The worth of a laboratory test depends on its ability to answer a clinical question of consequence for patient management. Although the need for formal prospective trials for the evaluation of therapeutic modalities has been widely recognized, relatively little attention has been given to the evaluation of diagnostic modalities. Inadequate test evaluations may result in biased conclusions about test performance. Evaluations of the clinical efficacy of laboratory tests should be formalized and include the following steps: (1) choose subjects who are representative of the clinical population to whom the test is ultimately to be applied; (2) perform all tests being evaluated on all the subjects and all tests on an individual subject at the same point in his clinical course; (3) determine the true answer to the clinical question by rigorous and complete means independent of the test or tests being studied; and (4) evaluate and compare test performance at all decision levels using receiver operating characteristic (ROC) curves.