Identification of Risk-Related Prenatal Medication Exposure Profiles for Childhood Asthma Using the Bayesian Profile...

To identify prenatal medication exposure profiles associated with childhood asthma.

This population-based cohort study used data from the POMME (PrescriptiOn Médicaments Mères Enfants) cohort, including 18 851 children born in the Haute-Garonne area (France) in 2010 and 2015, and followed until 2023. Prenatal exposure was defined as medications dispensed to mothers during pregnancy, classified by Anatomical Therapeutic Chemical subclasses, assessed separately for each trimester as none, one dispensing, or at least two dispensings. Childhood asthma from Age 4 years onward was identified using an algorithm based on dispensing of asthma medications. Bayesian profile regression was applied to simultaneously cluster children according to prenatal medication exposure profiles and asthma occurrence. High-risk profiles were characterised by predominant medication classes, timing of exposure and maternal and infant characteristics.

Seventeen prenatal medication exposure profiles were identified. Six profiles were associated with a higher average risk of childhood asthma compared with the overall population risk (15.2%). High-risk profiles were characterised by repeated exposure throughout pregnancy to medications for acid-related disorders, thyroid hormones, antibiotics, asthma medications and antihistamines, as well as by combinations involving antibiotics, nasal medications and cold medicines in late pregnancy. Mothers in high-risk profiles more frequently presented chronic conditions and pregnancy-related complications.

This exploratory study identified prenatal medication exposure profiles associated with childhood asthma using a semi-supervised Bayesian approach. This method highlights combinations and critical windows that may warrant further investigation and supports Bayesian profile regression as a useful tool for signal detection in perinatal pharmacoepidemiology.