Employment pathways in a large cohort of adult cancer survivors

Abstract

BACKGROUND

Employment and work-related disability were investigated in a cohort of adult cancer survivors who were working when they were diagnosed from 1997 to 1999 with a variety of cancers. Employment from the time of diagnosis through the early years of survivorship was studied, self-reported effects of cancer survival on disability and employment were quantified, and risk factors associated with cancer-related disability and withdrawal from employment were identified.

METHODS

One thousand four hundred thirty-three cancer survivors were interviewed by telephone from 1 year to nearly 5 years after diagnosis. They were asked retrospectively about employment from the time of diagnosis to follow-up and about work-related disability at follow-up. They also were asked whether disabilities or reasons for quitting work were cancer-related. Return to work and quitting work were projected over time in a life-table analysis. Risk factors were identified from logit analyses.

RESULTS

One of five survivors reported cancer-related disabilities at follow-up. Half of those with disabilities were working. A projected 13% of all survivors had quit working for cancer-related reasons within 4 years of diagnosis. More than half of survivors quit working after the first year, when three-quarters of those who stopped for treatment returned to work. Survivors of central nervous system, head and neck, and Stage IV blood and lymph malignancies had the highest adjusted risk of disability or quitting work.

CONCLUSIONS

Cancer survival sometimes has long-term effects on employment and the ability to work. Employment outcomes can be improved with innovations in treatment and with clinical and supportive services aimed at better management of symptoms, rehabilitation, and accommodation of disabilities. Cancer 2005. © 2005 American Cancer Society.

The employment consequences of surviving cancer have potential importance both for society and for cancer survivors. With roughly half of adult cancer survivors age < 65 years, many are at an age at which the effects of cancer and its treatment could alter their employment opportunities and choices.1 Cancer survivors who quit or cut back on work may suffer financially if their lost earning are not replaced by other sources of income or if they lose access to employment-related health insurance. Survivors who leave work may lose in psychological and social terms as well, considering the social connections afforded by work and its ties to self concept, self esteem, and life roles and satisfaction.2, 3 Some survivors may work in spite of cancer-related disabilities, perhaps to retain employer-sponsored health insurance, to replace income lost during treatment, or to cover expenses and protect against financial uncertainties associated with survivorship. For society, the total economic burden of cancer is increased by the lost productivity of survivors who quit working, reduce their hours, or have disabilities that limit the kind of work they can do.

Previous research points to a variety of ways that cancer survival may affect employment adversely over the long term, including physical limitations,4, 5 fatigue,6 emotional problems,6 difficulties with concentration and memory,5–7 awkward or negative interactions with coworkers,6, 8 and changing personal priorities.8, 9 Employment discrimination is also an ongoing concern, although the Americans with Disabilities Act (ADA) prohibits discrimination against cancer survivors, and there is little published evidence of widespread discrimination after the passage of the ADA in 1993.5

A recent review concluded that efforts to quantify employment concerns associated with cancer survival have been hampered by small sample sizes, lack of longitudinal data, and an inability to account for differences by cancer site.6 We address those weaknesses in this report on employment during the 4 years after initial diagnosis in a cohort of > 1400 survivors who were working when they were diagnosed with cancer. The survivors were identified in a larger, longitudinal survey of (working and nonworking) survivors who were diagnosed over 3 calendar years with most types of cancer. The survey asked retrospectively about employment from diagnosis through the day of interview, and survivors who quit their jobs were asked whether the reasons were cancer-related. We also report on limitations in ability to work that were cancer-related, including limitations of survivors who still were working when they were interviewed. The study's three main goals were 1) to describe the time path of employment in a cohort of survivors who were working at the time of diagnosis, 2) to quantify work disability and reductions in employment that survivors associated with cancer and its treatment, and 3) to identify risk factors associated with cancer-related disability and withdrawal from work.

MATERIALS AND METHODS

Sample

Surviving patients who were first diagnosed with cancer from January, 1997 through December, 1999 were identified from the tumor registries of 4 medical centers in Pennsylvania (Milton S. Hershey Medical Center, Geisinger Medical Center, and Lehigh Valley Hospital) and Maryland (The Johns Hopkins Hospital [Johns Hopkins]). Eligibility was limited to adults ages 25–62 years at diagnosis. The lower age limit was designed to avoid the complications of schooling for the labor force participation of younger adults. The upper limit was the earliest age at which retirees can receive Social Security benefits. All survivors who met these criteria were recruited for the larger study, except as noted below.

