Author + information
We welcome the opportunity to respond to the letter from Dr. Stein regarding our article (1). Most of the considerations of this letter are related to Table 3 in our study. These data were only provided for general information about the patients with sudden death and those with arrhythmic events or free of events. Because the duration of follow-up is not taken into account, definitive conclusion based on these data would certainly be inadequate. Some of these patients may have low HRV, and may therefore be considered to have a poor prognosis, but were free of events because of a very short follow-up, as shown by the large standard deviation of the duration of follow-up in this group (53 ± 41 months). Therefore, analysis of variance was not performed on these data. Moreover, we do not consider that rmsSD is really higher in patients with sudden death (29 ± 12 ms) as compared with patients free of events (25 ± 15 ms), and it is exactly similar in patients with arrhythmic events (25 ± 8 ms). The nonparametric Mann-Whitney Utest on these data was also inadequate, but found a nonsignificant p value when comparing rmsSD between each group (p = 0.4 to 0.9), because the difference is low as compared with the standard deviation of this variable.
It may be surprising that SDA5 was superior to SDNN in patients with outcomes. However, from a mathematic point of view, it is certainly not impossible to see segments with a standard deviation of their means superior to the standard deviation of the overall series, particularly when the number of normal to normal intervals is different in each segment, whether the heart rate is different or after exclusion of ectopic beats or artifact. Table 1gives a simplified numeric example for confirmation.
Dr. Stein pertinently noticed that VLF was relatively low in each group. Data were log-transformed after reviewing in the last version of the article to achieve normal distribution. Calculation was performed on the data in ms instead of ms2for VLF, and we apologize for this mistake. In fact, VLF was 6.9 ± 1.1 ln ms2in patients free of events, 6.1 ± 1.5 ln ms2in patients with sudden death and 6.6 ± 1.3 ln ms2in patients with arrhythmic events; therefore, VLF was greater than LF and HF in each group. The ratio of VLF to total power was calculated on data in ms2before the log transformation, as in the report of Mortara et al. (2).
Several complex phenomena may indeed participate to HRV in clinical practice, in particular, in patients with congestive heart failure. We would be happy to collaborate with Dr. Stein to study “erratic sinus rhythm” with a specific research software, because the analogic tape of the ambulatory electrocardiograms of our patients is still available. However, from a practical standpoint, we consider that an important issue of our article is that our results were obtained with a commercially available Holter monitoring system, and with demonstration that SDNN, a simple and reproducible measurement of HRV, is probably the most powerful for risk stratification. These findings may have more direct implications for identifying patients at risk and may guide subsequent investigations and therapy, either on an individual basis or as part of a controlled clinical trial.
- American College of Cardiology
- Fauchier L.,
- Babuty D.,
- Cosnay P.,
- Fauchier J.P.
- Mortara A.,
- Sleight P.,
- Pinna G.D.,
- et al.