Fuzzy cluster analysis off positive stress tests, a new method of combining exercise test variables to predict extent of coronary artery disease

Abstract

Fuzzy set theory is useful in the analysis of data having a graded degree of abnormality. Previous studies using sharp cutoff points between normality and abnormality have resulted in general guidelines for the interpretation of positive stress tests, but do not enable the clinician to simultaneously combine several stress test variables, each having a range of abnormality. In this study, positive stress test results from 109 patients were reviewed. An angiogram recorded within 1 month of the stress tests showed that 100 patients had coronary artery disease (CAD) (15 had left main CAD, and 27 had 3-vessel, 30 had 2-vessel, and 28 had 1-vessel disease) and 9 were normal. Six variables were selected for fuzzy cluster analysis: ST-segment change, difference between resting systolic ana peak exercise systolic blood pressure, total treadmill time, peak exercise heart rate as a percentage of 100% predicted maximum for age, time to onset of angina, and duration of repolarization abnormalities. The analysis used a similarity measure to compute how closely each stress test resembled a prototypical mildly, moderately, or severely abnormal stress test. Stress tests classified by this method showed better correlation with the extent of CAD than the degree of ST-segment depression alone. Unlike tests with mild degrees of ST depression (0.5 to 1.5 mm), tests classified as mild by the method virtually excluded high-grade CAD. Tests classified as severe were associated with severe CAD, were better in detecting left main and 3-vessel CAD than tests with ≥3 mm ST depression, and were useful in detecting high-grade CAD in patients with mild degrees of ST depression. Combining stress test variables by fuzzy cluster analysis can be useful in managing patients with positive exercise test results.