Low-Frequency/High-Frequency Ratio Is a Predictor of Death and Hospitalization in Patients on Maintenance Hemodialysis: A Prospective Study

Abstract

Introduction

This study aimed to evaluate the predictive value of the low-frequency/high-frequency (LF/HF) ratio in all causes of death and hospitalizations in maintenance hemodialysis (MHD) patients.

Methods

This is a single-center prospective study with a 48-h electrocardiograph (ECG) recording. A total of 110 patients were enrolled in the study from October 1, 2021, to September 30, 2022. ECG recordings started before initiation of the hemodialysis (HD) session and lasted for 48 h, covering the intra- as well as inter-HD period. We divided our participants into two groups based on the median value of LF/HF, one of the frequency domain parameters of heart rate variability (HRV). Patients with LF/HF <1.33 were categorized as group A and those with LF/HF ≥1.33 were group B. The endpoint of the study was a composite event of death or hospitalization. We followed all patients until the composite endpoint or the end of the study on February 28, 2023. Multivariate Cox regression was used to assess the adjusted effect of LF/HF on the composite endpoint.

Results

Patients in group A were older and the number of patients with diabetes was more than that of group B. With regards to the laboratory data, group A had lower serum creatinine and uric acid and higher ferritin and NT-ProBNP. In the index HD session, systolic blood pressure was higher but diastolic blood pressure was significantly lower in group A. During the median follow-up period of 8.8 (7.6–9.8) months, 27 hospitalizations and 10 deaths were documented. Increased LF/HF ratio was an independent protective factor of composite endpoint events (HR = 0.357, 95% CI: 0.162–0.790, p = 0.011).

Conclusion

Risks of mortality and hospitalizations are higher among HD patients having decreased LF/HF ratios. LF/HF in the 48-h recording can be considered as a prognostic factor for risk stratification in HD patients.

Introduction

Hemodialysis (HD), the prime therapy for end-stage renal disease (ESRD) patients, has largely improved the quality of patients, but the hospitalization rate and mortality rate are still high. The major cause of death observed in HD patients is cardiovascular disease (CVD) [1]. According to the 2022 United States Renal Data System (USRDS) annual report, there were 807,920 patients suffering from ESRD. In 2020, the rate of hospitalizations was 1.48 per person-year, although the rate has decreased in recent years [2]. ESRD is a known predictor of a short life span. Evaluation of risk for CVD events and hospitalizations would be a beneficial approach to improve the prognosis of maintenance hemodialysis (MHD) patients.

Heart rate variability (HRV), including time domain and frequency domain parameters, is a noninvasive measure to assess cardiac autonomic function in ESRD patients [3]. An analysis of frequency domain parameters demonstrated that both sympathetic and parasympathetic functions were impaired in patients suffering from chronic kidney failure [4]. LF/HF, a ratio of low frequency to high frequency, is thought to operationalize parasympathetic and sympathetic balance [5]. LF/HF has been observed to be more strongly associated with sudden cardiac death in HD patients with left ventricular hypertrophy, compared with the other frequency parameters [6]. As for patients receiving peritoneal dialysis, decreased LF/HF was also a predictor of mortality even with adjusted clinical parameters [7]. However, the value of LF/HF for predicting the mortality and CVD events did not reach an agreement [8, 9]. Chandra et al. [8] indicated that lower LF/HF values were the predictors of adverse cardiovascular and renal outcomes in the non-dialysis CKD patients. On the contrary, a prospective study of dialysis cohort conducted by Kuo et al. [9] demonstrated that a lower LF/HF value was associated with better survival.

Hospitalized dialysis patients are associated with a poor prognosis and also an increased clinical and economic burden [10]. Few HRV data are available on hospitalization events in HD patients. The evaluation of LF/HF as a predictive variable for future hospitalizations remains unclear. Above all, additional research investigating the association between LF/HF and adverse outcomes defined as all causes of death and hospitalizations are warranted. There are limited data available on HRV with regards to HD sessions and interdialytic period. Our study aimed to explore the independent predictive values of LF/HF on death and hospitalization events in MHD patients with 48-h electrocardiograph (ECG) recording.

