Study Design for the Evaluation of Diagnostic Tests

Abstract

Diagnostic tests, like therapeutic procedures, require proper analysis prior to incorporation into clinical practice. In studying diagnostic tests, an evaluation should be made of the reproducibility, accuracy, variation among those without the disease, and variation among those with the disease. Both diseased and disease-free states should be identified using a gold standard, if available. Three main guidelines can be used in evaluating and applying the results of diagnostic tests: validity of the study, expression of the results, and assessment of the generalizability of the results. Validity requires an independent, blind comparison with a reference standard. Methodology should be fully explained. Results should include sensitivity, specificity, and a receiver operating characteristic plot. Several categories of results should be provided in the form of likelihood ratios. Management decisions can be made on the basis of the posttest probability of disease after including both the pretest probability and the likelihood ratio in the calculation.

Key Words:

clinical trials - diagnostic tests - study design