892 Journal of Pain and Symptom Management Vol. 40 No. 6 December 2010

Brief Report

Evaluation of Central Serotonin Sensitivity inBreast Cancer Survivors with Cancer-RelatedFatigue SyndromeSusanna Alexander, PhD, MRCPI, Patrick Stone, MA, MD, MRCP,Sarah White, BSc, Paul Andrews, PhD, Stephen Nussey, PhD, and Gul Bano, PhDDivision of Mental Health (S.A., P.S., S.W.), Division of Basic Medical Sciences (P.A.), and Division

of Cellular and Molecular Medicine (S.N., G.B.), St. George’s University of London, London, United

Kingdom

Abstract

Context. Increased central serotonin sensitivity is hypothesized to contribute

toward the development of cancer-related fatigue syndrome (CRFS).Objectives. To compare the responses of breast cancer survivors with or without

CRFS to the buspirone challenge test (an index of central serotonin sensitivity).Methods. Disease-free women who had successfully completed treatment for

early-stage breast cancer were assessed. On the basis of the diagnostic interview forCRFS and a structured psychiatric interview, women were classified as either‘‘cases’’ of CRFS or ‘‘controls.’’ Women with comorbid psychiatric diagnoses wereexcluded. Volunteers underwent a challenge test using buspirone (a serotonin-selective agonist) using a double-blind, randomized, placebo-controlled protocol.Cortisol and prolactin responses were assessed at hourly intervals for the fourhours after administration of buspirone.

Results. Fourteen cases of CRFS and 28 controls participated in the study. Therewere no significant differences in baseline or stimulated cortisol release afterbuspirone challenge. There were differences neither in basal prolactin levels in thetwo groups nor in the total prolactin response to buspirone (as measured using thearea under the curve). In patients with CRFS, peak prolactin response occurred at120 minutes and sustained until 180 minutes post buspirone. In controls, peakprolactin response occurred at 60 minutes and then began to decline.

Conclusions. This studyhasdemonstrated theutility andacceptability ofbuspironeas a probe of central serotonin sensitivity in this population. No evidence was foundfor alterations in central serotonin sensitivity in patients with CRFS. Conclusions aretentative, however, because poor recruitment resulted in a small sample and anunderpoweredcomparison. J PainSymptomManage2010;40:892e898.�2010U.S.Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved.

Key Words

Breast neoplasms, buspirone, fatigue, hydrocortisone, prolactin, serotonin

Address correspondence to: Patrick Stone, MA, MD,MRCP, Division of Mental Health, St. George’s Uni-versity of London, 6th Floor Hunter Wing, Cranmer

Terrace, London SW17 ORE, United Kingdom.E-mail: p.stone@sgul.ac.uk

Accepted for publication: March 25, 2010.

� 2010 U.S. Cancer Pain Relief CommitteePublished by Elsevier Inc. All rights reserved.

0885-3924/$ - see front matterdoi:10.1016/j.jpainsymman.2010.03.023

Vol. 40 No. 6 December 2010 893Serotonin Responses in Cancer-Related Fatigue

Introduction Methods

Fatigue is a common and distressing symp-

tom for patients with cancer at all stages oftheir illness, from diagnosis, through cancertreatment, and into long-term survivorship.1

A recent systematic review concluded thatthere was good evidence for the persistenceof fatigue up to five years after completion ofbreast cancer therapy.2 Prevalence figures cansometimes be difficult to interpret because ofa lack of agreement as to the level of fatiguethat should be considered to be clinically sig-nificant. To address this problem, Cellaet al.3 proposed criteria for the definition ofa cancer-related fatigue syndrome (CRFS).These criteria have subsequently been evalu-ated in a number of studies1 and have gener-ally been found to be a meaningful andreliable way to identify subjects with severe fa-tigue. We have recently reported that the prev-alence of CRFS in a population of disease-freebreast cancer survivors (n¼ 200) was 30%, be-tween three months and two years after com-pletion of cancer treatment.4