Materials and Methods

This prospective study was conducted at a single HD center. The Ethics Committee of Peking University People’s Hospital approved the study with the IRB number 2023PHB101-001, and informed consent was obtained from all participants.

Study Population

Adult patients who received regular HD during the period from October 1, 2021, to September 30, 2022, with regular follow-up until February 28, 2023, were included in the study. Subjects with cardiac arrhythmia and implanted cardiac pacemakers were excluded from the study. The patients received thrice weekly HD, with each session lasting for 3.5–4.5 h, the blood flow rate was 200–300 mL/min, and the dialyzate flow rate was 500 mL/min. In all HD sessions, the dialyzate potassium was 2 or 3 mmol/L, calcium was 1.25 or 1.50 mmol/L, and bicarbonate was 32–36 mmol/L. Demographic data were collected from medical records. Blood samples were obtained and analyzed to get the laboratory data before dialysis session. Single pool kt/Vurea (spKt/Vurea) was evaluated as a marker of dialysis adequacy and was calculated using the Daugirdas equation [11]. Charlson comorbidity index (CCI) was calculated to evaluate comorbid conditions [12].

Echocardiography

Echocardiograms were performed on the non-dialysis days. All echocardiographic measurements were conducted by a well-trained physician in accordance with the recommendations of the American Society of Echocardiography [13]. Left ventricular mass, which was calculated using the Devereux formula, was correlated with the body surface area [14].

Heart Rate Variability

HRV parameters which were calculated using a three-channel recorder were evaluated for 38–48 h in all participants. The recording commenced at the initiation of the HD session with the patient placed in a supine position and ended either at the beginning of the next HD session or after 48 h. The spectral components of the frequency domain were calculated using fast fourier transform [15]. LF/HF ratio was calculated based on the frequency parameters, which were low frequency (0.04–0.15 Hz, LF) and high frequency (0.15–0.40 Hz, HF). The participants were divided into two groups based on the median value of LF/HF. The group with the lower values was called group A, and the other participants were categorized into group B.

Outcomes

The end point was a composite of death or the first hospitalization. Hospitalization events included hospitalization due to any cause. Death was all causes of death. All participants were followed up until the composite end point or the end of the study.

Statistical Analysis

The Shapiro-Wilk test was used to test the normality of the data. Data were presented as mean ± standard deviation or median (25th–75th percentile). Group differences were compared using either Student’s t-test or Mann-Whitney U test. Categorical variables were compared using the χ² test. The relationship between LF/HF and the composite end point was examined using Kaplan-Meier method and the log-rank test. A multivariate Cox regression model was used to determine predictors of the composite end point. p < 0.05 was considered statistically significant. SPSS 27.0 was used for statistical analyses.

Results

A total of 110 MHD patients were enrolled in the study. Figure 1 depicts the flowchart for the study selection.

Fig. 1.

Flowchart of participants. ECG, electrocardiogram; HRV, heart rate variability.

Baseline Characteristics

The median value of LF/HF was 1.33, and the distribution of LF/HF values are shown in Figure 2. Baseline characteristics of all the participants have been described in Table 1. The mean age of patients in group A was older than patients in group B (63.57 ± 13.22 vs. 52.2 ± 14.13). The incidence of diabetes in group A, which was 42.6%, was significantly higher than in group B with 23.2% patients having diabetes. CCI was found to be significantly higher in group A. Compared with group B, pre-dialysis systolic blood pressure and post-dialysis systolic blood pressure were both higher in group A, on the contrary, both the pre-dialysis diastolic blood pressure and post-dialysis diastolic blood pressure were lower. Analysis of the results indicated that patients in group A had lower levels of serum creatinine and uric acid and significantly higher NT-ProBNP levels compared with the other group. Patients with LF/HF <1.33 had a higher degree of ferritin. Regarding the echocardiography variables, group A had higher left ventricular mass index, but the difference was not significant.

