A systematic review of trial-level meta-analyses measuring the strength of association between surrogate end-points and overall survival in oncology

Background: Surrogates are frequently used in cancer medicine as the end-point of clinical trials and as the basis of United States Food and Drug Administration approvals, but they do not always represent outcomes that are important for patients. We aim to build upon previous umbrella reviews of surrogate validation studies by identifying and examining all meta-analyses of randomised controlled trials that evaluate the strength of correlation between overall survival (OS) and surrogate markers. Methods: Google Scholar and PubMed were searched by two independent reviewers for all eligible meta-analyses of randomised controlled trials examining the correlation between a surrogate end-point and OS in medical oncology. Included studies were trial-level (level-1) meta-analyses of randomised controlled trials in cancer. Data abstracted include date of publication, tumour type, setting, trial set, number of studies included in the analysis, dates of included publications, correlation coefficients and method to determine the correlation coefficient. Results: Seventy-eight articles met the inclusion criteria and reported correlations in 89 settings. Eleven (12%) of these validation studies found only high correlation(s), while nine (10%) settings showed a moderate-only correlation. Thirty-four (38%) reported only low correlation(s). Thirty-five (39%) reported correlations of different strengths, depending on surrogate marker used and test of correlation. Conclusions: In this large, umbrella analysis of surrogate validation studies, we found most surrogates in oncology had low or modest correlation with OS, which suggests that caution should be used when making conclusions based on surrogate markers.