Erin Marcotte, Logan Spector
Erin Marcotte and Logan Spector
childhood cancer, parental smoking, cancer risk factors, bias analysis
Background: Though the carcinogenic properties of tobacco smoke and smoking are well documented, previous studies of maternal smoking and childhood cancer risk have been inconclusive in some cancers and null in others, such as acute lymphoblastic leukemia. However, case-control studies comprised the majority of prior research and thus were limited by the potential for recall bias, selection bias, and limited sample sizes.
Objective: This analysis aimed to elucidate this relationship through a large, pooled analysis (over 10,000 cases and 25,000 controls) of existing registry-based case-control studies with a probabilistic error correction of smoking misclassification.
Methods: The association between binary (yes/no) maternal smoking as reported on birth certificates and childhood cancer (defined by International Classification of Childhood Cancer category) was investigated using unconditional logistic regression controlling for parental education, parental race/ethnicity, and parental age. These analyses were also repeated stratified by potential mediators such as birth defects in children, birth weight, and maternal BMI. Where data was available, additional analyses were conducted for smoking by trimester and cigarettes per day also using unconditional logistic regression. A probabilistic error correction was enacted for smoking status using information on the sensitivity and specificity of birth record reported smoking in existing literature.
Results: Before adjustment, maternal prenatal smoking was statistically significantly associated with an increased risk of lymphoid leukemias. After adjustment, the magnitude of the observed association was lessened, but the association remained statistically significant.
Discussion: This study provides evidence that maternal prenatal smoking may increase the risk of lymphoid leukemia in children.
Project funded by a T32 (T32CA099936)