Surrogate endpoints in chemoprevention of breast cancer: Guidelines for evaluation of new biomarkers

Markers of early events in the development of breast cancer are potential candidates for surrogate endpoints in chemoprevention trials. There are many such markers and the challenge is to identify truly relevant markers. If successful, surrogate endpoints offer several potential benefits in the conduct of prevention trials, including: shorter latency and hence shorter trials; reductions in size and cost of trials; and the opportunity to study prevention measures where use of primary outcomes would be excessively invasive or unethical. Although there are currently no validated surrogate endpoints for breast cancer, criteria for the discovery and validation of surrogates have been proposed. Putative surrogate endpoints should be biologically plausible, represent an early event in the causal pathway, be measurable by a standardized and reliable assay, and exhibit a dose-response. Perhaps most importantly, surrogates should statistically capture the effect of the intervention on outcome. The identification and establishment of a biomarker as a valid surrogate for cancer is a stepwise process that involves both smaller “transitional” studies and larger second-generation chemoprevention trials in which both primary outcome and putative surrogates are measured. Transitional studies are used to move new markers from the laboratory into use in human populations, and are designed to address specific questions of assay validity, treatment/marker associations, marker/disease associations, and inter- and intra-individual variability. Promising markers should be added to current and planned, large, traditionally designed chemoprevention trials in order to definitively address the issues of optimal representation, and to test the adequacy with which the marker(s) captures the effect of treatment on outcome. Ancillary studies of markers attached to these second-generation prevention trials must be powerful enough to detect clinically important differences, to elucidate potentially complex multivariate markers, and to validate hypothesized relationships.