Critical Appraisal: Prognostic versus Predictive Surrogate Markers in Oncology

In this era of adoption of novel biomarkers such as circulating tumor DNA (ctDNA) to monitor the risk of relapse or detect early progression, and renewed enthusiasm for previously studied biomarkers such as pathological complete response rates (pCR), there is a growing confusion regarding the utility of these biomarkers to predict outcomes at both individual-level and trial levels. This distinction between prognostic and predictive utility of biomarkers is important because misunderstanding between the two can lead to incorrect clinical inferences and misguided regulatory decisions. In this commentary, we discuss how to distinguish between prognostic and predictive biomarkers, when to consider them strong enough to be a surrogate endpoint in clinical trials, and whether they should change clinical practice. We use examples from DESTINY-Breast11, SERENA-6, DYNAMIC-III and IMvigor011 trials to elucidate these points. Destiny-Breast 11, SERENA-6 and IMvigor011 all three have recently been topics of debate at the FDA with the FDA offering approvals based on both Destiny-Breast 11 and IMvigor011 while the Oncology Drug Advisory Committee voting against approval based on SERENA-6.