Diagnostic efficacy of optimised evaluation of planar MIBI myocardium perfusion scintigraphy: a probabilistic...

The Bayesian (probabilistic) approach to the results of a diagnostic test appears to be more informative than an interpretation of results in binary terms (having disease or not). The aim of our study was the analysis of the effect of an optimised evaluation of myocardium perfusion scintigrams on the probability of CAD in individual patients.

197 patients (132 males and 65 females) suspected of CAD, with no history of myocardial infarction were examined. Scintigraphic images were evaluated applying two methods of analysis: visual (semiquantitative) and quantitative, and the combination of both. The sensitivity and specificity of both methods (and their combination) in the detection of CAD were determined and optimal methods of scintigram evaluation, separately for males and females, were selected. All patients were subjected to coronary angiography. The pre-test probability of CAD was assessed according to Diamond and the post-test probability was evaluated in accordance with Bayes's theorem. Patients were divided, according to a pre-test probability of CAD, into 3 groups: with low, medium and high probability of the disease. The same subdivision was made in relation to post-test probability of CAD. The numbers of patients in respective subgroups, before and after the test, were compared. Moreover, in order to test the reliability of post-test probability, its values were compared with real percentages of CAD occurrence among the patients under study, as demonstrated by the angiography.

The combination of visual and quantitative methods was accepted as the optimal method of male scintigram evaluation (with sensitivity and specificity equalling 95% and 82%, respectively) and a sole quantitative analysis as the optimal method of female scintigram evaluation (sensitivity and specificity amounted to 81% and 84%, respectively). In the subgroup of males the percentage of individuals with medium pre-test CAD probability equalled 52 and after the scintigraphic study it decreased to 21 (p < 0.0001). In the subgroup of females it changed from 60 to 43 (p = 0.05). The verification of the values of post-test probability revealed its high concordance with the real frequencies of CAD, with correlation coefficient being 0.98 and the regression line differing only slightly from the line of identity.

The results confirm the high reliability of the values of post-scintigraphic probability of CAD obtained in this way, and support the Bayesian, probabilistic interpretation of the study outcome and its application in the diagnostic process.