Participants included 1,652 cancer survivors and a control group of 406 siblings of cancer survivors from the Childhood Cancer Survivor Study (CCSS). The CCSS is a cohort study designed to evaluate the impact of childhood cancer and its treatment on long-term health and function.[16
] Eligible participants were treated for one of eight childhood cancer diagnoses at 26 institutions between 1970 and 1986 when younger than 21 years of age. Cohort entry was limited to those individuals who survived for at least five years after their original diagnosis. Siblings were recruited from a randomly selected subset of adolescent and adult survivors. The human subjects committee at each of the collaborating institutions approved the study protocol before participant enrollment. Participants provided informed consent for the questionnaires and medical record abstraction. Study participants completed a 24-page Baseline questionnaire between 1995 and 1996, at least five years post their original diagnosis. This questionnaire surveyed sociodemographic information, medical history, and functional limitations. A second Follow-Up questionnaire was initiated in 2003 (the full survey questionnaires are available at http://www.stjude.org/CCSS
The study population for the current analyses included (1) all cancer survivors and siblings whose parents participated in the Baseline survey when the survivor or sibling was 12-17 years of age, and (2) survivors or siblings who also completed the 2003 Follow-Up survey as adults approximately seven years later. Exclusion criteria included survivors and siblings who were over 17 years of age at the Baseline survey. Participants who reported experiencing a cerebrovascular abnormality, cerebral palsy, paralysis, mental retardation, or epilepsy were also excluded as these neurological conditions were perceived to have the potential to impact health status and behaviors independent of the predictor variables (e.g. requiring parents to regulate health behaviors). In addition, given the inclusion of physical activity and body mass index (BMI) outcomes in this study, four additional survivors with lower limb amputations were also excluded. See for a schematic of patient recruitment.
Schematic of recruitment of adolescents used in the analyses of this longitudinal study.
The Behavior Problem Index (BPI) was completed by parents as part of the Baseline survey. The BPI is a standardized questionnaire developed for the National Health Survey, and has been normed on a large nationally representative sample.[19
] This rating scale was developed as a shortened version of the Child Behavior Checklist,[20
] and has been widely used in medical populations and epidemiological studies. For each of the 32 items, parents are asked to rate their child's behavioral and/or emotional functioning using a Likert scale ranging from 1 (“Not True”) to 3 (“Often True”). Factor analyses in the normative sample resulted in five scales: depression/anxiety, headstrong behavior, attention deficit, social withdrawal, and antisocial behavior.[19
] This original factor structure has been cross-validated in a large cohort of adolescent survivors of childhood cancer.[14
] For the current study, participants were classified according to whether or not they displayed impairment on each of these factors, with impairment defined as having a level of problematic symptoms that occur in ≤ 10% of the original normative standardization sample.[19
Adult weight status and health behavior outcomes were collected during the 2003 Follow-Up survey. Body mass index (BMI) was calculated to indicate weight status. In order to adjust for adolescent BMI, height and weight was also obtained during the Baseline survey and was compared to national norms,[21
] and classified as follows: underweight = BMI < 5th
percentile, normal weight = BMI 5th
percentile, overweight = BMI 85th
percentile; and obese = BMI ≥ 95th
percentile. During adulthood BMI classifications were defined as follows: underweight = BMI < 18.5, normal weight = BMI 18.5–24.9; overweight = BMI 25–29.9; obese = BMI ≥ 30. Health behavior variables included physical activity, smoking, and sunscreen use. Health behaviors were selected based on those behaviors associated with increased risk for chronic health impairment and/or secondary neoplasm. Physical activity was measured in weekly minutes of moderate and vigorous physical exercise, and was dichotomized according to whether or not the participant met national standards established by the Centers for Disease Control (CDC), with “Inactivity” defined as falling below the recommended guidelines.[22
]. Adult participants were identified as a smoker if they reported currently smoking and smoking at least 100 cigarettes over their lifetime. Poor sunscreen use was defined as reporting “never”, “rarely”, or only “sometimes” using sunscreen when exposed to the sun for more than 15 minutes.
Logistic regression models formed the basis for all statistical approaches. A multinomial logit model was used to assess the association between the five BPI factors and BMI categories as adults. Given the small number of survivors in the underweight category, this group was combined with the normal weight category to examine predictors of obesity. Logistic regression models were used to assess the association between the five BPI factors and each of the three health behavior outcomes (i.e. physical activity, smoking, and sunscreen). Models were constructed using both survivor and sibling data to compare participant status. Separate models were then constructed for survivors only in order to include diagnosis and treatment variables. Full models for the survivor group comprise the majority of the results discussed below. For all survivor analyses, covariates included cancer diagnosis (i.e. leukemia vs. CNS tumor vs. other cancers [Hodgkin's disease, non-Hodgkin's lymphoma, Wilms' tumor, neuroblastoma, and soft tissue sarcoma]), cancer therapy (i.e. cranial radiation vs. non-cranial radiation; CNS chemotherapy vs. no CNS chemotherapy), sex, and psychotropic medication history (i.e. antidepressants, stimulants). For the BMI outcome model, BMI at adolescence was also included as a covariate. The presence of adult chronic health conditions was included as a covariate for the physical activity outcome. Avoidance of sun exposure (e.g. wearing protective clothing, staying in the shade) was used as a covariate for use of sunscreen.