CLINICIAN’S CORNERCLINICAL REVIEW

Cancer Survivors and UnemploymentA Meta-analysis and Meta-regressionAngela G. E. M. de Boer, PhDTaina Taskila, PhDAnneli Ojajärvi, PhDFrank J. H. van Dijk, PhD, MDJos H. A. M. Verbeek, PhD, MD

IMPROVEMENT IN THE TREATMENT AND

prognosisofmanyformsofcancerhasresultedinincreasingnumbersofcan-cersurvivors.1 Theprevalenceofcan-

cer survivors is expected to increase inmostcountriesbecauseofanagingpopu-lation and continued improvements inearly detection and treatment of cancer.Therefore, it is important to understandtheadverselong-termeffectsofcancersur-vivorshiponmedical,psychological, andsocial outcomes.2

A significant proportion of cancer sur-vivors experience physical, emotional,and social problems such as fatigue, pain,cognitive deficits, anxiety, and depres-sion, all of which may become chronic.3

These long-term medical and psycho-logical effects of cancer or its treatmentmay cause impairments that diminish so-cial functioning including the obtain-ment or retention of employment.4,5 Al-most half of all cancer survivors areyounger than 65 years. Thus, many can-cer survivors are at an age at which can-cer and its treatment could alter their em-ployment opportunities.6,7

Many cancer survivors want and areable to return to work after diagnosisand treatment.8 Patients often regard re-turning to work as indicative of com-plete recovery4 and regained nor-malcy.9 Employment is also associatedwith a higher quality of life.4 The en-

couragement of cancer survivors to re-turn to work also benefits aging soci-eties economically.

Relatively few studies have assessedthe association of cancer survivorshipwith unemployment. Several mecha-nisms may operate to promote unem-ployment after the diagnosis and treat-ment of cancer. Job discrimination,7,10

difficulty combining treatment withfull-time work,11 and physical or men-tal limitations12 may be major causes ofunemployment. Additional factors suchas age, sex, and the prevailing unem-

CME available online atwww.jamaarchivescme.comand questions on p 784.

Author Affiliations: Coronel Institute of OccupationalHealth, Academic Medical Center, Amsterdam, theNetherlands (Drs de Boer, Taskila, van Dijk, and Ver-beek); Primary Care Clinical Sciences, School of Healthand Population Sciences, The University of Birming-ham, Birmingham, England (Dr Taskila); Finnish Insti-tute of Occupational Health, Good Practices and Com-petence, Statistical Services, Helsinki, Finland (DrOjajärvi); and Finnish Institute of Occupational Health,Knowledge Transfer Team, Cochrane OccupationalHealth Field, Kuopio, Finland (Dr Verbeek).Corresponding Author: Angela G. E. M. de Boer, PhD,Coronel Institute of Occupational Health, AcademicMedical Center, PO Box 22700, 1100 DE Amster-dam, the Netherlands (a.g.deboer@amc.uva.nl).Clinical Review Section Editor: Mary McGraeMcDermott, MD, Contributing Editor. We encour-age authors to submit papers for consideration as aClinical Review. Please contact Mary McGraeMcDermott, MD, at mdm608@northwestern.edu.

Context Nearly half of adult cancer survivors are younger than 65 years, but the as-sociation of cancer survivorship with employment status is unknown.

Objective To assess the association of cancer survivorship with unemployment com-pared with healthy controls.

Data Sources A systematic search of studies published between 1966 and June 2008wasconductedusingMEDLINE,CINAHL,EMBASE,PsycINFO,andOSH-ROMdatabases.

Study Selection Eligible studies included adult cancer survivors and a control group,and employment as an outcome.

Data Extraction Pooled relative risks were calculated over all studies and accord-ing to cancer type. A Bayesian meta-regression analysis was performed to assess as-sociations of unemployment with cancer type, country of origin, average age at di-agnosis, and background unemployment rate.

Results Twenty-six articles describing 36 studies met the inclusion criteria. The analy-ses included 20 366 cancer survivors and 157 603 healthy control participants. Studiesincluded 16 from the United States, 15 from Europe, and 5 from other countries. Over-all, cancer survivors were more likely to be unemployed than healthy control participants(33.8% vs 15.2%; pooled relative risk [RR], 1.37; 95% confidence interval [CI], 1.21-1.55). Unemployment was higher in breast cancer survivors compared with control par-ticipants (35.6% vs 31.7%; pooled RR, 1.28; 95% CI, 1.11-1.49), as well as in survivorsof gastrointestinal cancers (48.8% vs 33.4%; pooled RR, 1.44; 95% CI, 1.02-2.05), andcancers of the female reproductive organs (49.1% vs 38.3%; pooled RR, 1.28; 95% CI,1.17-1.40). Unemployment rates were not higher for survivors of blood cancers com-pared with controls (30.6% vs 23.7%; pooled RR, 1.41; 95% CI, 0.95-2.09), prostatecancers (39.4% vs 27.1%; pooled RR, 1.11; 95% CI, 1.00-1.25), or testicular cancer (18.5%vs 18.1%; pooled RR, 0.94; 95% CI, 0.74-1.20). For survivors in the United States, theunemployment risk was 1.5 times higher compared with survivors in Europe (meta-RR,1.48; 95% credibility interval, 1.15-1.95). After adjustment for diagnosis, age, and back-ground unemployment rate, this risk disappeared (meta-RR, 1.24; 95% CI, 0.85-1.83).

Conclusion Cancer survivorship is associated with unemployment.JAMA. 2009;301(7):753-762 www.jama.com

©2009 American Medical Association. All rights reserved. (Reprinted) JAMA, February 18, 2009—Vol 301, No. 7 753

Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013

ployment rate in a specific country orregion could further increase the riskof unemployment for cancer survi-vors. A prior study has shown thatyoung adult survivors of childhood can-cer in the United States are at a higherrisk of unemployment compared witha like population in Europe,13 perhapsdue to differences in social security sys-tems and health care insurance.

The purposes of this meta-analysisare to quantify the risk of unemploy-ment among adult cancer survivorscompared with healthy control partici-pants, to examine the influence of prog-nostic factors on unemployment, andto identify groups of cancer survivorsat highest risk for unemployment.

METHODSEligibility Criteria

Studieswere included if theymet the fol-lowingcriteria: (1) inclusionofacontrolgroup of healthy participants; (2) inclu-sion of patients diagnosed with cancerworkingatthetimeofdiagnosis,meanageat diagnosis 18 years or older, and meanaverageageattimeofstudy18to60years;and (3) assessment of employment sta-tusmeasuredinafollow-upstudydesign.

Search Strategy

A series of literature searches was con-ducted using the electronic databasesMEDLINE, CINAHL, EMBASE,PsycINFO, and OSH-ROM. Studiespublished from 1966 to June 2008 wereretrieved, with no language restrictions.

Search terms included employment, un-employment, absenteeism, work, sick leave,vocational rehabilitation, occupational, vo-cational guidance, job satisfaction, occu-pation(s), rehabilitation, work disability,return-to-work, sickness absence, disabil-ity pension, work ability, or job per-formance and were combined with sur-vivor(s), late effects or longevity and withneoplasm(s), cancer(s), carcinoma(s), on-cology, leuk(a)emia(s), sarcoma(s), lym-phoma, melanoma, radiotherapy, or che-motherapy. When available, subjectheading terms such as Medical SubjectHeadings terms were added in allsearches. Publications included in themeta-analysis and review articles on em-

ployment in cancer survivors were hand-searched for additional references.

