Towards a Further Understanding of Meta-Analysis Using Gestational Exposure to Cannabis and Birth Defects as a Case in Point.

About 5%-10% of pregnancies in the US are exposed to cannabis with highest use reported during the first trimester. Two recent meta-analyses presented estimates of the risk of birth defects associated with prenatal exposure to cannabis; the larger and more recent meta-analysis pooled data from 18 cohort and 18 case-control studies with a total sample size of >19 million subjects. The meta-analyses found that prenatal exposure to cannabis was associated with a small but statistically significant increased risk of any birth defect (pooled odds ratios [ORs], 1.25-1.33); ORs were also significantly elevated for cardiovascular, gastrointestinal, nervous system, genitourinary, and musculoskeletal but not orofacial birth defects. The ORs were smaller and less likely to be statistically significant in adjusted analyses. These meta-analyses had strengths but also shortcomings. The strengths and shortcomings are explained in detail so that readers obtain a better understanding of how to critically assess findings in meta-analyses. One strength was the presentation of both unadjusted and adjusted pooled estimates; the former allow an understanding of risks in the average real world patient and the latter allow an understanding of the unique contribution of the exposure to the outcomes. Another strength was the presentation of cumulative meta-analyses which demonstrated from which calendar year onwards a finding became consistently statistically significant in the scientific literature. One shortcoming, in analyses of subcategories of birth defects, was the repeated representation of the same sample in the same forest plot; the many reasons why this is problematic are explained. Another shortcoming was the pooling of ORs obtained from cohort studies with those obtained from case control studies; conceptual and numerical reasons why this is problematic are also explained.