Introduction

Chronic kidney disease (CKD) is characterized by a progressive and irreversible decrease in renal function1. According to data from the International Society of Nephrology, hemodialysis is the most prevalent renal replacement therapy, with an average rate of 343 individuals per million2. Patients with CKD have many adverse effects resulting from the disease, such as a reduction in lean mass, mineral bone density, physical exercise intolerance, arterial hypertension, and physical capacity reduction3,4,5.

Another consequence for these patients is the high incidence of cardiovascular disease (CVD), which has emerged as the leading cause of mortality2,6,7. Among the contributing factors to CVD incidence, autonomic dysfunction has a determinant role8. Autonomic dysfunction is characterized by an increase in sympathetic activity and/or a reduction in parasympathetic activity and is strongly correlated with the risk of all causes9,10,11 of mortality. In the past few years, the scientific community has expressed increased interest in implementing a physical exercise routine for patients with CKD to attenuate the deleterious effects caused by kidney failure12,13,14,15,16,17,18,19,20,21. Few studies have evaluated the benefits of physical exercise on cardiac autonomic modulation, given that this system functioning contributes to mortality22, understanding how these interventions impact the heart rate variability (HRV) of these patients becomes important.

One of the adaptations to physical exercise is the modification of HRV through improvements in cardiac autonomic modulation and cardiovascular function23,24,25. A recent review25 demonstrated the effects of physical exercise on cardiovascular outcomes for patients with CKD, however, only studies in which the intervention modality was performed during the intradialytic period were included. Understanding the role of physical exercise beyond the intradialytic period is important for this population because many clinical centers do not offer the possibility of implementing a physical exercise program during hemodialysis, so understanding the effects of exercise beyond this period can make it more widespread26,27. Therefore, this systematic review with meta-analysis aimed to clarify the relationship between physical exercise interventions and changes in cardiac autonomic modulation in hemodialysis patients.

Results

Searches

The search resulted in 392 articles. After duplicate removal (n = 32), 360 articles remained for title and abstract evaluation. Seventeen studies met the established criteria for full reading, with nine exclusions (Fig. 1).

Fig. 1
figure 1

Flow diagram of study eligibility according to the PRISMA guidelines.

Full size image

Characteristics of the studies

Table 1 summarizes the study’s methods and the HRV analyses. The studies ranged from 3 to 12 months of intervention, three studies exclusively used aerobic exercise28,29,30, four interventions utilized concurrent exercise (aerobic + resistance)31,32,33,34, and one used breath-based lower limb training35. In all the studies, the participants were trained at least three times a week. Most studies used 24-hour ambulatory electrocardiogram recordings obtained from an ECG Holter device on nondialysis days to document HRV data. Only the Pereira28 (Polar RS800CX; Polar Electro™, Kempele, Finland) and Huang35 (8Z11, Enjoy Researcher, Inc., Taiwan) studies used a heart rate monitor to assess HRV.

Risk of study bias

The PEDro scale classified one28 study as methodologically high quality, four were rated as moderate quality29,30,31,32,33,35, and one article as low quality34. Table 2 shows the checklist for this scale.

Certainty of evidence

Only the SDNN index was moderate certainty, while the others were low certainty (LF/HF and RMSSD) and a very low certainty to Mean RR (Supplementary Material 2).

Quality of the training protocol

After analysis with the CERT, none of the articles achieved a high-quality classification. The percentages obtained were the following: Deligiannis et al.34, 21%; Reboredo et al.29, 47%; Huang et al.35, 53%; Mitsiou et al.30, 58%; Pereira et al.28, 58%; Kouidi et al.33, 58%; Kouidi et al.32, 68%; and Michou et al.31, 68%.

Table 1 Synthesis of sample characteristics, intervention protocols and HRV analysis of eligible articles.
Full size table
Table 2 Results of risk of bias analysis by use of the PEDro scale.
Full size table

Analysis of HRV indices

The HRV data were divided into time domain (SDNN, MeanRR, and RMSSD) (Fig. 2) and frequency domain (LF/HF) data (Fig. 3), presented in milliseconds and milliseconds (n.u), respectively. Analyses that included the article by Kouidi et al.32 had the correlation coefficient calculated from this study and applied to the calculation of the standard deviation of the others.

Fig. 2
figure 2

Results of the meta-analysis of the selected studies considering indices in the time domain. Caption: (A) Mean RR; (B) SDNN – Standard Deviation of intervals NN; (C): RMSSD – Root Mean of Successive Differences.

