The Impact of SARS-CoV-2 Infection on Heart Rate Variability: A Systematic Review of Observational Studies with Control Groups

Autonomic nervous system (ANS) dysfunction can arise after severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection and heart rate variability (HRV) tests can assess its integrity. This review investigated the relationship between the impact of SARS-CoV-2 infection on HRV parameters. Comprehensive searches were conducted in four electronic databases. Observational studies with a control group reporting the direct impact of SARS-CoV-2 infection on the HRV parameters in July 2022 were included. A total of 17 observational studies were included in this review. The square root of the mean squared differences of successive NN intervals (RMSSD) was the most frequently investigated. Some studies found that decreases in RMSSD and high frequency (HF) power were associated with SARS-CoV-2 infection or the poor prognosis of COVID-19. Also, decreases in RMSSD and increases in the normalized unit of HF power were related to death in critically ill COVID-19 patients. The findings showed that SARS-CoV-2 infection, and the severity and prognosis of COVID-19, are likely to be reflected in some HRV-related parameters. However, the considerable heterogeneity of the included studies was highlighted. The methodological quality of the included observational studies was not optimal. The findings suggest rigorous and accurate measurements of HRV parameters are highly needed on this topic.


Introduction
The coronavirus disease of 2019 (COVID-19) resulting from a severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection has caused a serious public health crisis worldwide and has had harmful impacts on human health and quality of life [1]. SARS-CoV-2 infections can often lead to serious health consequences, including mortality, and are associated with multiple organ failures, including that of the respiratory, cardiovascular, nervous, hepatobiliary, immune, and blood systems, and mental health conditions such as brain fog [2][3][4].
Recent studies have reported an association between SARS-CoV-2 infection and autonomic dysfunction. For example, a systematic review by Scala et al. analyzed 22 studies reporting features of ANS involvement during acute COVID-19 and concluded that the disease had ANS involvement even at an early stage and the involvement was associated with poor prognosis of patients with COVID-19 [5]. As a mechanism to explain the relationship between SARS-CoV-2 infection and autonomic nervous system (ANS) function, Mohammadian et al. (2022) suggested that SARS-CoV-2 viral particles may disrupt the function of the ANS center in the brainstem by disrupting the homeostasis of the brain renin-angiotensin system [6]. Other researchers suggested the interplay of ANS function and inflammation [7]. Recently, ANS dysfunction in patients with COVID-19 has been shown to lead to the various symptoms seen in acute COVID-19 as well as long COVID [8]. In addition to SARS-CoV-2, ANS dysfunction may be part of the explanation for some symptoms in the context of other infectious diseases, including orthostatic intolerance in 2. 1

.2. Types of Participants
Only patients infected with SARS-CoV-2 were included. This review was intended to focus on the direct impact of SARS-CoV-2 infection in COVID-19 patients, and thus long COVID (i.e., the continuation or development of new symptoms 3 months after the initial infection with SARS-CoV-2) patients were excluded. This is because long COVID does not necessarily require the current presence of SARS-CoV-2, and the condition is not dependent on the severity of acute SARS-CoV-2 infection [21]. Therefore, the association between long COVID and HRV would be worthy of being investigated as a separate topic. No restrictions were imposed on clinical condition, language, sex/gender, age, or race/ethnicity.

Types of Exposures
SARS-CoV-2 infection. There were no restrictions on the test for confirming the infection. However, the accuracy of the test was reflected in the quality assessment of the included studies.

Types of Controls
Non-infected individuals, healthy controls, or infected patients of different severity of COVID-19 were included in the control group. For longitudinal studies, the time of infection, including pre-infection, was included. As a relationship between SARS-CoV-2 infection and HRV-related parameters has not yet been well established, this review focused on the impact of SARS-CoV-2 infection on HRV parameters. And comparative studies of COVID-19 patients and individuals with other infectious diseases were outside the scope of this review. Specifically, comparisons between COVID-19 patients and non-infected individuals or healthy controls investigated the association between SARS-CoV-2 infection and HRV parameters. On the other hand, comparisons among COVID-19 of different severities investigated the clinical usefulness of HRV in the context of the assessment of patients with COVID-19.

