Bayesian analysis of stress thallium-201 scintigraphy.

The variation of the diagnostic value of stress Tl-201 scintigraphy with prevalence of coronary heart disease (CHD) in the population has been investigated using Bayesian reasoning. From scintigraphic and arteriographic data obtained in 100 consecutive patients presenting with chest pain, the sensitivity of stress Tl-201 scintigraphy for the detection of significant CHD was 90% and the specificity was 88%. From Bayes' Theorem, the posterior probability of having CHD for a given test result was calculated for prevalences of CHD ranging from 1% to 99%. The discriminant value of stress Tl-201 scintigraphy was best when the prevalence of CHD lay between 30% and 70% and maximum for a prevalence of 52%. Thus, stress Tl-201 scintigraphy would be an unsuitable diagnostic test where the prior probability of CHD is low, e.g., population screening programmes, and would add little where the clinical probability of having CHD is already high. However, where the prior probability of having CHD is intermediate stress Tl-201 scintigraphy may provide valuable diagnostic information.