Clinical Trials: A New Method for Assessing Accuracy of Diagnostic Tests

Abstract

Background: Clinical trials of disease management require accurate tests for making a diagnosis/patient follow-up. Currently, accuracy of diagnostic tests is hard to quantify, because it is generally assessed by two estimators, sensitivity and specificity, that are severely dependent on one another. If one estimator is high, the other is, as a rule, low.

Objective: To propose that the current dual approach to accuracy be replaced with a new method, called the overall accuracy level.

Methods and results: The overall accuracy level is defined as the proportion of test results that are correct. Usage of this level, unlike sensitivity and specificity levels, enables (1) to adjust for differences between numbers of patients with and without the disease, (2) to readily identify tests that give no information at all, (3) to provide the entire amount of information given by the test relative to the gold standard test, (4) to draw receiver operating characteristic (ROC) curves adjusted for the differences between numbers of patients with and without the disease. The method is less adequate for the assessment of diagnostic tests for extreme diseases like incurable deadly diseases and treatable but untreated deadly diseases for which diagnostic tests with either optimal sensitivity or optimal specificity are required.

Conclusions: Due to the dual sensitivity/specificity approach to accuracy of diagnostic tests it is, currently, difficult to find the most accurate diagnostic test for a given disease. The overall accuracy level is more appropriate to that aim.

Key words: :

• clinical trials
• diagnostic tests
• accuracy
• sensitivity
• specificity
• receiver operating characteristic curves