Heart Rate Variability as a Biomarker of Functional Outcomes in Persons with Acquired Brain Injury: Systematic Review and Meta-Analysis

Abstract

This review aimed to quantify correlations between heart rate variability (HRV) and functional outcomes after acquired brain injury (ABI). We conducted a literature search from inception to January 2020 via electronic databases, using search terms with HRV, ABI, and functional outcomes. Meta-analyses included 16 studies with 906 persons with ABI. Results demonstrated significant associations: Low frequency (LF) (r = −0.28) and SDNN (r = −0.33) with neurological function; LF (r = −0.33), High frequency (HF) (r = −0.22), SDNN (r = −0.22), and RMSSD (r = −0.23) with emotional function; and LF (r = 0.34), HF (r = 0.41 to 0.43), SDNN (r = 0.43 to 0.51), and RMSSD (r = 0.46) with behavioral function. Results indicate that higher HRV is related to better neurological, emotional, and behavioral functions after ABI. In addition, persons with stroke showed lower HF (SMD = −0.50) and SDNN (SMD = −0.75) than healthy controls. The findings support the use of HRV as a biomarker to facilitate precise monitoring of post-ABI functions.

1. Introduction

Acquired brain injury (ABI) refers to any brain damage that occurs after birth (O’Rance and Fortune, 2007). Common causes of ABI include both traumatic (e.g. accidents) and non-traumatic factors (e.g. stroke, cerebral infections, and degenerative neurological diseases) (O’Rance and Fortune, 2007; Lo et al., 2020; McGrath and King, 2004). ABI is a leading cause of long-term disability affecting multiple functional domains including cognition, emotions, and behaviors (O’Rance and Fortune, 2007; Lo et al., 2020; Winstein et al., 2016). Therefore, identifying accurate and objective biomarkers for measuring functional outcomes after ABI will build a strong foundation for advances in research and clinical management such as monitoring treatment responses or identifying populations for a trial (Selleck et al., 2017).

Heart rate variability (HRV) is a potential biomarker of functional outcomes. HRV is a simple and widely used noninvasive measure of the cardiac autonomic nervous system (ANS) (Berntson et al., 1997; Lewis, 2005). HRV refers to changes in the heart’s beat-to-beat intervals that are mainly regulated by the dynamic interplay of the sympathetic (SNS) and parasympathetic nervous systems (PNS) of the ANS (Malik et al., 1996). As a general rule, HRV decreases when the SNS activates in response to stressors (i.e. the flight-or-fight response), whereas HRV increases when the PNS (i.e. the rest-and-digest response) activates during resting periods (Appelhans and Luecken, 2006; Berntson et al., 1997). HRV is related to multiple functional outcomes including cognitive (e.g. executive functions and cognitive control), emotional (e.g. depression, stress, and anxiety), and behavioral (e.g. physical activity) functions (Beauchaine and Thayer, 2015; Ernst, 2017a; Fatisson et al., 2016; Jung et al., 2019).

The mechanisms underlying the relationship between HRV and functional outcomes are heart–brain interactions (Chand et al., 2020; Fatisson et al., 2016; Jung et al., 2019; Thayer et al., 2012). The PNS and the SNS are primarily influenced by the central autonomic network (CAN) (Benarroch, 1993). The CAN consists of prefrontal cortical (anterior cingulate, insular, orbitofrontal, and ventromedial cortices), limbic (the central nucleus of the amygdala, hypothalamus), and brainstem (periaqueductal gray matter, the nucleus of the solitary tract, nucleus ambiguus, ventrolateral medulla, ventromedial medulla) structures (Benarroch, 1993) that play critical roles in emotional, cognitive, and behavioral functions (Fatisson et al., 2016). As a functional unit, the CAN is a crucial component of an internal regulation system by which the brain controls the heart through visceromotor, neuroendocrine, and behavioral responses that are adaptive and flexible for various environmental demands (Benarroch, 1993; Thayer and Lane, 2000). Specifically, prefrontal cortical activity, which is the main brain structure of cognitive controls and emotional functions, affects the PNS, which in turn influences cardiac activity indicated by HRV (Benarroch, 1993; Thayer and Lane, 2000). In general, increases in HRV are related to better health outcomes (Ernst, 2017a; Kemp and Quintana, 2013).

