Pregnancy Registries Nested in International Pooled Health Care Databases

Harvard School Of Public Health, Boston MA

Investigators

Linked publications & trials

Abstract

ABSTRACT Since pregnant women are excluded from clinical trials, the safety of medications in pregnancy is largely unknown. Post-marketing, commonly used study designs such as exposure pregnancy registries rarely include enough pregnant women to detect rare adverse pregnancy outcomes. Moreover, these registries are expensive and usually limited to a few drugs. In the absence of trials, one of the most efficient and valid approaches to assess the unintended effects of drugs is a careful analysis of large health care databases. Yet, even these resources may have insufficient power. We propose to develop an international collaboration to study the safety of drugs in pregnancy. By pooling health care information from population-based databases in the Nordic Countries and the US we will be able to identify global nested exposed pregnancy cohorts to study in-utero exposure to individual drugs in relation to rare outcomes such as specific birth defects. In the US, using data from both the Medicaid Analytic eXtract (MAX) and MarketScan, we have identified a cohort of over 2 million pregnancies with information on prescription medication use and medical conditions from 3 months prior to pregnancy to one month post-delivery, along with the records of the offspring of these pregnancies for at least the first 3 months of life. This population based national cohort includes commercially insured women as well as Medicaid beneficiaries, a racially diverse and vulnerable population typically understudied. The Scandinavian investigators have already merged data from their national registers among all five Nordic countries to form a cohort of 3.1 million pregnancies. The proposed International Pregnancy Safety Study (InPreSS) network would add the US population; our goal is to grow this collaboration, using the approaches that will be optimized in this pilot study, incorporating data from other North American, European, Oceanic, and Asian databases. Since this multi-center study requires both data privacy and multivariate adjustment, we will use a privacy-maintaining propensity score (PS) approach: Each data-contributing center will create datasets with harmonized definitions for exposures, outcomes, and covariates. Then, a set of PSs will be computed in each center and PS-adjusted relative risk estimates will be pooled with random-effects meta-analytic techniques. To apply the methods and test the feasibility of InPreSS we will perform pilot studies on 1) antipsychotics during pregnancy and cardiac defects in the newborn and 2) beta-blockers and calcium channel blockers during pregnancy and risk of birth defects in the newborn. Their aim is to provide evidence on the comparative safety of specific strategies used to manage psychosis and hypertension in pregnancy. Between the US health care data and the five Nordic registers, we will identify cohorts of over 10,000 women on antipsychotics, 15,000 on beta-blockers and 5,000 on calcium channel blockers at conception. This initiative will represent a world-unique and important resource to study the comparative safety of prescription medications during pregnancy in relation to obstetric and neonatal outcomes.