Impact of Heart Rate Fragmentation on the Assessment of Heart Rate Variabilityâ€
Round 1
Reviewer 1 Report
General comments:
The present study reports on the age and sex dependence of HRV and HRF metrics. Though overlooked for many years, the high frequency (HF) component of the sinus rhythm is now being employed to identify also trends in cardio-electrophysiology as traditionally done with LF bands Within this regard, the authors of the current manuscript presented a well-documented study in terms of ECG recordings selection and methodology for data analysis, and across a large dataset spanning entirely the human age population. The authors were able to correlate some metrics of HRF with their initial assumptions, from where they draw a conclusion that HRF has a larger impact at ages 60 – 90. Additionally, the authors also point some physiological aspects that may explain this behaviour on the Discussion section, as well as some literature references to other studies on the field of HRF.
Before acceptance of the current manuscript, I would like to invite the authors to reply to some of the following questions, in the hope that they can still improve the quality of this work:
Major Questions:
1 – The data from this study was solely based on a database from a 24-h ECG Holter system. Can it be also applied to other equipment/systems? Would the results be similar if, for instance, shorter ECG recordings (less than 24-h) were used? This question is related to the practicability/convenience of patient’s wearing ECG recording systems for such a long time.
2 – The authors should also discuss hypothetical strategies for preventing HF increase in the elderly (lifestyle).
3 – Any thoughts or insights about the influence of HRF for patients suffering from heart conditions?
Minor Questions:
- Line 32: “Question 1” should be replaced by “Question 1:”
- Line 33: “Decreased HRV is increased post-infarction mortality risk…” Please review the sentence because currently it does not make sense.
- Line 37: “Question 2” should be replaced by “Question 2:”
- Figure 1: The frequency interval for HF is incorrect. Should it be 0.15 – 0.4 Hz?
- Line 90: “per” should be in italic
- Line 101: “fast Fourie transforms” replace by “Fast Fourier Transform”
- Line 116: Why 8 ms? Only because it is related to the sample rate (125 Hz) or is there a more scientific reason?
- Line 128: “24 h was calculate” replace by “24 h was calculated”
- Table 2: “The variables are explained in Table I and caption of Figure 7”
- Table 3: “The variables are explained in Table I and caption of Figure 7”
- Line 169: “was greater” by “were greater”
- Line 176 – 177: This sentence needs to be reviewed, for instance: “To estimate the impact of PIP on the interindividual differences in LF and HF power at each age, the correlations of HRV power with PIPh and PIPs are shown in Table 3.”
- Line 238: “All ECG data was recorded for some clinical purpose, although only data without ECG abnormalities was selected …”
- Line 244: “is more pronounced as the higher…” replace by “is more pronounced the higher…”
Author Response
Major Questions:
1 – The data from this study was solely based on a database from a 24-h ECG Holter system. Can it be also applied to other equipment/systems? Would the results be similar if, for instance, shorter ECG recordings (less than 24-h) were used? This question is related to the practicability/convenience of patient’s wearing ECG recording systems for such a long time.
> We added a discussion about this point in the limitation (lines 287-293).
Another potential limitation is that the present findings were obtained from 24-h ECG data. Thus, they may not apply to other equipment or systems. HRF itself, however, could occur in short ECG recordings such as those used for autonomic function assessment by short-term (typically 5 min) HRV [3]. The LF and HF powers obtained from such analyses also need to consider the influence of HRF, especially in elderly subjects. On the other hand, the present observations may not apply to heartbeat interval data estimated from pulse waves. The evaluation of HRF depends on the accuracy of interval measurement, which may not be enough with pulse wave signals.
2 – The authors should also discuss hypothetical strategies for preventing HF increase in the elderly (lifestyle).
> We added a paragraph to discuss the underlying mechanisms of increased HF and PIPh in the elderly along with their clinical significances (lines 251-265).
