Are Adverse Reproductive Outcomes Associated With the Illness or Its Treatment (or Both)?

To the Editor Sheehy et al¹ report an association between benzodiazepine use and spontaneous abortion. We are concerned that the authors did not sufficiently consider factors that affect their conclusions. First, the effect of the diseases for which benzodiazepines are prescribed impact outcomes. Pregnancy stress and stress-associated disorders are linked to adverse birth outcomes.² Long-term disturbances in hypothalamic-pituitary-adrenal axis function that may underlie the heightened vulnerability to adverse birth outcomes and off-spring psychopathology have been reported.³

Second, the authors do not adequately address the limitations of their data set. An advantage of an administrative database is its large sample size; however, incomplete data capture is common. Psychiatric diagnoses are underreported yet likely to be higher in the benzodiazepine group, which leads to residual confounding.⁴ Substance misuse and use disorders are particularly vulnerable to underreporting because of stigma attached to these conditions. The authors do not have information on substance use but rely on tobacco, alcohol, and other drug disorder diagnoses as proxies to adjust for exposures, which is likely to be inadequate. Substance misuse in pregnancy is strongly associated with both spontaneous abortion and benzodiazepine treatment–a clear setup for confounding.

Third, both a history of pregnancy loss and psychiatric disorders are linked with subsequent benzodiazepine use. The authors report that women with miscarriages had a significantly higher rate of pregnancy within the 12 months before their last menstrual period, which may reflect a greater likelihood of pregnancy loss prior to the index miscarriage and benzodiazepine exposure. While the authors’ reported E-value for the association between benzodiazepine exposure and miscarriage was favorable, caution in interpreting its accuracy when quantifying residual confounding is recommended.⁵

In associative studies, multiple analytic approaches control for variables that are difficult to measure, such as life-style, genetic, dietary, and environmental factors. For example, investigators who used between-person exposure designs reported an association between antidepressant use in pregnancy and autism spectrum disorder. However, it did not persist when compared with pregnancies that were or were not exposed to antidepressants in individual women, sibling designs, and prepregnancy compared with within- pregnancy methods.

Sheehy et al¹ discuss the US Food and Drug Administration D rating for benzodiazepines; however, these letter categories were eliminated in 2015 because they did not effectively communicate risk. The reasons we outline here reflect the complexity in interpreting data in this field. While the study by Sheehy et al¹ is an important contribution, it does not confirm a causal connection between benzodiazepine exposure and miscarriages.