Although CRF is increasingly recognized asa significant problem, the underlying patho-physiology is largely unknown.1,5 Recently, at-tention has focused on the potential role ofproinflammatory cytokines and their effectson central serotonin (5-hydroxytryptamine [5-HT]) and the hypothalamic-pituitary-adrenal(HPA) axis.6,7 Serotonin is thought to havewidespread effects on human behavior, includ-ing regulation of sleep, appetite, and mood.7

In support of the serotonin hypothesis offatigue, some studies in patients with chronic fa-tigue syndrome (CFS) have demonstratedraised plasma levels of free tryptophan8,9 or in-creased central serotonin sensitivity.10e12 Sero-tonin sensitivity has never previously beenassessed in subjects with CRFS.

Buspirone is a full agonist of the 5-HT1A au-toreceptors and a partial postsynaptic 5-HT1Areceptor agonist but also exerts a dopamineD2 receptor antagonism.13 It has previouslybeenused as ameans to assess the serotonin sen-sitivity of the HPA in patients with CFS.10,11 Wehave used buspirone as a probe to investigatethe sensitivity of the HPA axis in breast cancersurvivors with CRFS compared with nonfa-tigued survivors, using a double-blind, random-ized, placebo-controlled protocol.

The study was approved by the WandsworthResearch Ethics Committee and theSt. George’s National Health Service (NHS)Trust Research and Development committee.

Subjects for the buspirone challenge testwere a subgroup of patients recruited froma larger study of fatigue in breast cancer survi-vors, the results of which have been previouslypublished.4 Subjects for the larger study wererecruited from a nurse-led breast cancerfollow-up clinic at St. George’s Hospital. Aftercompletion of the first phase of the study, eligi-ble subjects were invited to participate in thisextension study using the buspirone challengetest. Separate written informed consent wasobtained for this phase of the study.

Subjects were cancer free at the time of studyentry and had completed cancer therapybetween three months and two years previ-ously. ‘‘Cases’’ of fatigue were identified usingthe diagnostic interview for CRF.3 The pres-ence of comorbid psychiatric disorders wasidentified using the Structured Clinical Inter-view for the Diagnostic and Statistical Manual ofMental Disorders, 4th edition.14 Of the 200women interviewed as part of the first phaseof the study, 60 were found to be cases ofCRFS, 104 did not fulfill the diagnostic criteria,and 36 women were found to have comorbidpsychiatric diagnoses. The patients with comor-bid psychiatric conditions were excluded fromfurther analysis. It was necessary to excludethese subjects because the presence of psychiat-ric disorders (or psychotropic medication)would have been a major confounder to theinterpretation of the challenge test. The otherwomen (provided they fulfilled certain addi-tional entry criteria, see below) were invitedto participate in the buspirone challenge test.

Buspirone Challenge Test ProtocolExclusion criteria for the buspirone test

were as follows: pregnancy or breast-feeding;patients with epilepsy; subjects taking psycho-tropic medication; patients who had difficultveins to cannulate; patients with significantlymphedema or pain in the arm; moderateor severely impaired renal or liver function(two times the upper limit); and patients un-able to fast (e.g., insulin-requiring diabetics).

894 Vol. 40 No. 6 December 2010Alexander et al.

Participants had an overnight fast (no foodor fluid other than water from midnight)and attended the Endocrine InvestigationUnit at St. George’s Hospital NHS Trust.Menstruating women were investigated in theluteal phase of their cycle. All tests were doneat 9 AM. An intravenous cannula was insertedand remained in place throughout the test pe-riod. After two baseline blood samples weretaken 15 minutes apart for the measurementof prolactin and cortisol, women were giveneither oral buspirone at a dose of 0.5 mg/kgup to a total of 45 mg or a placebo (thiamine)in a double-blind randomized protocol. Thisdose of buspirone has previously been success-fully used to test the HPA.15

After either buspirone or placebo adminis-tration, prolactin and cortisol responses weremeasured at hourly intervals for four hours.Women then attended for the second part ofthe test at least one week later (buspirone orplacebo).

