Psychosocial Outcomes and Health-Related Quality of Life in Adult Childhood Cancer Survivors: A Report from the Childhood Cancer Survivor Study

Adulthood is now attainable for most children treated for cancer with 5-year survival rates at 80% (1). Increased survival may be accompanied by long-term burden for some individuals related to the unique characteristics of their cancer diagnoses and treatment and the effect of treatment on their educational, psychological, and social development. Identification of subgroups of childhood cancer survivors at risk for poor health-related quality of life (HRQOL) is important for development of intervention strategies. As treatments evolve, determination of treatments promoting long-term survival and reduced risk for poor HRQOL outcomes will be a goal.

Findings in survivors are often derived from single-institution oncology programs or long-term follow-up clinics and suggest that psychosocial status and HRQOL are relatively good for most survivors, yet less favorable outcomes have been reported (2-6). Childhood Cancer Survivor Study (CCSS) reports have examined psychosocial outcomes and HRQOL within specific diagnostic groups (7-11). However, it is necessary to examine the entire cohort of childhood cancer survivors, compare them with siblings, and identify demographic and treatment risk factors so that risk-based treatments to enhance HRQOL of life can be developed.

This report describes psychosocial status, HRQOL, and life satisfaction in the CCSS cohort, compares findings to sibling controls and population norms, and examines broad demographic and treatment factors associated with poor outcomes among survivors. We have used siblings as the major comparison group for survivors because the demographic match to the survivors is closer in the sibling group than in cohorts from which the population norms were derived (12). This article, by painting the big picture, will provide the background for future, more detailed, within-diagnosis analyses.

The CCSS cohort, a resource funded by the National Cancer Institute, represents the largest and most comprehensively characterized group of childhood cancer survivors in North America. Details of study design and cohort characteristics have been described elsewhere (13, 14). Subjects were recruited from persons treated for an initial diagnosis of leukemia, central nervous system (CNS) malignancy, Hodgkin's disease, non–Hodgkin's lymphoma (NHL), kidney cancer, neuroblastoma, soft tissue sarcoma, or malignant bone tumor at 1 of 26 institutions across the United States and Canada, survivors >5 years, and diagnosed between 1970 and 1986 when younger than 21 years of age. Among the 20,267 eligible individuals, 17,273 were located; 14,024 (81.2%) survivors and 3,846 siblings were enrolled. Comparison of participants, nonparticipants, and those who were lost to follow-up showed similar demographic, disease, and treatment characteristics (14, 15). Of eligible survivors and siblings 18 years of age or older at contact and available for the second follow-up survey, 9,307 (84%) survivors and 2,875 (80%) siblings participated. The psychosocial portion of this survey was sent to all survivors and a randomly selected subsample of 500 siblings, with 7,147 (77%) survivors and 388 (78%) siblings participating. The study was approved by Institutional Review Boards of all participating institutions with participants providing informed consent.

Emotional health was evaluated by the 18-item Brief Symptom Inventory-18 (BSI-18), which includes symptoms over the previous 7 days. The BSI-18 has been validated in healthy volunteers (16), in cancer patients (17), and in an earlier administration with this cohort of cancer survivors (18). The BSI-18 has a summary scale (the global distress index) and three subscales (depression, anxiety, and somatization). Raw scores were converted to T-scores based on U.S. population norms and dichotomized using a cutpoint of 63. Those with T-scores ≥63 were classified as having poor emotional health (17, 18).

The Medical Outcomes Short Form-36 (SF-36) was used to evaluate HRQOL. Participants answered 36 questions about general health, well-being, and quality of life over the previous 4 weeks. The SF-36 has two summary scales and eight individual subscales representing different aspects of well-being (19). Data are presented as T-scores, with a general population mean of 50 and SD of 10. Higher scores indicate “better” HRQOL (19). T-scores were dichotomized for the multiple variable models. Participants with T-scores at least 1 SD below the population mean (≤40) were classified as reporting poor HRQOL.

Life satisfaction was determined by having participants complete the Cantril Ladder of Life (LOL). The LOL assesses respondents' life satisfaction with three self-report items that indicate life satisfaction in the past, present, and future. Ratings are made on a 10-point scale ranging from “best possible life” to “worst possible life” (20), providing a global rating of life satisfaction that has been used in both population studies and clinical survivor samples (21-23).

