Improving Biomarker Development and Assessment: Standards for Study Design

摘要：Background: The Early Detection Research Network (EDRN), NCI funded and investigator driven, has the mission to evaluate biomarkers for their clinical utilities in cancer risk prediction, diagnosis, early detection, and prognosis. Abundant cancer biomarkers reported in literature yet few are used in clinics. Therefore, the emphasis of the EDRN is biomarker validation. Although schema for a phased approach to development exists and guidelines are available for study reporting, a coherent and comprehensive set of guideline for a definitive biomarker validation study design have not been delineated.Methods: We proposed PROBE study design, Prospective specimen collec-tion and Retrospective Blinded Evaluation, for pivotal definitive evaluation of the accuracy of a classification biomarker. A detailed formulation of all aspects of the design is provided. Four tables itemize aspects that relate to (i) the Clinical Context; (ii) Performance Criteria; (iii) the Biomarker test; and (iv) Study power and termination. Alternative designs and strategies were contrasted to illustrate the merit of PROBE design.Results: The ideas are applied to studies of biomarkers the intended use of which is for disease diagnosis, screening, or prognosis. Two EDRN valida-tion studies (breast cancer and prostate cancer) were used as examples to elucidate PROBE design.Conclusion: Common biases that pervade the biomarker research literaturewould be eliminated if these rigorous standards were followed closely. We urge the adoption of the design as standard of practice for pivotal evaluation of the classification accuracy ofbiomarkers.