Diagnosing liver metastases: a Bayesian analysis.

Clinicians frequently perform tests to determine whether patients have liver metastases. Optimal use of a laboratory test requires that the clinician know the test's operating characteristics (its sensitivity and specificity) and have an estimate of the pretest probability that disease is present. We have surveyed studies that examined the value of four biochemical and three imaging tests in establishing a diagnosis of hepatic metastases in patients who underwent an invasive procedure to establish the presence or absence of disease. We have pooled the data from these studies to arrive at values for the sensitivity and specificity of each of these tests, and calculated the predictive values for these tests over a wide range of pretest probabilities of disease. Several examples illustrate how this information may be used clinically. We provide a framework for the optimal interpretation of these commonly ordered tests and indicate the data needed for their complete analysis.