HRV has been suggested as a useful biomarker for persons with ABI (Galea et al., 2018; Hasen et al., 2019; Lees et al., 2018; Seltzer et al., 2019; Yperzeele et al., 2015). Previous investigations have reported that persons with ABI can experience autonomic dysregulation represented in HRV (Constantinescu et al., 2019; Takahashi et al., 2015). Indeed, persons with ABI have lower HRV compared to healthy controls (Galea et al., 2018; Keren et al., 2005; Vistisen et al., 2014). Additionally, a previous qualitative review has suggested that HRV can be associated with functional outcomes including neurological functions and functional independence in persons with moderate-to-severe traumatic brain injury (TBI) (Hasen et al., 2019). Two review studies noted that HRV is a potential biomarker that is predictive of stroke and post-stroke complications, as well as functional outcomes including neurological severity and functional independence after stroke (Lees et al., 2018; Yperzeele et al., 2015). Although these previous studies have qualitatively suggested the association of HRV with functional outcomes in persons with ABI (Galea et al., 2018; Hasen et al., 2019; Lees et al., 2018; Seltzer et al., 2019; Yperzeele et al., 2015), none of these studies has attempted to quantitatively synthesize the results. Therefore, the main objective of this study was to quantify the effect sizes of correlations between HRV and functional outcomes in persons with ABI using meta-analysis to evaluate the use of HRV as a biomarker for functional outcomes in persons with ABI. We hypothesized that associations of HRV with cognitive, emotional, and behavioral domains would be statistically significant and have moderate effects. Our second objective was to compare differences in HRV between persons with ABI and healthy controls. We hypothesized that persons with ABI would generally show lower HRV indices than healthy controls.

2. Methods

2.1. Protocol Registration

We followed PRISMA (Preferred Reported Items for Systematic Reviews and Meta-Analyses) guidelines (Moher et al., 2009) and pre-registered the protocol in PROSPERO (CRD42020177327).

2.2. Literature Search

We performed a systematic literature search from inception to January 2020 in Ovid Medline, Embase, Scopus, Cochrane Central Register of Controlled Trials, and Clinicaltrials.gov. A medical librarian created search strategies with a combination of keywords and standardized terms (Appendix A). The search yielded 4,932 results: Ovid Medline n = 1,932, Embase n = 1,842, Scopus n = 975, Cochrane Central Register of Controlled Trials n = 180, and Clinicaltrials.gov n = 3. We exported results to EndNote and removed 992 duplicates, yielding 3,940 unique citations.

2.3. Study Selection

We examined 3,940 studies for eligibility through title/abstract screening followed by full-text screening. The inclusion criteria were studies that: (1) focused on ABI (e.g. stroke, traumatic brain injury, brain tumor, encephalitis); (2) examined associations between HRV and functional outcomes (e.g. neurological, cognitive, emotional, motor, activity of daily living/instrumental activity of daily living functioning) (Lo et al., 2020) measured using established and standardized instruments; (3) computed HRV indices following the guidelines of the Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology (Malik et al., 1996); (4) included participants aged 18+ years; (5) were written in English; and (6) were published in peer-reviewed journals. We excluded studies that examined the effects of interventions on either functional outcomes or HRV (e.g. effects of HRV biofeedback on functional outcomes or effects of physical exercise on HRV). Moreover, we excluded nonexperimental studies such as reviews, meta-analyses, comments/letters, case reports, protocols, as well as abstracts from conference proceedings.

We excluded 3,754 irrelevant studies in the title/abstract screening. Of the remaining 186 studies, we excluded 164 via full-text review. Following multiple meetings and email communications, we included 22 studies satisfying inclusion and exclusion criteria in the qualitative systematic review. We screened reference lists of the included studies and earlier relevant reviews to check for any additional studies and found no studies. Of these 22 studies, we included 16 in the meta-analyses. Six studies appeared only in the qualitative systematic review because (1) four studies used independent t-tests, not correlation analyses (Bassi et al., 2007; Rapenne et al., 2001; Tang et al., 2020; Tsai et al., 2019), (2) one study reported insufficient data (Beer et al., 2017), and (3) one study did not follow the guidelines (Malik et al., 1996) to estimate the HRV index (i.e., High frequency [HF]) used in the study (Graff et al., 2013). Two independent researchers (YL and RW) executed study selection procedures, and a third reviewer (AW) resolved disagreements. Figure 1 illustrates the study selection process.

Figure 1. Flow diagram of the study selection process.

This figure explains the study selection process, from the literature search to the full-text screening, to identify the included studies satisfying the inclusion and exclusion criteria.

