Biomarker Development, Methodological Challenges

Biomarker development is a common endeavor in medical research.  The purpose is to find indicators of disease occurrence or prognostic markers for response.   The process of development of biomarkers often starts with showing mean differences between responders and non-responders  or those with a disease or condition versus those without.  However, these statistically significant mean differences, while necessary are not sufficient to validate a biomarker.  Sensitivity, specificity, positive and negative predictive value are at least as important and the relative increase in performance using the biomarker over the usual clinical variables should be demonstrated.  This paper discusses the various assessments in the context of use for the biomarker,  the need for characteristics in addition to mean differences and the importance of independent validation of putative biomarkers.  Lastly, it is hoped that the process and thoroughness be considered with recognition that the task is at best a difficult task.