Full size image
Fig. 3
figure 3

Results of the meta-analysis of indices in the frequency domain Caption: LF/HF: Low frequency and High-Frequency ratio.

Full size image

Time domain

The SDNN index was the most used (seven studies), totaling 310 participants (160 intervention and 150 control). A summary measure of random effects was applied (as well as all the other variables in the time domain), revealing significant differences between groups (20.71 ms CI 95% [9.55, 31.87], p < 0.001, I²=95%), which were favorable for the group that received physical training. Mean RR (35.57 ms CI 95% [14.56, 56.57], p = 0.91, I²=0%), and RMSSD (10.55 ms CI 95% [6.75, 14.34], p = 0.37, I²=4%) also were favorable to intervention but not reached significant difference between the groups (p > 0.05). The absolute difference between means served as the summary measure in three analyses. Although this measure yielded a significant value, the SDNN analysis displayed high heterogeneity, indicating substantial variability (Fig. 2).

Frequency domain

Analyses in the frequency domain were the least common in the studies found. The Huang et al.35 and Reboredo et al.29 studies were eligible for the LF and HF analyses but were not included in the meta-analysis because they did not present data that would allow obtaining the standard deviation as recommended by the Cochrane Handbook and described in our methods. Thus, only LF/HF (0.28 ms (n.u) [0.11, 0.44], p = 0.18, I²=39%) was included in the meta-analysis with a summary measure of absolute differences and the use of random effects. LF/HF showed favorable to the intervention group and moderate heterogeneity (Fig. 3).

Discussion

This systematic review and meta-analysis examined the relationship between physical exercise interventions and alterations in cardiac autonomic modulation among hemodialysis patients. Our analysis included eight clinical trials with a total of 396 patients. One of the main findings is that all the time and frequency domain indices evaluated were favorable for the intervention group, evidenced by an increase in HRV activity. It is important to emphasize that even an index of observed parasympathetic activity (RMSSD) increased significantly in the intervention groups, representing a signal of exercise benefits for cardiovascular health in this population. CVDs are responsible for the majority of deaths in patients undergoing hemodialysis36. Estimations have indicated that 843.6 million people worldwide are affected by CKD, and in 2030, there will be an increase of 49% for individuals in the final disease stage2,37. Therefore, evaluating the autonomic nervous system and treating CVD are valuable, and HRV and physical exercise may complement each other due to their cost-effectiveness38,39.

Patients with CKD are a population that presents autonomic dysfunction, commonly represented by increased sympathetic activity and reduced HRV, as well as suppressed parasympathetic system action even in situations of a marked increase in it, such as at rest28,40,41,42. However, looking at the GRADE analyses, even though it is favorable to the intervention group in the Mean RR analyses, we are uncertain whether intervention can increase this outcome. The RMSSD and LF/HF may increase slightly, but this finding shows low certainty. In contrast, there is moderate certainty that physical exercise interventions lead to a probably increase in the SDNN index for these patients.

Our review elucidates some questions concerning the potential effects of training on the cardiac autonomic modulation of patients on hemodialysis. Among the studies found, Deligiannis et al.34, Kouidi et al.32, and Tourkantonis et al.34 reported an increase in the HRV expressed by the SDNN and a reduction in the vulnerability to arrhythmias in the training group compared to the control group. Kouidi et al.32 reported the same results regarding the SDNN index. Kouidi et al.33,] in addition to the beneficial changes in the SDNN, reported an improvement in parasympathetic activity indices, the RMSSD, and pNN50. Michou et al.31 recently showed that in diabetic kidney disease patients, similar results revealed significant improvements in the SDNN, RMSSD, and LF, as well as improvements in values that represent a better cardiorespiratory system, such as METs and VO2 peak. In this case, however, it is worth noting that, for both studies, the intervention strategy was concurrent training (aerobic + resistant).

Reboredo et al.29 and Huang et al.35 did not find significant changes in the comparison between groups; however, unlike the ones mentioned above, both studies received isolated training intervention, only aerobic training and lower limb exercises with respiratory control. Contrary, Pereira et al.28 reported an increase in HRV in the intervention group for both global (SDNN and SD2) and parasympathetic (RMSSD and SD1) indices. Mitsiou et al.30. reported increases in HRV after the training protocol for SDNN and LF (global indices) and parasympathetic indices (RMSSD and pNN50), as well did the studies by Reboredo et al.29 and Huang et al.35, which involved aerobic training alone for three and six months, respectively. These studies differently obtained a favorable outcome for the training group compared to their control peers, highlighting the gaps found in the current literature in elucidating the role of physical exercise in the HRV of patients on hemodialysis.

