PDG74 AVAILABILITY OF FDA-APPROVED SURROGATE MARKERS IN REAL-WORLD DATA

Abstract

Validated surrogate markers that predict patient-relevant clinical outcomes can support more efficient clinical trials and regulatory decisions. While real-world evidence is increasingly used in decision-making, the availability of validated surrogates in real-world data (RWD) sources is unclear. We reviewed FDA-approved surrogates and determined the extent to which surrogate marker results are measurable within a representative claims and electronic health records (EHR) source. We extracted surrogates that were the basis of FDA drug approvals for common therapeutic areas from the FDA’s surrogate endpoints table. Using Optum’s de-identified Integrated Claims-Clinical dataset, we evaluated the percent of patients with surrogate marker results (365 days follow-up) within indication-specific cohorts. Cohorts included prevalent and incident users of drugs for indications corresponding to each surrogate approval, as specified by the FDA. Among the 23 FDA-approved surrogate markers identified for 16 indications, 16 (70%) were laboratory tested molecular biomarkers (e.g. HbA1c, HIV-RNA), 5 (22%) required radiographic assessments (e.g. tumor progression, bone density), and 2 (9%) were physiological assessments (FEV1, blood pressure). Overall, 16 (70%) of 23 surrogate markers could be operationalized in the representative database. The percent of patients with surrogate results across indication-specific cohorts ranged from 4% (FEV1) to 86% (blood pressure). Molecular biomarkers were the most common surrogate type, due to cardiometabolic markers collected in regular clinical practice (LDL, triglycerides, HbA1c). Surrogates from unstructured sources (e.g. clinician notes) and endpoints requiring a sequence of assessments to measure changes over time were available in a subset of indication-specific cohorts with necessary data. This study demonstrates opportunities to use FDA-approved surrogates in RWD studies and monitoring of disease status in real-world settings. Further research is needed to develop and validate algorithms for extracting surrogate outcomes from structured and unstructured health data sources.