OBJECTIVES: To determine the validation status of surrogate endpoints accepted by the European Medicines Agency (EMA) under the conditional marketing authorization pathway (CMA), and the U.S. Food and Drug Administration (FDA) across all therapeutic areas.

METHODS: A systematic review was undertaken to identify human clinical studies that evaluated the statistical association between an accepted surrogate endpoint and any patient-relevant endpoint. Accepted surrogate endpoints were obtained through the FDA Table of Surrogate Endpoints and European Public Assessment Reports of CMA-approved drugs. Articles published between April 1989 and August 2021 were found through PubMed, Embase, Web of Science, Scopus and Cochrane Library. Diagnostic and drug dependence studies were excluded, as were statistical publications and narrative or systematic reviews without sufficient statistical analyses. Data extracted included study design, statistical methods used, individual-level or trial-level associations and authors’ conclusions of validity. The level of validity was assessed through three surrogacy evaluation frameworks: IQWiG (Institute for Quality and Efficiency in Health Care), BSES3 (Biomarker-Surrogate Evaluation Schema) and a three-level “Ciani hierarchy”. A surrogate was declared validated if a high overall grade was achieved in each framework.

RESULTS: From an initial screening of 18,904 articles, 246 articles were included in the data extraction, most of which were in oncology (71%), nephrology (10%) and respiratory disorders (5%). Observational studies, randomised controlled trials and meta-analyses accounted for 26%, 29% and 38% of included articles respectively. The coefficient of determination (R²) was the most common statistical metric used. High correlation (r>0.85) as specified in IQWiG was rarely attained. Only 5/30 (17%) regulatory-accepted surrogate endpoints included were confirmed to be valid.

CONCLUSIONS: Many of the FDA/EMA-accepted surrogate endpoints reviewed did not meet the criteria for validity. Coupled with the variation in assessment criteria across the three frameworks, this review highlights the difficulty of validating a surrogate endpoint for regulatory and health technology assessment decision-making.