A recent, large, case-control study found that, after adjusting for measured confounds, early gestational exposure to benzodiazepines was associated with a nearly doubled odds of spontaneous abortion. The risks differed little between pregnancies exposed to short- and long-acting benzodiazepines. The risks were elevated for all benzodiazepines for which data were adequate for analysis. The risks were higher with higher daily doses. These findings indicate that early gestational exposure to benzodiazepines is a marker for the risk of spontaneous abortion. These findings are not evidence for a cause-effect relationship because analyses in studies such as this can adjust for only measured confounds; therefore, unmeasured, inadequately measured, and unknown confounds will continue to contaminate the interpretation of findings. It follows, therefore, that replication of the study with the same study design using data extracted from other databases will serve no purpose. So, innovative research designs should be considered, such as the examination of risk associated with benzodiazepine exposure in the year before pregnancy, examination of risk in previously unexposed pregnancies, examination of risk in discordant sibling pairs, and examination of risk associated with paternal exposure. Such research designs have the potential to indirectly partially address unmeasured and unknown confounds, including those related to genetics and the family environment.