All types of cancer were included, with three exceptions. First, skin cancers other than melanoma were excluded. Second, most patients with Stage IV cancers at diagnosis were excluded.10 Only leukemias, plasma cell cancers, and lymphomas were included that commonly are classified as Stage IV at diagnosis but have relatively favorable survival rates. Finally, because urologic cancers were not under the administrative control of the registry at Johns Hopkins, patients with prostate and testicular cancers were excluded there. Table 1 shows that 5150 patients were targeted for recruiting, including ≈ 3000 from the Pennsylvania hospitals and 2000 from Johns Hopkins.

Table 1. Recruiting and Interviewing Outcomes in the Main Study

Recruiting/interviewing variable	No. of participants
Selected for recruiting	5150
Discovered ineligible for study (includes deaths)	340
Eligible, but no consent	2734
Unable to locate or contact	230
Refusal	2157
Recruiting ended before consent was obtained	347
Consent obtained	2076
Ineligible for interview	75
Discovered ineligible for study	71
Died before interview	4
Eligible, but no interview	238
Language problem	6
Physically incapacitated	11
Unable to locate or contact	103
Refusal	118
Interview completed	1763

The research protocol was approved by human subjects committees at each hospital and at Penn State's main campus. Each hospital obtained informed consent, verbally at three sites and in writing at one site, with support and training from the research team. Survivors were not identified to the research team until consent was obtained. Recruiting began in September, 2000 and ended at 3 sites by May, 2001. Recruiting at Johns Hopkins ended with 347 potential participants still in process, just before the second round of interviews began. Recruiting identified 340 patients at all sites who were ineligible because of age, a cancer diagnosed before 1997, or death. Consent was obtained from 2076 participants (43% of eligible patients).

This article focuses on the survivors who were working at the time of diagnosis and is based on data collected in the first of four annual interviews. In total, 1763 patients (1433 who were working at diagnosis) were interviewed initially from October, 2000 to December, 2001, from 1 year to nearly 5 years after their first diagnosis of cancer. The median length of follow-up for the working sample was 32 months from diagnosis. Twenty-three percent of the working sample was interviewed in the second year of survivorship, 35% in the third year, 32% in the fourth year, and 10% in the fifth. After accounting for deaths and other ineligible patients discovered in the interviews, the overall survey response rate (conditioned on eligibility and consent) was 88%.

Data

Many of the questions in the computer-assisted telephone interview were drawn from the Health and Retirement Study (HRS). HRS is an ongoing national cohort study of older Americans, originally ages 51–61 years, that began in 1992 and is conducted by the University of Michigan with funding primarily from the National Institute on Aging.11 In addition to asking about each individual's situation at the time of the interview, the cancer questionnaire asked retrospectively about employment at the time of diagnosis and during treatment and ascertained the month and year of changes in employment status. Participants who left jobs after diagnosis were asked whether the reasons were cancer-related. Participants who could not be interviewed in English were excluded from the study.

Clinical details regarding the site and stage of cancer were obtained from the registries. The date of diagnosis initially was obtained from the registries and then was updated in the interview, primarily to correct for lags between initial diagnosis and referral to the medical centers. Participants were asked about the status of their cancer and answered a limited set of questions to identify those still in treatment for their first cancer (the one that was diagnosed on the date in question). The prevalence of a small group of chronic illnesses at the time of interview was determined by asking whether a physician had ever told participants that they had diabetes, chronic lung disease, heart problems, a stroke, or arthritis or rheumatism. Like in the HRS, work disability was determined by asking participants whether they had any impairments or health problems at the time of interview that limited the kind or amount of paid work they could do. If so, then they were asked whether the problem was related to their cancer and whether they were able to work at all. Physically demanding jobs at the time of diagnosis were identified from the job title or description of work. Physically demanding jobs are those that rely predominantly on physical activity or strength rather than on cognitive or social abilities and skills.

Univariate statistics for the survivors who were working at the time of diagnosis are provided in Table 2. With some initial interviews occurring early in the second year of survivorship, the research design allowed the possibility of interviewing some participants before the end of routine treatment. Although that phenomenon could not be measured directly, the survey responses suggested that it occurred infrequently. Twenty-one percent of the participants reported that they still were being treated for their first cancer, but the minority had active cancer (Table 2). Only 57 interviewees in the second year of survivorship still were being treated for their first cancer, including 25 patients with active cancer. Fifty-eight percent of participants who said that they still were being treated for their first cancer and had active cancer also reported a metastasis or new cancer.

