There are two popular statistical approaches to biomarker evaluation. One models the risk of disease (or disease outcome) using, for example, logistic regression. A marker is useful if it has a strong effect on risk. The second evaluates classification performance using measures such as sensitivity, specificity, predictive values and ROC curves. There is controversy about which approach is most appropriate. Moreover, the two approaches often give contradictory results on the same data. We present a new graphic, the predictiveness curve, that complements the risk modeling approach. It assesses the usefulness of a risk model when applied to the population. In addition, the predictiveness curve relates to classification performance measures. The predictiveness and classification performance of a marker, displayed together in an integrated plot, provide a comprehensive and cohesive assessment of a risk marker or model. We demonstrate using data on PSA and risk factors from the Prostate Cancer Prevention Trial.



Included in

Biostatistics Commons