Prolactin was analyzed using a Roche Elecsys1010/2010 analyzer (Roche Diagnostics Lim-ited, West Sussex, UK) with interassay coeffi-cients of variation of 4% and intra-assaycoefficients of variation of 5.0%, respectively.

Statistical MethodsThis was the first study using the buspirone

challenge test in patients with CRFS. Basedon a previous study in patients with CFS,10

which found a significant difference in peakprolactin between CFS cases and controlswith an effect size equal to 0.9, it was estimatedthat 20 subjects per group would be requiredto detect a similar magnitude difference(with 80% power at a 5% significance level)in subjects with CRFS.

Statistical analysis was done using the Statis-tical Package for the Social Sciences (SPSS v15;SPSS, Inc., Chicago, IL). Repeated-measureanalysis of variance was used to assess changein the cortisol and prolactin levels over timeafter administration of placebo/buspirone in-cluding a between-groups factor of case or con-trol. The area under the curve (AUC) wascalculated using the trapezoid method16 forboth CRFS cases and controls. Results were ex-pressed as mean� standard deviation (SD) un-less otherwise indicated. An unpaired samplet-test was used to test for significance (set at

P< 0.05) of the difference between the AUCsin the CRFS cases and controls.

ResultsDescription of the PopulationA total of 44 of 164 (27%) of the potentially

eligible women from the larger study agreed toparticipate in the buspirone challenge test.There were no significant differences in ageor treatment history between the women whoagreed to participate in the buspirone chal-lenge test and those eligible women who de-clined to participate. A greater proportion ofthe participants (39 of 42, 93%) than the non-participants (90 of 120, 75%) were white, andparticipants had a higher platelet count(273� 109/L vs. 248� 109/L, P¼ 0.03). Therewere no other significant differences in hema-tological or biochemical parameters.Of the 44 volunteers, 16 women were

‘‘cases’’ of CRFS and 28 were ‘‘controls.’’ Twoof the cases subsequently withdrew from thestudy after initially consenting to participate(one woman could not be cannulated andthe other subject withdrew for family reasons).Basic demographic and disease-related vari-ables for the remaining 42 subjects are shownin Table 1. There were no significant differ-ences between the cases and the controlswith respect to age, lymph node positivity, ortreatment history. The CRFS cases had signifi-cantly lower serum sodium than the controls(138.6 vs. 140.1 nmol/L, P¼ 0.04), but therewere no other differences in hematologicalor biochemical parameters between the twogroups.

Basal CortisolBasal cortisol levels were within the normal

range, and there was no significant differencebetween the CRFS cases and the controls. Themean basal cortisol level for CRFS cases atT¼ 0 minutes was 375.2 (199.1) nmol/L andfor the controls was 378.5 (115.5) nmol/L(t¼ 0.07, P¼ 0.946).

Stimulated CortisolThe placebo did not cause a differential in-

crease in cortisol concentration between theCRFS or the control subjects (F¼ 0.12,P¼ 0.977). There was a significant change in

