial fibrillation in

the Netherlands. Plos one, 2014.

van Leent MWJ, Stevanović J, Jansman FG, Beinema MJ, Brouwers JRBJ, Postma MJ.

Cost-effectiveness of dabigatran compared to vitamin-K antagonists for the

treatment of deep venous thrombosis in the Netherlands using real-world data. Plos

one, 2015.

Stevanović J, Pechlivanoglou P, Kampinga MA, Krabbe PFM, Postma MJ. Multivariate

meta-analysis of preference-based quality of life values in coronary heart disease.

Submitted.

Stevanović, Postma MJ, Le HH. Budget Impact Analysis of Current- and Increasing

Market-share Scenarios for Point-of-Care Devices in Anticoagulation. Submitted.

Stevanović J, Dvortsin E, Kappelhoff B, Voorhaar M, Postma MJ. Dabigatran for the

treatment and secondary prevention of venous thromboembolism. A cost-

effectiveness analysis for the Netherlands. Submitted.

Conference presentations

Stevanović J, Pechlivanoglou P, Kampinga MA, Krabbe PFM, Postma MJ. Multivariate

meta-analysis of preference-based quality of life values in coronary heart disease.

Podium presentation at 36th Annual North American Meeting MDM.

Stevanovic J, Denee L, Koenders JM, Postma MJ. Incidence Description and Costs of

Acute Heart Failure in the Netherlands. Podium presentation at 17th European ISPOR


206

congress, 2014, Amsterdam, the Netherlands. Abstract published: Value in Health

2014; 17(7): A328.

Stevanović J, Pechlivanoglou P, Postma MJ. Meta-analysis of preference-based quality

of life values in heart failure. Poster presentation at 19th International ISPOR

congress, 2014, Montreal, Canada. Abstract published: Value in Health 2014; 17(3):

A197–A198.

Stevanović J, Pechlivanoglou P, Kampinga MA, Krabbe PFM, Postma MJ. Meta-

Analysis of Quality of Life in Coronary Heart Disease. Poster presentation at 16th

European ISPOR congress, 2013, Dublin, Ireland. Abstract published: Value in Health

2013; 16(7): A600.

Stevanović J, Pompen M, Le HH, Rozenbaum MH, Tieleman RG, Postma MJ. Economic

evaluation of apixaban for the prevention of stroke in non-valvular atrial fibrillation in

the Netherlands. Poster presentation at 16th European ISPOR congress, 2013, Dublin,

Ireland. Abstract published: Value in Health 2013; 16(7): A529.

Stevanovic J, Postma MJ, Pechlivanoglou P. Systematic Review on the Application of

Cardiovascular Risk Prediction Models in Pharmacoeconomics, With a Focus on

Primary Prevention. Poster presentation at 15th European ISPOR congress, 2013,

Berlin, Germany. Abstract published: Value in Health 2012; 15(7): A467–A468.

Stevanović J, O’Prinsen AC, Postma MJ, Pechlivanoglou P. Economic Evaluation of

Primary Prevention of Cardiovascular Diseases in Mild Hypertension: A Scenario

Analysis for the Netherlands. Poster presentation at 14th European ISPOR congress,

2012, Madrid, Spain. Abstract published: Value in Health 2011; 14(7): A376.

Acknowledgements

207

Acknowledgements

Work presented here would not be possible without collaboration, sharing of

knowledge and support from my supervisors, colleagues, friends and family.

First and foremost, I would like to thank my promotor and copromotors. Maarten,

thank you for giving me the opportunity to do research in your group, for sharing

your pharmacoeconomics expertise and for supporting and stimulating my

independence as a researcher. I have learned a great deal from you not only about

pharmacoeconomics but also people-managing and putting things into a broader

perspective - I could not have imagined a better promotor!

Petros, thank you for sharing with your passion for statistics with me, teaching me

always to question and encouraging me to use R. Thank you for our long discussions

and putting many hours in our projects!

Hoa, your knowledge, sharp eye for detail, your enthusiasm and support were of

great value for this work!

Long work hours may have not been always stimulating if it was not for my

colleagues and friends. Job and Stefan, thanks for keeping the spirits high and

supporting me learning and speaking Dutch. Hao, Hong-Anh, Koen, Maarten,

Elisabetta and Giedre, thank you for bringing joy in many conversations. Dinners with

you were never dull.