Table 2. Characteristics of Survivors Who Were Working at Diagnosis

Characteristic	Column %			Characteristic	Column %
	Total	Male	Female		Total	Male	Female
CNS: central nervous system.
Total no.	1433	516	917	Age at interview (yrs)
Cancer site				25–44	24.0	19.8	26.4
Blood	5.3	8.1	3.7	45–52	27.0	23.8	28.8
Breast	31.4	0.2	49.0	53–57	21.5	22.9	20.7
CNS	3.0	3.7	2.6	58–61	13.5	14.1	13.2
Colorectal	6.6	10.9	4.1	≥62	11.0	16.7	7.9
Head and neck	4.1	6.8	2.6	Undetermined	2.9	2.7	3.1
Lymph	5.5	8.9	3.6	Race
Prostate	8.3	23.1	0.0	White	90.1	92.2	88.9
Respiratory	3.6	3.5	3.7	Nonwhite	6.9	5.0	8.0
Melanoma	4.0	5.2	3.4	Undetermined	3.0	2.7	3.2
Thyroid	6.9	5.6	7.6	Married at diagnosis
Uterus	5.3	0.0	8.3	Married/partner	79.7	86.6	75.8
Other	15.9	24.0	11.3	Single	20.3	13.4	24.2
Stage (diagnosis)				Education
I	41.7	29.5	48.5	<High school	4.0	4.7	3.7
II	32.4	36.2	30.2	High school	30.9	31.2	30.8
III	15.5	18.2	14.0	Some college	21.9	17.1	24.6
IV	8.0	12.6	5.5	College	20.4	23.6	18.5
Unstaged	2.4	3.5	1.9	≥College	19.9	20.9	19.3
Follow up (mos)				Undetermined	2.9	2.5	3.1
12–23	22.7	22.5	22.8	Children <6 yrs
24–35	33.6	34.1	33.4	Yes	6.3	8.1	5.2
36–47	29.6	30.6	29.9	No	93.7	91.9	94.8
≥48	9.4	10.3	8.9	Occupation
Undetermined	2.8	2.5	2.9	Physically demanding	17.4	30.8	9.8
Treatment status				Other	81.4	67.8	89.0
Finished treatment	80.2	82.2	79.1	Undetermined	1.3	1.4	1.2
Initial treatment, active cancer	7.5	7.6	7.5	Chronic conditions
Initial treatment, inactive cancer	12.3	10.3	13.4	Yes	42.4	42.1	42.5
Any new cancer				No	57.3	57.9	56.9
Yes	14.0	14.3	13.8	Undetermined	0.3	0.0	0.5
No	85.3	84.9	85.6	Medical center
Undetermined	0.6	0.8	0.5	Geisinger	17.8	21.7	15.6
Undetermined	0.6	0.8	0.5	Hershey	16.7	18.2	15.8
Undetermined	0.6	0.8	0.5	Johns Hopkins	39.8	32.6	43.8
Undetermined	0.6	0.8	0.5	Lehigh Valley	25.8	27.5	24.8

Analysis

Because of the right-censoring associated with differences in the length of follow-up after diagnosis, we made life-table estimates (with PROC LIFETEST in SAS software) of the time path that characterized return to work from treatment. We also used this technique for time paths that described the probability of quitting work because of cancer separately for survivors who continued working during treatment and for survivors who returned to work within the first year. Log-rank tests showed no significant differences in the time paths according to gender.

Multivariate analyses that controlled for length of follow-up were undertaken to identify independent predictors of 1) cancer-related disability and 2) quitting between diagnosis and follow-up for reasons related to cancer. Ordered logit was used to analyze disability, because there were three ordered categories (no limitations in ability to work, limitations but able to work, unable to work at all). Standard dichotomous logit was used to analyze the probability of not working because of cancer at follow-up, corresponding to the time when disability was measured. The latter model analyzed the combined effect of the two dynamic processes described in the life-table analyses: returning to work (or failing to return to work) after stopping for treatment and quitting after continuing or returning to work.