Table 1Comparison Between CRFS Cases and Non-Fatigued Controls

CRFS (n¼ 14) Controls (n¼ 28) t-Test P-value

Age in years (SD) 56.5 (8.0) 58.9 (10.8) 0.43Lymph node positive 6/14 13/28 0.83Previous chemotherapy 9/14 15/28 0.52Previous radiotherapy 10/14 20/28 1.0Hormone therapy 10/14 27/28 0.07Hemoglobin, g/dL (SD) 13.3 (1.1) 13.2 (1.0) 0.82White blood count, �109/L (SD) 6.1 (2.0) 5.8 (1.5) 0.69Platelets, �109/L (SD) 280.1 (69.8) 269.4 (60.1) 0.63Sodium, mmol/L (SD) 138.6 (2.0) 140.1 (2.4) 0.04Potassium, mmol/L (SD) 4.3 (0.2) 4.5 (0.5) 0.17Urea, mmol/L (SD) 4.2 (0.9) 4.7 (1.2) 0.10Creatinine, mmol/L (SD) 67.4 (14.4) 73.7 (11.1) 0.17Glucose, mmol/L (SD) 5.8 (1.5) 5.3 (0.9) 0.31Calcium, mmol/L (SD) 2.4 (0.1) 2.4 (0.1) 0.37Phosphate, mmol/L (SD) 1.2 (0.2) 1.2 (0.2) 0.78Albumin, mg/dL (SD) 38.9 (2.9) 38.9 (7.1) 1.0Alkaline phosphatase, IU/L (SD) 65.4 (25.2) 61.6 (19.6) 0.63Alanine transaminase, IU/L (SD) 22.6 (6.2) 21.6 (7.0) 0.67Gamma glutamyl transferase, IU/L (SD) 30.2 (15.6) 26.2 (10.9) 0.41Bilirubin, mmol/L (SD) 10.9 (2.1) 11.6 (3.4) 0.41C-reactive protein, mg/dL (SD) 4.7 (4.8) 3.5 (4.9) 0.47Thyroid stimulating hormone, mU/L (SD) 7.1 (15.5) 1.7 (1.0) 0.22Free T4, pmol/L (SD) 13.9 (2.1) 15.3 (2.2) 0.09Serum osmolarity, mOsmol/L (SD) 291.8 (5.1) 294.3 (5.8) 0.18Urine osmolarity, mOsmol/L (SD) 398.9 (232.3) 330.8 (200.6) 0.36

Vol. 40 No. 6 December 2010 895Serotonin Responses in Cancer-Related Fatigue

cortisol levels over time after buspirone admin-istration (F¼ 7.13, P< 0.001) in both controlsand cases of CRFS, but there were no significantdifferences between the groups (F¼ 0.78,P¼ 0.536). The AUCs for cortisol after buspir-one stimulation were not significantly differentbetween the groups (t¼ 0.18, P¼ 0.86).

Basal ProlactinBasal prolactin levels were within the normal

range, and there was no significant differencebetween the CRFS cases and the controls. Themean basal prolactin level for CRFS cases atT¼ 0 minutes was 153.8 (57.6) mU/L and forthe controls was 170.3 (81.6) mU/L (t¼ 0.67,P¼ 0.508).

Stimulated ProlactinPlacebo administration did not cause a dif-

ferential change in prolactin level in eitherthe CRFS subjects or the control subjects(F¼ 0.23, P¼ 0.920). There was a slight reduc-tion in prolactin after placebo administration,which was statistically significant (F¼ 15.5,P< 0.001). Buspirone administration resultedin a significant (F¼ 8.4, P< 0.001) increasein prolactin in both groups, but this increasedid not differ between the groups (F¼ 1.9,P¼ 0.112). There was, however, no significant

difference in the AUC between the groups(cases AUC¼ 2029 mU/L min, controlsAUC¼ 1392 mU/L/min; t¼ 1.07, P¼ 0.29).Nonetheless, there did appear to be a differ-ence in the temporal profile of the prolactinresponse to buspirone between the twogroups, with the CRFS cases showing a delayedrise with a peak at 120 minutes (sustained until180 minutes) and the controls showing a peakat 60 minutes (Fig. 1). However, on statisticaltesting, the only significant difference betweenthe curves was the continued fall in prolactinlevels between 180 and 240 minutes in theCRFS cases compared with the slight rise inprolactin levels over the same time period inthe controls (P¼ 0.04). There was no signifi-cant difference (P¼ 0.19) in peak prolactin re-sponse between the cases (1037 mU/L) andthe controls (642 mU/L).