Thanks to all my colleagues and friends from the unit of PE2 for making a great and

inspiring working environment - Aulya, Bob, Bert, Christiaan, Dianna, Didik, Eelko,

Jens, Josta, Jannie, Lan, Marcy, Mark, Nynke, Pepijn, Pricillia, Peter, Thea, Ury and

Yugo. I enjoyed your company!

Milica, my dearest and closest friend, not only do you possess all the qualities one

can wish for in a friend, but you were also always the most inspiring, outgoing and

enjoyable person I know. I am incredibly grateful for our friendship!

Bianka, my dear paranimf, your vivid spirit and positive attitude made the working

hours (and after…) always fun!

The group of Balkanians in Groningen made me feel at home. Dear Vibor, Kaca &

Sara, Jelena & Slobodan, Ivana & Ivan, Jelena & Dule, Deki & Sonja Maja, Ena, Faris,

Igor, Jelena J., Jelena G., Igor, Ivan V., Visnja, Milica – thank you for all the parties,

dinners and going out. The time spent together meant a lot to me! My friend and

Acknowledgements

208

colleague, Jovan, dedicated to the same goal – I always enjoyed your visits to the

department in particular for bringing fresh jokes from Serbia.

Clement, thank you for being a housemate to Mici and me, I can’t imagine the

combination to be more colorful. Your witty comments turned every situation

hilarious! Additionally, I must thank you for introducing us to a great group of people!

Katia, Manuel, Helgi, Marta, Marcelo, Susan, Rebeca, Sara, Jaakko, Thomas – all

amazing people in their own way made every gathering fun.

Even when you are thousands of kilometres apart true friendship will not be harmed

by distance and whenever you meet again you can pick up where you left. Andrijana,

Tijana, Gaga, Jovana, Milena and Marija, thank you for our enduring friendship!

My dear family, tata, mama, and Nikola, thank you for your love, support,

encouragement and always believing in me! This would not be possible without you.

My dearest Henri, I am grateful that you are part of my life. All professional

challenges would have been more of a burden to deal with if it hadn’t been for your

love, support, luminosity and knowledge. Thank you for making my life complete!

SHARE

209

Research Institute for health research (SHARE)

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 internet

site: http://www.share.umcg.nl/.

More recent theses can be found in the list below. ((co-) supervisors are between brackets)

Potijk MR

Moderate prematurity, socioeconomic status, and neurodevelopment in early childhood;

a life course perspective (prof SA Reijneveld, prof AF Bos, dr AF de Winter)

Beyer AR

The human dimension in the assessment of medicines; perception, preferences, and

decision making in the European regulatory environment (prof JL Hillege, prof PA de

Graeff, prof B Fasolo)

Alferink M

Psychological factors related to Buruli ulcer and tuberculosis in Sub-saharan Africa (prof

AV Ranchor, prof TS van der Werf, dr Y Stienstra)

Bennik EC

Every dark cloud has a colored lining; the relation between positive and negative affect

and reactivity to positive and negative events

(prof AJ Oldehinkel, prof J Ormel, dr E Nederhof, dr JACJ Bastiaansen)

Stallinga HA

Human functioning in health care; application of the International Classification of

Functioning, Disability and Health (ICF) (prof RF Roodbol, prof PU Dijkstra, dr G Jansen)

Papageorgiou A

How is depression valued? (prof AV Ranchor, prof E Buskens, dr KM Vermeulen, dr MJ

Schroevers)

Spijkers W

Parenting and child psychosocial problems; effectiveness of parenting support in

preventive child healthcare (prof SA Reijneveld, dr DEMC Jansen)

Ark M van

Patellar tendinopathy; physical therapy and injection treatments (prof RL Diercks, prof JL

Cook)

Boven JFM van

Enhancing adherence in patients with COPD: targets, interventions and cost-

effectiveness

(prof MJ Postma, prof T van der Molen)

Gerbers JG

Computer assisted surgery in orthopedaedic oncology; indications, applications and

surgical workflow (prof SK Bulstra, dr PC Jutte, dr M Stevens)

SHARE

210

Niet A van der

Physical activity and cognition in children (prof C Visscher, dr E Hartman, dr J Smith)

Meijer MF

Innovations in revision total knee arthroplasty (prof SK Bulstra, dr M Stevens, dr IHF

Reininga)

Pitman JP

The influence and impact of the U.S. President’s Emergencey Plan for AIDS Relief

(PEPFAR) on blood transfusion in Africa; case studies from Namibië (Prof MJ Postma, prof

TS van der Werf)

Scheppingen C van

Distress and unmet needs in cancer patients; challenges in intervention research

(prof R Sanderman, dr G Pool, dr MJ Schroevers)

Bouwman II

Lower urinary tract symptoms in older men: does it predict the future?

