(A-238) Enhancing Post-marketing Quantitative Signal Detection at GSK: A Comprehensive Review

Background: Quantitative Signal Detection (QSD) is crucial for identifying post-marketing adverse events requiring further investigation. Traditionally, based on Disproportionality Analyses (DPA) within Spontaneous Reporting Systems (SRS), QSD has inherent limitations. GSK has deployed QSD for over 20 years but recently moved from separate systems for Pharmaceuticals (Rx) and Vaccines (Vx) to a unified platform to improve alignment and efficiency.

Objectives: Our work aimed to optimize GSK’s QSD in SRS by harmonizing alert strategies for Rx and Vx, focusing on reducing false positive (FP) alerts and improving resource allocation.

Methods: First, a retrospective analysis of alerts from Dec20 to Dec22 was conducted to identify inefficiencies and duplication of work. This analysis helped refining alert rules while ensuring no signals were missed when applying these rules retrospectively and was supported by internal experiments and peer-reviewed literature. Efficiency gains were then assessed over a 7-month period post-implementation, comparing pre- and post-implementation metrics for therapies with stable review frequencies to assess the impact of the revised QSD strategy.

Results: Over the 2-year period, 58 958 alerts were generated, with 123 investigated for validation, resulting in a low positive predictive value of 0.002, akin to literature benchmarks. Further alignment in DPA approaches was possible because of the significant heterogeneity between Rx and Vx: Vx limited Multi-item Gamma Poisson Shrinker (MGPS) to unlisted events, and Rx applied an EB05 ≥ 2 threshold across all events, estimating a 12% reduction in FP alerts (4 487 alerts/year). Alerts for medication errors, overdoses, and trends shifted to bespoke visualization reviews, anticipating a 10% decrease in FP alerts (3 484 alerts/year). Subgroup alerts were streamlined for secondary evaluation in clinically relevant contexts, estimating a 14% reduction (5 104 alerts/year). Implementation of GSK's revised QSD strategy, resulted in an actual 31% reduction in FP alerts and a 16% reduction at the therapy-event pair level, confirming the strategy’s effectiveness.

Conclusions: This work reduced duplicative reviews and aligned QSD across the portfolio, streamlining processes to prioritize important signals effectively and decrease false positives, allowing better resource allocation. Next steps include integrating qualitative elements and additional data streams into the signal detection process to move beyond the current reliance on SRS data, for furthering efficacy of QSD within a broader surveillance framework.