Multivariate receiver-operating characteristic curve analysis: prostate cancer screening as an example.

The evolution of test performance analysis should include the long-term costs and benefits associated with testing. Evolutionary laboratory techniques to achieve this include introduction of a new methodological technique, a multivariate extension to a current analytical technique, receiver-operating characteristic (ROC) curve analysis (MultiROC analysis). This extension to ROC methodology allows the comparison of composite test rules in a format similar to that of ROC curves. Statistical properties, guidelines for use, and a detailed example are described. MultiROC is used in the outcomes analysis of the value of screening for prostate cancer. The effect of age and different test decision thresholds are examined in an extension of a previously published outcomes analysis. The results indicate that the variations in test performances caused by these components are important in assigning a final cost:benefit ratio of screening for prostate cancer.