Due to the cross-sectional nature of this study, the causal links between changes in HRF and HRV metrics are not completely clear. In short, the increase in PIPh and Wh3 in the elderly could be the result of increased HF, although this is not consistent with weak correlations between PIPh and HF before age 20. Additionally, aging is accompanied by degenerations of the regulatory network, which has been thought to reduce physiological parasympathetic modulations, such as those controlling heart rate. This is expected to reduce HF and other short-term HRV metrics. Thus, the counterintuitive increase in HF in the elderly seems the result but not cause of increased HRF. Several mechanisms can be considered for the genesis of HRF [5], which include sinus node exit block, subtle atrial bigeminy originating near or within the sinoatrial (SA) node, modulated SA node parasystole caused by multiple interacting pacemaker sites in the SA node [16], and modulated periodicity of pacemaker clock in the SA node [17,18]. Thus, increased HRF may be a marker reflecting the age-related degeneration or pathologic impairment of the SA-node-atrial network for cardiac pacemaker function. Increased HRF in patients with coronary artery disease [5] and its association with adverse cardiovascular events [7] support this contention. Although it is interesting if HRF itself also plays a proarrhythmic role, particularly in the development of atrial fibrillation, it remains to be clarified.
3 – Any thoughts or insights about the influence of HRF for patients suffering from heart conditions?
> We discussed this point in the newly added paragraph mentioned above (lines 251-265).
Minor Questions:
- Line 32: “Question 1” should be replaced by “Question 1:”
> Line 32: We corrected the descriptions.
- Line 33: “Decreased HRV is increased post-infarction mortality risk…” Please review the sentence because currently it does not make sense.
> Line 34: We revised the sentence.
Decreased HRV is increased mortality risk after acute myocardial infarction.
- Line 37: “Question 2” should be replaced by “Question 2:”
> Line 37: We corrected the descriptions.
- Figure 1: The frequency interval for HF is incorrect. Should it be 0.15 – 0.4 Hz?
> Line 50: We corrected the error. Thank you!
- Line 90: “per” should be in italic
> Line 91: We changed it into italic.
- Line 101: “fast Fourie transforms” replace by “Fast Fourier Transform”
> Line 104: We corrected the error.
- Line 116: Why 8 ms? Only because it is related to the sample rate (125 Hz) or is there a more scientific reason?
> Line 124: Yes, it depends on the sampling rate. To clarify this, we wrote “… taking 125 Hz sampling frequency of the Holter ECG into consideration (1000 ms/125 Hz = 8 ms).”
- Line 128: “24 h was calculate” replace by “24 h was calculated”
> Line 135: We corrected the error.
- Table 2: “The variables are explained in Table I and caption of Figure 7”
> Line 186: We corrected the error.
- Table 3: “The variables are explained in Table I and caption of Figure 7”
> Line 190: We corrected the error.
- Line 169: “was greater” by “were greater”
> Line 176: We corrected the error.
- Line 176 – 177: This sentence needs to be reviewed, for instance: “To estimate the impact of PIP on the interindividual differences in LF and HF power at each age, the correlations of HRV power with PIPh and PIPs are shown in Table 3.”
> Line 192: We revised the sentence.
To estimate the impact of PIP on the interindividual differences in LF and HF power at each age, the correlations of HRV power with PIPh and PIPs were calculated (Table 3).
- Line 238: “All ECG data was recorded for some clinical purpose, although only data without ECG abnormalities was selected …”
> Line 284: We revised the sentence.
All ECG data were recorded for some clinical purpose(s), although only data without ECG abnormality were selected for this study.
- Line 244: “is more pronounced as the higher…” replace by “is more pronounced the higher…”
> Line 297: We corrected the error.
Thank you very much for the fine review.
Reviewer 2 Report
In the Data analysis part, the custom-made software is declared to use for the analysis with the reference to the recommended standard. However, there is no note if this software was somehow validated.
In the statistical analysis, there would be better to mention also a statistical test used for calculation results in Table S1. I did not find any notes about the type of test in the whole text. In connection with that, there would be proper to mention some test about the distribution of the tested groups. Further, there should be added the name of a tool where correlation coefficients and tests were calculated.
In the results and discussion part, it is very unclear the use of age-adjustment. There is no description, how this adjustment is used. It is very important because there are many results depended on this adjustment in Table S1. In other words, if the results in Table S1 are calculated for some specific ages (e.g. age 75 years and higher), that would be appropriate to mention this fact directly. This comment is connected with the previous one about used the statistical test.
There are differences between the text and the results in Table S1. Specifically, in the 3.1 is written: "Although there was no significant sex difference in age-adjusted mean in rMSSD or ULF, VLF, and LF were greater in men than in women and HF and VHF were greater in women than in men (Table S1)." However, if I find results for VLF and LF in Table S1, there is the p-value < 0.0001. Maybe, the sentence is intended otherwise, but in this form is confusing.
