Development and comparison of single-parameter indices characterizing severity of acute myocardial ischemia

Our aim was to assess the sensitivity/specificity of 2 indices of ischemia severity (∑|ST| and ST_RMS) for 5 triplets of leads: X, Y, Z; V2, V5, aVF; V2, V5, III; V3, V5, III; and V3, V6, III, each derived from 7 subsets of Mason-Likar 12-lead ECG with limb leads and dual precordial leads (V1, V3; V1, V4; V1, V5; V2, V4; V2, V5; V3, V5; V3, V6). Coefficients for deriving lead triplets were developed from the design set (n = 892). The test set comprised 12-lead ECGs (n = 99) acquired before and during ischemia induced by balloon-inflation angioplasty. We compared the ability of tested indices todetect ischemia by constructing their receiver operating characteristics (ROCs) and measuring a percentage area under the entire ROC curve (AUC) and in specificity range 0.8-0.9 (AUC_0.8-0.9). The mean performance for 7 predictor sets in terms of AUC (%) for ∑|ST|/ST_RMS was: 88.4/88.8 for X, Y, Z; 88.6/86.5 for V2, V5, aVF; 89.4/87.5 for V2, V5, III; 90.4/90.3 for V3, V5, III; and 90.5/90.2 for V3, V6, III. Mean values of AUC0.8-0.9 (%) for ∑|ST|/ST_RMS were: 78.1/80.5 for X, Y, Z; 80.7/71.6 for V2, V5, aVF; 77.6/74.5 for V2, V5, III; 77.7/80.8 for V3, V5, III; and 85.7/82.4 for V3, V6, III. Thus we conclude that the currently used indices using "pseudoorthogonal" leads V2, V5, aVF, and orthogonal leads X, Y, Z performed in ischemia detection nearly as the best indices based on V3, V6, III. These results should be corroborated on a larger study population.