DiscussionMorrow et al.5 hypothesized that persistently

elevated cytokines in patients with cancer (orin response to cancer treatment) could resultin alterations in central serotonin levels and/or an upregulation of a population of seroto-nin receptors leading to changes in HPA func-tioning and fatigue. Ryan et al.17 reviewed the

Fig. 1. Graph shows mean prolactin concentrations(�standard error of the mean) in the control group(n¼ 28) and in the CRFS group (n¼ 14) after bus-pirone administered at T¼ 0.

896 Vol. 40 No. 6 December 2010Alexander et al.

evidence from studies of volunteers withexercise-induced fatigue and patients withchronic fatigue syndrome and concludedthat there is ‘‘increasing evidence for a rolefor 5-HT in the genesis of central fatigue.’’The evidence, however, is conflicting,7 withsome studies suggesting that exercise-inducedfatigue is associated with central serotoninoveractivity,18e21 and other studies suggestingthat fatigue is unrelated to serotonin activity.22

Jager et al.7 have proposed that the relation-ship between fatigue and 5-HT levels may beU-shaped, with both decreased levels andraised levels being associated with increasedfatigue.

Most of the evidence from human (as op-posed to animal or in vitro) studies for a spe-cific relationship between cancer fatigue andaltered central serotonin sensitivity has beenderived (by extrapolation) from physiologicalchallenge tests in patients with CFS. Bakheitet al.10 reported increased prolactin respon-siveness to buspirone in patients with CFS(n¼ 15) compared with patients with depres-sion (n¼ 13) or healthy controls (n¼ 13).Sharpe et al.11 reported increased prolactin re-sponses to buspirone in 10 male patients withCFS and 11 controls. In a separate but similarstudy, they12 also reported enhanced prolactinsecretion in response to D-fenfluramine (aserotonin-releasing agent) in patients withCFS (n¼ 10) in comparison to a control groupwithout fatigue (n¼ 10).

Despite this circumstantial evidence and ex-pert opinion, no previous study has specificallytested the hypothesis that subjects with CRFShave altered central serotonin sensitivity com-pared with nonfatigued controls. We foundno evidence to support this hypothesis ina carefully controlled study of breast cancersurvivors with and without fatigue. Our studyhad a number of strengths. All the subjectshad undergone a robust assessment process,and women with potentially confounding psy-chiatric disorders had been excluded. Bothcases and controls had the same cancer diag-nosis and treatment history and were balancedwith respect to most study variables.The principal limitation is that this study was

underpowered. On the basis of previous stud-ies in CFS, we had estimated that a sample of20 cases and 20 controls would be adequateto detect differences between the groups witha similar magnitude. However, after three yearsof recruitment, we only managed to completethe buspirone challenge test in 14 cases and28 controls. One reason for this disappointingaccrual is probably that the study was perceivedas being rather too burdensome for volunteersto complete. Women had to be prepared togive up two mornings of their time and fastfrom midnight before the assessment. Anotherbarrier to accrual was the public reaction to anunrelated U.K. Phase I study of a monoclonalantibody, in which six previously fit youngmen suffered serious (life-threatening) ad-verse events.23 Unfortunately, the widespreadbad publicity that this trial attracted in the Brit-ish press undoubtedly had some effect on thewillingness of women to volunteer for ourown study. However, our data suggest thateven if we had recruited the originally esti-mated sample size, this would still have beeninadequate to demonstrate that any observeddifferences were statistically significant. Ex-trapolating from our data, we have calculatedthat we would have needed 98 subjects pergroup to detect (with 80% power, 5% signifi-cance) a difference in AUC between the twogroups of 637 mU/L min (the magnitude ofthe difference that we observed in the presentstudy). Or, if we had powered the study usingthe difference between peak and baseline pro-lactin response as the primary endpoint, wewould have needed 50 subjects per group todetect (80% power, 5% significance) the

Vol. 40 No. 6 December 2010 897Serotonin Responses in Cancer-Related Fatigue

observed difference in our present study(395 mU/L). In light of the accrual rates thatwe actually achieved, this would indicate thata multicenter study would be required toachieve adequate numbers in any future re-search using the buspirone challenge test incancer survivors.

