Vol. 40 No. 2 August 2010 Journal of Pain and Symptom Management 217

Original Article

Age-Associated Differences in Fatigue AmongPatients with CancerZeeshan Butt, PhD, Arati V. Rao, MD, Jin-Shei Lai, PhD, Amy P. Abernethy, MD,Sarah K. Rosenbloom, PhD, and David Cella, PhDDepartment of Medical Social Sciences (Z.B., J.-S.L., S.K.R., D.C.), Comprehensive Transplant Center

(Z.B.), Institute for Healthcare Studies (Z.B., J.-S.L., D.C.), Department of Pediatrics (J.-S.L.), and

Department of Psychiatry and Behavioral Sciences (S.K.R., D.C.), Northwestern University Feinberg

School of Medicine, Chicago, Illinois; and Division of Medical Oncology (A.V.R., A.P.A.), Department

of Medicine, Duke University Medical Center, Durham, North Carolina, USA

Abstract

Context. There has been some suggestion that the fatigue experienced by older

cancer patients is more severe than that of younger cohorts; however, there is littleempirical evidence to support this claim.

Objectives. The goal of the present study was to determine the differentialimpact of age and cancer diagnosis on ratings of fatigue using a validated self-report instrument.

Methods. The Functional Assessment of Chronic Illness Therapy-Fatigue subscaleconsists of 13 items measuring fatigue experience and its impact on daily life, withscores ranging from 0 (severe fatigue) to 52 (no fatigue). Fatigue data were availablefrom the U.S. general population (n¼ 1075; 51.3% female, 45.9� 16.5 years) anda sample of mixed-diagnosis cancer patients (n¼ 738; 64.3% female, 58.7� 13.6years). General population participants were recruited using an Internet-basedsurvey panel; patients with cancer were recruited from Chicago-area oncology clinics.

Results. On average, the cancer patient group reported more severe fatiguethan the general population group (36.9 vs. 46.6; F[1,1797]¼ 271.95, P< 0.001).There was evidence for increased fatigue with age (F[6,719]¼ 2.56, P< 0.02)among patients with cancer, but not in the general population (P¼ 0.06).Furthermore, the group� age interaction was not significant (P¼ 0.44).Hemoglobin (Hgb) was treated as a covariate for 430 patients with available data;there was no main effect for age in this analysis.

Portions of these findings were presented at the2008 meeting of the Society for BehavioralMedicine.

Funding source: A subset of the data analyzed for thisstudy was collected as part of National Institutes ofHealth (NIH) R01 CA60068 (Principal Investigator:David Cella, PhD). Manuscript preparation sup-ported in part by grant UL1RR025741 from theNational Center for Research Resources, NIH.

Disclosures: Dr. Butt has received grant supportfrom and served as a consultant for Ortho Biotech

and has served as a consultant for Johnson &Johnson. Dr. Cella has received grant supportfrom Ortho Biotech. The other authors have nodisclosures.

Address correspondence to: Zeeshan Butt, PhD, Depart-ment of Medical Social Sciences, Northwestern Uni-versity Feinberg School of Medicine, 750 N. LakeShore Drive, 10th Floor, Chicago, IL 60611, USA.E-mail: z-butt@northwestern.edu

Accepted for publication: January 19, 2010.

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

0885-3924/$esee front matterdoi:10.1016/j.jpainsymman.2009.12.016

218 Vol. 40 No. 2 August 2010Butt et al.

Conclusion. Older adults, whether they had a cancer diagnosis, reported morefatigue than younger adults. These differences may be explained, in part, by Hgblevel. Future research would be helpful to explore longitudinal changes in fatiguein the general population and guide fatigue management for the older cancerpatient. J Pain Symptom Manage 2010;40:217e223. � 2010 U.S. Cancer PainRelief Committee. Published by Elsevier Inc. All rights reserved.