2.4. Data Extraction

We developed a standardized data coding form in Microsoft Excel to extract information regarding (1) characteristics of study samples (i.e. types of ABI, age, sex, and sample size), (2) HRV measures (i.e. device, conditions, HRV domains, indices, and HRV units), (3) functional outcomes (i.e. names and purposes of measures, and time point for measures), and (4) major findings. To compute correlation effect sizes (the main analysis), we extracted correlation coefficients between functional outcomes and HRV indices. We also extracted means and SDs from the ABI and healthy control groups to compute effect sizes of differences between the two groups (the secondary analysis).

Five included studies reported insufficient information for effect size computations, thus, we contacted the corresponding authors via email to request the necessary information. Three provided missing information, and two did not respond. Two researchers performed data extraction independently (YL and RW). Any discrepancies between the researchers were resolved via discussion and consensus with a third researcher (AW).

2.5. Statistical Analysis

We performed all statistical analyses with R statistical software version 3.2.4 using the Metaphor package (Viechtbauer, 2010). We performed meta-analyses when at least two studies were available with common HRV indices and functional outcomes (Higgins et al., 2019). In meta-analyses, we did not include the HRV indices of frequency domain when the included study did not strictly follow the frequency bands suggested by the guidelines (Malik et al., 1996): HF (0.15–0.6 Hz) in Graff et al’s study (2013); LF (0.04–0.14 Hz), HF (0.15–0.5 Hz), and LF/HF ratio in Hilz et al.’s study (2011), and HF (0.15–0.5 Hz) and LF/HF ratio in Hilz et al.’s study (2017) (Table 1). Although the samples of three studies overlapped (Liao et al., 2016; Sung, Chen, et al., 2016; Sung, Lee, et al., 2016), each reported separate results obtained with different HRV indices (i.e. frequency and time domains), measurements (i.e. Beck Anxiety Inventory [BAI] and Beck Depression Inventory [BDI]), and stages (i.e. acute and chronic stages). Thus, we included a unique study in separate meta-analyses so that each sample only contributed one effect size to ensure independence and minimize the conflation of effect-size estimates. We used a random-effects model in all meta-analyses to account for heterogeneity in the included studies (Borenstein et al., 2010).

Table 1.