A further limitation of our study was the useof buspirone as a selective probe for 5-HTfunction. In healthy subjects, buspirone doesindeed stimulate prolactin via 5-HT1A receptoractivation; however, buspirone also has dopa-mine receptor antagonist properties. As therelative contributions of 5HT1A receptor ago-nism and dopamine receptor antagonism tothe prolactin response were not studied, itwas not possible to determine which receptorsystem was primarily responsible for the ob-served patterns in prolactin response. It alsomay be argued that the dose of buspirone ad-ministered in our study was inadequate to fullyactivate the brain serotonergic systems. Someprevious studies in other patient groups haveused a uniform oral dose of 60 mg,10,24 whereasother studies have used a dose of 0.5 mg/kg ei-ther with no maximum dose15 or with an upperlimit of 45 mg11 as in our own study. One of thereasons that we chose the more conservativedosing regimen was the observation that inhigher doses buspirone can cause nausea,10,11

and this in itself may be responsible for prolac-tin release as a nonspecific stress response tosubjective side effects. Unfortunately, there isno ideal serotonin-specific probe available forhuman studies, and despite its limitations, bus-pirone was shown to be a safe, accessible, andwell-tolerated agent for this purpose.

We chose to include women who had com-pleted breast cancer treatment but who werenot experiencing clinically significant fatigueas our control group. We believe that this wasthe most appropriate control group to usefor this study because they resembled the casesin all regards with the exception of sufferingfrom severe fatigue. It might have been inter-esting to also include a ‘‘healthy control’’group of age-matched women with no historyof cancer. However, although this would haveallowed us to investigate differences betweenthe nonfatigued survivors and healthy con-trols, it would not have helped to investigatewhy some women experience fatigue post-treatment and other women do not. Moreover,

it would have significantly added to the accrualproblems with the study.

In summary, we have attempted to test thehypothesis that CRF is associated with in-creased central serotonin sensitivity and havefailed to find evidence that this is, in fact, thecase. Our study can be criticized for being un-derpowered; however, it has provided valuablepreliminary data for other investigators wish-ing to study this potential mechanism for fa-tigue in greater detail and has set boundarieson the magnitude of the expected differencesthat are likely to be detected (if any such differ-ences exist).

Disclosures and AcknowledgmentsThis study was funded by Cancer Research

UK Grant no. C11075/A7143. The authors de-clare no conflicts of interest.

The authors thank Dr. Janine Mansi,Dr. Laura Assersohn, Mr. Anup Sharma, Mr.Dibeyshwar Banerjee, Mrs. Sue Lownes,Mr. Joe Diffley, Ms. Lisa Wynn, Ms. FionaAntopoulous, and Dr. Catherine Coleman.

References1. Stone PC, Minton O. Cancer-related fatigue.Eur J Cancer 2008;44:1097e1104.

2. Minton O, Stone P. How common is fatigue indisease-free breast cancer survivors? A systematic re-view of the literature. Breast Cancer Res Treat 2008;112:5e13.

3. Cella D, Peterman A, Passik S, Jacobsen P,Breitbart W. Progress toward guidelines for themanagement of fatigue. Oncology (Williston Park)1998;12:369e377.

4. Alexander S, Minton O, Andrews P, Stone P. Acomparison of the characteristics of disease-freebreast cancer survivors with or without cancer-related fatigue syndrome. Eur J Cancer 2009;45:384e392.

5. Morrow GR, Andrews PLR, Hickok JT,Roscoe JA, Matteson S. Fatigue associated with can-cer and its treatment. Support Care Cancer 2002;10:389e398.

6. Bower JE. Cancer-related fatigue: links with in-flammation in cancer patients and survivors. BrainBehav Immun 2007;21:863e871.

7. Jager A, Sleijfer S, van der Rijt CCD. The patho-genesis of cancer related fatigue: could increasedactivity of pro-inflammatory cytokines be the com-mon denominator? Eur J Cancer 2008;44:175e181.