I suppose that in Tables 2,3, correlation coefficients presented in bold are statistically significant. But, if it is true, this fact is not mentioned in the description of tables. If the bold text represents something else then the statistical significance of the correlation should be added to results.
I missed some deeper interpretation of the correlation between PIPs, and LF and HF, respectively. There is not clear why this correlation is negative and stronger than the correlation of the PIPh. Maybe the correlation is not significant, but this is mentioned in the comment above. In any case, this different behavior of the PIPh and PIPs seems to be interesting.
Author Response
Comments and Suggestions for Authors
In the Data analysis part, the custom-made software is declared to use for the analysis with the reference to the recommended standard. However, there is no note if this software was somehow validated.
> Line 95: The software has been validated extensively with simulated R-R interval data including HRV components of known amplitude and frequency. We added a description of that.
In the statistical analysis, there would be better to mention also a statistical test used for calculation results in Table S1. I did not find any notes about the type of test in the whole text. In connection with that, there would be proper to mention some test about the distribution of the tested groups. Further, there should be added the name of a tool where correlation coefficients and tests were calculated.
> Line 136: We revised the paragraph of 2.3 statistical analysis.
SAS program package (SAS Institute, Cary, NC) was used for statistical analyses.
The gender effects on variables were evaluated after adjusting for age effects by the analysis of covariance with the SAS General Linear Model procedure.
The distributions of variables have been examined and the power of ULF, VLF, LF, and HF was transformed into natural logarithmic values to normalize the distributions (line 106).
In the results and discussion part, it is very unclear the use of age-adjustment. There is no description, how this adjustment is used. It is very important because there are many results depended on this adjustment in Table S1. In other words, if the results in Table S1 are calculated for some specific ages (e.g. age 75 years and higher), that would be appropriate to mention this fact directly. This comment is connected with the previous one about used the statistical test.
> Line 141: We added a description of the method for the adjustment of age effects. We performed an analysis of covariance with the SAS GLM procedure.
There are differences between the text and the results in Table S1. Specifically, in the 3.1 is written: "Although there was no significant sex difference in age-adjusted mean in rMSSD or ULF, VLF, and LF were greater in men than in women and HF and VHF were greater in women than in men (Table S1)." However, if I find results for VLF and LF in Table S1, there is the p-value < 0.0001. Maybe, the sentence is intended otherwise, but in this form is confusing.
> Line 163: We revised the sentence.
There was no significant gender difference in rMSSD or ULF after adjusting for age effects, but VLF and LF were greater in men than in women and HF and VHF were greater in women than in men (Table S1).
I suppose that in Tables 2,3, correlation coefficients presented in bold are statistically significant. But, if it is true, this fact is not mentioned in the description of tables. If the bold text represents something else then the statistical significance of the correlation should be added to results.
> Lines 185 and 189: We added the descriptions about the statistical significances of correlation coefficients and explained the meaning of bold-faced values.
I missed some deeper interpretation of the correlation between PIPs, and LF and HF, respectively. There is not clear why this correlation is negative and stronger than the correlation of the PIPh. Maybe the correlation is not significant, but this is mentioned in the comment above. In any case, this different behavior of the PIPh and PIPs seems to be interesting.
> Lines 236-250: We added a paragraph to discuss this point.
In the present study, we observed a negative correlation between PIPh and PIPs (Table 2). Also, while PIPh showed positive correlations with LF and HF in all age groups, PIPs showed strong negative correlations with LF and HF (Table 3). These indicate that the HRF metric shows the opposite behavior depending on the definition of the inflection point. Although the soft inflection point is defined as the point at which a change in consecutive NN intervals is preceded or followed by unchanged consecutive NN intervals, even changes in consecutive NN intervals if they were below the detection threshold determined by ECG sampling frequency are not detected and are judged unchanged. Conversely, if the change exceeds the threshold, the points that had been defined as a soft inflection point could change to a hard inflection point or a non-inflection point. Therefore, increasing the amplitude of HRF could increase the points where the judgment changes from soft to hard, creating a complementary relationship between PIPs and PIPh. Similarly, increasing the amplitude of LF and HF could increase the points where the judgment changes from soft to non-inflection, creating negative correlations between PIPs and these HRV metrics. Although PIPs may help characterize the dynamic features of NN interval inflections, PIPs itself does not seem to be a measure of the degree of fragmentation, i.e., HRF.
Thank you very much for your fine review.