For these analyses, we considered sex, age at diagnosis and interview, length of follow-up, race/ethnicity, marital status, educational attainment, annual household income, and health insurance as independent variables in our multiple variable models. Major medical condition was included as a covariate in the models and was derived from information provided by participants about medical late effects. Participants who reported complete deafness, kidney dialysis, congestive heart failure, myocardial infarction, angioplasty, bypass surgery, stroke, liver cirrhosis, a heart, lung, or kidney transplant, amputation, joint replacement or second cancer, and/or current use of seizure medications, medications for heart problems or high blood pressure, chemotherapy, immune suppressants, or oxygen were classified as having a major medical condition (24). Diagnosis, treatment, and related information were obtained from medical records.

Descriptive statistics were calculated for demographic and treatment variables and compared between survivors and siblings with generalized estimating equations (25). To account for sibling-survivor pairs, mixed models were also used to calculate and compare age- and gender-adjusted mean scores between siblings and survivors overall and, by diagnosis, on subscales and summary scales for the BSI, the SF-36, and the LOL (26). Mean scores on the BSI-18 and SF-36 were compared among study participants and age- and gender-specific population norms with one-sample t tests. Frequencies and percents of dichotomized outcomes for the psychosocial, HRQOL, and life satisfaction scales were calculated and compared among survivors using both demographic and treatment characteristics as predictors in multiple variable logistic regression models. Statistical Analysis System version 9.1 was used for analysis. Bonferroni corrections (n = 17) were used to reduce errors related to multiple tests, with α levels set at 0.003.

Participants were more likely than nonparticipants to be female, older than 24 years of age, white, college educated, married, and employed than were nonparticipants (data not shown). Survivor participants did not differ from nonparticipants by cancer diagnosis, survival time, or baseline BSI scores. As expected, based on their random selection for participation in the psychosocial portion of the questionnaire, sibling participants did not differ from nonparticipants by sex, age, race, educational attainment, employment, marital status, or scores on the BSI at baseline.

Demographic characteristics of participating survivors and siblings are shown in Table 1. Survivors had a median age of 32 (18-54) years and siblings had a median age of 33 (18-58) years. Among survivors, median age at diagnosis was 7 (0-20) years and median survival time was 23 (15-34) years. Cases were more likely than siblings to be younger, white, single, not employed, and not insured; have lower income; and have a major medical condition.

Age- and gender-adjusted means and SEs from generalized linear mixed models on each of the BSI, SF-36, and LOL scales are shown in Table 2A to C.

Emotional Health (BSI-18). Compared with siblings, survivors reported more symptoms of global distress, depression, anxiety, and somatization, although scores are lower (better) than population norms for both groups. Survivors of acute lymphoblastic leukemia, astrocytomas, CNS tumors other than medulloblastoma, Hodgkin's disease, NHL, Wilm's tumor, soft tissue sarcoma, Ewing's sarcoma, and osteosarcoma reported higher levels of global distress. Survivors of astrocytoma, NHL, Ewing's sarcoma, and osteosarcoma reported significantly higher depressive symptoms, anxiety, and somatization. Generally, survivors and siblings had fewer symptoms than the general population, with survivors of astrocytoma reporting more depression and survivors of Hodgkin's disease reporting more somatization than population norms.

HRQOL (SF-36). Whereas survivors and siblings scored higher than norms in mental health, survivors scored lower than population norms on all other aspects of HRQOL, except for pain and the mental component summary, and lower than siblings on the physical component summary of the SF-36; however, sibling and survivor means did not differ significantly on the mental component summary, a pattern persisting for survivors across diagnostic groups, except for survivors of other or unspecified leukemia and bone cancer other than Ewing's sarcoma or osteosarcoma. General health subscale means were also lower for survivors than siblings. Survivors of CNS tumors, lymphoma, soft tissue, or bone malignancies reported more problems in physical function, role physical, general health, and social function domains than did siblings. Bone cancer survivors also reported significant bodily pain. In addition, survivors of osteosarcoma had lower mean scores than siblings on the role emotional and mental health subscales. Survivors of astrocytoma also scored lower on the mental health subscale, and survivors of NHL scored lower on the vitality subscale when compared with siblings. Compared with age-relevant U.S. population norms, survivors had poorer outcomes on physical function, role physical, general health, vitality, role emotional, and social function domains as well as physical component but had better outcomes on the mental health subscale. This pattern persisted for survivors across diagnostic groups, except for those with leukemia, Wilm's tumor, neuroblastoma, and bone cancer other than Ewing's sarcoma or osteosarcoma. Compared with norms, siblings had better outcomes on physical function, general health, and mental health domains and poorer outcomes on the vitality domain.