An important objective of the multivariate analysis was to compare adjusted odds ratios (ORs) by cancer site. Sites were defined according to topologic groups of International Classification of Diseases for Oncology (2nd revision) diagnostic codes assigned by the registries. Colorectal cancer, which was the most prevalent cancer that was well distributed across genders, was chosen as the reference site. ORs and their confidence intervals (CIs) were graphed to allow readers to visualize the relative magnitudes by site. A logarithmic scale was used to make the CIs symmetric. Sites with risks that differed significantly from the reference site, colorectal cancer, were identified by CIs > 1 or < 1 in the graphs. Significant differences between other pairs of sites could not be identified from nonoverlapping CIs in the graphs, because such ORs are not independent statistically in logit models. Consequently, a Wald chi-square (with 1 degree of freedom) was used to test for significant differences in all other pair-wise comparisons.

Because Stage IV diagnoses were associated only with blood and lymph cancers as a consequence of the study design, survivors of Stage IV cancers were identified by the interaction of stage and site in the models. Separate indicators were assigned to Stage IV blood cancers, Stage IV lymph cancers, and other lymph cancers. Virtually all of the blood cancers were diagnosed as Stage IV; therefore, we assigned other stages to the residual category of “other cancers” and refer to the remaining Stage IV blood cancers as simply “blood cancers.”

Likelihood ratio tests rejected pooling males and females in the disability model (P = 0.0384), but not in the model for quitting work (P = 0.1797). A pooled disability model with female interactions for all variables was used to test for differences in specific coefficients by gender.

In view of the relatively low rate of participation in the study, we assessed potential participation biases with a logit model of participation that was estimated from patient characteristics available from the registries for the 4735 survivors who were eligible for the main study. Participation varied significantly by facility and was significantly higher for females compared with males, whites compared with nonwhites, and for some cancer sites (breast, Stage IV lymph, blood, prostate, thyroid) compared with others (melanoma, head and neck, uterus, central nervous system), as detailed in materials available from the authors. Although they were statistically significant in such a large sample, the differences in participation were not associated strongly with study outcomes. Consequently, two different tests detected no significant participation biases related to employment, disability, or other quality-of-life outcomes.

RESULTS

Eighty-eight percent of the male survivors and 78% of the female survivors in the main survey were working at the time of diagnosis. Similar percentages of males and females who were working at the time of diagnosis stopped during cancer treatment (41% and 39%, respectively). According to life-table projections through 4 years of survivorship, most of the survivors who would return to work did so in the first year (Fig. 1). The projected rate of return to work after 4 years was 84%. Of the survivors who returned to work in the first year, 11% would quit for cancer-related reasons in the next 3 years (Fig. 2). Nine percent of the survivors who continued working during treatment would quit for reasons related to cancer within 4 years, with 2–3% continuing to leave work in each year of survivorship after the first year (Fig. 2).

Figure 1. This chart illustrates the cumulative percentage of survivors who returned to work based on life-table projections of data collected in different years of survivorship.

Figure 2. This chart illustrates the cumulative percentage of survivors who quit working for reasons related to cancer based on life-table projections of data collected in different years of survivorship.

One survivor out of 5, 21% of the females and 16% of the males who were working at diagnosis, reported limitations in ability to work that were related to cancer (Fig. 3). The higher total rate of cancer-related disability for females compared with males was statistically significant (P = 0.0047). With follow-up of ≥ 1 year after diagnosis, the disability rate did not vary significantly by year of follow-up (P = 0.2862). Among survivors with any work-related disabilities, almost half again as many were able to work as could not work at all. Fifty-three percent of survivors had completed their initial treatment and were disease-free. Forty-eight percent of those with any cancer-related disability were working at follow-up.

Figure 3. This chart illustrates the percent distribution of cancer survivors who were working at the time of diagnosis by disability at follow-up.

Figure 4 shows adjusted ORs by cancer site from multivariate logit models of cancer-related disability and quitting work. The sites are ordered in each graph by the estimated ORs. Controlling for other variables shown in Table 3, the 4 sites with the highest disability rates among women (central nervous system, blood, head and neck, and Stage IV lymph) differed significantly from the 4 sites with the lowest rates (uterus, female breast, melanoma, and thyroid) in all pair-wise comparisons. Among men, significant differences in disability rates mainly involved the two sites with the highest rates (central nervous system cancers and blood cancers) and the two sites with the lowest rates (prostate cancer and head and neck cancers). The gender difference in disability rates for head and neck cancers was statistically significant. The disability rate for prostate cancer was significantly lower than the rates among men for all sites except melanoma, thyroid, and head and neck.