Fig. 2.

Distributions of LF/HF value.

Table 1.

Baseline characteristics

	Group A (n = 54)	Group B (n = 56)	p value
Age, years	63.57±13.22	52.2±14.13	<0.001
Male, n (%)	34 (63)	39 (69.6)	0.459
Dialysis vintage, months	92 (42.25, 144.75)	57.5 (24.75, 119.75)	0.106
Hypertension, n (%)	40 (74.1)	45 (80.4)	0.432
Diabetes, n (%)	23 (42.6)	13 (23.2)	0.03
CAD, n (%)	10 (18.5)	6 (10.7)	0.246
CCI	4.31±1.53	3.3±1.24	<0.001
BMI, kg/m2	22.98±3.31	23.22±4.77	0.76
spKt/Vurea	1.50±0.24	1.47±0.25	0.446
LF/HF	0.85±0.04	2.36±0.16	<0.001
Pre-HD SBP, mm Hg	160.54±19.14	148.06±18.88	0.001
Pre-HD DBP, mm Hg	72.65±17.43	78.7±13.479	0.046
Post-HD SBP, mm Hg	157.85±18.79	150.3±16.65	0.029
Post-HD DBP, mm Hg	73.11±12.75	84.44±11.63	<0.001
Hb, g/L	110.35±11.49	112.13±8.20	0.352
Alb, g/L	39.02±2.26	39.73±2.69	0.135
Urea, mmol/L	26.42±8.20	26.46±5.27	0.976
Scr, µmol/L	879.09±179.51	1,089.18±227.19	<0.001
UA, µmol/L	352.76±77.18	387.66±93.39	0.035
Na, mmol/L	136.82±3.12	137.85±2.59	0.062
K, mmol/L	4.70±0.75	4.81±0.64	0.397
Ca, mmol/L	2.26±0.17	2.28±0.21	0.57
P, mmol/L	1.59±0.46	1.64±0.41	0.658
PTH, pg/mL	209.55 (130.32, 422)	238.85 (126.6, 357.65)	0.938
25(OH)D, nmol/L	73.98 (32.27, 123.6)	73.53 (37.34, 113.55)	0.724
Ferritin, ng/mL	320.6 (168.35, 456.15)	228.85 (159.02, 379.55)	0.047
lnNT-ProBNP, pg/mL	8.91±1.16	7.82±0.97	<0.001
LVEF, %	64.59±9.74	65.78±9.50	0.55
LVMI, g/m2	110.87±35.30	98.32±28.02	0.058

Group A, LF/HF <1.33; group B, LF/HF ≥1.33; CAD, coronary artery disease; CCI, Charlson comorbidity index; BMI, body mass index; pre-HD SBP, pre-hemodialysis systolic blood pressure; pre-HD DBP, pre-hemodialysis diastolic blood pressure; post-HD SBP, post-hemodialysis systolic blood pressure; post-HD DBP, post-hemodialysis diastolic blood pressure; Hb, hemoglobin; Alb, albumin; Scr, serum creatinine; UA, uric acid; Na, sodium; K, potassium; Ca, calcium; P, phosphorus; PTH, parathyroid hormone; 25(OH)D, 25-hydroxy vitamin D; NT-ProBNP, N-terminal pro B-type natriuretic peptide; LVEF, left ventricular ejection fraction; LVMI, left ventricular mass index.

Outcomes

During a median 8.8-month (7.6–9.8) follow-up period, 37 composite endpoint events were documented in the participants. The details which included 10 deaths and 27 hospitalizations have been listed in the Table 2. All deaths in group B were caused due to CVD events. In group A, five out of eight deaths were caused by CVD events, and remaining three were caused by infections. Hospitalizations occurred in 15 patients (27.78%) in group A and 12 patients (21.43%) in group B. A higher incidence of CVD and peripheral arterial disease was found in group A.