Characteristics of the included studie

No	Author	Year	Population	Sample size (m:f)	Age, mean±SD (median, range)	HRV device	HRV recording	Domain measures of HRV	Major findings
1	Arad et al*	2002	IS	16 (10:6)	73±10	ECG recording using lead II or V3	Short-term (5 min) during routine physiotherapy effort	F: LF (0.04–0.15 Hz), HF (0.15–0.4Hz) T: SDNN	Higher LF, HF, and SDNN were correlated with better functional independence
2	Bassi et al	2007	Acute IS	85 (46:39)	60±12.4	24-hour Holter monitoring using a 3-channel bipolar recorder	24 hr	T: SDNN, RMSSD, SDANN, pNN50	Patients with lower ADL independence (BI score <75) showed significantly lower SDNN and SDANN than those with higher ADL independence (BI score ≥75)
3	Beer et al	2017	Subacute IS	13 (10:3)	(60, 42–71)	Polar Advanced Heart Rate Monitor (RS800CX)	Short-term in resting, breathing, grip, and cognitive stimulus	F: HF (0.15–0.4Hz) T: SDNN, RMSSD	HRV of stroke patients showed less sensitivity to changes in testing conditions and failed to show correlations with cognitive performance
4	Chen et al*	2012	Acute IS	50 (27:23)	64.2±11.8	ECG recording using lead I	Short-term (5 min) in the resting supine position	F: LF (0.04–0.15 Hz), HF (0.15–0.4Hz), LF/HF ratio, VLF (0.003–0.04 Hz), TP (0.003–0.4 Hz) T: RRI	Lower TP and HF were correlated with higher stroke severity but not so other indices
5	Francis et al*	2016	Chronic TBI	30 (23:7)	45.73±13.68	ECG recording using 3 Ag/AgCl sensors	Short-term (5 min) in the resting sitting position	F: LF (0.04–0.15 Hz), HF (0.15–0.4Hz), LF/HF ratio T: SDNN, RMSSD	Lower SDNN, RMSSD, LF, and HF were correlated with higher social cognition, empathy, and alexithymia but not so LF/HF ratio
6	Graff et al	2013	IS	63 (44:19)	(62, 30–87)	ECG recordings using a PowerLab system with Lab Chart software	Short-term in the resting position	F: LF (0.04–0.15Hz), HF (0.15–0.6Hz), LF/HF ratio, VLF (0–0.04Hz) T: SDNN, RMSSD, RRI mean, pNN50, NN50 Non-linear: SD1, SD2, DFAαl, DFAα2, ApEn, SampEn, FuzzyEn	Lower LF/HF ratio was correlated with higher stroke severity and lower functional independence A group with poor neurological recovery showed lower LF, HF, LFnu, LF/HF ratio, ApEn, SampEn, and FuzzyEn but showed higher HF(%) and HFnu compared to the group with good neurological recovery
7	Hilz et al*	2011	Acute IS	50 (25:25)	65.8±12.7	ECG recording using a 3-lead	Short-term (5 min) in the resting position	F: LF (0.04–0.14 Hz), HF (0.15–0.50Hz), TP (unclear), LF/HF ratio T: RRI, SDNN, RMSSD, RRI-CV	Higher stroke severity was correlated with higher LF and LF/HF ratio but with lower SDNN, RMSSD, HF, and TP
8	Hilz et al*	2017	Chronic TBI	40 (30:10)	33.8	ECG recording using a 3-lead	Short-term (2 min) in the resting supine and standing position	F: LF (0.04–0.15 Hz), HF (0.15–0.5Hz), LF/HF ratio T: RRI, SDNN, RMSSD, RRI-CV	Higher LFnu, LF/HF ratio, and HF were correlated with higher neurological functions in the standing position No correlations between HRV and neurological functions were found in the supine position
9	Katz-Leurer et al*	2005	Acute IS	39 (19:20)	63±10	Holter monitoring using ECG with a Marquette 8500 recorder	Short-term (10–12 min) in the resting supine position with free breathing	F: LF (0.04–0.15 Hz), HF (0.15–0.4Hz) T: SDNN, RMSSD	Lower SDNN was correlated with higher stroke severity and lower functional independence Higher SDNN, RMSSD, LF, and HF were correlated with higher motor function
10	Korpelainen et al*	1996	Acute IS	31 (21:10)	52.2±11.1	24-hour ECG recording	24 hr	F: LF (0.04–0.15Hz), HF (0.15–0.4Hz), VLF (0.005–0.04 Hz) T: SDNN, RMSSD	Higher SDNN, TP, VLF, and LF were correlated with lower stroke severity and better ADL independence
11	Liao et al*	2016	Acute TBI	165 (65:100)	40.08±14.75	ECG	Short-term in the resting position	F: LF (0.04–0.15Hz), HF (0.15–0.4Hz), LF/HF ratio, VLF (0.0033–0.04Hz), TP (0–0.4 Hz) T: SDNN	Lower TP, VLF, LF, and HF were correlated with higher anxiety
12	Rapenne et al	2001	TBI	13 (Unclear)	Unclear	24-hour Holter monitoring using 2-lead ECG	24 hr	F: LF (0.04–0.15Hz), HF (0.15–0.4Hz), LF/HF ratio, TP (unclear) T: RMSSD, pNN50	Patients with lower neurological function (GCS <10) had significantly lower RMSSD, pNN50, and LF/HF ratio than those with higher neurological function
13	Robinson et al*	2008	IS, HS	49 (24:25)	63±13.3	24-hour Holter monitoring using a 3-channel bipolar recorder	24 hr	T: SDNN, RMSSD, SDANN, pNN50	SDNN was not correlated with depression, anxiety, functional independence, or cognitive function People with depression had significantly lower SDNN than those without depression but not so other indices
14	Sethi et al*	2016	Acute IS, HS	13 (7:6)	61±12	ECG	24 hr	T: SDNN	Higher SDNN was correlated with higher motor function measured using FMA
15	Shapira-Vadler et al*	2015	Subacute IS, HS	23 (12:11)	58±10.5	Polar Advanced Heart Rate Monitor (RC800CX)	Short-term (6 min) in the resting prone position	F: LF (0.04–0.15Hz), HF (0.15–0.4Hz), LF/HF ratio T: SDNN, RMSSD	Higher SDNN and RMSSD were correlated with higher motor function (6MWT)
16	Sung, Chen et al*	2016 (a)	TBI	125 (48:77)	40.5±25.5	ECG recording using 1st lead I	Short-term (5 min) in the resting position	F: LF (0.04–0.15Hz), HF (0.15–0.4Hz), LF/HF ratio, VLF (0.0033–0.04Hz), TP (0–0.4 Hz) T: SDNN	Lower LF was correlated with higher anxiety while higher LF/HF ratio was associated with higher depression
17	Sung, Lee, et al*	2016 (b)	TBI	109 (0:109)	35.59±15.63	ECG recording using 1st lead I	Short-term (5 min) in the resting position	F: LF (0.04–0.15Hz), HF (0.15–0.4Hz), LF/HF ratio, VLF (0.0033–0.04Hz), TP (0–0.4Hz) T: SDNN	Negative correlations between LFnu, HFnu, LF/HF ratio, and depression
18	Szabo et al*	2018	Acute HS	47 (27:20)	60.8±16.5	Finometer device	Short-term (10 min) in the resting position	F: LF (0.04–0.15Hz), HF (0.15–0.4Hz), LF/HF ratio, TP (0.003–0.4Hz)	Lower HF, LF, and LF/HF ratio were correlated with higher stroke severity. Patients with poor outcomes (mRS ≥4) had higher HFnu, lower LF/HF ratio, lower LF(nu), and lower HF
19	Tang et al	2020	Acute IS	142 (125:17)	63.9±10.2	ECG	Short-term (5 min) in the resting position	F: LF (0.04–0.15Hz), HF (0.15–0.4Hz), LF/HF ratio T: RRI	No difference in HRV between people with poor outcomes (mRS >3) and those with good outcomes
20	Tessier et al*	2017	Acute IS	56 (39:17)	51.7±13	HR monitor Polar RS 800	Short-term in the resting supine position	F: LF (0.04–0.15Hz), HF (0.15–0.4Hz), LF/HF ratio T: RMSSD	Lower RMSSD, LF, HF, and LF/HF ratio were correlated with severe depression. No correlations between HRV and neurological function. Higher RMSSD, LF, and HF were associated with better cognitive function
21	Tsai et al	2019	Acute IS	34 (26:12)	63.2±8.7	Standard 3-lead ECG (Ivy Biomedical, Model 3000; Branford, CT, USA)	Short-term	F: LF (0.04–0.15Hz), HF (0.15–0.4Hz), LF/HF ratio	No difference in HRV between people with good outcomes and poor outcomes (mRS ≥4)
22	Xu et al*	2016	Acute IS	63 (38:25)	70.7±11.92	24-hour Holter monitoring using ECG	24 hr	T: RRI, SDNN, RMSSD	Lower SDNN was negatively associated with higher stroke severity