Study SelectionThe literature search was conducted in-dependentlyand induplicateby2 inves-tigators (A.dB. and J.V.). Each abstractwas evaluated independently and in du-plicateby2investigators(A.dB.andJ.V.).Onlyabstracts reportingempirical stud-ies were selected. Full reports of poten-tially relevant articles were reviewed by2investigators(A.dB.andJ.V.).Disagree-mentswereresolvedwithathirdreviewer(F.vD.) through consensus.

Data Extraction

Data extraction and quality assessmentwere performed independently by 3 re-viewers (A.dB., J.V., and T.T.) with 2 re-viewers per article. Unemployment datawereextractedusingseveralmethods.Stu-dents, homemakers, and individuals re-tired for reason of age were excluded (ifpossible) so thatonlypatientseligible forworkwereincluded.Ifapplicable,patientswhoweremenandwomen,working fulltimeandpart time,unemployed, andre-ceiving a disability pension were com-bined. If reported, thenumberofpatientsnotworkingbecauseofdisabilitywas re-corded.Next, ifnecessary,percentagesofunemployed and employed individualswereconverted into frequencies.Finally,the number of individuals unemployedand the total number of participants eli-gible for work in the patient and controlgroupswereentered into thedataextrac-tionform.Authorswerecontactedincaseof uncertainty about the data.

The diagnoses were sorted into diag-nostic groups. If at least 50% of the pa-tients in a study had a specific diagno-sis, the study was included in a diagnosticgroup for that specific cancer—if not, itwas categorized under mixed diagnoses.

The following information was inde-pendentlyextractedby2authors (A.dB.,J.V. or T.T.) and entered in the data ex-traction form:countryoforigin,diagno-sis, average age at diagnosis, sex, follow-uptimesincediagnosis,characteristicsofthe control group, and source of data(registry-based or not). Disagreementswere resolved by discussion.

Methodological QualityAssessmentMethodological quality was assessedusing the Methodological Index forNon-randomized Studies (MINORS).14

Quality assessment was conducted in-dependentlyand induplicateby2 inves-tigators for each article (A.dB., T.T. orJ.V.). Studies were scored on 12 items:aim of the study, inclusion of consecu-tive patients and participation rate, pro-spectivedatacollection, inclusionofem-ploymentmeasure,unbiasedassessmentof study end points, appropriateness offollow-uptimeafterdiagnosis, inclusionof loss to follow-up, prospective calcu-lationof thestudysize, comparablecon-trolgroup,contemporarycontrolgroups,baseline equivalence of groups on sev-eral factors, and adequate statisticalanalysis. Studies received 0 to 2 pointsfor each of these 12 components. Thetotal score ranged from 0 to 24 points.Low-quality and high-quality studieswere defined as earning fewer than 16points and 16 points or greater, respec-tively, on the MINORS test. An adjustedcutoff point was calculated for a sensi-tivity analysis. The adjusted score sum-marized the 4 items most important toour review: prospective data collection,loss to follow-up, comparable controlgroup, and basic equivalence of groupson several factors. The adjusted scoreranged from 0 to 8 with a cutoff scorefor high quality defined as 5 or greater.

The criteria were tested on a separatesetofarticlestoensureagreementbetweenassessors.

Data Analysis

Theresultsof all controlled studieswereplotted as relative risks (RR) and thenpooledusingtheinversevariancemethod.To avoid unnecessary heterogeneity, weformed homogeneous groups of studiesaccording to cancer diagnosis and usedtheseassubcategories.Asubgroupmeta-analysis was performed for the outcomeof unemployment due to disability.

TheRRwasused tosummarize thedi-chotomousunemploymentdata.Theran-dom effects model, described by DerSi-monianandLaird,15 wasselectedoverthefixedeffectsmodelbecauseitincorporates

CANCER SURVIVORS AND UNEMPLOYMENT

754 JAMA, February 18, 2009—Vol 301, No. 7 (Reprinted) ©2009 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013

within- and between- study variability,which is applicable to this meta-analysisthat involves observational studies withinherentlymorevariabilitythanrandom-ized trials. Statistical heterogeneity wasassessedwith�2 testsandquantifiedwiththe I2 statistic, which describes the per-centage of total variation across studiesthat isattributabletoheterogeneityratherthanchance.16 I2 valuesof25%,50%,and75%havebeensuggestedas indicatorsoflow, moderate and high heterogeneity,respectively.16

To explain heterogeneity betweenstudies and to examine the influence ofprognostic factors on unemployment,we performed a Bayesian meta-regression. The following factors werestudied: average age of study group par-ticipants (�50 vs �50 years), diagno-sis (grouped into 5 categories: testicu-lar, breast, prostate, blood, and otherand mixed), geographic area related tosocial security systems and health careinsurance policies (Europe, UnitedStates, or other countries), and back-ground unemployment rate (controlgroup unemployment rate at fol-low-up minus mean unemploymentrate from all studies at follow-up). Sexwas not included in the analysis be-cause of the high correlation betweencertain diagnoses and male or femalesex. An additional analysis was per-formed for the studies from the UnitedStates vs all other countries.

We analyzed the influence of thesevariables first in a univariate analysisand next in a multivariate analysis usinga Bayesian meta-regression model. TheBayesian analysis yields posterior dis-tributions for the meta-RRs for whichthe median values are used. These meta-RRs indicate the change in the study RRfor studies with the characteristic of in-terest compared with studies withoutthis characteristic. A Bayesian ana-logue of a confidence interval is calleda credibility interval (CrI), which is aposterior probability interval and alsoincludes knowledge of a prior distri-bution in addition to the data.17-19 Non-informative normal priors were used forlog RRs and noninformative � priors forthe corresponding precisions.

A sensitivity analysis was performedto account for methodological qualitydifferences. The multivariate analysisusing a Bayesian meta-regression modelwas repeated for the high-quality stud-iesonlywith the2cutoffpoints forqual-ity. Statistical meta-analyses were per-formed using Review Manager software(RevManversion4.2.8,theCochraneCol-laboration, Oxford, England). WinBugsversion1.3(MRCBiostatisticsUnit,Cam-bridge,England)wasusedinallBayesianmeta-regressionmodels.20,21 WeusedtheEgger test toassess thepossibilityofpub-licationbias.22 Thepowerof therandom-effects meta-analysis was assessed withthe Hedges and Pigott procedure.23

Unless stated otherwise, � was set at.05 and all tests were 2-sided.

RESULTSA total of 1766 abstracts were retrievedfrom the electronic databases(FIGURE 1). Of these, 33 original ar-ticles provided data on the employmentstatus of cancer survivors comparedwith healthy control participants.1,24-55

Duplicate publication was identifiedin 7 articles and for these we useddata from only the first published ar-ticle, which resulted in 26 originalarticles.1,24-48 Data on 2 to 8 cancertypes were reported in each of 3 ar-ticles25,26,36 and the separate results foreach cancer diagnosis are presented asindividual studies. Thus, this meta-analysis included 26 articles reportingresults from 36 studies of different can-cer diagnoses.

Figure 1. Selection for Studies of Cancer Survivors, Control Participants, and EmploymentOutcomes

26 Articles included (36 studies ondifferent diagnoses)

33 Articles included

99 Screened for more detailed assessmentof inclusion and exclusion criteria88 MEDLINE11 EMBASE

7 Articles excluded (identical oroverlapping results of includedarticles)

27 Articles included 6 Articles included from hand searcha

72 Publications excluded10 Insufficient employment data62 No healthy control group

1667 Articles excluded based ontitle and abstract124 No cancer227 Childhood cancer671 No employment data44 No control group

442 Duplicates

1766 Potentially relevant articles identifiedand screened for inclusion1006 MEDLINE360 EMBASE124 CINAHL

9 OSH-ROM267 PsycINFO

159 Occupational disease

OSH-ROM denotes Occupational Safety and Health-Read Only Memory.aOf hand-searched articles, all were from included articles and none were from relevant reviews.

