“Need to Account for Familial Confounding in Systematic Review and Meta-Analysis of Prenatal Tobacco Smoke Exposure and Schizophrenia”, 2020-04-02 (; similar):
In the context of continued uncertainty regarding the long-term mental health effects of prenatal exposure to maternal smoking during pregnancy, we read with great interest the recently published meta-analysis of smoking and schizophrenia by Hunter and colleagues. Although the meta-analysis found that “exposure to prenatal smoke increased the risk of schizophrenia by 29%” (pg3), the authors noted that “familial confounding may explain some of the observed association” (pg8). We agree with the importance of this alternative hypothesis. In fact, we were surprised that the review did not consider the results of sibling comparison studies that have directly addressed it, particularly given that the review had the opportunity to do so using data from articles included in the meta-analysis.
…The two studies of interest represented over 60% of the weighted sample. They yielded covariate-adjusted hazard ratios of 1.33 (95% confidence interval [CI], 1.23–1.45) and 1.13 (95% CI, 1.05–1.23). However, their sibling comparison results, which were excluded from the review, were weaker and not statistically-significant (hazard ratios, 1.21 [95% CI, 0.96–1.52] and 1.09 [0.84–1.42], respectively). These results suggest that familial confounding, rather than a true casual effect, explains much of the observed associations. The weaker associations from the sibling comparisons may thus dampen enthusiasm regarding a potentially meaningful role of exposure to maternal smoking during pregnancy in offspring schizophrenia. An ~10%–20% relative difference in rates of what is a rare outcome would suggest that modifying maternal smoking would have only a limited impact on the incidence of offspring schizophrenia.