Key Words

Fatigue, cancer, assessment, patient-reported outcome, hemoglobin

IntroductionFatigue is the most prevalent symptom

among individuals with cancer and may be be-cause of the disease itself, its treatment, and/or psychosocial variables.1 Depending on thepatient population and means of measuring fa-tigue, prevalence estimates among cancer pa-tients are generally high, ranging from 60%to more than 90%.1 Furthermore, in a largesample of patients with advanced cancer whohave received chemotherapy, fatigue was spon-taneously endorsed and ranked as the most im-portant symptom that should be monitored.2

Although common, cancer-related fatigue(CRF) remains poorly understood.3 Patientsmay describe their experience of fatigue interms of being exhausted, tired, weak, orslowed. In clinical practice, fatigue may be ne-glected or underdetected because of the factthat it is a subjective experience that is assessedby patient self-report. Treatment of CRF is fur-ther complicated by its multifactorial clinicalmanifestations, involving both psychologicaland physical components.

One common cause of fatigue in the contextof cancer is anemia. The decrease in hemoglo-bin (Hgb) leads to patient weakness, pallor,dyspnea, and fatigue. Given their nonspecificnature, symptoms of anemia are often difficultto attribute directly to anemia itself. However,the impact of low Hgb can be far reaching.4,5

Important heath-related outcomes such asquality of life (QOL) can be enhanced withproper evaluation and treatment of fatigueand other anemia-related symptoms,6 espe-cially for the older adult.7

There has been some suggestion that thefatigue experienced by older cancer patients ismore severe than that of younger cohorts; how-ever, there is little empirical evidence to support

this claim. This is especially relevant in the caseof the elderly, who may be more likely to con-sider fatigue as a normal part of the diseasecourse with which they must suffer. In fact,our group has found no effect of age on CRF,when comparing patients older than 50 withthose younger.3 That said, few data are availableto guide the assessment and treatment of CRFin older cancer patients. This lack of informa-tion is of considerable concern given that CRFcan seriously compromise patients’ QOL andability to function on a daily basis.8

Given the general aging of the cancer popula-tion, and the importance of addressing fatiguein this population, we conducted a closer exam-ination of the potential impact of age on CRF.Specifically, available cross-sectional data setswere used to investigate whether fatigue variedsystematically as a function of both age and di-agnosis. We hypothesized that cancer patientswould report more fatigue than the generalpopulation and that this difference would bemore pronounced for the older sample. Thatis, a statistical interaction between age and diag-nosis was expected with respect to fatigue. Hgbvalues were available for a subset of patients,and it was hypothesized that differences infatigue may be at least partly explained by thisimportant clinical variable.

Patient and MethodsSample

Two existing data sets for this cross-sectionalsecondary data analysis were used. Full detailson recruitment of the samples are found inthe original publications.3,9e12 Briefly, the firstdata set consisted of a large sample (n¼ 1075)of individuals from the general population,randomly drawn to complete a series of

Vol. 40 No. 2 August 2010 219Age-Associated Differences in Fatigue

questionnaires from more than 100,000 indi-viduals enrolled in an Internet-based surveypanel.9 Data from several mixed-diagnosis can-cer samples3,10e12 comprised the second dataset (n¼ 738). In this second data set, patientswere recruited from Chicago-area oncologyclinics for studies on health-related QOL. Allpatients received some treatment for theircancer.

EthicsParticipants provided informed consent be-

fore data collection. Original data collectionand informed consent procedures were ap-proved by the appropriate institutional reviewboard.

AssessmentsAll participants completed the Functional As-

sessment of Chronic Illness Therapy-Fatigue(FACIT-F) subscale.10 The FACIT-F is a 13-itemscale that asks respondents to rate statements re-garding their fatigue experience and its impacton their daily lives. Sample items include ‘‘I feelfatigued,’’ ‘‘I feel weak all over,’’ and ‘‘I feel list-less (washed out).’’ All items are rated ona 0 (not at all) to 4 (very much) scale. By scoringconvention, after appropriate reverse scoring of11 items, lower scores on the FACIT-F subscaleindicate greater levels of fatigue. (A scoringtemplate is available at www.facit.org.) Origi-nally developed for use with cancer patients,9,13

the scale has been successfully tested for use inthe general population3,9 and chronic anemiaof aging.7 To enhance the clinical usefulness ofthe FACIT-F subscale, Cella et al.13 estimateda minimum clinically meaningful difference of3 points by using both anchor- and distribution-based methods. Additionally, Eastern Coopera-tive Oncology Group (ECOG) performanceratings were available for all cancer patientsand Hgb values, obtained within 30 days of fa-tigue ratings, were available for a subset of 430cancer patients.