Table 2.

Incidence of endpoint events

	Total	LF/HF <1.33	LF/HF ≥1.33
Patients, n	110	54	56
Composite endpoint events, n (%)	37 (33.64)	23 (42.59)	14 (25)
Death, n (%)	10 (9.09)	8 (14.81)	2 (3.57)
CVD events, n (%)	7 (6.36)	5 (9.26)	2 (3.57)
Infection, n (%)	3 (2.73)	3 (5.56)	–
Hospitalizations, n (%)	27 (24.55)	15 (27.78)	12 (21.43)
CVD events, n (%)	14 (12.73)	9 (16.67)	5 (8.93)
PAD, n (%)	6 (5.45)	4 (7.41)	2 (3.57)
Infection, n (%)	4 (3.64)	1 (1.85)	3 (5.36)
Others, n (%)	3 (2.73)	1 (1.85)	2 (3.57)

Others included cancer and gastrointestinal hemorrhage.

PAD, peripheral arterial disease.

Figure 3 shows the Kaplan-Meier curves for outcome of the composite end point in participants stratified based on the median value of LF/HF. Kaplan-Meier analysis indicated that the estimated cumulative composite endpoint event probability was significantly higher in the LF/HF <1.33 group.

Fig. 3.

Kaplan-Meier curves for outcome of death and hospitalizations in participants stratified according to the median value of LF/HF.

The univariate Cox proportional hazards regression model analysis demonstrated that, LF/HF (HR = 0.422, 95% CI: 0.198–0.901), coronary artery disease (CAD) (HR = 4.136, 95% CI: 1.368–12.504), age (HR = 0.967, 95% CI: 0.943–0.99), and post-HD diastolic blood pressure (HR = 0.963, 95% CI: 0.94–0.987) were the factors associated with the composite end point of death or hospitalization. Diabetes (HR = 1.487, 95% CI: 0.769–2.876) and dialysis vintage (HR = 1.003, 95% CI: 0.999–1.007) were not associated with risk of the composite end point. Detailed results of univariate analysis were shown in Table 3. Multivariate analysis demonstrated that the history of CAD and LF/HF, after being controlled for other risk factors, were independent predictors of composite endpoint events in MHD patients. In multivariate Cox model, coronary artery disease was associated with a death and hospitalization risk of 3.275 (95% CI, 1.329–8.075). Lower LF/HF was an independent predictive factor of mortality or hospitalization events (Table 3).

Table 3.

Results of univariate and multivariate Cox models

	Univariate			Multivariate
	HR	95% CI	p value	HR	95% CI	p value
Age, years	0.967	0.943–0.99	0.006			0.813
Male	0.973	0.488–1.941	0.939
Dialysis vintage, months	1.003	0.999–1.007	0.147			0.114
CCI scores, n	0.923	0.634–1.345	0.677
Hypertension	0.998	0.47–2.118	0.996
Diabetes	1.487	0.769–2.876	0.239
CAD	4.136	1.368–12.504	0.012	3.275	1.329–8.075	0.01
LF/HF	0.422	0.198–0.901	0.026	0.357	0.162–0.790	0.011
Pre-HD SBP, mm Hg	1.01	0.994–1.027	0.215
Pre-HD DBP, mm Hg	0.977	0.949–1.006	0.123			0.52
Post-HD SBP, mm Hg	1.000	0.982–1.017	0.976
Post-HD DBP, mm Hg	0.963	0.94–0.987	0.002			0.995
spKt/v	1.03	0.275–3.86	0.965
Hb, g/L	1.027	0.993–1.062	0.122			0.365
Alb, g/L	0.907	0.792–1.039	0.158			0.831
UA, µmol/L	0.997	0.992–1.002	0.312
Ca, mmol/L	0.688	0.149–3.181	0.632
P, mmol/L	1.234	0.58–2.627	0.585
PTH, pg/mL	1.001	0.999–1.002	0.371
lnNT-ProBNP, pg/mL	1.27	0.973–1.658	0.079			0.493
LVEF, %	0.977	0.948–1.007	0.128			0.337
LVMI, g/m2	1.003	0.993–1.013	0.53

