Drug exposure: inclusion of dispensed drugs before pregnancy may lead to underestimation of risk associations.

ABSTRACT

OBJECTIVES: To assess the impact of exposure misclassification on risk associations when using prescription databases as the source for drug exposure in pregnancy by applying results from a validation analysis of exposure classification. STUDY DESIGN AND SETTING: Linkage of data on 27,656 participants in the Norwegian Mother and Child Cohort Study (MoBa) with the Norwegian Prescription Database (NorPD). Exposure to selective serotonin reuptake inhibitors (SSRIs) was defined by dispensed drugs during pregnancy including different time windows before pregnancy. The validity of NorPD data was estimated using self-reported use in MoBa as the reference standard. We applied the results from the validation analysis on data from a Nordic study on SSRI use in pregnancy and risk of persistent pulmonary hypertension in the newborn. RESULTS: Sensitivity increased and specificity decreased when the time window in NorPD was expanded before pregnancy. Using the same time window as in the Nordic study (+90 days before pregnancy), for use in early pregnancy, the odds ratio (OR) corrected for misclassification was 2.6 compared with the OR of 1.6 in the Nordic study. CONCLUSION: Expansion of the time window to include intervals before pregnancy can lead to lower specificity and underestimation of risk associations.