(232) Addressing Over-Representation of Early Pregnancy Outcomes in Database Pregnancy Safety Studies

Background: In most database studies, study populations are defined during a specified calendar period without regard for duration of enrollment after baseline. Database pregnancy safety studies, however, identify study populations by pregnancy outcomes, which can introduce certain biases. One problem in pregnancy studies is that early pregnancy outcomes such as spontaneous abortion (SAB) are over-represented near the end of the study period if there is insufficient time to observe pregnancies with live births that occur after the calendar end date (“fixed cohort” bias). A similar, but less widely appreciated, problem occurs throughout a study when women having early pregnancy outcomes disenroll before sufficient time has elapsed to observe live births.

Objectives: To describe the impact of enrollment criteria on early pregnancy outcomes and propose criteria that address a relative excess of early pregnancy outcomes over the study period.

Methods: We used a Healthcare Integrated Research Database (HIRD) retrospective cohort of commercially-insured women in the U.S. with pregnancy start dates between 01August2021 and 31July2023. We identified a group of new users of a medication, and we compared two study inclusion criteria: Method 1 required continuous enrollment from pregnancy start to end of pregnancy (EOP); Method 2 required enrollment from pregnancy start until a sufficient time to capture live births (start of pregnancy + 305 days). The outcome was SAB identified using ICD-10-CM and CPT codes at < 20 weeks gestation. Proportions of SAB were calculated for both inclusion criteria.

Results: Using Method 1, among 10,636 pregnancies with enrollment from pregnancy start date through EOP, SAB events were identified in 26.7% (95% CI: 25.9%-27.5%). Using Method 2, among 8,654 pregnancies with enrollment through pregnancy start date plus 305 days, SAB events were identified in 18.6% (95% CI: 17.8%-19.4%).

Conclusions: In a database study, including pregnancies with early pregnancy outcomes that are not observed for sufficient enrollment time to capture a live birth showed a relative excess of SAB. By requiring minimum study enrollment, some early pregnancy outcomes are excluded so there is a small loss of precision, but validity is improved. Not taking into account the opportunity to observe all EOP events can create an impression of a misleadingly high proportion of early EOP outcomes. Impact on comparative studies will also be discussed.