Prescribed, non-obstetric-specific drugs are commonly used during pregnancy, with some studies indicating that more than 80% of women use at least one prescribed drug during pregnancy (1). The risks and benefits of drugs used in pregnancy are primarily established through post-marketing surveillance, due to limited pre-approval drug studies during pregnancy (2) and because evaluation of rare outcomes, such as stillbirth or birth defects, requires large numbers of observations (3), which will only occur after a drug is introduced into populations including women of reproductive potential.   

Thus, active surveillance is the safety net that we rely upon to evaluate potential effects of a drug used during pregnancy on the woman, fetus and child, especially rare outcomes. However, historically, pharmacovigilance of drugs in pregnancy has lacked standardization or a systematic approach and been fragmented between different data sources, resulting in inadequate data for health care providers or pregnant persons to assess risk and ensure maternal medical conditions are adequately treated during pregnancy. 

Guidance for optimizing surveillance data collection to standardize outcome measures and reporting criteria has been limited.  This results in variable data quality and impedes the ability to rapidly combine data and/or to integrate evidence generated from different studies to decrease the time taken to provide reliable conclusions about medicinal product safety. This can be facilitated by harmonized definitions and data collection tools for a core set of exposure and outcome variables, sparing local resources from repeated development of these tools and supporting adequate scientific standards in collection of these data.   

The outcomes with the greatest public health impact include birth outcomes such as preterm and very preterm birth, small-for-gestational age and very small-for-gestational age, stillbirth, neonatal death (prior to hospital discharge if available), birth defect surface examination findings at birth (4), and maternal outcomes including pregnancy complications (e.g., pre-eclampsia) and death.  

Critical to pharmacovigilance is the quality, interpretability and comparability of data and the need for longitudinal data collected over time. Standardization both within and across surveillance sites is paramount. Having an adequate comparison group and obtaining denominator data is critical to understanding any potential signal, and surveillance efforts will fail if only abnormal findings are recorded. The prevalence of each outcome has a large impact on the denominator required to understand a signal; for example, rare events such as neural tube defects require prolonged surveillance of large numbers of exposures to gain precision.

A critical component of pharmacovigilance surveillance is being able to link maternal antenatal/obstetric records to birth/infant records. This can be difficult in many settings where electronic records are not yet available, unique patient identifiers are not in place, and name changes can occur. Electronic health records can potentially link maternal data, birth records, prescription records, birth defect records (when such registries exist) and other relevant databases. These databases can be linked together if there are unique patient identifiers or an equivalent, and the resulting linked and subsequently anonymized databases can be accessed in secure data environments to maintain patient confidentiality. 

Pharmacovigilance models for birth outcome surveillance include the Tsepamo study in Botswana and Eswatini birth outcome study. Both studies use sentinel sites to provide surveillance of outcomes in a standardized fashion with appropriate comparison groups. They collect at one timepoint (delivery) standard demographic data and outcomes such as gestational age and birth weight on all births at a sentinel site, allowing concurrent comparison groups that can be stratified by maternal HIV/treatment status, with more intensive, consented data collection from the mother if a birth defect is identified and infant photos, permitting evaluation of potential risk factors for defects and independent evaluation/confirmation of the defect. 

The Western Cape Province of South Africa provides a model for electronic medical record linkage to obtain pregnancy/infant safety data. Approximately 15 years ago the Western Cape Provincial Government invested in unique patient identifiers used across all public-sector health services in the province; with the support of research and other partners, the Western Cape Provincial Health Data Centre (WCPHDC) integrates multiple electronic sources of data at an individual level to identify healthcare system encounters as well as disease episodes, such as pregnancy or HIV (5,6).

This toolkit provides pharmacovigilance guidance for key maternal and birth/infant outcomes (see Outcome measures section), focusing on standardization, data quality and harmonization of outcomes to allow collaboration and linkage between surveillance networks, and enable rapid and collaborative evaluation of a safety signal should one be identified.