TBI: Traumatic brain injury, IS: ischemic stroke, HS: hemorrhagic stroke, M: male, F: female, ECG: electrocardiogram, F: frequency, T: time, LFA: low frequency area, LF: low frequency, LFnu: normalized low frequency, HF: high frequency, HFnu: normalized high frequency, VLF: very low frequency, RRI: RR interval, RRI-CV: coefficient of variation of RRIs, BI: Barthel Index, ADL: activities of daily living, FMA: Fugl-Meyer Assessment, GCS: Glasgow Coma Scale, 6MWT: 6-Minute Walk Test, mRS: modified Rankin Scale

^*Studies included in meta-analyses

The main objective of this study was to synthesize correlation coefficients between functional outcomes and HRV indices to evaluate the use of HRV as a biomarker of functional outcomes after ABI. To account for differences in scale direction, when measurements had a scale direction opposite the other tools under the same functional domain (i.e. Scandinavian Stroke Scale [SSS] under neurological functions), we multiplied correlation coefficients by −1 (Higgins et al., 2019). We calculated overall effect sizes based on Pearson’s r correlations. Among 16 studies included in the meta-analyses, nine reported Pearson’s r and seven reported Spearman’s ρ correlations. We first transformed raw effect sizes reported as Spearman’s ρ to Pearson’s r according to Gilpin’s conversion table (Gilpin, 1993). Then we transformed all Pearson’s r correlations into Fisher’s z scores for variance stabilization. We used Fisher’s z scores and their variances in the meta-analyses. Then we back-transformed all pooled estimates to Pearson’s r correlations for interpretation and forest plot generation (Borenstein et al., 2011). Correlations of 0.10, 0.30, and 0.50 represented small, medium, and large correlations, respectively (Cohen, 2013).

We performed a secondary analysis to compare HRV values in persons with ABI, including subgroups of stroke and TBI, with healthy controls. To synthesize effect sizes, we calculated standardized mean differences (SMDs) using Hedge’s adjusted g with a 95% confidence interval (CI). We used Hedge’s adjusted g because it accommodates the low sample sizes of the included studies (Hedges and Olkin, 2014). SMD values of 0.2, 0.5, and 0.8 were defined as small, medium, and large effect sizes, respectively (Cohen, 2013).