Types of Outcome Measures
HRV-related outcomes included the following time and frequency domains. The time domains included the mean standard deviation of the normal-to-normal (NN) interval (SDNN), SDNN index, standard deviation of the average NN interval (SDANN), square root of the mean squared differences of successive NN intervals (RMSSD), standard deviation of the differences between adjacent NN intervals (SDSD), number of pairs of adjacent NN intervals differing by more than 50 ms in the entire recording (NN50 count), proportion of NN50 divided by the total number of NN intervals (pNN50), and HRV triangular index. The frequency domains included the mean total power (TP), powers in ultra-low frequency range (ULF), very low frequency range (VLF), low frequency (LF), coefficient of component variance for LF (CCVLF), high frequency (HF), coefficient of component variance for HF (CCVHF), normalized unit of LF power (LFnorm), normalized unit of HF power (HFnorm), and LF/HF ratio [10].

Search Strategy
Four electronic bibliographic databases were comprehensively searched by one researcher (C.-Y.K.

Study Selection
Two independent researchers (C.-Y.K. and B.L.) screened the titles and abstracts of documents to identify potential studies which meet the inclusion criteria. After the initial screening, the two independent researchers assessed the full texts of the screened studies. Any disagreements between the researchers were resolved through their discussion. End-Note20 (Clarivate Analytics, Philadelphia, PA, USA) was used to manage citations from included articles.

Data Extraction
Among the included studies, two independent researchers (C.-Y.K. and B.L.) extracted variables for analysis. These variables included first author, publication year, publication type, country, study type, comparison type, characteristics of participants, assessment tools for HRV, measured HRV parameters and findings, relationships between HRV parameters and other clinical variables, and author conclusions. Among the characteristics of the participants, in particular, the cardiovascular conditions of participants, which can be considered as an important covariate in the association between HRV parameters and COVID-19, was extracted. The extracted data from the included studies were entered into a Microsoft Excel file (Microsoft, Redmond, WA, USA). The two independent researchers performed the data extraction processes, and any inconsistencies were resolved through their discussion.

Quality Assessment
Depending on the study type, corresponding methodological quality assessment tools were used, developed by the National Heart, Lung, and Blood Institute group [22]. The Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies was used for observational cohorts and cross-sectional study types. The Study Quality Assessment Tools Quality Assessment of Case-Control Studies was used for case-control studies. Two independent researchers (C.-Y.K. and B.L.) performed the quality assessment, and any disagreements were resolved through their discussion.

Data Analysis
Considering the heterogeneity of the population, times from the COVID-19 outbreak, and potential comorbid diseases, quantitative analysis was not planned in the protocol of this systematic review. A quantitative analysis can only be performed if sufficient homogeneity between the studies and outcomes used is ensured. However, the clinical heterogeneity between the included studies was considerable. Therefore, the impact of SARS-CoV-2 infection on HRV was analyzed qualitatively. Among the HRV parameters, clinically relevant vagally mediated HRV (vmHRV) parameters were of interest, which include RMSS, HF power, and HFnorm.

Assessment of Heterogeneity
The heterogeneity of included studies was investigated both qualitatively and quantitatively. Qualitatively, the potential causes of heterogeneity of the included studies according to study design, clinical characteristics of participants, and HRV measurement method were analyzed. Moreover, the authors performed a meta-analysis for the purpose of confirming the justification of performing only qualitative analysis. Command 'metan' in Software Stata version 13.1 (StataCorp, College Station, TX, USA) was used to perform meta-analysis. A meta-analysis was conducted on HRV parameters reported in two or more studies, in which value of HRV parameters were presented in the form of mean and standard deviation, and the number of subjects in each included group was specified. For the meta-analysis, a random-effect model was used, and the results were presented as standardized mean difference (SMD) and 95% confidence intervals (CIs). Instead of being interpreted clinically, the results of meta-analysis were interpreted for the purpose of visualizing and quantifying the heterogeneity of the study results. I-square statistic was used to estimate heterogeneity statistically, and if the values were greater than 50% and 75%, they were considered to have substantial and considerable heterogeneity, respectively.