Life Satisfaction (LOL). Overall, survivors and siblings reported high levels of current and predicted life satisfaction, except for CNS tumor survivors who predicted lower levels of satisfaction 5 years into the future than did siblings. Both survivors and siblings reported better current and future life satisfaction than did the general population.

Frequencies and percents of having poor psychosocial or quality of life among survivors by sociodemographic characteristics are shown in Table 3A and B with odds ratios (OR) and 95% confidence intervals (95% CI).

Emotional Health (BSI-18). In adjusted models, female gender, lower educational attainment, unemployment, annual household income <$20,000, lack of health insurance, being African-American, and having a major medical condition were associated with increased risk for reporting symptoms of depression, anxiety, somatization, and global distress. Current age over 35 years was also associated with higher risk of reporting depressive symptoms. Compared with those who were married/living as married, unmarried survivors were more likely to report depressive symptoms and global distress. Divorced or separated individuals also reported more symptoms of anxiety when compared with those who were married/living as married. Hispanics were more likely than Whites to report symptoms of somatization and global distress.

HRQOL (SF-36). Female gender, unemployment, annual household income <$20,000, and lack of health insurance were associated with poor HRQOL across all summaries and subscales of the SF-36. Participants with a major medical condition were more likely than those without to report poor HRQOL across most domains, with the exception of mental health subscale and mental component summary scale. Older age and lower educational attainment were associated with increased risk of reporting poor physical HRQOL (physical component summary, general health, physical function, role physical, and bodily pain subscales), less vitality, and poor social function.

Compared with Whites, Hispanics were at increased risk of reporting poor outcomes on role physical and social function subscales. Other ethnic minorities (non-Black and non-Hispanic) were at increased risk of reporting poor HRQOL in role emotional and mental health domains. Compared with those who were married/living as married, unmarried survivors were at increased risk of reporting poor outcomes on role emotional, social function, and mental health subscales. When compared with those who were married/living as married, divorced or separated individuals also reported an increased prevalence of problems in role physical, general heath, and vitality domains.

Frequencies and percents of having a poor psychosocial or quality of life outcome among survivors by treatment characteristics are shown in Table 4A and B with ORs and 95% CIs from unconditional logistic regression models.

Emotional Health (BSI-18). In adjusted models, the only predictor of poor emotional health was previous treatment with cranial radiation, with only modest strength. Survivors treated with cranial radiation were at a slightly higher risk of reporting depressive symptoms compared with those without cranial radiation.

HRQOL (SF-36). After adjusting for gender, age at diagnosis and survival time, and treatments with surgery, chemotherapy, and cranial radiation therapy, were each associated with higher risk of reporting poor HRQOL in physical function, role physical, and general health subscales and on the physical component summary. In multiple variable models, history of surgery was associated with elevated risk of poor HRQOL in physical function, role physical, bodily pain, vitality, role emotional, and social function subscales and in physical and mental components when compared with those without surgical history, although association magnitude was modest, except for physical function (OR, 1.9; 95% CI, 1.5-2.5). Survivors who received chemotherapy compared with those without reported poor HRQOL only in general health. Younger age at diagnosis was associated with decreased risk of poor HRQOL in physical function, role physical, bodily pain, and general health subscales and in the physical component summary. Survivors >25 years reported better HRQOL in physical function, role physical, bodily pain, general health, and physical component summary compared with survivors >30 years.

Whereas adult survivors of childhood cancer from the CCSS report more symptoms of psychological distress compared with siblings, both groups report better psychological adjustment than that of population norms. Survivors report poorer HRQOL in physical and social but not in mental health domains compared with siblings. Compared with U.S. population norms, survivors have increased risk of poor HRQOL in physical and social domains, with the largest effect size in vitality. Compared with population norms, siblings fare better in physical function and general health, and both survivors and siblings report better mental health than does the general population, as well as high levels of current life satisfaction. Except for CNS tumor survivors, both groups also expect high life satisfaction in 5 years. Our findings add to information from previous CCSS estimates of health status (24) and chronic disease (27). Despite their increased risk for poor health status, and the fact that >70% of childhood cancer survivors report a chronic health condition, survivors' expectation of future life satisfaction is excellent.

