Summary of Validation Procedures for Diagnostic Tests

Clinical studies are impossible without adequate diagnostic tests, and diagnostic tests can, therefore, be considered the real basis of evidence-based medicine. In 1995 Reid et al¹ stated after a search of 1302 diagnostic studies that most diagnostic tests are inadequately appraised. Efforts to improve the quality of diagnostic tests are given by initiatives like those of the CONSORT² (Consolidated Standard Randomized Trials) movement and the STARD³ (Standards for Reporting Diagnostic Accuracy) group launching quality criteria statements for diagnostic tests in 2002 and 2003. In spite of such initiatives the evaluation of diagnostic tests prior to implementation in research programs, continues to be lacking.⁴ A diagnostic test can be either qualitative, e.g., the presence of an elevated erythrocyte sedimentation rate to demonstrate pneumonia, or quantitative, e.g., the ultrasound flow velocity to estimate the invasive electromagnetic flow velocity. For both qualitative and quantitative diagnostic tests three determinants of validity have been recommended by working parties: Assess accuracy: the test shows who has the disease and how severe it is. Assess reproducibility: when a subject is tested twice, the second test produces the same result as the first test.

Assess precision: there is a small spread in a random sample of test results. The methods of assessment have, however, not been defined so far. The current paper reviews correct and incorrect methods and new developments.