CANCER SURVIVORS AND UNEMPLOYMENT

©2009 American Medical Association. All rights reserved. (Reprinted) JAMA, February 18, 2009—Vol 301, No. 7 755

Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013

Description of Included StudiesTABLE 1 and TABLE 2 summarizecharacteristics of the 26 includedarticles. There were no articles from1966 to 1993, 4 articles from 1994-

2001, and 22 articles from 2002-2008. There were 10 studies of breastcancer, 7 of blood cancers (eg, leuke-mia and [non]-Hodgkin disease), 3 ontesticular cancer, 3 on prostate cancer,

2 on cancers of the female reproduc-tive system (cervical and ovarian can-cer), 2 on gastrointestinal cancers, 1each on melanoma, nervous systemtumors, thyroid cancer, nasopharyn-

Table 1. Characteristics of Included Studies for Breast, Prostate, Testicular, and Blood Cancers

Source by Diagnostic Group (Location)

No. ofPatients

(% Women)a

PatientAge at

Follow-up,Mean(SD or

Range), ya

Start ofFollow-up

AfterDiagnosis,

Mean(Range), ya

Characteristicsof Control Group

Matching Study Design, Patient/Control

Breast cancerFarley Short et al,25 2008 (United States) 289 (100) NR (55-65) 4 (3-5) Sex Hospital-based cross-sectional/registry

Gudbergsson et al,26 2007 (Norway) 269 (100) 53 (7) 1-5 Sex, age, andresidency

Hospital-based cross-sectional survey/population survey

Helgeson and Tomich,31 2005 (United States) 267 (100) 54 6 (5-7) Sex, age, ethnicity,and residency

Prospective cohort/cross-sectionalsurvey of neighborhood controlparticipants

Bradley et al,32 2005 (United States) 445 (100) 51 (8) 0.75 Sex and residency Registry-based cohort/populationsurvey

Taskila-Abrandt et al,36 2004 (Finland) 4098 (100) 2-3 Sex and age Registry-based retrospective cohort/population registry

Maunsell et al,37 2004 (Canada) 567 (100) 47 (7) 3 (3-4) Sex and age Registry-based retrospective cohort/population survey

Servaes et al,40 2002 (the Netherlands) 150 (100) 46 (6) 2 Sex and age Hospital-based cross-sectional survey/acquaintance survey

Bradley et al,41 2002 (United States) 156 (100) 55 (5) 7 (1-32) Sex and age National registry-based cross-sectionalsurvey (for both groups)

Dorval et al,45 1998 (Canada) 54 (100) 9 (8-9) Sex, age, andresidency

Hospital-based retrospective cohort/cross-sectional survey

Weitzner et al,46 1997 (United States) 60 (100) 54 (9) �5 Sex Hospital-based cross-sectional survey/non–cancer patients survey

Blood cancersMajhail et al,27 2007 (United States)

[non-Hodgkin disease]276 (45) 46 (22-73) 6 (2-17) Sex, age, and

ethnicityHospital-based cross-sectional survey/

sibling survey

Syrjala et al,34 2005 (United States) [leukemia] 126 (52) 46 (9) 11 Sex, age, race,and education

Hospital-based cross-sectional survey/sibling survey

Andrykowski et al,35 2005 (United States)[leukemia and non-Hodgkin disease]

589 (62) 49 (10) 7 (2-23) Sex, age, andeducation

Registry-based cross-sectional survey/acquaintances survey

Taskila-Abrandt et al,36 2004 (Finland)[blood cancer]

902 (46) 2-3 Sex and age Registry-based retrospective cohort/population registry

Wettergren et al,39 2003 (Sweden)[Hodgkin disease]

90 (45) 47 (12) 14 (6-26) Age and residency Registry-based cross-sectional survey(for both groups)

Joly et al,47 1996 (France) [Hodgkin disease] 93 42 (23-85) 10 Sex, age, andresidency

Registry-based cross-sectional survey(for both groups)

van Tulder et al,48 1994 (the Netherlands)[Hodgkin disease]

81 (48) 44 (11) 14 Sex and age Hospital-based cross-sectional survey/non-cancer patients survey

Prostate cancerFarley Short et al,25 2008 (United States) 215 (0) NR (55-65) 4 (3-5) Sex Hospital-based cross-sectional/registry

Bradley et al,33 2005 (United States) 243 (0) 56 (6) 1 Sex and residency Registry-based cohort/populationsurvey

Taskila-Abrandt et al,36 2004 (Finland) 240 (0) 2-3 Sex and age Registry-based retrospective cohort/population registry

Testicular cancerGudbergsson et al,26 2007 (Norway) 244 (0) 45 (10) 1-5 Sex, age, and

residencyHospital-based cross-sectional survey/

population survey

Taskila-Abrandt et al,36 2004 (Finland) 206 (0) 2-3 Sex and age Registry-based retrospective cohort/population registry

Joly et al,42 2002 (France) 70 (0) 47 (29-67) 11 (5-20) Sex, age, andresidency

Registry-based cross-sectional survey(for both groups)

Abbreviation: NR, not reported.aBlank cells or cells not showing % women, SD, or range denote that data were not reported or available.

CANCER SURVIVORS AND UNEMPLOYMENT

756 JAMA, February 18, 2009—Vol 301, No. 7 (Reprinted) ©2009 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013

geal tumors, sarcoma, and 4 on mixeddiagnoses. Fourteen included articles*are from the United States; 2 of each arefrom Canada,37,45 the Netherlands,40,48

and France42,47; and 1 of each is fromSweden,39 Finland,36 Korea,24 Nor-way,26 Taiwan,44 and Australia.28 Theaverage age at diagnosis ranged be-tween 40 and 56 years. Mean follow-up time since diagnosis varied from 9months to 15 years. Excluding articleson breast cancer and cancers of themale and female reproductive system,the percentage of female patientsranged from 26% to 69%. Patients in12 articles were identified through ahospital-based registry,24-27,30,34,40,43-46,48

and patients in 13 articles were identi-

fied from a national or regional regis-try.† In 1 article, the patients werefollowed-up in a prospective cohort.31

In 12 articles,‡ students, homemak-ers, and retired individuals were iden-tified and excluded from analyses.

Study methodological quality isshown in TABLE 3. The MINORS qual-ity score ranged from 10 to 21 points.Only 10 (38%) of the articles includedemployment outcomes as part of themain study aim, while it was a second-ary outcome in the other studies. Acomparable control group with no his-tory of cancer and similar unemploy-ment risk was present in 7 (27%) of thearticles. Seventeen (65%) of the ar-

ticles included a control group withbaseline equivalence on age and sex.

Meta-analysesThe overall meta-analysis of the 36studies included 177 969 participants,composed of 20 366 cancer survivorsand 157 603 healthy control partici-pants. Cancer survivors were morelikely to be unemployed than healthycontrol participants (33.8% vs 15.2%;pooled RR, 1.37 [95% CI, 1.21-1.55];FIGURE 2). Additional meta-analysis bydiagnosis showed an increased risk ofunemployment for survivors of breastcancer (35.6% vs 31.7%; pooled RR,1.28 [95% CI, 1.11-1.49] Figure 2), gas-trointestinal cancers (48.8% vs 33.4%;pooled RR, 1.44 [95% CI, 1.02-2.05]),and cancers of the female reproduc-*References 1, 25, 27, 29-35, 38, 41, 43, 46.

