Broussard et al utilize a case-control study to conclude there is a relationship between early pregnancy opioid analgesic use and certain birth defects. However, 3 concerns severely limit its future research and clinical usefulness. First, drug use data are collected retrospectively “between 6 weeks and 2 years” postdelivery. This method risks selective recall or confirmation bias, and this extended recall window (with unknown duration between cases and controls) compromises the study’s internal validity. Second, the choice and measurement of statistical control variables in the logistic regression analyses generates concern. Measuring tobacco use as “periconceptional smoking status (no-smoking from 1 month before to 1 month postconception, smoking at least once in the same period)” equates mothers smoking 1 cigarette once to mothers smoking 2 packs daily . This crude measure may lack meaningful statistical control for the possible differential impact of smoking on outcomes between cases and controls. Furthermore, failure to measure and control for alcohol use, a known teratogen, is startling. These 2 measurement lapses seriously threaten the study’s statistical conclusion validity. Third, respondents (66%) reported common prescription opioid use reasons were “surgical procedures (41%), infections (34%), chronic diseases (20%), and injuries (18%).” The extent that these issues are independently related to the birth defects described is unknown. Also unknown is whether or not the cases and controls differed in their respective prevalence of these events. These are stressful events, and stress can influence fetal and neonatal outcomes. Thus, omitting the measurement and statistical control of stressful events potentially biases the results. Finally, the potential relationship between birth defects and other medications (eg, acetaminophen) that are present in opioid analgesic medications was not addressed.

Broussard et al conclude that an “association between early pregnancy maternal opioid analgesic treatment and certain birth defects” exists and that, “This information should be considered by women and their physicians who are making treatment decisions during pregnancy.” Such communications must be targeted only to the pregnant patient population in need of opioids during early pregnancy and include information that the relative risk of such defects in the presence of opioid use is very modest, given the low base rates. Further, the communication of these findings should be only one small part of a complete risk:benefit discussion with these women. Importantly, results have no known applicability to opioid-dependent women seeking drug treatment, in whom no relationship has been found between maternal substance abuse and birth defects.