Higher heart rate variability as a predictor of atrial fibrillation in patients with hypertension

Abstract

Introduction

Atrial fibrillation (AF) is the most common cardiac arrhythmia requiring medical therapy¹. Various pathophysiological mechanisms for the development of AF have been studied². Recently, there has been an increasing evidence that the dysfunction of the autonomic nervous system (ANS) including the sympathetic and parasympathetic nervous systems, and the interaction between sympathetic and parasympathetic nervous systems are involved in the pathogenesis of AF³.

Despite the increasing evidence of an association between dysfunction of ANS and AF, whether abnormalities in the ANS can predict the development of AF remains unclear. Several population-based studies showed that lower heart rate variability (HRV) was associated with an increased risk of new-onset AF^4–6. However, another study reported that higher HRV value was associated with the incident AF⁷. The electrocardiography (ECG) recording time used to analyze HRV and analysis methods differed among the studies, leading to inconsistent results.

Hypertension is the most common cardiovascular risk factor in patients with AF⁸. Previous studies reported that HRV was not only associated with cerebrovascular diseases such as stroke and myocardial infarction but also relatively mild disease such as hypertension and anxiety disorders^9–11. The objective of this study was to investigate whether HRV using 24-h Holter monitoring could predict the development of AF in patients with hypertension.

Methods

Results

Among the 2100 patients who underwent 24-h Holter monitoring between May 2018 and April 2019, 480 patients were excluded due to the following: 120 patients were younger than 18 years old, 191 patients had persistent AF, 73 patients had paroxysmal AF, 5 patients had no HRV data, 3 patients had permanent pacemaker, 14 patients had complete atrioventricular block, 2 patients had insufficient recording, and 72 patients performed Holter monitoring repeatedly (Fig. 1). After excluding 480 patients who met the exclusion criteria and 838 patients without hypertension, 782 patients with hypertension were finally analyzed.

Open in a separate window

Figure 1

Enrollment of patients in the study. AF, atrial fibrillation; HRV, heart rate variability.

Risk factors for AF occurrence

During an average follow-up of 1.1 years, 44 patients developed AF. Among the 44 patients who developed AF during the follow-up period, 27 (61.4%) patients were found to have paroxysmal AF and 17 (38.6%) patients were found to have persistent AF at the initial AF identification. Of the 27 patients found to have paroxysmal AF, 10 patients had a history of AF, and 10 of the 17 patients found to have persistent AF had a history of AF.

The univariate analysis showed that the traditional risk factors such as age, coronary artery disease, hemodialysis, and chronic heart failure were associated with an increased risk of AF (Table ​(Table4).4). Several HRV parameters including VLF (P = 0.002), LF (P < 0.001), HF (P < 0.001), rMSSD (P < 0.001), and pNN50 (P < 0.001) were associated with AF occurrence. History of AF, the mean heart rate, and PAC count was also associated with AF occurrence. ROC analysis for HF, rMSSD, pNN50, and PAC count as predictors of AF occurrence revealed areas under the curve of 0.7033, 0.7045, 0.6911, and 0.7060, respectively (all P < 0.001, Fig. 2). The best HF cut off value of 11.1 for AF occurrence resulted in a sensitivity of 65.9% and a specificity of 62.9%. The sensitivity and specificity of the rMSSD at a cut off value of 29.5 for AF occurrence were 61.4% and 62.9%, respectively. In the case of pNN50 at a cut off value of 7.0, the sensitivity and specificity were 61.4% and 58.5%, respectively. Satisfying any one of these three HRV cut off values was associated with the occurrence of AF (P < 0.001, Table ​Table5).5). The Kaplan–Meier estimates of AF occurrence according to any risk of HRV parameters are presented in Fig. 3 (log rank P < 0.001). Cox regression analysis showed that older age, higher PAC burden, hemodialysis, coronary artery disease, history of AF, and any risk of HRV parameters were independent predictors for the occurrence of AF (Table ​(Table66).

Table 4

Univariate analysis of the occurrence of atrial fibrillation.

Variable	HR (95% CI)	P value
Age	1.050 (1.018–1.082)	0.002
Female gender	0.841 (0.463–1.527)	0.569
DM	1.490 (0.812–2.733)	0.198
CAD	2.219 (1.142–4.309)	0.019
CKD	1.540 (0.550–4.310)	0.411
Hemodialysis	6.298 (1.934–20.057)	0.002
History of CVA	1.467 (0.705–3.054)	0.306
Acute CVA	0.622 (0.307–1.260)	0.188
Chronic heart failure	2.700 (1.064–6.853)	0.037
Acute heart failure	3.257 (0.788–13.464)	0.103
History of AF	9.921 (5.478–17.969)	< 0.001
VLF	1.019 (1.007–1.032)	0.002
LF	1.031 (1.015–1.047)	< 0.001
HF	1.076 (1.044–1.108)	< 0.001
LF/HF ratio	0.509 (0.242–1.070)	0.075
SDNN	1.004 (0.998–1.010)	0.208
ASDNN	1.017 (1.008–1.025)	< 0.001
rMSSD	1.036 (1.023–1.049)	< 0.001
pNN50	1.057 (1.035–1.080)	< 0.001
BB50	1.000 (1.000–1.000)	0.003
Mean heart rate	0.960 (0.935–0.986)	0.003
PAC counts	1.000 (1.000–1.000)	< 0.001