898 Vol. 40 No. 6 December 2010Alexander et al.

8. Badawy AAB, Morgan CJ, Llewelyn MB,Albuquerque SRJ, Farmer A. Heterogeneity of se-rum tryptophan concentration and availability tothe brain in patients with the chronic fatiguesyndrome. J Psychopharmacol 2005;19:385e391.

9. Castell LM, Yamamoto T, Phoenix J,Newsholme EA. The role of tryptophan in fatiguein different conditions of stress. Adv Exp Med Biol1999;467:697e704.

10. Bakheit AM, Behan PO, Dinan TG, Gray CE,O’Keane V. Possible upregulation of hypothalamic5-hydroxytryptamine receptors in patients with post-viral fatigue syndrome. BMJ 1992;304:1010e1012.

11. Sharpe M, Clements A, Hawton K, et al. In-creased prolactin response to buspirone in chronicfatigue syndrome. J Affect Disord 1996;41:71e76.

12. Sharpe M, Hawton K, Clements A, Cowen PJ. In-creased brain serotonin function in men withchronic fatigue syndrome. BMJ 1997;315:164e165.

13. Depoortere R, Boulay D, Perrault G, et al.SSR181507, a dopamine D2 receptor antagonistand 5-HT1A receptor agonist. II: behavioral profilepredictive of an atypical antipsychotic activity. Neu-ropsychopharmacology 2003;28:1889e1902.

14. Spitzer M, Gibbon M, Williams J. StructuredClinical Interview for DSM-IV-TR Axis I Disorders,Research Version, non-patient edition (SCID-I/NP). New York: New York State Psychiatric Institute,2002.

15. Bridge MW, Marvin G, Thompson CE, et al.Quantifying the 5-HT1A agonist action of buspironein man. Psychopharmacology 2001;158:224e229.

16. Altman DG. Serial measurementsdthe area un-der the curve. In: Altman DG, ed. Practical statisticsfor medical research. Boca Raton, FL: Chapman &Hall/CRC, 1999.

17. Ryan JL, Carroll JK, RyanEP, et al.Mechanisms ofcancer-related fatigue. Oncologist 2007;12(Suppl 1):22e34.

18. Blomstrand E, Hassmen P, Ekblom B,Newsholme EA. Administration of branched-chainamino acids during sustained exercisedeffects onperformance and on plasma concentration ofsome amino acids. Eur J Appl Physiol 1991;63:83e88.

19. Mittleman KD, Ricci MR, Bailey SP. Branched-chain amino acids prolong exercise during heatstress in men and women. Med Sci Sports Exerc1998;30:83e91.

20. Struder HK, Hollmann W, Platen P, et al. Influ-ence of paroxetine, branched-chain amino acidsand tyrosine on neuroendocrine system responsesand fatigue in humans. Horm Metab Res 1998;30:188e194.

21. Wilson WM, Maughan RJ. Evidence for a possi-ble role of 5-hydroxytryptamine in the genesis of fa-tigue in man: administration of paroxetine, a 5-HTre-uptake inhibitor, reduces the capacity to performprolonged exercise. Exp Physiol 1992;77:921e924.

22. Morrow GR, Hickok JT, Roscoe JA, et al. Differ-ential effects of paroxetine on fatigue and depres-sion: a randomized, double-blind trial from theUniversity of Rochester Cancer Center CommunityClinical Oncology Program. J Clin Oncol 2003;21:4635e4641.

23. Kenter MJH, Cohen AF. Establishing risk of hu-man experimentation with drugs: lessons fromTGN1412. Lancet 2006;368:1387e1391.

24. Malt EA, Olafsson S, Aakvaag A, et al. Altereddopamine function in fibromyalgia patients: a neu-roendocrine study with buspirone in women withfibromyalgia compared to female population basedcontrols. J Affect Disord 2003;75:77e82.