Figure 4. Adjusted odds ratios and 95% confidence intervals for cancer sites from logit models of work disability and quitting work related to cancer (logarithmic scale). Adjusted odds ratios are from the models shown in Table 3 and control for all of the variables listed there. CNS: central nervous system.

Table 3. Adjusted Odds Ratios from Logit Models of Disability and Quitting Work Related to Cancer

Variablea	Work-related disability				Quit working: Both male and female (n = 1372)
	Male (n = 494)		Female (n = 875)		OR	95%CI
	OR	95%CI	OR	95%CI
OR: odds ratio; 95%CI: 95% confidence interval; CNS: central nervous system; dx: diagnosis. a The reference group was colorectal cancer survivors with Stage I disease at diagnosis, interviewed 12–23 months after diagnosis, age <45, male, white, not married at diagnosis, no children age <6 years, less than a high school education, job at diagnosis not physically demanding, and no chronic conditions. b P < 0.01. c P < 0.05. d P < 0.10. e Female > male: P < 0.05. f Female < male: P < 0.05.
Blood	2.181	0.645–7.376	5.303b	1.706–16.487	3.030b	1.238–7.416
Female breast			0.621	0.249–1.552	0.440c	0.203–0.952
CNS	3.574d	0.922–13.861	5.765b	1.629–20.400	2.202	0.785–6.179
Head and neck	0.197ce	0.046–0.833	3.392ce	1.018–11.299	1.713	0.684–4.293
Lymph other stages	0.784	0.174–3.528	0.909	0.192–4.308	0.899	0.268–3.014
Lymph Stage IV	1.284	0.268–6.156	3.254	0.773–13.700	0.963	0.260–3.566
Prostate	0.107b	0.031–0.365			0.485d	0.181–1.300
Respiratory	0.603	0.146–2.488	1.468	0.468–4.607	1.218	0.473–3.139
Melanoma	0.581	0.128–2.642	0.623	0.155–2.509	0.763	0.259–2.250
Thyroid	0.372	0.065–2.115	0.752	0.242–2.337	0.602	0.216–1.673
Uterus			0.583	0.184–1.850	0.377	0.117–1.212
Other cancers	0.875	0.362–2.116	1.272	0.484–3.343	0.789	0.370–1.679
Stage II	2.654c	1.18–5.947	2.169b	1.389–3.386	1.529d	0.969–2.413
Stage III	2.473c	1.052–5.815	2.982b	1.765–5.040	2.444b	1.487–4.019
Unstaged	1.279	0.324–5.055	0.684	0.182–2.570	0.927	0.310–2.773
24–35 mos after dx	0.416c	0.203–0.849	0.745	0.460–1.208	0.830	0.522–1.320
36–47 mos after dx	0.179bc	0.077–0.414	0.987a	0.613–1.588	0.750	0.464–1.212
≥48 mos after dx	0.398d	0.147–1.073	0.736	0.372–1.456	1.078	0.582–1.997
In initial treatment, active cancer	3.298b	1.352–8.044	5.498b	2.954–10.236	3.496b	2.025–6.038
In initial treatment, inactive cancer	1.914	0.768–4.767	1.396	0.783–2.490	1.118	0.613–2.037
Any new cancer	3.522bf	1.700–7.293	1.367f	0.823–2.270	2.101b	1.340–3.294
Age
45–52 yrs	2.903c	1.129–7.463	1.979c	1.177–3.329	1.081	0.631–1.853
53–57 yrs	1.931	0.689–5.410	1.289	0.716–2.320	1.080	0.603–1.935
58–61 yrs	1.673	0.515–5.439	1.435	0.760–2.710	1.268	0.677–2.376
≥62 yrs	2.270	0.688–7.495	0.859	0.365–2.019	1.631	0.819–3.245
Female					1.715c	1.084–2.713
Nonwhite	1.700	0.569–5.075	0.801	0.410–1.564	0.793	0.409–1.539
Married at dx	0.545a	0.244–1.218	1.497de	0.969–2.311	1.015	0.657–1.569
Children age <6 yrs	1.564	0.491–4.983	0.689	0.267–1.777	1.693	0.797–3.598
Education
High school	1.558	0.439–5.523	0.677	0.286–1.601	0.664	0.312–1.411
Some college	1.857	0.472–7.308	1.092	0.453–2.627	0.883	0.403–1.936
College	1.334	0.314–5.674	1.013	0.405–2.532	0.570	0.247–1.316
≥College	1.584	0.367–6.843	0.660	0.255–1.708	0.211b	0.082–0.542
Physically demanding job at dx	3.987b	1.910–8.322	2.588b	1.440–4.651	1.296	0.801–2.096
Any chronic conditions	1.855c	1.023–3.361	1.781b	1.220–2.601	2.011b	1.385–2.920