Figure 1.
PRISMA flow diagram of this study. Abbreviations: CINAHL, Cumulative Index to Nursing and Allied Health Literature; COVID, coronavirus disease; HRV, heart rate variability.

Methodological Quality Assessment
In the four case-control studies [49,55,59,65], all of the research questions or objectives were clearly described. Three of these studies [55,59,65] clearly described the study population criteria, whereas the other study [49] did not. Of the case-control studies [49,55,59,65], none of them justified their sample size. In addition, all the included studies [49,55,59,65] did not describe whether the case group and control group were recruited from the same or similar population, including the timeframe. However, the case and control groups were clearly defined and separated. Among the studies [49,55,59,65], whether the protocol was previously registered was not clear. Therefore, it could not be confirmed whether the predefined inclusion and exclusion criteria were applied in these studies. Three studies [49,59,65] did not describe the sampling method, and the other study [55] noted that a convenient sampling method was used, but it was unclear whether randomization was applied or not. Among the studies [49,55,59,65], there was no mention of the use of concurrent controls. These studies included participants infected with SARS-CoV-2, but only three studies [55,59,65] clearly described the confirmation method. Although the assessment was not reported as being blind in these studies [49,55,59,65], since the confirmation of SARS-CoV-2 infection may have come from medical records, it was concluded that the identification of the participants may have been possible. All these studies reported that matched controls were included, but one study [49] did not describe the matching conditions. Regarding the overall quality, three [57,59,65] were evaluated as fair and the other study [49] as poor ( Table 2).
Regarding the overall quality, three [50,58,64] were evaluated as fair and the remaining ten studies [51][52][53][54]56,57,[60][61][62][63] as poor (Table 3).   Were key potential confounding variables measured and adjusted statistically in the analyses? If matching was used, did the investigators account for matching during study analysis? adopted from The National Heart, Lung, and Blood Institute [22]. Were inclusion and exclusion criteria for being in the study prespecified and applied uniformly to all participants? Q5. Was a sample size justification, power description, or variance and effect estimates provided? Q6. For the analyses in this paper, were the exposure(s) of interest measured prior to the outcome(s) being measured? Q7. Was the timeframe sufficient so that one could reasonably expect to see an association between exposure and outcome if it existed? Q8. For exposures that can vary in amount or level, did the study examine different levels of the exposure as related to the outcome (e.g., categories of exposure, or exposure measured as continuous variable)? Q9. Were the exposure measures (independent variables) clearly defined, valid, reliable, and implemented consistently across all study participants? Q10. Was the exposure(s) assessed more than once over time? Q11. Were the outcome measures (dependent variables) clearly defined, valid, reliable, and implemented consistently across all study participants? Q12. Were the outcome assessors blinded to the exposure status of participants? Q13. Was loss to follow-up after baseline 20% or less? Q14. Were key potential confounding variables measured and adjusted statistically for their impact on the relationship between exposure(s) and outcome(s)? adopted from The National Heart, Lung, and Blood Institute [22].

