(C-150) Review of Post-Authorization Safety Studies Submitted to the HMA-EMA Catalogue of RWD Studies Evaluating Cancer as an Outcome

Background: In 2014, the FDA and NCI held a two-day meeting to discuss current strategies for evaluation of carcinogenic effects of medications used to treat non-cancerous conditions (published in 2016). Notable themes included evaluation of all cancers in aggregate vs. specific cancers; defining risk window and latency; estimating risk; identifying a comparison cohort; and evaluating rare exposure and rare cancers. Ten years later, it is unclear whether the recommendations shared after the meeting are reflected in actual practice.

Objectives: Summarize key design elements used in post-authorization safety studies (PASS) evaluating cancer as an outcome and submitted to the HMA-EMA Catalogue of RWD Studies.

Methods: Systematic review of studies listed in the HMA-EMA Catalogue of RWD Studies as of October 9, 2024 using the following search criteria: status- planned, ongoing, or finalized; study type- observational; population age- adolescent and adults age 18+; scope of study- risk assessment. From the pool of studies, we limited evaluation to those with malignancy as an outcome. Two independent reviewers collected data from each study on: year of start of study, whether it was a regulatory requirement, exposure definition, follow-up duration, outcome definition, analytic approach, and other study characteristics.

Results: 70 studies had malignancy as an outcome of interest; of these, 49 had sufficient information available. Start of data collection ranged from 2005 to 2025, with 69% of studies starting in 2017 or later. Most studies (80%) were conducted to satisfy a regulatory requirement. Follow-up duration for evaluating cancer risk was >5 years (75% of 40 with available information). 39% consider a composite cancer outcome only, as opposed to inclusion of specific cancer types. None of the studies considered time-varying cancer risk in their primary analysis. 59% of studies included a latency period when defining risk windows. Most studies (73%) used a once-exposed always at-risk definition for their primary approach (OEAR, including: intention to treat (ITT) (20%), overlapping exposures (OL) (20%), OEAR-drug of interest (DOI) (20%), and not specified (12%)), citing the potential for malignancies to occur after discontinuation of their putative causal agent. When considering both primary and sensitivity approaches (categories are not mutually exclusive), on-treatment/per protocol definitions were most common (41%), followed by ITT (24%), OL (22%), duration/dose (22%), and OEAR-DOI (20%).

Conclusions: There is considerable heterogeneity in approaches to evaluating cancer as an outcome in PASS, with limited utilization of the recommendations identified by the FDA-NCI meeting.