FDA Demonstration Projects to Inform Development of a Pregnancy Safety Study Framework
Background on Demonstration Projects
FDA established five demonstration projects to inform development of the framework, guide optimal design and use of postapproval pregnancy safety studies, and meet FDA’s PDUFA VII commitments. Projects A through D fall under CDER’s purview and will be conducted in the Sentinel System. Project E falls under CBER’s purview and will be conducted in the Biologics Effectiveness and Safety (BEST) System.
- Project A. Assessing the performance of pregnancy registries versus electronic healthcare database studies to detect a signal when the exposure to medication in pregnancy is relatively common.
- Project B. Assessing the performance of single-arm descriptive safety studies versus signal identification methods using electronic healthcare data to detect a signal when the exposure to medication in pregnancy is anticipated to be low.
- Project C. Assessing the performance of pregnancy registries versus electronic healthcare database studies to evaluate a signal when the exposure to medication in pregnancy is relatively common.
- Project D. Assessing the performance of major congenital malformations (MCM) as a composite outcome in signal detection and evaluation when there is true risk for some but not all specific malformations.
- Project E. Using BEST capabilities to apply validated algorithms from EHR and claims-linked healthcare data to evaluate pregnancy-related outcomes in people administered vaccines in pregnancy.
Goals of Demonstration Projects
Demonstration projects A, B, and C compare study designs under differing product exposure levels (rare to common) to generate data to guide the selection of postapproval study approaches. The broader goal is to better understand in what context a registry study, a single-arm descriptive safety study, and/or an electronic healthcare database study will provide data needed to answer FDA’s regulatory questions about safety signal identification and safety signal evaluation for drug and biological products administered during pregnancy.
The goal of project D is for FDA to better understand the performance, interpretation, and usefulness of the composite outcome of major congenital malformations (MCMs) for safety signal detection and safety signal evaluation of drug and biological products administered in pregnancy. The MCMs that make up the composite outcome vary widely in terms of location, nature, and severity. A specific MCM risk may only be present in some subcomponent of the composite outcome and associated with drug exposure during a particular critical period in gestation – potentially masking the ability to detect the specific risk when the broader MCM risk is assessed. The study goal includes describing, assessing, and categorizing the specific malformations that are included in MCM to provide data-driven recommendations that also consider biologic plausibility. This information helps identify when it is appropriate to use MCM as a composite outcome when conducting safety signal detection and safety signal evaluation for drug and biological products administered during pregnancy.
The goal of project E is to use BEST capabilities to apply validated algorithms from EHR and claims-linked healthcare data to evaluate pregnancy-related outcomes in individuals administered vaccines in pregnancy.
Considerations for Selecting the Demonstration Projects
FDA selected the demonstration projects A, B, and C with the intent to compare performance characteristics across the different study approaches. The studies will assess the ability and timeliness of signal detection and evaluation considering the outcome of interest; trade-off between missing a true signal and identifying too many false signals; the study’s technical capability; and sample size over time for the exposed populations for each study design.
The studies can assess the capture of maternal, fetal, and infant outcomes for each study approach by using commonly used methods for signal detection (leveraging both untargeted and targeted analyses) and signal evaluation (targeted analysis). For electronic database methods, signal detection will be conducted using a series of analyses for hypothesis generation and signal evaluation using refined methods for hypothesis testing for targeted outcomes. For signal detection, the set of outcomes being assessed might be limited to major congenital malformations, may include non-live birth outcomes, or may encompass a wider variety of pregnancy related outcomes. For the signal evaluation methods that study targeted exposure-outcome pairings, it is possible to do a one-time, end-of-study analysis after a particular sample size has accrued.
Demonstration project D could potentially be addressed through a simulation study – simulating circumstances where several different analytic approaches are tested with respect to each other, all using identical datasets. A simulation environment allows for more sensitivities or variables to be tested, providing more information to FDA with respect to optimal approach(es). A key finding of a simulation study would be to understand when one approach may be preferred over another. This likely depends on a) the underlying incidence rate of the specific malformation and its contribution to the composite; b) the effect size (strength) of the increased risk; c) the potential for outcome misclassification of the specific risk; and d) the timing and length of the critical period of exposure during pregnancy and the potential for exposure misclassification. A simulation study is ideal for varying these parameters and then testing signal detection (targeted and untargeted) and signal evaluation (targeted) to analyze the performance, interpretation, and usefulness of MCM as a composite outcome in different scenarios. If an empiric example is available from real-world data, the conclusions of the simulation study can be compared against the real-world data.
Demonstration project E builds on the capabilities in BEST. To conduct safety surveillance of biological products in pregnant individuals, BEST developed the capability to: (1) identify pregnancy outcomes and gestational age using validated algorithms from standard coding systems in claims data in the ICD-10 era and (2) link pregnant individuals with live deliveries to live-born infants in administrative claims databases.
CBER uses BEST for active surveillance of CBER-regulated biological products including vaccines, blood components, and other biological products. The BEST System infrastructure includes access to electronic health record (EHR) data sources and large administrative healthcare claims databases. Some of the BEST databases are linked to local and state Immunization Information Systems, which enhance vaccine capture. Some BEST databases are linked to medical records which can be used to validate outcomes in claims databases. Data are stored in a distributed data network that is transformed into an expandable and adaptable common data model, the Observational Medical Outcomes Partnership (OMOP) common data model, and common protocols can be implemented consistently across multiple databases.
Timeline for Completing Demonstration Projects
The PDUFA VII Commitment Letter stated that FDA will initiate the demonstration projects before September 30, 2024. The Letter also stated that the results of demonstration projects will allow for updating the proposed framework and developing a guidance or MAPP/SOPP by September 30, 2027.