Threshold criteria to identify clinically remarkable levels of ECG T-wave alternans

Automatic detection of T-wave alternans (TWA) from Holter ECG recordings is a non-invasive, clinically useful tool to identify patients at risk of malignant arrhythmias. A previously proposed time-domain approach, based on correlation method (CM), provides TWA characterization in terms of local measurements of single T-wave oscillation amplitudes, and global measurements of TWA-episode duration, amplitude, and magnitude (duration times amplitude). A local threshold criterion (LTC) implemented in CM algorithm is supposed to avoid false positive TWA (TWA+) due to noise. The present study was designed to set up a new global threshold criterion (GTC), based on TWA magnitude, to be contrasted to LTC. To this aim, the CM was applied to a population of 150 coronary artery diseased patients (CAD-group), who are known to show increased levels of TWA, compared with a control population of 150 healthy subjects (H-group). Our new GTC identified 28 TWA+ cases among CAD-patients and 1 among H-subjects, whereas, much higher numbers (119 CAD-patients and 61 H-subjects) were detected by LTC. High number of TWA+ among H-subjects is clinically unacceptable. Thus, involvement of GTC in the CM algorithm is decidedly more reliable for identification of TWA+ cases than LTC, which appears prone to false TWA+ prediction.