To examine the association between maternal smoking during pregnancy and autism spectrum disorders (ASD) of children through Meta-analysis.

We searched data on relative risk (RR) and 95% confidence interval (CI) on cohort studies published between January 2000 and July 2019 from PubMed, Embase, Web of Science, Cochrane Library, China National Knowledge Infrastructure, and Wanfang database. We used Stata software 15.1 to perform the Meta analysis with random effect model applied to pool RRs according to the results of heterogeneity test through subgroup analysis and Meta regression analysis to explore the potential heterogeneity, publication bias and sensitivity.

A total of eleven cohort studies involving 1 631 618 samples and 9 276 ASD cases were included in this Meta-analysis. Results showed that maternal smoking was associated with the increased risk of autism spectrum disorder (RR=1.16, 95%CI: 1.02-1.32). For subgroup analysis, the pooled RR for prospective studies (RR=1.16, 95%CI: 1.10-1.23) appeared higher than that in the retrospective studies (RR=0.92, 95%CI: 0.83-1.06). The pooled RR for studies with adjusted confounding factors (RR=1.13, 95%CI: 1.04-1.23) was higher than that without (RR=1.12, 95%CI: 1.04-1.20). In studies that exposure to smoking assessed before delivery, inter-study heterogeneity appeared higher than those after delivery. Sample size and time of assessment on smoking seemed the sources of heterogeneity. No significant publication bias was observed in this study, and the results were quite stable.

Maternal smoking was associated with the increased risk of autism spectrum disorder. However, value of the combined effect seemed low. High-quality, large-sample, and prospective cohort studies should be conducted to further verify the causal relationship, based on the correction of potential confounding factors.

Pregnancy; Smoking; Autism spectrum disorders; Cohort study; Meta-analysis