Survivors of CNS tumors, lymphoma, bone, and soft tissue sarcomas have the lowest mean HRQOL scores. Female gender, lower educational attainment, unmarried status, annual household income <$20,000, unemployment, lack of medical insurance, and having a major medical condition are all associated with a higher risk of psychological distress and poor outcomes in some aspects of HRQOL. Previous treatment with cranial radiation and surgery are associated with poor HRQOL, particularly in physical domains. Participants diagnosed <10 years of age or treated in the past 25 years report better HRQOL than those diagnosed >10 years of age or treated >30 years ago.

Our findings support the low mean psychological distress found in a study of 161 childhood cancer survivors (28, 29) on the SCL-90-R (30), the parent instrument of the BSI-18, and yet are consistent with another investigation (31) of 101 survivors where ∼32% screened positive for psychological distress on the SCL-90-R, with a history of cranial radiation and poor physical health as predictors. Our results extend earlier CCSS findings of increased depression and somatization in survivors of leukemia and lymphoma (9), brain tumor (8), and solid tumors (10) compared with siblings. Our findings are also consistent with earlier reports of associations between poor HRQOL and female gender (32-35), older age (35), lower educational attainment (33), unemployment (33), and medical limitations (33). Others found being survivors >26 years (32) and treatment with cranial radiation or surgery (34, 35) as risks for poor physical HRQOL. Younger age at diagnosis and decreased risk of poor physical HRQOL in our study contradicts another study of 2,152 childhood cancer survivors (35) that used the Health Utilities Index Mark III, a discrepancy possibly explained by differences in the focus of the Health Utilities Index Mark III versus SF-36.

Although our findings are based on a large sample, varied diagnoses, and participants from 26 children's cancer centers throughout North America, they should not be considered without taking into account potential study limitations. First, we used only self-report data to characterize outcomes. Clinical interviews may have improved the precision of our estimates, particularly those related to psychological distress. Second, the SF-36 is a generic HRQOL measure. To investigate the physical and psychological functioning of these survivors more systematically, more specific measures may be needed. Change over time with longitudinal repeated measures assessments is needed to determine stability of our findings and the developmental trajectory of HRQOL in this population. In addition, the survivor participants in our study were more likely than nonparticipants to be female, white, married, employed, and college graduates. It is possible that these successful survivors were more aware of their levels of achievement and thus either more or less likely to express dissatisfaction with their outcome.

Based on our data, we could interpret the findings of our study as good news. Overall, most of the measures indicate that survivors and siblings both report lower psychological distress and better HRQOL, as well as life satisfaction, in comparison with population norms. Where there were differences, generally effect sizes were small, except for the vitality subscale as noted above. In fact, our sample seems to report significantly better mental health than that of population norms, with a large effect size. We could explain our findings as “posttraumatic growth” or the psychological resilience that develops in coping with adverse circumstances (36, 37). That is, experiencing childhood cancer or being a young sibling of a sister or brother with cancer inoculates individuals to other negative life experiences and provides them with feelings of life satisfaction and overall psychological well being. In one study, lower level of worry during cancer treatment was related to ultimate life satisfaction (37). On the other hand, there are some data indicating that most people show life satisfaction even with very poor health states, such as amputation (38, 39), and several studies indicating that survivors of cancers in general and childhood cancers in particular may be biased reporters and tend to deny difficulties and overestimate their positive health and satisfaction (40-44). O'Leary et al. (45) reported from a sample of Dana-Farber patients participating in CCSS that survivors had a strong tendency toward an enhanced self-appraisal and such bias had a differential effect on self-reported physical and emotional quality of life.

Unlike smaller cohort studies, our large sample permits analyses that identify vulnerable subgroups and suggest specific diagnostic/treatment pathways to psychological symptoms and poor HRQOL. Understanding reasons for pain in bone tumor survivors or poor future outlook in CNS tumor survivors might lead to earlier preventive strategies, such as better pain management during bone cancer treatment or early educational interventions for children with CNS tumors (46). Rather than implying poor HRQOL in all survivors, we have identified certain survivor groups with more severe effect and suggest the need for prospective, longitudinal, in-depth examination of pathways and mediators of psychological distress and poor HRQOL. Studies of these high-risk subgroups can inform strategies designed to reduce psychosocial and HRQOL sequelae.