Jing-Yang Huang, Yueh-Pin Lin, Ming-Hong Hsieh, Ming-Chih Chou, Hua-Pin Chang
These findings suggest that previously reported associations are likely attributable to confounding by indication rather than direct medication effects. However, because the residual confounding inherent to observational designs cannot be entirely ruled out, these findings should be interpreted cautiously. Further studies using designs that explicitly account for covariate timing, medication initiation, and familial confounding are warranted.
PURPOSE: Sedative-hypnotics are commonly prescribed for anxiety and insomnia during pregnancy; however, concerns persist regarding their potential long-term neurodevelopmental effects on offspring. Therefore, this study aimed to examine the associations between prenatal exposure to benzodiazepines or Z-hypnotics and subsequent diagnoses of attention-deficit/hyperactivity disorder (ADHD) and autism spectrum disorder (ASD) using a restricted cohort design.
METHODS: We analyzed nationwide data from Taiwan's National Health Insurance and Birth Registry from 2009 to 2016, restricting the study population to live-born offspring whose mothers had diagnosed insomnia, depression, or anxiety in order to minimize confounding by indication. A total of 25,159 exposed offspring were matched in a 1:2 ratio with 50,318 unexposed controls based on child sex and birth date ± 6 months. Hazard ratios (HRs) were estimated using Cox proportional hazards models, with adjustment for maternal comorbidities and sociodemographic factors.
RESULTS: Among 75,477 mother-child pairs, the crude incidence rates of ADHD were 6.98 and 6.09 per 10,000 person-months in the exposed and unexposed groups, respectively. For ASD, the rates were 1.34 and 1.15, respectively. Following multivariable adjustment, prenatal exposure to sedative-hypnotics was not significantly associated with an increased risk of ADHD (adjusted HR 1.08; 95% CI 0.99-1.18) or ASD (adjusted HR 1.08; 95% CI 0.88-1.32). Associations observed in crude analyses were attenuated after adjusting for maternal psychiatric severity and familial factors.
CONCLUSION: These findings suggest that previously reported associations are likely attributable to confounding by indication rather than direct medication effects. However, because the residual confounding inherent to observational designs cannot be entirely ruled out, these findings should be interpreted cautiously. Further studies using designs that explicitly account for covariate timing, medication initiation, and familial confounding are warranted.