(prof K van der Meer, prof JM Nijman, dr WK van der Heide)

Geurts, MME

Integrated pharmaceutical care; cooperation between pharmacist, general practitioner,

and patient and the development of a pharmaceutical care plan

(prof JJ de Gier, prof JRBJ Brouwers, prof PA de Graeff)

Sun J

Developing comprehensive and integrated health system reform policies to improve use

of medicines in China (prof HV Hogerziel, prof SA Reijneveld)

Wyk, L van

Management of term growth restriction; neonatal and long term outcomes

(prof SA Scherjon, prof JMM van Lith, dr KE Boers, dr S le Cessie)

Vos FI

Ultrasonography of the fetal nose, maxilla, mandible and forehead as markers for

aneuploidy

(prof CM Bilardo, prof KO Kagan, dr EAP de Jong-Pleij)

Twillert S van

Linking scientific and clinical knowledge practices; innovation for prosthetic

rehabilitation

(prof K Postema, prof JHB Geertzen, dr A Lettinga)

Loo HM van

Data-driven subtypes of major depressive disorder

(prof RA Schoevers, prof P de Jonge, prof JW Romeijn)

Raat AN

Peer influence in clinical worokplace learning; a study of medical students’use of social

comparison in clinical practice (prof J Cohen-Schotanus, prof JBM Kuks)

Standaert BACGM

Exploring new ways of measuring the economic value of vaccination with an application

to the prevention of rotavirus disease (prof MJ Postma, dr O Ethgen)

SHARE

211

Kotsopoulos N

Novel economic perspectives on prevention and treatment: case studies for paediatric,

adolescent and adult infectious diseases (prof MJ Postma, dr M Connolly)

Feijen-de Jong, EI

On the use and determinants of prenatal healthcare services

(prof SA Reijneveld, prof F Schellevis, dr DEMC Jansen, dr F Baarveld)

Hielkema M

The value of a family-centered approach in preventive child healthcare

(prof SA Reijneveld, dr A de Winter)

Dul EC

Chromosomal abnormalities in infertile men and preimplantation embryos

(prof JA Land, prof CMA van Ravenswaaij-Arts)

Heeg BMS

Modelling chronic diseases for reimbursement purposes

(prof MJ Postma, prof E Buskens, prof BA van Hout)

Vries, ST de

Patient perspectives in the benefit-risk evaluation of drugs

(prof P Denig, prof FM Haaijer-Ruskamp, prof D de Zeeuw)

Zhu L

Patterns of adaptation to cancer during psychological care

(prof AV Ranchor, prof R Sanderman, dr MJ Schroevers)

Peters LL

Towards tailored elderly care with self-assessment measures of frailty and case

complexity

(prof E Buskens, prof JPJ Slaets, dr H Boter)

For 2014 and earlier theses visit our website

Curriculum Vitae

213

Curriculum Vitae

Jelena Stevanović was born in Niš, Serbia on the 13th July, 1983. She completed

studies of Pharmacy at Medical faculty, University of Niš in 2008. After gaining her

degree, she worked for a medical wholesale company as a sales manager and later as

a pharmacist.

In 2011, she moved to the Netherlands to start her PhD program at University of

Groningen at department of Pharmacoepidemiology and Pharmacoeconomics; her

promotor was Prof. dr. M.J. Postma and copromotors were Dr. P. Pechlivanoglou and

Dr. H.H. Le. Her research topics ranged from exploring the possible solutions to the

methodological challenges in simulating and synthesizing pharmacoeconomic

evidence in cardiovascular disease prevention to deriving pharmacoeconomic

evidence on the use of oral anticoagulants of direct societal relevance in the Dutch

setting. The projected date of obtaining her PhD degree is December 22nd, 2015.

During her PhD research, Jelena authored and co-authored several scientific

publications and gave podium and poster presentations at internationally renowned

congresses, including the International Society for Pharmacoeconomics and

Outcomes Research (ISPOR) and Medical Decision Making congress. From March to

September 2015 Jelena worked as a research associate at KU Leuven and ERASMUS

MC on a project exploring budget impact of introducing infliximab biosimilar in the

EU5.

Since mid-September 2015, Jelena is employed by Bristol-Myers Squibb as an

Associate Pricing & Reimbursement manager.

university of groningen pharmacoeconomics of cardiovascular … · prediction models in...

Documents