The results of Pereira et al.28 and Mitsiou et al.30 are similar to others found in the literature with patients with cardiovascular impairments who received aerobic training between nine and 12 weeks, such as patients with recent coronary events (increased SDNN as well as nocturnal HRV)43, heart failure (increase in SDNN and HF, respectively)44,45, acute myocardial infarction (improvement in LF and HF indices)46 and coronary artery disease (increased HRV)47. The results of long-term interventions involving six months of aerobic exercise were similar to the findings of Deligiannis, Kouidi, and Tourkantonis34 and Kouidi et al.32,33. Mazzuero et al.48 reported an increase in the SDNN index in postinfarcted patients, as did Villafaina and colleagues49; however, patients with fibromyalgia were the subject of the study (this population also presented autonomic impairment). Finally, Pietila et al.50 verified an increase in HRV and the HF index, an index of predominantly vagal activity.

In a recently published review, Verrelli et al.25 aimed to evaluate the cardiovascular outcomes of intradialytic physical exercise interventions. Among the main findings, the role of exercise in promoting pulse wave velocity, diastolic blood pressure, left ventricular ejection fraction and even heart rate variability stands out. These data are in the same direction as our findings and with another reviews40,51,52, suggesting that physical exercise carried out inside or outside the hemodialysis session has a beneficial effect on the cardiovascular health of these individuals. However, the review did not set out to analyze indices of parasympathetic predominance, an important marker of cardiac autonomic modulation, in addition, HRV ends up being a secondary analysis. As mentioned in the introduction, understanding physical exercise beyond the intradialysis moment is important, given that it has several limitations for it to occur during the session, being limited to a few modalities and failing to cover a significant portion of these. patients, as this modality still requires greater implementation in dialysis centers27.

Some of the possibilities for the absence of significance in the other indices of the studies included in this review are related to the low/moderate quality of most studies; intervention; and control groups, which had a small sample size; training strategy; intervention time; overload control; and a low number of articles on the subject included in the analysis. The methods analyses utilizing the CERT in different studies indicate the difficulty of replicating the protocol regarding the choice of overload and training protocol. Thus, the inferences about overload are difficult, even though this is a relevant training parameter to obtaining significant results53. However, this characteristic is not exclusive to the technical literature of this review54.

Additionally, when analyzing the training protocols, it was possible to note that the studies31,32,33,34 that used concurrent training (aerobic + resistance) presented at least one index with an improved response in the training group compared to others that only utilized aerobic exercise. Mitsiou et al.30. and Pereira et al.28. obtained similar results using an isolated training strategy (aerobic), and Reboredo et al.29. did not observe significant improvement in aerobic training adopted in isolation, despite the vast body of literature indicating that this modality can improve cardiorespiratory function, as well as autonomic indices24,55,56,57. Therefore, there is a clear need for more controlled studies using physical exercise in isolation.

Our group recommends conducting new clinical trials due to the limited availability of studies on the topic, which hinders the extrapolation of results, such as subgroup analyses. We observed a lack of studies for all training modalities, whether concurrent, resistant, or aerobic. In addition, the training protocols are often poorly described, and the study methods lack the necessary academic rigor and clarity in their design and presentation. The sample size was sometimes not calculated due to the complexity of this type of intervention in recruiting patients, but adding this parameter to the experiment is necessary for advancing the technical quality of the intrinsic findings of studies and those extrapolated by a meta-analysis. Finally, we suggest the standardization of HRV analysis and the use of the reunited indices, at least concerning the most used to assess the parasympathetic activity (RMSSD and HF), as these indices are closely linked to the occurrence of CVD in this population and are well characterized as being reduced to the detriment of the sympathetic activity.

Our results indicated significant improvements in the autonomic response for the mean RR, SDNN, RMSSD, and LF/HF indices that occurred after the training protocol. However, extrapolation of these findings needs caution, since the level of certainty observed was low and very low for most of them. Nevertheless, these findings suggest a promising direction for physical activity programs for patients with CKD on hemodialysis, due to the cardiac risks in this population. Thus, improvements in the autonomic nervous system may indicate greater longevity for these individuals. Despite these findings, the magnitude of the impact of physical training on changing cardiac autonomic modulation in patients undergoing hemodialysis, as well as the degree of recommendation and certainty of evidence, still need attention.