Impact on vmHRV Parameters
(1) SARS-CoV-2 infection (vs. negative control): Mixed results were observed for RMSSD. That is, two studies [55,61] found that the RMSSD (ms) in COVID-19 patients was statistically significantly lower than that of negative controls, while the other two studies [49,54] found the opposite significant result (all, p < 0.05). The other two studies [51,64] did not find a statistically significant difference on the parameter. In Risch et al. (2022) [64], no significant differences on RMSSD from baseline were found in the incubation, presymptomatic, symptomatic, and recovery stages of SARS-CoV-2 infection (all, p > 0.05). In Sari et al. (2020) [49], no difference in RMSSD between COVID-19 patients and negative controls was observed (p > 0.05), but the values of symptomatic COVID-19 patients were significantly lower than that of negative controls (p < 0.05).
Regarding SARS-CoV-2 infection, four studies [49,55,59,61] reported statistically significant reductions in HF power (ms 2 ) compared to that of negative controls (all, p < 0.05). Although three studies [49,51,65] did not find a significant difference in this parameter between COVID-19 patients and negative controls, among the studies, Sari et al. (2020) [49] found that HF was significantly lower in symptomatic COVID-19 patients compared to negative controls (p < 0.05). Moreover, in Milovanovic et al. (2021) [59], mild COVID-19 patients showed lower HF values compared to negative controls (p < 0.05), but no significant difference was observed between severe COVID-19 patients and negative controls (p > 0.05). Regarding HFnorm (nu), Milovanovic et al. (2021) [59] found no statistically significant difference associated with SARS-CoV-2 infection (p > 0.05), but this parameter was rarely investigated in the included studies ( Figure 2).  power; HRV, heart rate variability; NS, not significant; RMSSD, the square root of the mean squared differences of successive NN intervals; vmHRV, vagally mediated heart rate variability. Note: The number (N) in the upper right corner of each cell means the number of participants included in the analysis. '↑' (green shading) and '↓' (orange shading) mean significantly higher or lower than that of COVID-19 negative controls, respectively, while 'NS' (yellow shading) indicates no statistically significant difference was found.
Mixed results were observed with respect to LF/HF ratio and LF power (ms 2 ). Specifically, two included studies [59,61] found significantly higher LF/HF ratio in COVID-19 patients compared to negative controls, while another two [49,64] found a significantly lower ratio in COVID-19 patients (all, p < 0.05). For LF power, Topal et al. (2021) [61] found significantly higher LF in COVID-19 patients compared to negative controls, but two other studies [55,59] found the opposite significant result (all, p < 0.05).

Qualitative Analysis
The heterogeneity of the included studies was analyzed qualitatively in terms of the study design, clinical characteristics of participants, and HRV measurement method. The study designs of the included studies varied to a total of four (i.e., case-control study, retrospective analysis, cross-sectional study, and prospective cohort study), and only four included studies [50,53,62,64] (4/17, 23.53%) were prospective cohort studies with relatively strict study designs. In addition, only five studies [55,58,59,64,65] (5/17, 29.41%) were identified that statistically adjusted key potential confounding variables in their analysis.

Findings of This Review
This review was conducted to systematically investigate the effect of SARS-CoV-2 infection on HRV-related parameters, and a total of 17 observational studies [49][50][51][52][53][54][55][56][57][58][59][60][61][62][63][64][65] were included in this review. The main findings of this review are as follows: (1) Methodological quality of included studies: The methodological quality of the included observational studies was not optimal. Among the included studies, only two [50,64] justified the sample size, and in all studies, the blinding of analysis was not guaranteed. In addition, in studies other than case-control studies, there were only two studies [58,64] that measured potential confounding variables and adjusted statistically for their impact on the outcome. (2) HRV parameters investigated: The most frequently investigated HRV parameter in relation to SARS-CoV-2 infection was RMSSD, followed by LF/HF ratio, LF power, HF power, and pNN50. (3) Impact on vmHRV parameters: Among the significant differences found, compared to negative controls, a consistent finding of the vmHRV parameter associated with SARS-CoV-2 infection was low HF power. Mixed results were observed for RMSSD, and studies on HFnorm were lacking. In relation to the different severity or prognosis of COVID-19, it was reported that the RMSSD and HF power were significantly lower in symptomatic COVID-19 patients compared to asymptomatic COVID-19 patients, and that the RMSSD was significantly lower and the HFnorm was significantly higher in died COVID-19 patients than in survived patients. However, consistent findings were rare. (4) Impact on other HRV parameters: Some included studies have found significantly lower SDNN and pNN50 in patients with COVID-19 compared to negative controls, but mixed results were observed for LF/HF ratio and LF power. Significantly lower SDNN was observed in symptomatic patients compared to asymptomatic COVID-19 patients and in severe patients compared to mild COVID-19 patients. In the comparison of died and survived COVID-19 patients, no significant difference in SDNN was found, but low SDNN was significantly associated with worse prognosis of COVID-19 patients, including fewer survival days. (5) Heterogeneity of included studies: Included studies were heterogeneous in terms of study design, clinical characteristics of participants, and HRV measurement method. Also, in the quantitative analysis, substantial heterogeneity was observed for RMSSD, HF power, SDNN, LF power, and LF/HF ratio.