We examined Q and I² statistics to evaluate heterogeneity among effect sizes of the included studies. We considered heterogeneity present when the Q-value was significant (p < 0.10). I² statistics provide the amount of variance among the included studies; I² values of 25%, 50%, and 75% indicated low, moderate, and high heterogeneity, respectively (Higgins et al., 2003). When we observed moderate-to-high heterogeneity (I² > 50%) in the meta-analysis with more than two studies, we adopted univariate meta-regression analyses with random-effects models to examine moderators of heterogeneity. We included the following potential moderators in the meta-regression analyses: age, sex, HRV unit, type of injury, type of measurement, and HRV recording duration.

3. Results

3.1. Qualitative Systematic Review

3.1.1. Study Sample

Among the 22 studies that met the inclusion and exclusion criteria (Figure 1, Table 1), 16 reported persons with stroke, and six reported persons with TBI. Of 16 studies of persons with stroke, types of stroke included hemorrhagic stroke (n = 1), ischemic stroke (n = 12), and both hemorrhagic and ischemic stroke (n = 3). The total sample size of the included studies was 1,256 (stroke = 774, TBI = 482). Most studies contained mixed male and female samples; only one focused on a female sample (Sung, Lee, et al., 2016).

3.1.2. HRV Measures

HRV recording: Duration, position, and devices.

Durations of HRV recording were short-term recording (i.e. 5–15 minutes) (n = 16) and long-term recording (i.e. 24 hours) (n = 6). The 16 studies using short-term recording measured HRV: in a resting position (n = 11); mixed resting positions with cognitive stimulus, walking, and standing positions (n = 3); during a routine physiotherapy session (n = 1); and in an unclear position (n = 1). Short-term recording studies frequently used electrocardiograms (ECGs) (n = 12), wrist-worn Polar heart rate monitors (n = 3), and a Finometer device (n = 1). Long-term recording studies measured HRV with ECGs (n = 4) and three-channel bipolar recorders (n = 2) (see Table 1).

HRV domains and indices.

HRV parameters consisted of linear parameters, including time and frequency domains, and nonlinear parameters (Malik et al., 1996; Shaffer and Ginsberg, 2017). Time domain analysis measured variations in heart rate over time or intervals between successive normal cardiac cycles. Frequency domain analysis assessed periodic oscillations of heart rate signals comprising different frequencies and amplitudes, such as low frequency (LF) (0.04–0.15 Hz) and HF (0.15–0.40 Hz) (Malik et al., 1996). Nonlinear parameters such as a Poincaré plot assessed the dynamics of HRV by plotting every R–R interval against the prior interval, creating a scatterplot of hidden correlation patterns of a time series signal (Shaffer and Ginsberg, 2017). Most of the included studies measured both time and frequency domain indices of HRV (n = 16). Two studies measured only frequency domain indices (Szabo et al., 2018; Tsai et al., 2019). Four studies measured only time domain indices (Bassi et al., 2007; Robinson et al., 2008; Sethi et al., 2016; Xu et al., 2016). Only one study applied nonlinear indices of HRV, in addition to time and frequency domain indices (Graff et al., 2013). Time domain studies frequently used SDNN and RMSSD. Frequency domain studies often used Total Power (TP), LF, HF, and LF/HF ratio. Among these six indices, HF and RMSSD reflect parasympathetic activity (Ernst, 2017b; Shaffer and Ginsberg, 2017). SDNN, which is the square root of the variance, reflects all the cyclic components responsible for variability in the recording period; thus, it may be a more general index (Malik et al., 1996). Both sympathetic and parasympathetic activities can contribute to SDNN(Malik et al., 1996; Shaffer and Ginsberg, 2017). LF can reflect both sympathetic and parasympathetic activities in addition to baroreflex activity (Ernst, 2017b; Goldstein et al., 2011; Shaffer and Ginsberg, 2017). The LF/HF ratio has been considered an autonomic balance, this concept has been extensively challenged (Billman, 2013). TP is the sum of HRV indices of frequency domains (e.g. LF, HF, and very low frequency [VLF]) (Ernst, 2017b; Shaffer and Ginsberg, 2017). See Table 2 for descriptions of HRV domains and indices.

Table 2.

