Revised estimates of diagnostic test sensitivity and specificity in suspected biliary tract disease.

ABSTRACT

BACKGROUND: The purpose of this study was to estimate the sensitivity and specificity of diagnostic tests for gallstones and acute cholecystitis. METHODS: All English-language articles published from 1966 through 1992 about tests used in the diagnosis of biliary tract disease were identified through MEDLINE. From 1614 titles, 666 abstracts were examined and 322 articles were read to identify 61 articles with information about sensitivity and specificity. Application of exclusion criteria based on clinical and methodologic criteria left 30 articles for analysis. Cluster-sampling methods were adapted to obtain combined estimates of sensitivities and specificities. Adjustments were made to estimates that were biased because the gold standard was applied preferentially to patients with positive test results. RESULTS: Ultrasound has the best unadjusted sensitivity (0.97; 95% confidence interval, 0.95 to 0.99) and specificity (0.95; 95% confidence interval, 0.88 to 1.00) for evaluating patients with suspected gallstones. Adjusted values are 0.84 (0.76 to 0.92) and 0.99 (0.97 to 1.00), respectively. Adjusted and unadjusted results for oral cholecystogram were lower. Radionuclide scanning has the best sensitivity (0.97; 95% confidence interval, 0.96 to 0.98) and specificity (0.90; 95% confidence interval, 0.86 to 0.95) for evaluating patients with suspected acute cholecystitis; test performance is unaffected by delayed imaging. Unadjusted sensitivity and specificity of ultrasound in evaluating patients with suspected acute cholecystitis are 0.94 (0.92 to 0.96) and 0.78 (0.61 to 0.96); adjusted values are 0.88 (0.74 to 1.00) and 0.80 (0.62 to 0.98). CONCLUSIONS: Ultrasound is superior to oral cholecystogram for diagnosing cholelithiasis, and radionuclide scanning is the test of choice for acute cholecystitis. However, sensitivities and specificities are somewhat lower than commonly reported. We recommend estimates that are midway between the adjusted and unadjusted values.