Clinical Interpretation
Although some included studies did not find a significant association between HRVrelated parameters and SARS-CoV-2 infection, these studies did not deny the effect of SARS-CoV-2 infection on ANS function. For example, one study [51] found no significant differences in the HRV parameters between non-critically ill COVID-19 patients and healthy volunteers. However, this study [51] found significant differences between the two groups in Sudoscan and automated pupillometry results. Specifically, the patient group had significantly higher pupillary dilatation velocities (p = 0.040), baseline pupil diameter (p = 0.039), and incidence of feet sudomotor dysfunction (p = 0.038). The investigators [51] concluded the presence of ANS dysfunction in the early stage of COVID-19. Their findings may suggest a difference in sensitivity between the HRV test and other tests to observe changes in the ANS associated with SARS-CoV-2 infection.
The current review focused on the vmHRV parameter in relation to ANS function, to explore the association of parasympathetic activity to SARS-CoV-2 infection or COVID-19. As results, a significant drop in HF power related to SARS-CoV-2 infection was consis-tently observed [49,55,59,61]. This may be understood in the context of vagal invasion by SARS-CoV-2 and the important role of the nerve system in neuroimmunometabolism [67]. A potential association with the significant reduction in HF power observed in patients with long COVID can also be assumed [68], as the development of postural orthostatic syndrome after COVID-19 involves mechanisms such as increased sympathetic activity and decreased parasympathetic activity due to SARS-CoV-2 infection [69]. But the scope of this review limits the extension of the findings to long COVID. According to our findings, HFnorm showed no significant difference according to the presence or absence of SRAS-CoV-2 infection [59]. However, HFnorm was significantly higher in died COVID-19 patients, and the increase of this parameter was associated with a worse prognosis, higher mortality, and higher IL-6 levels [50]. They discussed this association by considering a high HFnorm a depletion of sympathetic activity and proportionally greater vagal activity [50]. The association of increased cardiac vagal and decreased cardiac sympathetic activities with death in patients with critically ill conditions such as acute respiratory distress syndrome was previously reported by the same research team [50]. The underlying mechanism explaining the association between poor prognosis and increased vagal modulation in critically ill patients needs further investigation, and may involve the use of anti-cholinergic and sympathomimetic agents used to correct ANS dysfunction in critically ill patients. RMSSD was significantly lower in symptomatic COVID-19 patients and deceased COVID-19 patients compared to controls [49]. Also, the decrease of this parameter was related to sudden cardiac death in hospitalized COVID-19 patients [63]. RMSSD, a parameter reflecting the integrity of vagus nerve-mediated autonomic regulation of the heart, is known to be negatively related to sudden death such as epileptic sudden death [70]. A lowered parameter suggests a reduced cardioprotective effect of the vagus nerve, possibly increasing the risk of myocardial damage and fatal arrhythmias [71].
Among the HRV parameters other than vmHRV, the significant differences observed were decreased SDNN, LF, HF, pNN50, and SDANN associated with SARS-CoV-2 infection or the poor severity and prognosis of COVID-19. Decreased SDNN and SDANN are consistent predictors of cardiac death [72,73], and this review also found that SARS-CoV-2 infection or severity was associated with reduced HRV [49,60,64]. In patients with sepsis, the SDNN was also classified as an effective parameter for predicting mortality in a previous meta-analysis [74]. Although not included in this review because of the lack of a control group, Mol et al. [43] concluded that a higher SDNN predicted survival, especially in elderly patients, in their retrospective cohort study.
These findings on HRV parameters suggest a potential involvement of vagal tone and ANS function in SARS-CoV-2 infection and the clinical course of COVID-19 patients, including symptom onset and death. However, since other confounding factors such as shifts in respiratory rate and volume may affect these parameters [10], and there were no included studies that strictly controlled for respiratory variables, the involvement of ANS is still tentative.