Open in a separate window

DM, diabetes mellitus; CKD, chronic kidney disease; CAD, coronary artery disease; CVA, cerebrovascular accident; VLF, very low frequency; LF, low frequency; HF, high frequency; SDNN, standard deviation of the NN interval; ASDNN, average standard deviation of all 5-min RR intervals; rMSSD, square root of the mean squared differences of successive NN interval; pNN50, the percentage of RR intervals that are more than 50 ms different from the previous interval; BB50, the count of intervals that are more than 50 ms different from the previous interval.

Open in a separate window

Figure 2

ROC curve of heart rate variability and premature atrial contractions to predict the occurrence of atrial fibrillation. ROC, Receiver operating characteristic; HF, high frequency; rMSSD, square root of the mean squared differences of successive NN interval; pNN50, the percentage of RR intervals more than 50 ms different from the previous interval; PAC, premature atrial contraction; AUC, area under the curve.

Table 5

Relationship between HRV risk score according to cut–off value and occurrence of atrial fibrillation.

Variable	HR (95% CI)	P value
HF ≥ 11.1 (ms)	3.084 (1.653–5.753)	< 0.001
rMSSD ≥ 29.5 (ms)	2.528 (1.378–4.638)	0.003
pNN50 ≥ 7.0 (%)	2.261 (1.223–4.179)	0.009
One or more risk of HRVa	3.261 (1.648–6.452)	0.001
Risk score of HRVb
0 (n = 405)	1.0 (reference)
1 (n = 65)	4.249 (1.647–10.963)	0.003
2 (n = 54)	0.688 (0.086–5.173)	0.699
3 (n = 258)	3.597 (1.770–7.310)	< 0.001
PACs ≥ 145 beats for 24 h	3.788 (2.031–7.068)	< 0.001

Open in a separate window

HRV, heart rate variability; HF, high frequency; rMSSD, square root of the mean squared differences of successive NN interval; pNN50, the percentage of RR intervals that are more than 50 ms different from the previous interval; PAC, premature atrial contraction.

^aIndicates that at least one of the above mentioned abnormal HRV parameters (HF, rMSSD, and pNN50) were present.

^bSum of the above mentioned abnormal HRV parameter (HF, rMSSD, and pNN50).

Open in a separate window

Figure 3

Kaplan–Meier estimate of AF-free survival according to the presence of any risk of HRV. AF, atrial fibrillation; HRV, heart rate variability. Any risk of HRV meant that at least one of the abnormal HRV parameters (HF, rMSSD, and pNN50) were present. AF, atrial fibrillation; HRV, heart rate variability.

Table 6

Multivariate analysis of the occurrence of atrial fibrillation.

Variable	HR (95% CI)	P value
Age	1.032 (1.001–1.065)	0.043
Female gender	0.667 (0.359–1.240)	0.201
PAC counts	1.000 (1.000–1.000)	< 0.001
Hemodialysis	3.853 (1.130–13.139)	0.031
CAD	2.074 (1.015–4.241)	0.045
Chronic heart failure	0.710 (0.261–1.932)	0.502
History of AF	7.738 (4.202–14.250)	< 0.001
One or more risk of HRVa	2.396 (1.167–4.919)	0.017

Open in a separate window

PAC, premature atrial contraction; CAD, coronary artery disease; HRV, heart rate variability.

^aIndicates that at least one of the abnormal HRV parameters (HF, rMSSD, and pNN50) were present.

Discussion

The main finding of this study was that higher HRV in patients with hypertension could predict the development of AF independent of demographics or known cardiovascular risk factors. HF, rMSSD, and pNN50 were notably associated with the occurrence of AF, along with age, hemodialysis, and high PAC burden, which are traditional risk factors for AF. Our findings suggest that higher HRV representing abnormal autonomic fluctuation is associated with a higher risk of AF development. Separately, patients with hypertension had more comorbidities and lower HRV than those without hypertension.

The ANS plays an important role in the initiation and maintenance of AF through atrial electrical remodeling^12,13. A previous study showed that increased vagal activity could cause shortening of the atrial effective refractory periods¹². Another study revealed that AF could be initiated by premature beats during vagal stimulation¹⁴. Moreover, autonomic fluctuations preceding AF are common^15,16. Most studies evaluating these phenomena used HRV parameters to estimate dysfunction of ANS^15,16. HRV is a quantitative method used to measure the balance between the sympathetic and parasympathetic nervous system^17,18. HRV indicates fluctuations in autonomic inputs to the heart rather than the mean level of autonomic tones. Therefore, not only lower HRV values, but also higher HRV values might indicate abnormal conditions.

