Gestational exposure to benzodiazepines, 2: The risk of congenital malformations examined through the prism of compatibility intervals

Abstract

A recent meta-analysis of the major malformation risk after gestational exposure to benzodiazepines identified 8 prospective cohort studies with 5,195 exposed and 2,082,467 unexposed women. Benzodiazepine exposure was not associated with a statistically significant increase in the risk of total malformations (8 studies) or cardiac malformations (4 studies). The malformation risk was not significantly increased after specifically first trimester exposure, either (5 studies). However, there was a significant increase in risk associated with combined first trimester exposure to benzodiazepines and antidepressants (3 studies). The authors of the meta-analysis interpreted their findings based on the conventional P < .05 cutoff for statistical significance. However, when the confidence intervals (CIs) associated with the pooled odds ratios (ORs) were examined, it was evident that almost the entire range of values in the CIs for total malformations and for cardiac malformations was compatible with a population value for the OR that indicated increased risk (OR > 1.00). Importantly, the CI for the first trimester exposure analysis was somewhat better distributed around the null (OR = 1.00), suggesting a lower likelihood of increased risk. All ORs were very low to low in value, indicating a very small increase in the absolute risk. Besides explaining how CIs may be interpreted as compatibility intervals, this article reminds readers that associations identified even in meta-analyses of cohort studies do not indicate a causal effect because confounding by indication can never be ruled out in observational research designs. A reasonable conclusion, therefore, is that gestational exposure to benzodiazepines is a marker of risk for cardiac and total malformations, and, importantly, that first trimester exposure to benzodiazepines may not be associated with much increase in risk, if at all.