Data AnalysisSociodemographic and clinical comparisons

between the general population and patientgroups were conducted using independentsamples t-tests or c2 tests, as appropriate. Anal-yses of variance (ANOVAs) and analysis of co-variance were used to assess differences infatigue across age categories. Statistical

significance was set at P< 0.05. All analyseswere conducted using PASW Statistics 17(SPSS, Chicago, IL).14

ResultsSample Characteristics

As summarized in Table 1, the general popula-tion sample was 45.9� 16.6 years, 51% female,and primarily (84%) Caucasian. When askedto rate their health status on a 1 (‘‘good’’) to 5(‘‘bad’’) scale, 56% of the sample respondedwith either a 1 or 2 rating. Cancer patientswere 58.7� 13.6 years, 64% female, 88% Cauca-sian, and 79% self-reported ECOG performancestatus ratings of 0 or 1 (no symptoms to symp-tomatic, but ambulatory).15 Breast (33%) andcolorectal (12%) were the most common tumortypes. Patients were older (t[1802]¼ 17.3,P< 0.001) and more likely to be female(c2[1]¼ 29.6, P< 0.001), compared with thegeneral population sample.

FatigueAs expected, cancer patients reported more

fatigue (FACIT-F subscale¼ 36.9� 11.4) thanthe general population sample (46.6� 7.2;F[1,1804]¼ 329.2, P< 0.001). This finding wasnot because of the difference in gender ratiosbetween the samples. Females reported morefatigue than males in the general populationsample (t[1057]¼ 3.7, P< 0.001), whereasmales and females reported comparable levelsof fatigue (t[724]< 1, P> 0.5) in the cancersample. Additionally, when the entire sample(i.e., patients and general population) was sub-divided using a binary age split (<65 vs. $65),older individuals reported more fatigue(40.2� 11.0) than those less than 65 years(43.4� 10.0; F[1,1797]¼ 33.9, P< 0.001). Be-cause an age cut off of 65 is arbitrary, the firstage analysis was followed up with a mean com-parison of fatigue by age, divided by decade.This analysis confirmed the association of agewith fatigue (F[6,1797]¼ 15.5, P< 0.001).Table 2 presents the mean fatigue scores by de-cade across the entire sample. Considered inde-pendently, diagnosis status and age wereassociated with increased fatigue, but the agetrend becomes more consistent when the twosubsamples are combined. Given the relativelylow representation of the very young and the

Table 1Sociodemographic and Clinical Summary of Samples

General Population (n¼ 1075) Cancer Sample (n¼ 738)

Age (mean� standard deviation), in years 45.9� 16.6 58.7� 13.6

% Female 51 64

Ethnicity, %Caucasian 84.1 88.0African American 10.1 6.0Hispanic 3.0 3.0Other 2.8 3.0

Self-reported health status, %1 (‘‘good’’) 19.4 37.6 ECOG PSR 02 36.5 41.6 PSR 13 31.0 17.0 PSR 24 10.7 3.8 PSR 35 (‘‘bad’’) 2.5 0.0 PSR 4

FACIT-F subscale score 46.6� 7.2 36.9� 11.4

Hgb d 12.0� 1.9 g/dL

Tumor type, % dBreast 33.4Colorectal 11.9NHL 9.1Ovarian 7.3Prostate 5.1

Cancer stage, %I d 10.2II d 24.6III d 26.9IV d 19.8

Note: ECOG Performance Status Rating: 0¼ fully active without restriction; 4¼ completely disabled, no self-care.Lower scores on the FACIT-F subscale indicate greater levels of fatigue.Cancer patients reported more severe fatigue than the general population (F[1,1804]¼ 329.2, P< 0.001).NHL ¼ non-Hodgkin’s lymphoma; PSR ¼ performance status rating.

220 Vol. 40 No. 2 August 2010Butt et al.

very old in our cancer sample, we also tested theassociation of age separately for each group. Inthe cancer sample, there was a significant associ-ation (F[6,719]¼ 2.56, P< 0.02) suggesting anincrease in fatigue with age. The association offatigue with age in the general population sam-ple suggested a nonsignificant trend(F[6,1065]¼ 2.03, P¼ 0.06).

