Eligible participants were identified from the records of the CCG. All survivors were diagnosed with AML at less than 21 years of age, received treatment on one of the four protocols from 1979 to 1995 (CCG 251, 213, 2861, or 2891) and were alive at least 5 years following diagnosis. Treatment details for each protocol have been previously reported [9
]. Study protocols and materials were approved by the human subjects committee at each participating institution. Eligible survivors were contacted by mail and asked to participate. Adolescent (age at interview 12–17 years) and young adult survivors (≥18 years) without Down syndrome and without active AML or secondary malignant neoplasms are the focus of this analysis. Survivors were asked to complete a Childhood Cancer Survivor Study (CCSS) baseline questionnaire (see ccss.stjude.org) and a Youth Risk Behavior Survey (YRBS). Data regarding treatment and acute complications were abstracted from the CCG/COG registry in a standardized fashion. For determination of total body irradiation (TBI) exposure, patients were grouped by intent to treat. Patients were categorized by treatment group using the a priori hypothesis that chemo and autoBMT were more similar to one another than to alloBMT, particularly with regard to risk of late effects and CHCs.
CCSS Baseline Questionnaire
Each survivor or proxy was asked to complete the CCSS baseline questionnaire [13
] from which CHC variables were constructed as described by Oeffinger et al. [14
]. Grade III and IV CHCs were categorized as severe. Mental health was coded as described by Hudson et al. [15
Youth Risk Behavior Survey
The national YRBS is administered biannually to a sample of high school students. In this study, the YRBS was administered to survivors directly, not to proxy caregivers, by a trained telephone interviewer. Substance use was assessed both qualitatively and quantitatively (e.g., “Have you ever tried cigarette smoking?” and “During the past 30 days, on how many days did you smoke cigarettes?”) We assigned each question a potential value of 1 or 2 points based on potential risk to childhood cancer survivors (ever smoked = 1, smoked last 30 days = 2, drank alcohol in last 30 days = 1, >5 drinks per day in last 30 days = 2, tried marijuana = 1, marijuana use in last 30 days = 2, use of chewing tobacco/snuff = 1, cocaine use in last 30 days = 2, and “ever tried” cocaine, heroin, methamphetamines, illicit steroids, and glue/aerosols sniffing = 1 each) creating a substance exposure (SubExp) score. Potential scores ranged from 0 to 17. Given the significant medical risks faced by survivors, a relatively low threshold of substance exposure was considered clinically significant. Scores were also calculated using the 2001 national YRBS cohort to ensure an appropriate distribution and that literature-documented predictors (age, gender, race, and household income) were valid predictors of our SubExp score.
In the nationally administered YRBS and in our study, survivors were also asked if during the past 12 months, they “felt so sad or hopeless almost every day for two weeks or more in a row, that you stopped doing some usual activities?” Suicidality was assessed based on answers to “during the past 12 months, did you ever seriously consider attempting suicide?” If the answer was yes, survivors were then asked if they had ever made a plan and how many times they had attempted suicide. Each positive answer was assigned a single point to create a sadness/suicidality score with potential scores ranging from 0 to 4.
Demographic and health variables were compared in young adults between treatment groups using t-tests for continuous outcomes and chi-squared or Fisher’s exact tests for categorical variables. Results for each substance exposure, high (≥3), and very high (≥6) SubExp score were analyzed between treatment groups with chi-square or Fisher’s exact tests for statistical significance. Results of the mean SubExp score were compared using a t-test. Results described remained robust regardless of the definition of high exposure. Missing values were assumed to be missing at random and were excluded prior to analysis.
Results of the SubExp score and sadness score were then analyzed in linear regression models. First, factors known from the literature to be significant in the general population (sex, age, race, household income) were analyzed in multivariate models. Age at diagnosis was included due to the difference between treatment groups. Other possible predictors (cancer-related pain or anxiety, sadness, multiple and severe CHCs, and TBI exposure) were entered into the model separately based on a priori hypotheses. Multivariate analyses were not performed in the adolescent cohort due to sample size limitations.