†References 1, 28, 29, 32, 33, 35-39, 42, 47.‡References 1, 24, 28, 30, 34, 35, 39, 42-46.

Table 2. Characteristics of Included Studies for Other and Mixed Cancer Diagnoses

Source (Location)Cancer

Diagnosis

No. ofPatients

(% Women)

Patient Ageat Follow-up, Mean

(SD orRange), ya

Start ofFollow-up

AfterDiagnosis,

Mean(Range), ya

Characteristicsof Control

GroupMatching Study Design (Patient/Control)

Hewitt et al,1 2003 (United States) Mixed 2317 (63) 46 10 None National registry-based cross-sectional survey (for bothgroups)

Lee et al,24 2008 (Korea) Stomach 408 (26) 2 (2-3) None Hospital-based cross-sectionalsurvey/population survey

Eakin et al,28 2006 (Australia) Mixed 421 (68) 46 (18-) b Sex and age National registry-based cross-sectional survey (for bothgroups)

Sabatino et al,29 2006 (United States) Mixed 1710 (69) 47 (18-64) 5c None National registry-based cross-sectional survey (for bothgroups)

Bradley et al30 2006 (United States) Cervical 54 (100) 55 (12) 11 (6-20) Sex Hospital-based cross-sectionalsurvey/non–cancer patientssurvey

Taskila-Abrandt et al,36 2004 (Finland) Nervous system 878 (54) 2-3 Sex and age Registry-based retrospectivecohort/population registry

Taskila-Abrandt et al,36 2004 (Finland) Melanoma 853 (51) 2-3 Sex and age Registry-based retrospectivecohort/population registry

Taskila-Abrandt et al,36 2004 (Finland) Digestive tract 1153 (48) 2-3 Sex and age Registry-based retrospectivecohort/population registry

Taskila-Abrandt et al,36 2004 (Finland) Female reproductive 1265 (100) 2-3 Sex and age Registry-based retrospectivecohort/population registry

Taskila-Abrandt et al,36 2004 (Finland) Thyroid 629 (82) 2-3 Sex and age Registry-based retrospectivecohort/population registry

Yabroff et al,38 2004 (United States) Mixed 897 (62) 50 7 Sex, age, andeducation

National registry-based cross-sectional survey (for bothgroups)

Hoffman et al,43 2002 (United States) Sarcoma 28 (42) 43 15 (2-30) Sex and age Hospital-based retrospectivecohort/cross-sectionalsurvey

Fang et al,44 2002 (Taiwan) Nasopharyngeal 137 (34) 7 (2-) Sex and age Hospital-based cross-sectionalsurvey/non–cancer patientssurvey

aBlank cells or cells not showing SD or range denote that data were not reported or available.bReported as long-term survivors.cValue reported as median.

CANCER SURVIVORS AND UNEMPLOYMENT

©2009 American Medical Association. All rights reserved. (Reprinted) JAMA, February 18, 2009—Vol 301, No. 7 757

Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013

tive organs (49.1% vs 38.3%; pooled RR,1.28 [95% CI, 1.17-1.40]). The highestrisk for unemployment was identifiedamong survivors of nervous system can-cer (RR, 1.78 [95% CI, 1.58-1.99]) andnasopharyngeal cancer (RR, 2.47 [95%CI, 1.67-3.66]), but these involvedsingle studies only. Higher risks of un-employment compared with healthycontrol participants were not shownamong survivors of blood cancer(30.6% vs 23.7%; pooled RR, 1.41 [95%CI, 0.95-2.09]), prostate cancer (39.4%vs 27.1%; pooled RR, 1.11 [1.00-1.25]), and testicular cancer (18.5% vs18.1%; pooled RR, 0.94 [95% CI,0.74-1.20]).

Seven studies reported1,27,30,38-40,42 un-employment because of disability. Sub-group meta-analysis showed a higherrisk for unemployment because of dis-ability for cancer patients comparedwith control participants (RR, 2.84[95% CI, 1.91-4.20]).

Meta-regression

TABLE 4 summarizes results of the uni-variate and multivariate Bayesian meta-regression analyses. The crude univari-ate meta-RR for the United Statescompared with Europe shows a higherunemployment risk for cancer survi-vors in the United States (meta-RR, 1.48[95% CrI, 1.15-1.95]). The crude uni-

variate meta-RR for cancer survivors inthe US studies compared with all otherstudies in Europe, Canada, Australia, andAsia was 1.36 (95% CrI, 1.05-1.79). Thecrude meta-RRs of unemployment riskwere higher for all cancer diagnoses com-pared with testicular cancer (meta-RRs,1.21-1.58), but findings were not statis-tically significant. No differences in un-adjusted unemployment risk or unem-ployment risk adjusted for country, age,diagnosis, and background unemploy-ment rate were identified for older pa-tients (�50 years of age) compared withyounger ones (meta-RR, 0.99 [95% CrI,0.70-1.40] and meta-RR, 1.08 [95% CrI,0.70-1.69], respectively). Studies with a

Table 3. Methodological Quality Assessment Based on MINORSa

Source Aimb Ratec Datad Measuree Biasf Timeg Lossh Sizei Controlj Contempk Factorl Analysism TotalnAdjusted

Totalo

Hewitt et al1 1 0 0 1 1 2 0 0 1 2 2 2 12 3Lee et al24 2 2 1 2 2 2 1 0 0 2 0 2 16 2Farley Short et al25 2 1 1 2 2 2 1 0 2 0 1 2 16 5Gudbergsson et al26 2 2 2 2 2 2 1 0 2 2 2 0 19 7Majhail et al27 1 1 2 1 2 2 0 0 1 2 1 2 15 4Eakin et al28 1 0 0 2 2 0 0 0 2 2 2 2 13 4Sabatino et al29 1 0 0 2 2 2 0 1 0 2 0 0 10 0Bradley et al30 1 2 2 2 2 2 1 0 1 2 1 2 18 5Helgeson and Tomich31 2 1 2 2 2 2 2 0 1 2 1 0 17 6Bradley et al32 0 2 1 2 2 0 1 0 1 2 2 2 15 5Bradley et al33 2 2 1 2 2 0 1 2 1 2 2 2 19 5Syrjala et al34 2 2 2 2 2 2 1 0 1 2 2 2 20 6Andrykowski et al35 1 2 2 2 2 2 1 0 1 2 2 2 19 6Taskila-Abrandt et al36 2 2 1 2 2 2 2 0 2 2 2 2 21 7Maunsell et al37 2 1 1 2 2 2 2 0 2 2 2 2 20 7Yabroff et al38 2 0 0 2 2 0 0 0 2 2 2 2 14 4Wettergren et al39 1 2 0 2 2 2 1 0 1 2 2 2 17 4Servaes et al40 1 0 1 2 2 2 0 0 2 2 2 0 14 5Bradley41 2 0 0 2 2 2 0 0 1 2 0 2 13 1Joly et al42 1 2 2 2 2 2 1 0 1 2 2 2 19 6Hoffman et al43 1 2 2 2 2 2 1 0 1 2 0 0 15 4Fang et al44 1 0 2 2 2 2 0 0 1 2 2 0 14 5Dorval et al45 1 2 2 2 2 2 1 0 1 2 2 2 19 6Weitzner et al46 1 0 0 2 2 2 0 0 0 2 1 0 10 1Joly et al47 1 2 2 2 2 2 1 0 1 2 2 2 19 6van Tulder et al48 1 1 2 2 2 2 2 0 1 2 2 2 19 7aAssessed with the adapted Methodological Index for Non-Randomized Studies (MINORS).14

bClearly stated aim (0,1,2 points).c Inclusion of consecutive patients and response rate (0,1,2).dProspective collection of data (0,1,2).e Inclusion of employment measure (0,1,2).fUnbiased assessment of study end points (0 or 2).gFollow-up time appropriate (0,1,2).hLoss to follow-up (0,1,2).iProspective calculation of the study size (0 or 2).jComparable control group (0,1,2).kContemporary control groups (0 or 2).lBaseline equivalence of groups on several factors (0,1,2).mAdequate statistical analysis (0 or 2).nTotal: minimum equals 0; maximum equals 24 points.oAdjusted total: minimum equals 0; maximum equals 8 points.