A test of the interaction between sample(i.e., general population vs. cancer sample)and age required a test of both effects simulta-neously. Considered conjointly, the main effectfor sample (F[1,1797]¼ 271.9, P< 0.001) andage (F[6,1797]¼ 3.5, P< 0.01) remained; how-ever, there was no support for a sample-age in-teraction (F[6,1707]¼ 0.98, P¼ 0.44). Thislack of statistical interaction suggests thatalthough cancer patients reported more fa-tigue than the general population, this differ-ence did not increase significantly with age.Across the total sample, results from a re-gression analysis suggested that, with each ad-ditional decade, fatigue scores worsened

(mean� standard error) by 1.27� 0.14 pointson the FACIT-F subscale (P< 0.001) (Fig. 1).

HemoglobinThe potential impact of Hgb on fatigue ratings

by age was investigated in the subset of cancer pa-tients with available data (n¼ 430). Of note,those patients with available Hgb values werenot different from those patients without thesedata with respect to age (F[1,730]¼ 0.001,P¼ 0.98), sex (c2[1]¼ 0.15, P¼ 0.70), orFACIT-F subscale scores (F[1,730]¼ 1.65,P¼ 0.20).

For those with available Hgb, the mean�standard deviation values were within normallimits (12.0� 1.7 g/dL) and ranged from 7.0to 19.3 g/dL. Hgb values were modestly associ-ated with FACIT-F subscale scores (r¼ 0.24,P< 0.001), in the expected direction. Menhad higher Hgb values (12.4� 2.0 g/dL)than women (11.8� 1.5 g/dL; t[428]¼ 3.15,P< 0.005). Interestingly, although there wasno direct association of Hgb with age category

Table 2FACIT-F Subscale Scores by Population, Age, and for the Total Sample

Population Age Category (Years) n % Mean Standard Deviation Minimum Maximum

Cancer population 18e30 18 2.5 35.5 11.7 6 5130e40 48 6.6 37.8 11.1 11 5240e50 133 18.3 36.0 11.6 7 5250e60 181 24.9 37.3 11.3 3 5260e70 172 23.7 38.5 11.4 3 5270e80 145 20.0 36.1 10.4 7 5280þ 29 4.0 30.5 13.3 3 52

Total 726 100.0 36.8 11.4 3 52

General population 18e30 224 20.9 46.2 7.6 15 5230e40 219 20.4 47.6 5.9 23 5240e50 218 20.3 46.2 7.8 16 5250e60 190 17.7 47.1 6.5 19 5260e70 128 11.9 46.1 7.3 19 5270e80 72 6.7 46.0 7.9 18 5280þ 21 2.0 43.4 8.6 19 52

Total 1072 100.0 46.6 7.2 15 52

Total 18e30 242 13.5 45.4 8.5 6 5230e40 267 14.8 45.8 8.1 11 5240e50 351 19.5 42.3 10.6 7 5250e60 371 20.6 42.3 10.4 3 5260e70 300 16.7 41.8 10.6 3 5270e80 217 12.1 39.4 10.7 7 5280þ 50 2.8 35.9 13.1 3 52

Grand total 1798 100.0 42.6 10.3 3 52

Note: Total n does not equal 1813 because of missing data.FACIT-F subscale scores have a possible range of 0e52. Lower scores indicate greater levels of fatigue.A three-point change on the FACIT-F subscale has been shown to indicate clinically significant change in fatigue over time.13

Vol. 40 No. 2 August 2010 221Age-Associated Differences in Fatigue

(F[6,429]¼ 0.75, P¼ 0.61), when Hgb wastreated as a covariate in an ANOVA of the ef-fect of age on fatigue ratings, Hgb explainedsignificant variance in fatigue ratings(F[1,426]¼ 24.15, P< 0.001) but age category(i.e., decade) did not (F[6,426]¼ 1.63,P¼ 0.14).

DiscussionThe present cross-sectional study is notable in

that it is the first to directly assess the statisticalinteraction between diagnosis and age on fa-tigue. Based on extant research, one might ex-pect no or limited impact of age on CRF.16

Cella et al.3 found that people older than 50years in the general population reported morefatigue than their younger counterparts onthe 13-item FACIT-F subscale, but found nosuch age effect in the sample of anemic cancerpatients. Similar analyses in other cancer clini-cal trials have failed to find a substantial effectof age on fatigue- and/or anemia-related QOLin cancer patients.17,18 In contrast, the present

analyses suggest that, as the decades accumu-late, ratings of fatigue increase. However, in-creases in fatigue are not more pronouncedfor older patients with cancer compared withthe general population. To the authors’ knowl-edge, this is the first published analysis that ad-dresses this topic by considering cancerdiagnosis and age concurrently.