Descriptions and prevalence of HRV indices in the included studies

Variable (units)	Description	No. of studies N (%)
Frequency domain (frequency range)
TP (ms2)	• Short-term: The variance of Normal–Normal (NN) intervals over the temporal segment (≈≤<0.4 Hz) • 24 hours: Variance of all NN intervals (≈≤0.4 Hz)	7 (29.17%)
VLF (ms2)	Power in VLF range (Short-term: ≤0.04 Hz, 24 hours: 0.003–0.04 Hz)	6 (25.00%)
LF (ms2)	Power in LF range (0.04–0.15 Hz)	18 (75.00%)
LF norm (nu)	LF power in normalized units: LF/(total power–VLF)×100	7 (29.17%)
HF (ms2)	Power in HF range (0.15–0.4Hz)	18 (75.00%)
HF norm (nu)	HF power in normalized units: HF/(total power–VLF)×100	7 (29.17%)
LF/HF ratio	Ratio LF (ms2)/HF (ms2)	15 (62.50%)
Time-domain: statistical measures
SDNN (ms)	The standard deviation of all normal RR (NN) intervals	17 (70.83%)
SDANN (ms)	The standard deviation of the averages of normal RR (NN) intervals in all 5-minute segments of the entire recording	2 (8.33%)
RMSSD (ms)	The square root of the mean of the sum of the squares of differences between adjacent normal RR (NN) intervals	14 (58.33%)
SDNN index (ms)	Mean of the standard deviations of all normal RR (NN) intervals for all 5-minute segments of the entire recording	0 (0.00%)
NN50 count (ms)	Number of pairs of adjacent NN intervals differing by more than 50ms in the entire recording	2 (8.33%)
pNN50 (%)	NN50 count divided by the total number of all normal RR (NN intervals)	5 (20.83%)
Nonlinear domain
SD1 (ms)	Poincaré plot standard deviation perpendicular to the line of identity	1 (4.17%)
SD2 (ms)	Poincaré plot standard deviation along the line of identity	1 (4.17%)
DFAα1	Detrended fluctuation analysis, which describes short-term fluctuations	1 (4.17%)
DFAα2	Detrended fluctuation analysis, which describes long-term fluctuations	1 (4.17%)
ApEn	Approximate entropy	1 (4.17%)
SampEn	Sample entropy	1 (4.17%)
FuzzyEn	Fuzzy entropy	1 (4.17%)

TP: Total Power, VLF: very low frequency, LF: low frequency, LF norm: low frequency norm, HF: high frequency, HF norm: high-frequency norm, LF/HF ratio: ratio of low frequency and high frequency

^*Descriptions of HRV indices were extracted from the guidelines of the Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology (Malik et al., 1996).

3.1.3. Functional Outcomes

We categorized functional outcomes into four domains: (1) neurological functions (n = 10), (2) cognitive functions (n = 3), (3) emotional functions (n = 6), and (4) behavioral functions (n = 10) (Table 3).

Table 3.

Measurements for functional outcomes in the included studies

No	Author	Year	Neurological functions (n = 10, 43.5%)	Cognitive functions (n = 3, 17.4%)	Emotional functions (n = 6, 26.1%)	Behavioral functions (n = 10, 43.5%)
1	Arad et al*	2002				FIM**
2	Bassi et al	2007				BI
3	Beer et al	2017		MoCA, Timed serial-2 subtraction task
4	Chen et al*	2012	NIHSS**
5	Francis et al*	2016			TAS**, DASS, BEES, IRI, BVAQ, TASIT+
6	Graff et al	2013	NIHSS			mRS+
7	Hilz et al*	2011	NIHSS**
8	Hilz et al*	2017	GOSE+
9	Katz-Leurer et al*	2005	SSS**+			MAS**, FIM**
10	Korpelainen et al*	1996	SSS**+			BI**
11	Liao et al*	2016			BAI**
12	Rapenne et al	2001	GCS+
13	Robinson et al*	2008		MMSE	HDRS**, HAS	BI
14	Sethi et al*	2016				FMA**
15	Shapira-Vadler et al*	2015				6MWT**, MMT, FR, FIM
16	Sung, Chen, et al*	2016a			BAI**, BDI**
17	Sung, Lee, et al*	2016b			BDI-II
18	Szabo et al*	2018	NIHSS**			mRS+
19	Tang et al	2020				mRS+
20	Tessier et al*	2017	NIHSS**	MoCA	BDI-II**, Global z-score of depression**, HADS, SADQ, VASdep,
21	Tsai et al	2019				mRS+
22	Xu et al*	2016	NIHSS**+

