Study population

Participants were members of the Childhood Cancer Survivor Study (CCSS) cohort, described in detail previously. Briefly, participants were at least 5 year survivors of childhood cancer, diagnosed when younger than age 21 years at 1 of 26 institutions in North America between 1970 and 1986. The protocol was approved by institutional review boards at all institutions. Consent was obtained from survivors older than 18 years and from parents of survivors younger than 18 years of age. Of 20,691 eligible survivors, 14,358 survivors were enrolled. Study participants completed a baseline questionnaire in 1995 and follow-up questionnaires thereafter ( http://www.stjude.org/ccss ). The baseline questionnaire gathered information on demographics, cancer type, medications, psychologic status, pregnancy history, and lifestyle factors. The medical records of those who consented were abstracted. The follow-up 2000 pregnancy questionnaire gathered information on parent’s age at pregnancy, time to pregnancy, fertility problems, infertility treatment such as medications or in vitro fertilization, chemotherapy or radiotherapy received by either parents during or a year before pregnancy, antenatal care, substance abuse, medical complications during pregnancy, and dates and outcomes of pregnancies. The follow-up 2003 questionnaire was similar to the baseline questionnaire and in addition collected information on neurocognitive functions, short form-36 health survey, dental and bone health, and posttraumatic stress. The follow-up 2007 questionnaire gathered information similar to the baseline questionnaire. Survivors who reported relapse or second neoplasms before pregnancy were excluded.

For the current study, we restricted our sample to survivors who consented to medical record abstraction for treatment information, and included pregnancies reported by female survivors 14 to 45 years of age who had previously completed a questionnaire reporting lifestyle factors and psychologic distress ( Figure ). Nonsingleton pregnancies, pregnancies resulting from in vitro fertilization, or reported pregnancies without a known outcome, were not included ( Figure ).

Flowchart illustrating the selection of eligible pregnancies from the Childhood Cancer Survivor Study

CCSS , Childhood Cancer Survivor Study; IVF , in vitro fertilization; P , pregnancies; Q , questionnaire.

Gawade. Pregnancy outcomes in childhood cancer survivors. Am J Obstet Gynecol 2015 .

Adverse pregnancy outcomes

Pregnancy outcomes including preterm birth, stillbirth, and miscarriage were obtained from the 2000 and 2007 CCSS questionnaires. Pregnancy outcomes were self-reported by survivors as live birth, stillbirth, miscarriage, or medical abortion. We defined preterm birth as a reported gestational age of less than 37 weeks for a live birth. Stillbirth was defined as fetal death occurring after 20 weeks of gestation or later and a miscarriage was defined as fetal death occurring before 20 weeks of gestation. Self-reported pregnancy outcomes of live birth, stillbirth, miscarriage, or medical (elective) abortion were further validated by expert review with additional clarification through a telephone interview.

Independent variables

Variables of interest, including BMI, smoking, alcohol consumption, physical inactivity, and psychologic distress were obtained from the baseline and 2003 questionnaires completed before pregnancy. Alcohol consumption was only captured on the baseline questionnaire and thus evaluated in separate multivariable models only including pregnancies reported on the 2000 questionnaire. We defined BMI as self-reported weight in kilograms (kg) divided by self-reported height in meters squared (m ² ) and categorized survivors as underweight (<18.5 kg/m ² ), normal (18.5-24.9 kg/m ² ), overweight (25-29.9 kg/m ² ), and obese (≥30 kg/m ² ). Pack-years of smoking were calculated by dividing the product of the average number of cigarettes smoked per day and the number of years smoked by 20. Smoking status was categorized based on pack-years of smoking as 0, 0.1 to 5, or >5 pack-years (0 was assigned for those who never smoked). Survivors who reported either >3 drinks/day or >7 drinks/week were classified as risky drinkers, per the National Institute on Alcohol Abuse and Alcoholism guidelines. Survivors were classified as physically inactive if they reported <150 minutes of moderate or <60 minutes of vigorous physical activity per week. Information from the Brief Symptom Inventory (BSI)-18 was used to identify survivors with global psychologic distress or depression. Survivors with T-score ≥63 for the total BSI-18 were classified as having global distress and those with T-score score ≥63 for the depression subscale as having depression.

Cancer and treatment variables were obtained from medical records. Analyses included organ-specific radiotherapy and chemotherapeutic exposures previously reported to influence adverse pregnancy outcomes. Organ-specific radiation dose was estimated by medical physicists and the sum of all radiation treatments was used as the total radiation dose to the uterus, ovaries, and pituitary, which was categorized as ≤2.5 Gy and >2.5 Gy for analysis. The alkylating agent score was calculated by dividing the cumulative sum of the tertile scores of all alkylating agents into tertiles and given a score of 1 to 3. Anthracycline treatment was also categorized in cumulative dose tertiles. Other variables including race, annual household income, level of educational attainment, marital status, and insurance status, were self-reported by survivors on questionnaires completed before pregnancy.

Statistics

Demographic factors and treatment exposures of survivors with eligible pregnancies were examined as frequencies and percentages. To evaluate participation bias, we compared the demographics, treatment characteristics, and exposure distribution of the participants of baseline to 2000 and 2003 to 2007 intervals to the survivors (nonparticipants) who did not complete 2000 and 2007 questionnaires. Data analysis was performed using SAS software, version 9.2 (SAS Institute, Cary, NC).

To evaluate associations between factors of interest and adverse pregnancy outcomes, multivariable regression models with a generalized estimating equation adjustment to account for correlated data from women who had more than 1 pregnancy were used. The COPY method (10,000 copies of original data) was applied to estimate the maximum likelihood when models did not converge. A minimally sufficient set of covariates including age at diagnosis, age at pregnancy, educational attainment, and parity was identified a priori. Potential interactions between lifestyle factors or psychological distress and age at start of pregnancy, parity, type of cancer, and treatment were evaluated in multivariable models. Results are reported as odds ratios (ORs) with 95% confidence intervals (CIs). P values < .05 were considered significant, except for interaction analyses where we used P < .15. When interaction terms were significant, ORs for each strata of the significant effect modifier are presented. Our results may be subject to selection bias because not all eligible survivors participated. To assess this potential bias, we performed a sensitivity analyses ( Appendix ; Supplementary Table ).