CANCER SURVIVORS AND UNEMPLOYMENT

758 JAMA, February 18, 2009—Vol 301, No. 7 (Reprinted) ©2009 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013

Figure 2. Meta-analysis of Cancer Survivors vs Control Participants and Employment Outcomes

FavorsSurvivors

FavorsControls

101.00.1

RR (Random) 95% CI

Cancer Survivors

No.Weight,

%Total No.

Healthy ControlParticipants

No. Total No.RR (Random)

(95% CI)Source

Nervous system cancer483 3.55878 272 878Taskila-Abrandt et al,36 2004 1.78 (1.58-1.99)

Thyroid cancer189 3.43629 189 629Taskila-Abrandt et al,36 2004 1.00 (0.84-1.18)

Sarcoma8 1.2128 6 28Hoffman et al,43 2002 1.33 (0.53-3.35)

Nasopharyngeal cancer62 2.68137 26 142Fang et al,44 2002 2.47 (1.67-3.66)

Melanoma273 3.51853 290 853Taskila-Abrandt et al,36 2004 0.94 (0.82-1.08)

Prostate cancer168 3.54240 158 240Taskila-Abrandt et al,36 2004 1.06 (0.94-1.20)45 2.63243 35 256Bradley et al,33 2005 1.35 (0.90-2.03)62

275

3.27215 466

659

1933Farley Short et al,25 2008 1.20 (0.96-1.50)

9.44698 2429Subtotal 1.11 (1.00-1.25)6.0

I2, %Breast cancer

13 1.5848 10 85Weitzner et al,46 1997 2.30 (1.09-4.85)34 2.8854 36 114Dorval et al,45 1998 1.99 (1.42-2.80)72 3.42156 2083 5818Bradley et al,41 2002 1.29 (1.08-1.53)53 2.69150 25 78Servaes et al,40 2002 1.10 (0.75-1.63)

113 3.26567 131 890Maunsell et al,37 2004 1.35 (1.08-1.70)1598 3.644098 1434 4098Taskila-Abrandt et al,36 2004 1.11 (1.05-1.18)140 3.12445 59 372Bradley et al,32 2005 1.98 (1.51-2.60)100 3.27267 82 187Helgeson and Tomich,31 2005 0.85 (0.68-1.07)53 2.92269 65 382Gudbergson et al,26 2007 1.16 (0.83-1.61)83

2259

3.36289 514

9439

1969Farley Short et al,25 2008 1.10 (0.90-1.34)

30.146343 13 993Subtotal 1.28 (1.11-1.49)77.9

Blood cancers29 2.7581 44 114van Tulder et al,48 1994 0.93 (0.64-1.35)24 2.6093 52 186Joly et al,47 1996 0.92 (0.61-1.40)2 0.5990 12 177Wettergren et al,39 2003 0.33 (0.07-1.43)

429 3.56902 314 902Taskila-Abrandt et al,36 2004 1.37 (1.22-1.53)106 1.71589 8 126Adrykowski et al,35 2005 2.83 (1.42-5.66)27 2.37126 26 128Syrjala et al,34 2005 1.05 (0.65-1.70)43

660

1.49276 7

463

319Majhail et al,27 2007 7.10 (3.25-15.53)

15.072157 1952Subtotal 1.41 (0.95-2.09)82.3

Mixed cancer diagnoses389 3.592317 3874 77 489Hewitt et al,1 2003 3.36 (3.05-3.70)153 3.38897 238 2746Yabroff et al,38 2004 1.97 (1.63-2.38)19 2.39421 122 2579Eakin et al,28 2006 0.95 (0.59-1.53)

657

1218

3.631710 12 178

16 412

50 023Sabatino et al,29 2006 1.58 (1.48-1.68)

12.995345 132 837Subtotal 1.83 (1.13-2.96)98.3

Female reproductive system cancers641 3.601265 502 1265Taskila-Abrandt et al,36 2004 1.28 (1.17-1.39)

7

648

0.8654 4

506

57Bradley et al,30 2006 1.85 (0.57-5.96)

4.461319 1322Subtotal 1.28 (1.17-1.40)0

Gastrointestinal cancers618 3.601153 504 1153Taskila-Abrandt et al,36 2004 1.23 (1.13-1.33)99

717

3.28317 137

641

768Lee et al,24 2008 1.75 (1.40-2.19)

6.881470 1921Subtotal 1.44 (1.02-2.05)88.4

Testicular cancer6 1.1159 9 95Joly et al,42 2002 1.07 (0.40-2.86)

58 3.02206 64 206Taskila-Abrandt et al,36 2004 0.91 (0.67-1.22)30

94

2.49244 39

112

318Gudbergson et al,26 2007 1.00 (0.64-1.57)

6.62509 619Subtotal 0.94 (0.74-1.20)0

6886 24 015 100.00 94.320 366 157 603Overall 1.37 (1.21-1.55)

CI denotes confidence interval; RR, relative risk.

CANCER SURVIVORS AND UNEMPLOYMENT

©2009 American Medical Association. All rights reserved. (Reprinted) JAMA, February 18, 2009—Vol 301, No. 7 759

Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013

low background overall unemploy-ment rate showed lower risks for unem-ployment among cancer survivors com-pared with studies with higherbackground unemployment rates ( meta-RR, 0.24 [95% CrI, 0.11-0.54]).

In the multivariate analysis, the un-employment risk for cancer survivors inUS studies was not different comparedwith European studies after adjustmentfor diagnosis, age, and background un-employment risk (meta RR, 1.24 [0.85-1.83]). After adjustment for country, di-agnosis, and age, the adjusted meta RRfor the effect of a low background un-employment rate was not statistically sig-nificant at 0.38 (95% CI, 0.11-1.27).

When 11 low-quality studies identi-fied with the MINORS score were omit-ted, the overall RR for unemploymentamong cancer survivors compared withhealthy control participants decreased to1.22 (95% CI, 1.11-1.34). In the sensi-tivity analysis, which included the 25high-quality studies only, none of the ad-justed meta-RRs were statistically sig-nificant (Table 4). Meta-RRs, similar tothe adjusted meta-RRs of the high-quality studies in Table 4, were found inasecondsensitivity analysiswith26high-

quality studies identified with the ad-justed MINORs score.

Results of the Egger test revealed therewas no publication bias (P=.89). Re-sults of the Hedges and Pigott proce-dure for power analysis showed that thepower of our random-effects meta-analysis ranges from 0.77 to 0.98 for de-tecting a mean effect size above zero, witha mean effect size of 0.31 (based on therelative risk of 1.37) of the 36 studies.