With respect to quitting work because of cancer, the differences between the three sites with the highest rates (blood, central nervous system, head and neck) and the four sites with the lowest rates (uterus, female breast, prostate, and thyroid) were significant in all pair-wise comparisons. Survivors of blood cancers were significantly more likely to quit work for cancer-related reasons than survivors of all cancers except for central nervous system, head and neck, respiratory, and Stage IV lymph cancers. Despite the gender difference in disability rates for head and neck cancers, females were no more likely than males to quit work because of head and neck cancers.

The rates of disability and quitting work also varied by other clinical factors (Table 3). Disability rates for both genders and the quitting rate were higher for survivors who were still in initial treatment for active cancer. Disability and quitting rates for survivors who described themselves as still in treatment for a first cancer that was inactive were similar to the rates for survivors who were no longer in treatment. New cancers or metastases increased the likelihood of quitting work (OR, 2.101; P = 0.0012) and disability among males (OR, 3.522; P = 0.0007), but the association with disability was weaker among females. Stage at diagnosis was significant in all three models, with more advanced disease subsequently presenting greater employment challenges for survivors. Cancer survivors with other chronic health conditions were more likely to report work disabilities related to cancer (male: OR, 1.855; P = 0.0417; female: OR, 1.781; P = 0.0028) and were more likely to quit working for cancer-related reasons (OR, 2.011; P = 0.0002).

Among survivors who were working at the time of diagnosis, there were relatively few socioeconomic or demographic differences in the risk of disability or quitting work related to cancer. Survivors of both genders who were ages 45–52 years at follow -up were more likely to report cancer-related disabilities than younger survivors, but that age group was no more likely to quit working. Married women were more likely to report cancer-related disabilities than married men (P = 0.0298), and women were more likely to quit work (OR, 1.715; P = 0.0212). Survivors in physically demanding jobs had higher disability rates (male: OR, 3.987; P = 0.0002; female: OR, 2.588; P = 0.0015) but were no more likely to quit work (P = 0.2914). Survivors with any postgraduate education were less likely to quit working than any other educational group.

DISCUSSION

In keeping with Fitzhugh Mullan's characterization of the “seasons of cancer survival,”12 a central objective of the current study was to describe the employment of cancer survivors in terms of a time path from diagnosis. Such a pathway explicitly allows for continuing risks to employment, encompassing not only return to work after cancer treatment but, for example, long-term and late-term effects on retirement. The pathways described here, which were projected through the first 4 years of survivorship from the initial interviews in a 4-year cohort study, eventually will be extended for 3 more years.

The projections imply that 13% of survivors who were working at diagnosis quit for cancer-related reasons in the first 4 years of survivorship. Those who failed to return to work after treatment accounted for 5 percentage points of the total (40% stopped working × 12% who did not return because of cancer = 4.8%). Those who quit after returning in the first year accounted for another 3 percentage points (40% stopped working × 73% returned in Year 1 × 11% later quit for cancer-related reasons = 3.2%). Those who quit after working through treatment accounted for 5 percentage points (60% continued working × 9% quit for cancer-related reasons = 5.4%). More than half of the reduction in employment occurred after the first year.

Information resources for cancer survivors and health professionals usually emphasize three employment topics: reintegration into the workplace after treatment, legal protections against discrimination, and legal rights to accommodation during treatment.13–15 They give relatively little attention to the other issues raised herein. Focusing on return to work and reintegration after treatment ignores survivors who quit working at a later stage of survivorship, including some who return to work after treatment. To emphasize that most cancer survivors suffer no long-term disability is important in combating employment discrimination but offers little support or comfort to the one worker out of five at diagnosis who indeed does have cancer-related disabilities. Some survivors confront problems related to employment that go beyond discrimination and may occur or persist over the long term.