Among the limitations of this review, we highlight the lack of subgroup analyses by type of physical exercise and moment of intervention (intradialytic and non-dialytic), the small number of included studies making some analyses unfeasible, and the inclusion of studies not being conditioned on their quality. Strengths include the meticulously designed search strategy for each database, methodological rigor in data extraction and analysis using different tools, and the inclusion of only randomized controlled studies with different physical exercise interventions. It is expected that this study can be used as a reference for the application and adoption of physical training programs in clinics and hemodialysis centers as well as to encourage public policies that promote physical exercise for these patients. Furthermore, other authors could understand the current gaps regarding physical training and cardiac autonomic modulation in patients with CKD who undergo renal replacement therapy.

In conclusion, the study reviewed the scientific literature to understand how physical exercise has been applied to improve the autonomic nervous system in patients with CKD undergoing hemodialysis. Results indicated that physical interventions can increase the HRV in these patients for indices in time and frequency domains, suggesting benefits for cardiovascular health.

Methods

Study design

The study is designed to meet the requirements of a systematic review with meta-analysis following the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) model58 and the Participants, Interventions, Comparisons, Outcomes, and Study Design (PICOS) strategy59. Following the PICOS model, we defined: (i) Participants: Patients with chronic kidney disease undergoing hemodialysis treatment; (ii) Interventions: Physical exercises; (iii) Comparisons: Control groups without physical exercises; (iv) Outcomes: Heart rate variability; (v) Study Design: Randomized clinical trials. This study was registered in the PROSPERO database (CRD42021255073) to obtain a low risk of bias and reduce the duplication chances of the reviewed topic.

Eligibility criterion

The inclusion criteria for eligible studies were as follows: randomized clinical trials with patients in the end-stage of CKD receiving hemodialysis treatment and a control group; any physical exercise intervention for at least 12 weeks; at least three months of treatment and older than 18 years of age; and HRV as an outcome. The exclusion criteria included commentary articles, summaries, animal model experiments, systematic reviews and/or meta-analyses, studies involving patients not receiving hemodialysis or other renal replacement therapy, and transplant recipients. The searches had no date restriction, prioritizing those published in electronic bibliographic databases, and contact was made with the authors for eventual clarifications about the absence of complete articles or analysis and data contained in your summary.

Search strategy

The search strategy was applied to the following databases: EMBASE, MEDLINE, Cochrane Library (CENTRAL), and PEDro. When necessary, the corresponding authors were contacted to answer questions about the indices used to analyze HRV and further crucial information. The selection of keywords was conducted by searching the MeSH (Medical Subject Headings) platform for the MEDLINE and The Cochrane Library databases, Emtree for EMBASE, and DeCS (Health Sciences Descriptors) for the PEDro database (Table 3). The PICOS59 strategy and a systematic review60 related to the main theme published at the CENTRAL were used to create the descriptor pattern. To the searches, certain terms were added because they were essential for this study, such as heart rate variability (Emtree was the only platform that contained), autonomic modulation, sympathetic hyperactivity, cardiac autonomic modulation, and autonomic dysfunction. The last search update was made on September 20, 2024.

Table 3 Platforms and respective descriptors found to be used following the PICOS strategy.
Full size table

Study selection

Two authors (HSD and PHM) independently conducted the searches on the databases (the search strategy is available in supplementary material 1) following a peer-review model. The authors imported the search results into the software ®Rayyan61 (http://rayyan.qcri.org). After duplicate removal, selections were made based on title and abstract reviews, with full-text assessments conducted when necessary. During the whole process, if the pair disagreed on the inclusion/exclusion of an article, a third author (CSCR) was asked to resolve any disagreements among the pairs.

Both reviewers independently read and included articles that met the eligibility criteria. For this purpose, a standardized form was used, and both reviewers conducted the process independently. Later bias analysis tools, as well as more rigorous methodological quality. It is worth noting that the methodological quality assessment form was solely employed for an initial evaluation of the methodological content within the respective articles, acting as a filter for the types of studies considered.