COMMENTResults of this meta-analysis show thatcancer survivors are 1.37 times morelikely to be unemployed than healthycontrol participants. Increased risks forunemployment were identified for sur-vivors of breast cancer, gastrointestinalcancers, and cancer of the female repro-ductive organs. Survivors of blood can-cer, prostate cancers, and testicular can-cer did not have a higher unemploymentrisk. Survivors in the United States wereat 1.5 times higher risk of unemploy-ment than survivors in Europe. After ad-justment for diagnosis, age, and back-ground unemployment risk or afteromitting the low-quality studies, this as-sociation disappeared. No differences in

unemployment risk related to age werefound, but in studies with a relatively lowbackground unemployment rate, the riskfor unemployment for cancer patientswas lower compared with healthy con-trol participants than in studies per-formed in countries with a relatively highbackground unemployment rate.

This is the first study tosystematicallysummarizetheriskofunemployment foradults who survived cancer. A strengthof our study was that we were able to ex-tractdistinctunemploymentfiguresfromstudiesmeetingour inclusioncriteria. Incontrasttoearlierstudies,wesummarizedresults in one overall estimate of unem-ployment risk. In addition, we assessedassociationsofprognostic factorssuchasdiagnosis andcountryoforiginwithun-employment risk in a meta-regressionanalysis. Finally, we used a valid and re-liable indexforassessingmethodologicalquality.

Alimitationofthismeta-analysisisthatthequalityof includedstudiesvariedde-pending on study design and objectives.In some studies, unemployment rateswere secondary outcomes. Other stud-ies aimed to find the best estimate of theunemployment risk in cancer survivorsand included an age- and sex-matchedcontrol group. Furthermore, heteroge-neity in the meta-analyses remained af-ter dividing studies according to cancerdiagnosesandcountry.Even thoughthemeta-regression pointed at a strong in-fluence of country of origin and diagno-sis, the association of country of originwith unemployment was not significantafter adjusting for covariates, and theas-sociationof cancer typewithunemploy-mentwasnotsignificantafter low-qualitystudies were omitted from analyses. Agedid not have a clear association with un-employmentrisk.Weassumethat there-maining heterogeneity is caused by fac-torsother than thosecharacteristics thatwe could measure with the informationgiven in the studies.For instance, accessto health insurance associated with em-ployment may vary between countries,attention to unemployment issues mayvarybetweentreatmentcentersandhos-pitals, and occupational health servicesmay not be available in all countries.

Table 4. Results of Univariate and Multivariate Bayesian Meta-regression Models With Crudeand Adjusted Meta-relative Risks for Prognostic Factors

FactorNo. of

Studies

All Studies (n = 36) High-Quality Studiesa

No.

AdjustedMeta-RR (95%CrI) Multivariate

(n = 25)b

Crude Meta-RR(95% CrI)Univariate

AdjustedMeta-RR (95%

CrI) Multivariateb

CountryEurope 16 1 [Reference] 1 [Reference] 15 1 [Reference]United States 15 1.48 (1.15-1.95) 1.24 (0.85-1.83) 7 0.98 (0.66-1.56)Other 5 1.47 (1.03-2.12) 1.16 (0.68-1.96) 3 1.34 (0.85-2.27)

Cancer diagnosisTesticular 3 1 [Reference] 1 [Reference] 3 1 [Reference]Breast 10 1.35 (0.76-2.37) 1.20 (0.65-2.22) 6 1.15 (0.67-1.85)Prostate 3 1.21 (0.62-2.39) 1.17 (0.55-2.47) 3 1.28 (0.67-2.31)Blood 7 1.42 (0.77-2.64) 1.38 (0.79-2.50) 6 1.27 (0.74-2.04)Other or mixed 13 1.58 (0.90-2.75) 1.48 (0.87-2.56) 7 1.35 (0.84-2.09)

Patient age18-50 y 23 1 [Reference] 1 [Reference] 15 1 [Reference]�50 y 8 0.99 (0.70-1.40) 1.08 (0.70-1.69) 6 1.03 (0.72-1.46)Not reported 5 1.19 (0.81-1.77) 1.10 (0.70-1.77) 4 1.09 (0.69-1.65)

Backgroundunemploymentrate

36 0.24 (0.11-0.54) 0.38 (0.11-1.27) 25 0.63 (0.21-1.99)

Abbreviations: CrI, credibility interval; RR, relative risk.aHigh quality denotes 16 points or greater on the MINORS test.bAdjusted for all the other variables in the model.

CANCER SURVIVORS AND UNEMPLOYMENT

760 JAMA, February 18, 2009—Vol 301, No. 7 (Reprinted) ©2009 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013

Even though the risk of unemploy-ment is higher in cancer survivors, theunemployment rate among survivorsacross all studies was 34%. Thus, mostcancer survivors remain employed af-ter their cancer diagnosis, but there isopportunity for improving unemploy-ment rates among cancer survivors.

Several possible mechanisms may ex-plain the higher unemployment ratesamong cancer survivors. First, cancersurvivors may be less available for thelabor market. To address this, we ex-cluded all homemakers, students, andretired individuals from our analyseswhen possible, in an attempt to in-clude only those who were available forand seeking employment. Second,health is an important determinant oflabor participation even though this re-lationship might be complex.56 Eighty-six percent of cancer patients (97/113) vs 76% of unemployed controlparticipants (100/131) stated that thedecision to stop working was their ownin the study of Maunsell et al.37 In thesame study, 47% (53/113) of the un-employed breast cancer survivors statedthat they were unemployed because ofhealth reasons compared with 18% (24/131) of the unemployed control par-ticipants. In 5 other studies,1,24,32,33,42 rea-sons for unemployment were provided.More often for patients than for con-trol participants, reasons for unem-ployment were physical limitations,cancer-related symptoms, or both. Fur-thermore, voluntary unemployment isnot likely unless patients have other re-sources for income,57 which is not thecase for most cancer survivors. Fi-nally, in the 7 studies1,27,30,38-40,42 in ourreview that measured unemploymentand disability, the relative risk of re-ceiving a disability benefit or other-wise being disabled for work was al-most 3 times higher for survivorscompared with control participants.Therefore, the mechanism behind thehigher unemployment rate among can-cer survivors is likely to be a higher dis-ability rate.

Except for blood cancers, the re-sults of our meta-analysis are similar tothose of earlier cohort studies on pre-

dictors of return to work,58-62 in whichpatients with breast cancer58,59 or tu-mors of the head and neck59,60 showedmore difficulty and men with testicu-lar cancer59-62 less difficulty returningto work than patients with other can-cer diagnoses.

An interesting finding was a higherunadjusted risk for unemploymentamong cancer survivors in the UnitedStates compared with Europe. The ef-fects of health on labor force partici-pation are likely to be socially deter-mined because they are the result of theinteraction of the labor market, jobsearch behavior, economic incentives,and health insurance.56 Almost all ofthese factors differ between the UnitedStates and Europe and could explain thedifference. However, after adjustmentfor diagnosis, age, and background un-employment risk, or after exclusion oflow-quality studies, the higher unem-ployment risk for cancer survivors inthe United States vs Europe disap-peared. Therefore, the difference canalso be due to other determinants orconfounding by low-quality studies.Half of the studies from the UnitedStates (8 of 15) and 1 from Europe (outof 16) were of low quality. Most of theexcluded American studies had higherrisk estimates and this might havecaused the reduction of the adjustedrisk in the sensitivity analysis.