CCI scores, Charlson comorbidity index scores; CAD, coronary artery disease; LF/HF, low-frequency/high-frequency ratio; pre-HD SBP, pre-hemodialysis systolic blood pressure; pre-HD DBP, pre-hemodialysis diastolic blood pressure; post-HD SBP, post-hemodialysis systolic blood pressure; post-HD DBP, post-hemodialysis diastolic blood pressure; spKt/v, single pool Kt/v; Hb, hemoglobin; Alb, albumin; UA, uric acid; Ca, calcium; P, phosphorus; PTH, parathyroid hormone; NT-ProBNP, N-terminal pro B-type natriuretic peptide; LVEF, left ventricular ejection fraction; LVMI, left ventricular mass index.

Discussion

To the best of our knowledge, this is the first prospective study that conducted 48-h HRV recording for evaluating the predictive value of LF/HF in death and hospitalization events in HD patients. The main finding of the study was that reduced LF/HF was an independent predictor of death and hospitalizations in MHD patients.

HRV, on behalf of the autonomic function, was conventionally measured over 24 h. Chen et al. [16] measured the frequency domain HRV parameters 30 min before and after HD session and found that HRV changes before and after HD were more effective predictive factors in predicting overall and cardiovascular mortality when compared with measurement of these parameters only before HD. Different from the measurement of HRV parameters only before and after HD, Chang et al. [17] in their study measured HRV parameters before HD and hourly during HD. When all parameters were put into the joint model, the results demonstrated that lower LF/HF values were independent factors associated with cardiovascular mortality. Osataphan et al. [18] measured HRV at different time points including post-dialysis. Their study indicated that lower post-dialysis HRV parameters were independent predictors of mortality in HD patients, but they did not find predictive values of HRV parameters at pre-dialysis and during the HD session. A study conducted by Yang et al. [19] also demonstrated that long-term HRV could predict all-cause mortality in HD patients, while short-term HRV failed. Above all, different time points had different predictive values. After excluding the comorbid conditions that could possibly affect HRV, the values of LF/HF decreased sharply following initiation of HD [20]. A detailed analysis of the HD session demonstrated that LF/HF increased gradually since the HD initiated and decreased after the therapy [21]. The HRV parameters were affected during the HD session, mainly in terms of a shift in sympathovagal balance toward sympathetic activation. As for HD patients, they took regular thrice-time weekly HD. During the HD interval, the volume, electrolytes, and uremia toxins underwent periodic alterations. A 48-h HRV analysis could record intradialytic as well as interdialytic data and offer better insights; hence, we observed the HRV with 48-h ECG measurement.

LF/HF ratio is considered to represent sympathovagal balance or to reflect sympathetic function [15]. Several studies have investigated the associations between LF/HF ratio and mortality or CVD; however, no conclusion could be drawn due to different results. A multiple center prospective study of CKD patients conducted by Chandra et al. [8] demonstrated that lowered LF/HF ratio could predict the risk of CVD and progression to ESRD. In another study, lower LF/HF ratio after 8-year follow-up was shown to be an independent predictor for cardiovascular mortality in MHD patients [17], which was similar to the results obtained by us. However, a study conducted by Kuo et al. [9], consisting of an analysis of a 41 HD patient cohort, reported contradictory results according to which higher LF/HF ratio group patients had poorer prognosis. We speculated that the possible reason for these results could be a small sample size and the fact that the analysis was based on short-term recordings, which was a challenge for interpretation [22]. In our study, LF/HF ratios are based on data that has been recorded over a longer period, which covered intradialytic as well as interdialytic period. LF/HF <1.33 was found to be a predictive factor of mortality and hospitalizations. The potential mechanism for explaining the relationship between reduced LF/HF and mortality remains unclear. Since a previously conducted study had demonstrated that with the progression of chronic kidney disease, LF/HF was significantly lower [8], and as discussed above, decreased LF/HF was associated with the poor prognosis in HD patients, we highlighted the evaluation of HRV when initiating HD in advanced chronic kidney disease patients.

