Response to Dr. Selvaraj's Comments on the “Assessment of Physiological Amplitude, Duration and Magnitude of ECG T‐Wave Alternans”

We thank Dr. Selvaraj for his interest in our recent paper and we appreciate the opportunity to further discuss our methodology. A preliminary point to be made for clarity is that the application of our adaptive match filter (AMF) to an ECG tracing provides an output consisting of an amplitude‐modulated sinusoidal signal, referred to as T‐wave alternans (TWA) signal, which allows estimation of not only TWA amplitude (TWAA; μV), but also TWA duration (TWAD; beat) and magnitude (TWAM = TWAA · TWAD; μV · beat). ¹ , ² , ³ This parameterization allows quantification of both transient (short TWAD) and stable (long TWAD) TWA. ¹ The TWA normality region defined in Ref. (2) relies on an appropriate definition of threshold (at the verge of abnormality) values for these parameters.

Cast doubt on the ability of our AMF method to control for noise, which might result in noise‐driven false identification of “physiological” TWA levels, may arise from our unclear description of Methods, for which we apologize. We clarify here that, beyond removal of both artifacts and non‐sinus beats during preprocessing, a further control for noise is accomplished in our AMF algorithm by a test requiring that a TWA‐signal amplitude persists in being greater than 5 μV for at least 3.5 cycles (i.e., 7 consecutive beats), before a positive‐TWA output is produced. ⁴ The ability of our AMF method to avoid noise‐driven false positive‐TWA receives further support from the supplementary analysis suggested by Dr. Selvaray as follows. White random noise, with amplitude ranging from 0 to 100 μV, was added to three synthetic tracings, obtained by 128‐fold repetition of a basic ECG complex from one healthy subject. Three TWA cases, simulated by changing T‐wave amplitudes, were considered: NO_TWA (TWAD = 0 beat, TWAA = 0 μV, TWAM = 0 beat ·μV); TRANSIENT_TWA (TWAD = 59 beat, TWAA = 35 μV, TWAM = 2065 beat ·μV, as found in our healthy population ² ); and STABLE_TWA (TWAD = 128 beat, TWAA = 35 μV, and a TWAM = 4480 beat ·μV). Results are shown in Figure 1. In all cases, the AMF provides an accurate estimation of the simulated TWA (8% maximum error, over the considered noise range).

Figure 1.

AMF‐estimated (thick solid line) and simulated (thin solid line) values of TWAD, TWAA, and TWAM, plotted as a function of the maximum amplitude of white random noise. In cases where estimated and simulated quantities superimpose, only the thick solid line is visible.

Beyond noise, respiration may also cause false TWA detections. ⁴ In particular, when an ECG is recorded in resting, supine condition, as it was the case for the participants in our study, ² respiration is expected to be a periodic (about 15/min), stationary signal. Accordingly, in the event of breathing period including an even number of cardiac cycles, a false detected TWA would be stable. Unlike this event, in our population, only transient TWA was observed.

In conclusion, on the basis of the observations reported here, our hypothesis on the existence of physiological TWA appears reinforced.