Future Research

More and better research on the workimpact of cancer is necessary to in-form the decisions of cancer survivorsand the clinicians who provide theirtreatment. Better estimates of the riskof unemployment from high-qualitystudies are needed to enhance the iden-tification of prognostic factors and vul-nerable subgroups for unemploymentamong cancer survivors. Future re-search should involve large patientsamples, include matched controlgroups, focus on different cancer diag-noses, compare differences betweencountries, and analyze a number ofprognostic variables including socio-economic and cultural factors on un-employment rates.

Apart from the effects on employ-ment, there are probably long-term ef-fects of cancer on work ability,63 workcapacity,64 and wage losses64,65 for a largegroup of survivors. Employment out-comes can be improved with innova-tions in treatment and with clinical andsupportive services aimed at better man-agement of symptoms, rehabilitation,and accommodation for disabilities.Moreover, workplace interventions areneeded that are aimed at realizing work-place accommodations and paid sickleave during treatment. The develop-ment and evaluation of such interven-tions is urgently needed because theycould mitigate the economic impact ofsurviving cancer and improve the qual-ity of life for survivors.

In conclusion, cancer survivors areat an increased risk of unemployment,especially survivors of breast cancer,gastrointestinal cancers, and cancers ofthe female reproductive organs and can-cer patients living in countries or timeswith relatively high unemploymentrates. Development of interventions in-volving clinicians and other profession-als to enhance participation in the worklife of cancer survivors is needed.Author Contributions: Dr de Boer had full access to allthe data in the study and takes responsibility for the in-tegrity of the data and the accuracy of the data analysis.Study concept and design: de Boer, van Dijk, Verbeek.Acquisition of data: de Boer, Taskila, Verbeek.Analysis and interpretation of data: de Boer, Taskila,Ojajärvi, van Dijk, Verbeek.Drafting of the manuscript: de Boer, Taskila, Verbeek.Critical revision of the manuscript: de Boer, Ojajärvi,van Dijk, Verbeek.Statistical analysis: de Boer, Ojajarvi, Verbeek.Administrative, technical, material support: van Dijk.Study supervision: de Boer, van Dijk, Verbeek.Data extraction and quality assessment: de Boer,Taskila, Verbeek.Financial Disclosures: Dr Taskila reports that her con-tribution was funded by the Finnish Work Environ-ment Fund. The other authors report no financial dis-closures.Role of the Sponsor: The Finnish Work EnvironmentFund is a nonprofit governmental funding organiza-tion. The Finnish Work Environment Fund had no rolein the design and conduct of the study; collection, man-agement, analysis, and interpretation of the data; andpreparation, review, or approval of the manuscript.

REFERENCES

1. Hewitt M, Rowland JH, Yancik R. Cancer survi-vors in the United States. J Gerontol A Biol Sci MedSci. 2003;58(1):82-91.2. Feuerstein M. Optimizing cancer survivorship. J Can-cer Surviv. 2007;1(1):1-4.3. Smith T, Stein KD, Mehta CC, et al. The rationale,design, and implementation of the American Cancer

CANCER SURVIVORS AND UNEMPLOYMENT

©2009 American Medical Association. All rights reserved. (Reprinted) JAMA, February 18, 2009—Vol 301, No. 7 761

Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013

Society’s studies of cancer survivors. Cancer. 2007;109(1):1-12.4. Spelten ER, Sprangers MAG, Verbeek JH. Factorsreported to influence the return to work of cancersurvivors. Psychooncology. 2002;11(2):124-131.5. Taskila T, Lindbohm ML. Factors affecting cancersurvivors’ employment and work ability. Acta Oncol.2007;46(4):446-451.6. Short PF, Vasey JJ, Tuncelli K. Employment path-ways in a large cohort of adult cancer survivors. Cancer.2005;103(6):1292-1301.7. Verbeek J, Spelten E. Work. In: Feuerstein M, ed.Handbook of Cancer Survivorship. New York, NY:Springer Science and Business Media; 2007.8. Hoffman B. Cancer survivors at work. CA CancerJ Clin. 2005;55(5):271-280.9. Kennedy F, Haslam C, Munir F, Pryce J. Returningto work following cancer. Eur J Cancer Care (Engl).2007;16(1):17-25.10. Main DS, Nowels CT, Cavender TA, EtschmaierM, Steiner JF. A qualitative study of work and workreturn in cancer survivors. Psychooncology. 2005;14(11):992-1004.11. Nachreiner NM, Dagher RK, McGovern PM, BakerBA, Alexander BH, Gerberich SG. Successful return towork for cancer survivors. AAOHN J. 2007;55(7):290-295.12. Park JH, Park EC, Park JH, Kim SG, Lee SY. Jobloss and re-employment of cancer patients inKorean employees. J Clin Oncol. 2008;26(8):1302-1309.13. de Boer AG, Verbeek JH, van Dijk FJ. Adult sur-vivors of childhood cancer and unemployment: ametaanalysis. Cancer. 2006;107(1):1-11.14. Slim K, Nini E, Forestier D, Kwiatkowski F, PanisY, Chipponi J. Methodological index for non-randomized studies (MINOR). ANZ J Surg. 2003;73(9):712-716.15. DerSimonian R, Laird N. Meta-analysis in clinicaltrials. Control Clin Trials. 1986;7(3):177-188.16. Higgins JP, Thompson SG, Deeks JJ, Altman DG.Measuring inconsistency in meta-analyses. BMJ. 2003;327(7414):557-560.17. Congdon P. Applied Bayesian Modelling.Chichester, England: John Wiley and Sons; 2003.18. Gelman A, Carlin JB, Stern HS, Rubin DB. BayesianData Analysis. 2nd ed. Boca Raton, FL: Chapman &Hall/CRC; 2003.19. Ojajärvi A, Partanen T, Ahlbom A, et al. Estimat-ing the relative risk of pancreatic cancer associated withexposure agents in job title data in a hierarchicalBayesian meta-analysis. Scand J Work Environ Health.2007;33(5):325-335.20. DuMouchel WH. Bayesian meta-analysis. In: BerryDA, ed. Statistical Methods in Pharmaceutical Sci-ences. New York, NY: Marcel Dekker; 1990:509-529.21. Spiegelhalter D, Thomas A, Best N. WinBUGS UserManual, Version 1.3. Cambridge, England: MedicalResearch Council Biostatistics Unit; 1999.22. Egger M, Davey Smith G, Schneider M, MinderC. Bias in meta-analysis detected by a simple, graphi-cal test. BMJ. 1997;315(7109):629-634.23. Hedges LV, Pigott TD. The power of statistical testsin meta-analysis. Psychol Methods. 2001;6(3):203-217.24. Lee MK, Lee KM, Bae JM, et al. Employment sta-tus and work-related difficulties in stomach cancer sur-vivors compared with the general population. Br JCancer. 2008;98(4):708-715.25. Farley Short P, Vasey JJ, Moran JR. Long-term ef-fects of cancer survivorship on the employment of olderworkers. Health Serv Res. 2008;43(1 pt 1):193-210.26. Gudbergsson SB, Fossa SD, Sanne B, Dahl AA.

