(171) Concordance Between Subgroup Disproportionality Scores and PRAC Discussions of Subgroup Risk

Publication Number: 1216

Background: The heterogeneity of an exposed population may result in differentiated therapeutic response and/or adverse reactions (AR). There is a lack of systematic assessment of the extent to which quantitative data mining on spontaneous reports correlates with subgroup safety risk differences.

Objectives: To assess the concordance between subgroup disproportionality scores and PRAC discussions of subgroup risk.

Methods: We implemented the subgroup methodology of Sandberg et al. [1] to statistically screen subgroup that may be at increased risk of adverse events. The FDA Adverse Event Reporting System (FAERS) data, cumulative through 2Q 2021, were used as the data source. Test cases to determine concordance were identified by reviewing EMA Pharmacovigilance Risk Assessment Committee (PRAC) minutes from 2015-2019 and any mention of a potential subgroup at a differentiated risk, where the subgroup overlapped with the Sandberg methodology, were included.
[1] Sandberg L., et al.: Drug Saf. 2020, 43(10): 999-1009.

Results: 28 PRAC subgroup examples were included in the analysis representing 1,866 subgroup product event (SPE) combinations in FAERS, out of which 1123, 395 and 346 SPEs considered patient’s age, underlying condition, and sex, respectively. Applying the subgroup methodology 12,951 SPEs for age and 3,021 SPEs for sex were detected, of which only 2 and only 1, respectively, were concordant with the PRAC SPEs. None of the 175,794 SPEs detected for underlying condition were concordant with the 395 PRAC SPEs.

Conclusions: Very poor concordance was found between subgroup disproportionality scores and PRAC discussions of differentiated subgroup risk. Several examples concerned subgroups overrepresented in the database or specific to the drug or event and could only be detected if the requirement on not being disproportionate overall was relaxed. Also, most examples were diluted among several active ingredients, MedDRA Preferred Terms and age subgroups; regrouping these at other or customed levels might have improved the performance. Other data sources might be beneficial particularly for underlying condition subgroups however additional work is needed to better understand what improvements are required in methods and/or data to enable more robust subgroup risk identification.