Outcomes

The investigated outcome, heart rate variability, was analyzed with the following indices: in the time domain, the standard deviation of all NN intervals (SDNN); the root mean square sum of squares of differences between NN intervals (RMSSD), which represents global activity and parasympathetic activity; and the mean RR interval between all regular beats (mean RR), which represents global activity. In the frequency domain, the LF/HF ratio analysis considered it an indicator of the balance between low frequency (LF – global activity) and high frequency (HF – parasympathetic activity).

Summarizing results measures

Effect measures utilized for group comparisons involved analyzing the absolute difference between means (mean difference), as the studies employed the same continuous scale. Following article selection, the application of a fixed or random effect model was also evaluated, depending on the internal disparity of studies.

The analyses were based on the Cochrane Handbook for Systematic Reviews of Interventions, where we computed the differences between post - and pre-mean for each index and group related to the HRV indices for the mean difference. To obtain the standard deviations in analyses with articles that provided sufficient information, we calculated the correlation coefficient for the intervention and control groups and used it to obtain the standard deviation in the other studies analyzed using the same index (Cochrane Handbook Chap. 6.5.2.8). However, in analyses without any study with sufficient information, we adopted the use of the p-value obtained in the comparison of the means between groups of the articles. After this, we obtained the t-value using the Cochrane online calculator. Finally, we calculate the standard error and standard deviation (Cochrane Handbook Chap. 6.5.2.3).

Furthermore, as suggested by the Handbook, we excluded from the analysis studies that did not provide sufficient data to enable the calculations. All the statistical computations for the meta-analysis utilized the Cochrane Collaboration’s Review Manager (RevMan) version 5.4.1 software.

Certainty of evidence

The GRADE assessment (The Grading of Recommendations Assessment, Developing and Evaluation) (https://www.gradepro.org)62 determined the certainty of the evidence for each HRV index. This tool verifies the certainty of evidence using the risk of bias, inconsistency, indirect evidence, imprecision, and publication bias. The PEDro scale (https://pedro.org.au) assessed the risk of bias; however, quality was not a condition for inclusion in this review, and articles with a score ≥ 7 were considered high quality, those with a score of 5 to 6 were moderate, and scores ≤ 4 were low quality.

Therefore, to assess the risk of bias by GRADE, the following criteria were adopted: the PEDro scale score, the weight of the studies with the highest risk of bias, and the proportion of studies included in the analysis. Additionally, the significance of the analysis was supported by studies of lower or higher quality, in which GRADE’s tools were used to assign risk. For inconsistency, the overlap of the 95% confidence intervals was evaluated, as was the heterogeneity and the significance of the heterogeneity. Indirect evidence was assessed using the PICOS strategy, taking into account the research questions and how each study was designed to gather information, with the CERT score serving as a reference. Imprecision was evaluated by considering the minimum information value (number of participants), alongside an analysis of the point estimate, observing the size of the confidence intervals (represented by the analysis diamond) and their position relative to the centerline. The funnel plot and other statistical analysis were not used, because The Cochrane Collaboration recommendations advise against their use in meta-analysis, with 10 or fewer eligible articles for the outcomes. For each of the topics, decisions were made on whether to downgrade each item, grading it as strong, moderate, low, or very low, with the help of the GRADEpro online platform (https://gdt.gradepro.org/app/) created for evaluation of GRADE63, also using the criteria already present in the platform itself in the stratification of the analyses.

Heterogeneity

The I² test was applied to assess the degree of heterogeneity, evaluating the percentage of variation in the estimate of the effect attributed to heterogeneity; interpreting this as 0–40%, heterogeneity might not be important, 30–60% may represent moderate heterogeneity, 50–90% may represent substantial heterogeneity, and 75–100% may indicate considerable heterogeneity63.

Training protocol quality

With the CERT tool, we evaluated the design of the training protocol used in the eligible studies. It consists of a scale of 16 topics totaling 19 items that gather the main information that a physical training protocol should contain (from the material used to the control of the effort intensity)64,65. The tool was applied by two independent evaluators (HSD and VSZ) using a model previously created for Microsoft Excel software (Microsoft Office, Microsoft Corporation, Redmond, Washington, USA) version 2019 based on the checklist of the supplementary material of the official article, which contains the entire description of the tool65. The CERT has a checklist that can reach a result of up to 100% for the present study, and a percentage ≥ 70% was considered high quality of the outlined protocol and less than low/moderate quality66.