Two previous long-term follow-up studies showed that lower HRV was associated with an increased risk of AF^4,5. However, another study reported that the association between lower SDNN and incident AF was not significant after adjusting for traditional risks factor for AF⁶. One recently published study showed that not only lower HRV but also higher rMSSD was associated with the development of AF⁷. These inconsistent results might be due to differences in ECG recording time and the study population used in HRV analysis. In our study, higher HRV parameters were associated with the development of AF, contrary to the results of some previous studies. There are several possible explanations for these differences. First, our study focused on hypertensive patients, unlike previous studies using general population^4–6. Hypertensive patients generally have lower HRV than healthy people¹⁰. In our study, patients with hypertension had more comorbidities and lower HRV, consistent with other studies (Supplementary Tables ^S1 and ^S2). Since the mean HRV value might be higher in the general population than in hypertensive patients, higher HRV values might not be significantly different enough to predict AF. Second, we used 24-h Holter monitoring to measure HRV. In general, the time domain methods are ideal for long-term ECG recording, whereas the frequency domain methods are preferred to the time domain methods for short-term recording¹⁷. In the two previous long-term follow-up studies, short-term (2-min⁴ and 45-min⁵) ECG recording was used. Therefore, the HRV parameters measured by time domain methods were more appropriate in our study. Third, a closer look at the previous Atherosclerosis Risk in Communities (ARIC) cohort study reported that lower HRV was associated with an increased risk of AF, a figure in that report showed that higher HRV values were also associated with the occurrence of AF. However, this association between higher HRV values and the occurrence of AF was not addressed in previous ARIC cohort study⁴.

Parasympathetic dominance is associated with an increased propensity for AF, and vagal-mediated paroxysmal AF is preferentially seen in young individuals who have structurally normal heart¹⁹. Generally, HF components are predominantly modulated by the parasympathetic nervous system, whereas LF components are considered a marker of sympathetic modulation^20,21. HRV represents the degree of autonomic fluctuations rather than the mean level of autonomic tones. Therefore, relatively higher HRV could be considered excessive fluctuations in the ANS rather than a physiologic condition. In other words, even though hypertensive patients generally had low HRV than those without hypertension, hypertensive patients with comparatively higher HRV might be considered to have disproportionate fluctuations in the ANS. A previous study⁷ showed that higher rMSSD was associated with incident AF could be interpreted from this point of view.

We enrolled patients with sinus rhythm at baseline 24-h Holter monitoring. Therefore, patients with paroxysmal AF were enrolled in this study. Sixty-six of 782 (8.4%) hypertensive patients had a history of AF. Perhaps more patients with a history of AF might be included due to its asymptomatic feature. Among 66 patients with a history of AF, AF recurred in 20 patients. A history of AF was an independent risk factor for AF occurrence as well as higher HRV in this study. When we examine a patient in an outpatient clinic, it is difficult to know whether the patient has a history of AF. Therefore, it is important to predict the occurrence of AF regardless of a history of AF, and the HRV value would be one of the factors that could predict the occurrence of AF.

This study had several limitations. First, this was a retrospective cohort study. We could not fully evaluate many other risk factors for AF, including obstructive sleep apnea, body weight, alcohol consumption, or physical activity. Second, the duration of follow-up was relatively short and the degree of effort to identify AF varied from patient to patient. Since most patients with AF are asymptomatic, it is difficult to detect AF with an intermittent ECG test. Regular 24-h Holter monitoring could not be performed due to the limitations of the retrospective study. Therefore, AF incidence in our study was low. Third, our study enrolled patients undergoing hemodialysis. Because blood pressure and autonomic tone change before and after dialysis, short-term ECG recordings are generally preferred to avoid environmental influences. However, since the number of dialysis patients in our study was very small, we analyzed HRV data using 24-h Holter recordings. Finally, since HRV parameters measured by frequency domain methods were made by 24-h Holter recording, there might be a problem of stationarity.

In conclusion, higher HF, rMSSD, and pNN50 in patients with hypertension as surrogate markers for excessive autonomic fluctuation could predict the occurrence of AF. Frequent AF screening should be considered for hypertensive patients with higher HRV values and other risk factors for AF. Further large scale, prospective studies are needed to verify our findings.

Supplementary Information

Author contributions

K.J.C., S.H.K., and K.R.L. conceptualized, designed the research and drafted the manuscript. S.H.K. and K.J.C. collected data. K.R.L. and K.J.C. analyzed data. K.J.C., J.H.S., D.R.R., B.K.L., and B.R.C. performed research and interpreted the data. S.H.K and K.R.L contributed equally to this work.

Funding

This study was supported by a research grant (Grant No. 2020-001) funded by Gangwon Heart Association.

Data availability

The datasets used and/or analyzed in the present study can be shared on reasonable request.

Competing interests

The authors declare no competing interests.

Footnotes

References