A controlled study of job strain in primary-treated can-cer patients without metastases. Acta Oncol. 2007;46(4):534-544.27. Majhail NS, Ness KK, Burns LJ, et al. Late effectsin survivors of hodgkin and non-hodgkin lymphomatreated with autologous hematopoietic cel ltransplantation. Biol Blood Marrow Transplant. 2007;13(10):1153-1159.28. Eakin EG, Youlden DR, Baade PD, et al. Healthstatus of long-term cancer survivors. Cancer Epide-miol Biomarkers Prev. 2006;15(10):1969-1976.29. Sabatino SA, Coates RJ, Uhler RJ, Alley LG, PollackLA. Health insurance coverage and cost barriers toneeded medical care among U.S. adult cancer survi-vors age�65 years. Cancer. 2006;106(11):2466-2475.30. Bradley S, Rose S, Lutgendorf S, Costanzo E,Anderson B. Quality of life and mental health in cer-vical and endometrial cancer survivors. Gynecol Oncol.2006;100(3):479-486.31. Helgeson VS, Tomich PL. Surviving cancer: a com-parison of 5-year disease-free breast cancer survi-vors with healthy women. Psychooncology. 2005;14(4):307-317.32. Bradley CJ, Neumark D, Bednarek HL, Schenk M.Short-term effects of breast cancer on labor marketattachment. J Health Econ. 2005;24(1):137-160.33. Bradley CJ, Neumark D, Luo Z, Bednarek H, SchenkM. Employment outcomes of men treated for pros-tate cancer. J Natl Cancer Inst. 2005;97(13):958-965.34. Syrjala KL, Langer SL, Abrams JR, Storer BE, MartinPJ. Late effects of hematopoietic cell transplantationamong 10-year adult survivors compared with case-matched controls. J Clin Oncol. 2005;23(27):6596-6606.35. Andrykowski MA, Bishop MM, Hahn EA, et al.Long-term health-related quality of life, growth, andspiritual well-being after hematopoietic stem-celltransplantation. J Clin Oncol. 2005;23(3):599-608.36. Taskila-Abrandt T, Martikainen R, Virtanen SV,Pukkala E, Hietanen P, Lindbohm ML. The impact ofeducation and occupation on the employment statusof cancer survivors. Eur J Cancer. 2004;40(16):2488-2493.37. Maunsell E, Drolet M, Brisson J, Brisson C, MasseB, Deschenes L. Work situation after breast cancer.J Natl Cancer Inst. 2004;96(24):1813-1822.38. Yabroff KR, Lawrence WF, Clauser S, Davis WW,Brown ML. Burden of illness in cancer survivors. J NatlCancer Inst. 2004;96(17):1322-1330.39. Wettergren L, Bjorkholm M, Axdorph U, BowlingA, Langius-Eklof A. Individual quality of life inlong-term survivors of Hodgkin’s lymphoma—a com-parative study. Qual Life Res. 2003;12(5):545-554.40. Servaes P, Verhagen CA, Bleijenberg G. Rela-tions between fatigue, neuropsychological function-ing, and physical activity after treatment for breastcarcinoma. Cancer. 2002;95(9):2017-2026.41. Bradley CJ, Bednarek HL, Neumark D. Breast can-cer survival, work, and earnings. J Health Econ. 2002;21(5):757-779.42. Joly F, Heron JF, Kalusinski L, et al. Quality of lifein long-term survivors of testicular cancer. J Clin Oncol.2002;20(1):73-80.43. Hoffman RD, Saltzman CL, Buckwalter JA. Out-come of lower extremity malignancy survivors treatedwith transfemoral amputation. Arch Phys Med Rehabil.2002;83(2):177-182.44. Fang FM, Chiu HC, Kuo WR, et al. Health-related quality of life for nasopharyngeal carcinomapatients with cancer-free survival after treatment.

Int J Radiat Oncol Biol Phys. 2002;53(4):959-968.45. Dorval M, Maunsell E, Deschenes L, Brisson J,Masse B. Long-term quality of life after breast cancer.J Clin Oncol. 1998;16(2):487-494.46. WeitznerMA,MeyersCA,StuebingKK,SaleebaAK.Relationship between quality of life and mood in long-termsurvivorsofbreastcancer treatedwithmastectomy.Support Care Cancer. 1997;5(3):241-248.47. Joly F, Henry-Amar M, Arveux P, et al. Late psy-chosocial sequelae in Hodgkin’s disease survivors.J Clin Oncol. 1996;14(9):2444-2453.48. van Tulder MW, Aaronson NK, Bruning PF. Thequality of life of long-term survivors of Hodgkin’sdisease. Ann Oncol. 1994;5(2):153-158.49. Bradley CJ, Neumark D, Oberst K, Luo Z, BrennanS, Schenk M. Combining registry, primary, and sec-ondary data sources to identify the impact of canceron labor market outcomes. Med Decis Making. 2005;25(5):534-547.50. Bradley CJ, Neumark D, Luo Z, Schenk M. Em-ployment and cancer. Cancer Invest. 2007;25(1):47-54.51. Bradley CJ, Bednarek HL, Neumark D. Breast can-cer and women’s labor supply. Health Serv Res. 2002;37(5):1309-1328.52. Drolet M, Maunsell E, Brisson J, Brisson C, MasseB, Deschenes L. Not working 3 years after breast cancer.J Clin Oncol. 2005;23(33):8305-8312.53. Drolet M, Maunsell E, Mondor M, et al. Work ab-sence after breast cancer diagnosis: a population-based study. CMAJ. 2005;173(7):765-771.54. Taskila-Abrandt T, Pukkala E, Martikainen R,Karjalainen A, Hietanen P. Employment status of Finn-ish cancer patients in 1997. Psychooncology. 2005;14(3):221-226.55. Short PF, Vasey JJ, Belue R. Work disability as-sociated with cancer survivorship and other chronicconditions. Psychooncology. 2008;17(1):91-97.56. Currie J. Health, health insurance and the labormarket. In: Ashenfelter O, Card D, eds. Handbook ofLabor Economics. Amsterdam, the Netherlands: ElsevierScience BV; 1999:3309-3415.57. Sawyer M, Spencer D. On the definition of in-voluntary unemployment. J Socio-Econom. 2008;37:718-735 doi:10.1016/j.socec.2006.12.049.58. Razavi D, Delvaux N, Bredart A, et al. Profes-sional rehabilitation of lymphoma patients. SupportCare Cancer. 1993;1(5):276-278.59. Weis J, Koch U, Geldsetzer M. Changes in occu-pational status following cancer. Soz Praventivmed.1992;37(2):85-95.60. van der Wouden JC, Greaves-Otte JG, GreavesJ, Kruyt PM, van Leeuwen O, van der Does E. Occu-pational reintegration of long-term cancer survivors.J Occup Med. 1992;34(11):1084-1089.61. Spelten ER, Verbeek JH, Uitterhoeve AL, et al. Can-cer, fatigue and the return of patients to work-a pro-spective cohort study. Eur J Cancer. 2003;39(11):1562-1567.62. Bloom JR, Hoppe RT, Fobair P, Cox RS, VargheseA, Spiegel D. Effects of treatment on the work expe-riences of long-term survivors of Hodgkin’s disease.J Psychosoc Oncol. 1988;6:65-80.63. de Boer AG, Verbeek JH, Spelten ER, et al. Workability and return-to-work in cancer patients. Br JCancer. 2008;98(8):1342-1347.64. Steiner JF, Cavender TA, Nowels CT, et al. Theimpact of physical and psychosocial factors on workcharacteristics after cancer. Psychooncology. 2008;17(2):138-147.65. Lauzier S, Maunsell E, Drolet M, et al. Wage lossesin the year after breast cancer. J Natl Cancer Inst. 2008;100(5):321-332.

CANCER SURVIVORS AND UNEMPLOYMENT

762 JAMA, February 18, 2009—Vol 301, No. 7 (Reprinted) ©2009 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013