Limitations
Given the profound and widespread impact of COVID-19 on human health and the importance of non-invasive, early indicators, this systematic review has a strength as a first comprehensive review of the impact of SARS-CoV-2 infection on HRV-related variables. However, the following limitations are recognized. First, the heterogeneity of the studies included in this review is a major limitation of this review. Our review qualitatively analyzed the heterogeneity of the included studies, and found considerable heterogeneity in the study design, clinical characteristics of participants, and HRV measurement method. Among the included studies, the limitation that information on the cardiovascular conditions of the participants could be confirmed in only nine studies [49][50][51]54,55,58,60,61,63] (9/17, 52.94%) may also contribute to the heterogeneity. As such, only 53% of included studies [49,51,[54][55][56][57]60,61,65] used ECG to measure HRV parameters, and it cannot be considered to have equal reliability as the HRV parameters obtained for commercially avail-able wearable devices such as Fitbits. In addition, it is still possible that the heterogeneity was due to the age or sex of the participants, the difference in their underlying diseases, the difference in the duration of COVID-19, the difference in the history of COVID-19 vaccination, and the change in HRV due to factors other than SARS-CoV-2 infection. As the meta-analysis also reaffirmed the substantial heterogeneity included (I-square values, 60.0% to 96.5%), the meta-analysis was not interpreted clinically. Second, due to limitations in the methodological quality of the included studies, the reliability of the findings obtained in this review is challenged. In particular, a risk of bias was observed in justification of sample size, the possibility of assessor blinding, and analysis accounting potential covariates. Third, most of the included studies [49][50][51]54,56,58,60,[63][64][65] were small, with a total sample size of fewer than 100 cases. The small sample sizes may have exaggerated the effect of SARS-CoV-2 infection on HRV-related variables. Moreover, the small sample sizes may explain the failure to find statistically significant differences between groups in most outcomes found in this review. Fourth, this review focused on SARS-CoV-2 infection and excluded long COVID; however, the clinical course of COVID-19 needs further refinement. Although one included study [64] divided the COVID-19 stage into five stages (i.e., baseline to recovery) and analyzed the change in HRV parameters for each stage, there is a limitation in that the other included studies did not specify a different period of COVID-19. Fifth, despite the well-known relationship between inflammation and some HRV parameters [75], there were no studies analyzing the relationship between HRV parameters and persisting inflammatory signs in patients with COVID-19. Sixth, studies on the threshold for detecting clinically meaningful changes, such as minimal clinically important differences in HRV parameters, are lacking. Therefore, although some changes of HRV parameters found in this review were statistically significant, they could not be rigorously interpreted as clinically meaningful changes. Finally, given that the scope of interest of this review was HRV parameters, suggesting some aspect of ANS function only, caution is needed in interpreting our findings to extend to the association between SARS-CoV-2 infection and ANS function.

Conclusions
The findings of this review show that SARS-CoV-2 infection, the severity of COVID-19, and its prognosis are likely to be reflected in some HRV-related parameters. Among vmHRV parameters, decreases in RMSSD and HF were associated with SARS-CoV-2 infection or the poor severity and prognosis of COVID-19, and decreases in RMSSD and increases in HFnorm were observed in died critically ill COVID-19 patients. However, this review highlights the considerable heterogeneity of the included studies. The findings suggest that rigorous and accurate measurements of HRV parameters are highly needed on this topic.