Admittedly, the employment pathways estimated in this study may not be representative of the experiences of recent cancer survivors in general. The sample was regional, not national, and was limited in the number of minorities. Educational attainment and the employment rate at follow-up were higher in the full cohort of nearly 1800 recent survivors than for the prevalent population of all adult survivors age < 65 years in the United States.16 In addition, the low consent rate in recruiting participants introduced the possibility of nonparticipation bias. However, a formal assessment identified no significant biases in employment and disability rates related to observable differences between participants and nonparticipants in the main study. With the positive effect of advanced education (over-represented in this study) and the negative effect of physically demanding occupations (under-represented) on employment, the findings herein probably are conservative with respect to the average effect of cancer survival on disability and employment.

The length of recall also is a potential study limitation in the current study, because some participants were asked nearly 5 years later about their employment at diagnosis and during treatment. The length of recall was a compromise between the years of survivorship that could be described from the study and the years of longitudinal data collection. The research design assumed that cancer diagnosis and treatment are such salient, life-altering events that most respondents would be able to recall their employment status in relation to them. Finally, despite a sample that was much larger than most in the literature, small samples and wide confidence intervals for subgroups were sometimes limitations in comparing employment risks across cancer types and other risk factors.

Nevertheless, the apparent effects of cancer survival on employment in our study generally are consistent with the literature. Two recent, cross-sectional studies that compared long-term breast cancer survivors with similar women without breast cancer found positive and significant, but somewhat smaller effects of cancer survival on employment than the average for all sites reported here, in keeping with the lower relative odds of quitting work for female breast cancer compared with some other cancers.4, 17 Our reported rates of return to work, 73% after 1 year and 84% after 4 years, are higher than the average reported in a 2002 literature review (62%).6 However, the 10 studies considered in that review varied with respect to length of follow-up and type of cancer, and the rates ranged widely from 30% to 93%. Above-average educational attainment and employment at diagnosis in our study likely contributed to a higher rate of return to work.

In addition to providing an aggregate view of employment in a survivor cohort, our study identified risk factors associated with differences in disability and quitting work within the cohort. Knowledge of these risk factors can help to prioritize efforts to develop effective treatments with fewer negative consequences for survivors. It also can help to target clinical and psychosocial services that address threats to their employment.

The statistical analyses identified blood and lymph cancers that were Stage IV at diagnosis, central nervous system cancers, and head and neck cancers as particularly problematic for employment. Nervous system cancers are likely to interfere with perception, cognition, or movement, which are critical functions in most jobs. Head and neck cancers have been associated with lower than average rates of return to work in earlier studies.6, 18 The much higher rate of disability for women with head and neck cancer compared with men warrants further investigation. Perhaps the long-term consequences of head and neck cancers—sometimes including difficulty speaking and highly visible effects of surgical excision on appearance—are more consequential for women than for men. The systemic treatment of Stage IV blood cancers and lymphomas involves more generalized and, potentially, more damaging treatment with radiation, chemotherapy, or high-dose chemotherapy and stem cell transplantation than the localized treatment of early-stage solid tumors.

In conclusion, the results of this study point to both short-term and long-term challenges for cancer survivors and the field of oncology. Survivors are at risk for disabilities and quitting work even after the first year of survivorship, when most of those who stop for treatment return to work. The short-term challenge for oncology is to identify survivors with employment problems at all stages of survivorship and then help them with a comprehensive range of clinical and supportive services aimed at better management of symptoms, rehabilitation, and accommodation of disabilities. The long-term challenge for oncology is to devise new treatments that will leave even fewer survivors with work-related disabilities.

Acknowledgements

The authors gratefully acknowledge the assistance of Witold Rybka and Margaret Davitt (Milton S. Hershey Medical Center), James Evans and Sharon Reeder (Geisinger Medical Center), Gregory Harper and Andrea Geshan (Lehigh Valley Hospital and Health Network), and James Zabora and Barbara Abdullah (The Johns Hopkins Hospital). CODA, Inc. of Silver Spring, Maryland, conducted the interviews. The authors also are indebted to the study's Patient Advisory Committee and to Frank Lawrence and Cathy Bradley for suggestions concerning the article.