Across the overall sample of cancer patientsand the general population, there is a statisti-cally significant but somewhat modest increasein fatigue with age (1.27 points per decade onthe FACIT-F subscale). Evaluation of the can-cer and general population subgroups re-vealed that the increase in fatigue was moreevident in the cancer sample. Hgb level ap-pears to be a meaningful covariate and maybe one of many factors that account for theage-associated differences in fatigue withinthe cancer sample.

Low Hgb is common in the elderly, and theincidence increases with age. A recent analysisof U.S. national data using the NationalHealth and Nutrition Examination Survey(1988e1994) suggests that the rate of anemia

Age

80+

71-8

0

61-7

0

51-6

0

41-5

0

31-4

0

18-3

0

FAC

IT-F

atig

ue S

ubsc

ale

52

48

44

40

36

32

28

24

20

16

12

8

40

LESS

fatigue

MORE

fatigue

cancer sample

general population

Fig. 1. Fatigue in the general population and among cancer patients with advancing age. Bars representmeansþ 95% confidence interval. Across both groups, there was evidence for increased fatigue with age(F[6,1797]¼ 3.53, P< 0.01) but no group� age interaction (P> 0.25).

222 Vol. 40 No. 2 August 2010Butt et al.

is 10%e11% for individuals older than 65years and greater than 20% for individuals old-er than 85 years.19 Thus, given that both ane-mia and cancer are common among theelderly, the question of age-associated fatiguechanges is important to consider.

Of course, CRF is not always because of lowHgb,1,20,21 as suggested by the modest correla-tion between Hgb and FACIT-F subscale scoresin the present analysis. CRF also may be broughton by the direct effects of the cancer, comorbidmedical conditions, psychosocial factors, and/or other treatment-related side effects1,22,23

that were not assessed as part of this study. We fo-cused on anemia as a potential explanatory vari-able in the present analysis because it wasavailable for a significant portion of patients inthis secondary data analysis. Additional dataare needed to characterize the fatigue of the old-est cancer patients; our own sample was deficientin this age range. It also would be helpful to rep-licate the present findings in a sample of patientsreceiving more aggressive treatment regimens,as this would likely have significant impact on pa-tient’s reported fatigue. Although we were lim-ited in terms of potential covariates and ourcell sizes for the oldest old, our secondary dataanalysis was successful in generating hypothesesfor future studies. Indeed, there is much

research to be done to improve our understand-ing of how ratings of fatigue change with age.

There remain questions to be answered withrespect to changes in fatigue over time and theconsequences of fatigue in aging adults. As in-dividuals age, there may be a shift in their re-port of fatigue, because of some level ofaccommodation for or acceptance of reducedactivity levels.24 Fatigue also may impact theability to participate in important life activities,which may be associated with inactivity, subse-quent loss of muscle tone and weakness, andincreased disability.25 A sense of decreasedability to engage in physical activity also mayhave psychological consequences that shouldbe monitored.26 Certainly, additional researchmay be useful to explore longitudinal changesin fatigue to guide fatigue management for theolder patient with and without cancer.

References1. Wagner LI, Cella D. Fatigue and cancer: causes,

prevalence and treatment approaches. Br J Cancer2004;91:822e828.

2. Butt Z, Rosenbloom SK, Abernethy AP, et al.Fatigue is the most important symptom for ad-vanced cancer patients who have had chemother-apy. J Natl Compr Canc Netw 2008;6:448e455.

Vol. 40 No. 2 August 2010 223Age-Associated Differences in Fatigue

3. Cella D, Lai JS, Chang CH, Peterman A,Slavin M. Fatigue in cancer patients comparedwith fatigue in the general United States popula-tion. Cancer 2002;94:528e538.

4. Beghe C, Wilson A, Ershler WB. Prevalence andoutcomes of anemia in geriatrics: a systematic re-view of the literature. Am J Med 2004;116(Suppl7A):3Se10S.

