Background: Maternal smoking during pregnancy has been often implicated in the development of childhood leukemia with ambiguous results. Hence, we conducted a meta-analysis aiming to summarize current evidence and quantify any tentative impact.
Procedure: We retrieved one cohort (553 leukemias compared to 1,440,542 children), 20 case-control studies and also analyzed the updated Greek case-control dataset with unpublished data, yielding in total 11,092 cases and 25,221 controls.
Results: Odds ratios reported in the studies included ranged from 0.70 to 2.20 for acute lymphocytic (ALL) and from 0.60 to 2.17 for acute myelocytic leukemia (AML). The combined effect regarding the association of maternal smoking (any vs. no) and leukemia risk was 1.03 for ALL (95% CI = 0.95-1.12, random effects model) and 0.99 for AML (95% CI = 0.90-1.09, fixed effects model). The results remained unchanged when sensitivity analyses were undertaken of studies reporting same maternal smoking periods, those focusing only on childhood leukemia deaths or investigations which did not clearly define AML subtype.
Conclusions: The findings of the meta-analysis challenge the limits of traditional epidemiology to provide sound inferences when point estimates of constituent studies range around the null. In particular, this study provides no support to a hypothesis linking maternal smoking during pregnancy with subsequent development of main childhood leukemia subtypes. Further investigations employing molecular and genetic epidemiology, however, might be needed in the hope to reveal even minimal risks pertaining individuals with specific susceptibility to tobacco compounds who sustain high environmental exposures prenatally or postnatally.
Copyright © 2011 Wiley Periodicals, Inc.