According to a previously conducted study, spKt/Vurea was correlated with LF/HF in MHD patients. Improved HD adequacy was beneficial in terms of autonomic dysfunction improvement in HD patients [20]. However, in this study, we did not find significant difference of spKt/Vurea between group A and group B, probably because majority of our participants had received relatively high HD adequacy. Another possible reason could have been different HRV recording periods. In a study conducted by Tong and Hou [20], the association of spKt/Vurea and LF/HF was evaluated during an HD session, but in our study, LF/HF was based on 48-h HRV data. Sympathetic activation and vagal withdrawal were related to left ventricular hypertrophy in ESRD patients, and the benefit of daily hemodialysis on left ventricular hypertrophy could have been accompanied by improvement in HRV [23]. Decreased HRV parameters were associated with higher values of fluid overload in HD patients. HRV indices improved after 3 months of fluid management therapy [24].

In addition, the history of CAD enhances the risk of all causes of death and hospitalizations more than threefold. In the study, we recorded the proportion of CAD based on medical records. Conversely, dialysis patients without symptoms of a classic angina still have a high prevalence of CAD [25]. Furthermore, when ESRD patients experienced myocardial infarction, they had a lower frequency of receiving revascularization [26]. Also, we urge the clinical screening for CAD in dialysis patients to further improve the prognosis following coronary angiography and intervention. A significant predictive value of diabetes on the composite end point could not be determined. On one hand, diabetes played a critical role in terms of autonomic dysfunction, which was associated with decreased LF/HF [3]. On the other hand, the proportion of diabetes in the patients was relatively small, and follow-up period was short, and we hope to analysis in a larger cohort.

Our study had some limitations. First, the median follow-up time was 8.8 months and the observation time was short; however, significant difference was observed in the two groups. Second, we did not analyze HRV during the HD. In our further studies, we hope to analyze the 4-h HRV parameters alone to test the reproducibility of short-term results in the same participants. Also, our study was based on a smaller cohort of MHD patients, the results need to be further examined in multicenter HD participants. Our study period included pandemic of COVID-19, which contributed to the relatively high rate of the composite end point. Last, we did not strictly limit recording time of HRV after a short/long interval, but previous study had demonstrated that there was no difference in HRV parameters between short and long interdialytic intervals [27].

Conclusion

Decreased LF/HF ratio is a significant predictive factor of death and hospitalization events in HD patients during the 48-h recording.

Acknowledgment

The authors would like to thank Guozhu Chen for data analysis.

Statement of Ethics

This study protocol was reviewed and approved by the Ethics Committee of Peking University People’s Hospital, approval number 2023PHB101-001. Written informed consent was obtained from all subjects who participated in the study.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

This research was supported by Beijing Municipal Natural Science Foundation (No. 7222201).

Author Contributions

Y.C., X.L., L.Z., and Y.W. contributed to data collection; Y.C. carried out the statistical analysis and prepared the first draft of the paper; L.G. and L.Z. designed this trial, reviewed the data, approved the final manuscript version, and took responsibility for the integrity and accuracy of the data, including any adverse effects. All authors have read and agreed with the published version of the manuscript.

Funding Statement

This research was supported by Beijing Municipal Natural Science Foundation (No. 7222201).

Data Availability Statement

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author.