5. Pujade-Lauraine E, Gascon P. The burden ofanaemia in patients with cancer. Oncology 2004;67(Suppl 1):1e4.

6. Vogelzang NJ, Breitbart W, Cella D, et al.Patient, caregiver, and oncologist perceptions ofcancer-related fatigue: results of a tripart assessmentsurvey. The Fatigue Coalition. Semin Hematol 1997;34:4e12.

7. Agnihotri P, Telfer M, Butt Z, et al. Chronicanemia and fatigue in elderly patients: results ofa randomized, double-blind, placebo-controlled,crossover exploratory study with epoetin alfa. J AmGeriatr Soc 2007;55:1557e1565.

8. Cella D, Kallich J, McDermott A, Xu X. The lon-gitudinal relationship of hemoglobin, fatigue andquality of life in anemic cancer patients: resultsfrom five randomized clinical trials. Ann Oncol2004;15:979e986.

9. Cella D, Zagari MJ, Vandoros C, et al. Epoetinalfa treatment results in clinically significant im-provements in quality of life in anemic cancer pa-tients when referenced to the general population.J Clin Oncol 2003;21:366e373.

10. Yellen SB, Cella DF, Webster K, Blendowski C,Kaplan E. Measuring fatigue and otheranemia-related symptoms with the Functional As-sessment of Cancer Therapy (FACT) measurementsystem. J Pain Symptom Manage 1997;13:63e74.

11. Lai JS, Cella D, Dineen K, et al. An item bank wascreated to improve the measurement of cancer-relatedfatigue. J Clin Epidemiol 2005;58:190e197.

12. Lai JS, Crane PK, Cella D. Factor analysis tech-niques for assessing sufficient unidimensionality of can-cer related fatigue. Qual Life Res 2006;15:1179e1190.

13. Cella D, Eton DT, Lai JS, Peterman AH,Merkel DE. Combining anchor anddistribution-based methods to derive minimal

clinically important differences on the Functional As-sessment of Cancer Therapy (FACT) anemia and fa-tigue scales. J Pain Symptom Manage 2002;24:547e561.

14. PASW for Windows. Rel 17.02. Chicago, IL: 2009.

15. Oken MM, Creech RH, Tormey DC, et al. Toxic-ity and response criteria of the Eastern CooperativeOncology Group. Am J Clin Oncol 1982;5:649e655.

16. Butt Z, Cella D. Relationship of hemoglobin, fa-tigue, and quality of life in anemic cancer patients.In: Nowrousian MR, ed. Recombinant human eryth-ropoietin (rhEPO) in clinical oncology: Scientificand clinical aspects of anemia in cancer. NewYork: Springer-Verlag, 2008; 369e392.

17. Thome B, Dykes AK, Hallberg IR. Quality of lifein old people with and without cancer. Qual LifeRes 2004;13:1067e1080.

18. Aapro MS, Cella D, Zagari M. Age, anemia, andfatigue. Semin Oncol 2002;29:55e59.

19. Guralnik JM, Eisenstaedt RS, Ferrucci L,Klein HG, Woodman RC. The prevalence of anemiain persons age 65 years and older in the UnitedStates: evidence for a high rate of unexplained ane-mia. Blood 2004;104:2263e2268.

20. Lipman AJ, Lawrence DP. The management of fa-tigue in cancer patients. Oncology 2004;18:1527e1534.

21. Jacobsen PB. Assessment of fatigue in cancerpatients. J Natl Cancer Inst Monogr 2004;93e97.

22. Mock V, Atkinson A, Barsevick AM, et al. Cancer-related fatigue. Clinical practice guidelines in oncol-ogy. J Natl Compr Canc Netw 2007;5:1054e1078.

23. Levy M. Cancer fatigue: a review for psychia-trists. Gen Hosp Psychiatry 2008;30:233e244.

24. Schwartz CE, Bode R, Repucci N, et al. Theclinical significance of adaptation to changinghealth: a meta-analysis of response shift. Qual LifeRes 2006;15:1533e1550.

25. Fried LP, Tangen CM, Walston J, et al. Frailty inolder adults: evidence for a phenotype. J Gerontol ABiol Sci Med Sci 2001;56:M146eM156.

26. Brown LF, Kroenke K. Cancer-related fatigue andits associations with depression and anxiety: a system-atic review. Psychosomatics 2009;50:440e447.