^*Studies included in meta-analyses

^**Measures used in meta-analyses

⁺Measures that have the opposite direction of scale with other tools under the same domain

Neurological functions: NIHSS; National Institutes of Health Stroke Scale (Higher=Poorer), GOSE; Extended Glasgow Outcome Scale (Higher=Better), SSS: Scandinavian Stroke Scale (Higher=Better), GCS: Glasgow Coma Scale (Higher=Better)

- Cognitive functions: MoCA; Montreal Cognitive Assessment (Higher=Better), MMSE; Mini-Mental Status Exam (Higher=Better)

-Emotional functions: TAS; Toronto Alexithymia Scale (Higher=Poorer), DASS; Depression Anxiety Stress Scale (Higher=Poorer), BEES; Balanced Emotional Empathy Scale (Higher=Poorer), IRI; Interpersonal Reactivity Index (Higher=Poorer), BVAQ; Bermond-Vorst Alexithymia Questionnaire (Higher=Poorer), TASIT; The Awareness of Social Inference Test (Higher=Better), HDRS; Hamilton Depression Rating Scale (Higher=More severe), HDAS; Hamilton Depression Anxiety Scale (Higher=More severe), BAI; Beck Anxiety Inventory (Higher=More severe), BDI; Beck Depression Inventory (Higher=More severe), HASD; Hospital Anxiety and Depression Scale (Higher=More severe), SADQ-10; Stroke Aphasic Depression Questionnaire (Higher=More severe), VASdep; Visual Analog Scale for depression (Higher=More severe)

-Behavioral functions: FIM; Functional Independence measure (Higher=Better), BI; Barthel Index (Higher=Better), mRS; Modified Rankin Scale (Higher=Poorer), MAS; Motor Assessment Scale (Higher=Better), FMA; Fugl-Meyer Assessment (Higher=Better), 6MWT; 6-Minute Walk Test (Higher=Better), MMT; Manual muscle testing (Higher=Better), FR; Modified functional reach test (Higher-Better)

Neurological functions.

Among the 10 studies, stroke studies used the National Institutes of Health Stroke Scale (NIHSS) (n=6) and the SSS (n=2), and TBI studies used the Extended Glasgow Outcome Scale (GOSE) (n=1) and the Glasgow Coma Scale (GCS) (n=1) (Table 3). The majority of studies (n = 6) reported that reduced HRV indices (i.e., TP, HF, LF, VLF, LF/HF ratio, SDNN, RMSSD, and pNN50) were associated with more severe neurological dysfunction (Chen et al., 2012; Katz-Leurer and Shochina, 2005; Korpelainen et al., 1996; Rapenne et al., 2001; Szabo et al., 2018; Xu et al., 2016). Three studies reported mixed findings (Graff et al., 2013; Hilz et al., 2011; Hilz et al., 2017). Graff et al. (2013) reported that the group with worse neurological function showed lower HRV (i.e., LF, HF, LF in normalized units (nu), LF/HF ratio, ApEn, SampEn, and FuzzyEn) but higher HFnu compared to the group with better neurological function. Conversely, Hilz et al. (2011) indicated that stroke neurological severity was associated with higher LF and LF/HF ratio but lower SDNN, RMSSD, HF, and TP. Hilz et al.’s study (2017) observed no correlations between HRV and neurological functions in the supine position, although higher LFnu, LF/HF ratio, and lower HF were related to higher neurological function in persons with TBI in a standing position (Hilz et al., 2017). Last, Tessier et al.’s study (2017) indicated no correlations between HRV indices and neurological functions (Table 1).

Emotional functions.

Emotional functions measured in the six studies were alexithymia (Francis et al., 2016), anxiety (Liao et al., 2016; Robinson et al., 2008; Sung, Chen, et al., 2016; Tessier et al., 2017), and depression (Liao et al., 2016; Robinson et al., 2008; Sung, Chen, et al., 2016; Sung, Lee, et al., 2016; Tessier et al., 2017). Studies of depression (n = 4) used the BDI (n = 3) most frequently (Table 3). Most studies (four of six) demonstrated that lower HRV, including SDNN, RMSSD, LF, LFnu, HF, HFnu, VLF, TP, and LF/HF ratio, were correlated with higher emotional dysfunction in empathy (Francis et al., 2016), anxiety (Liao et al., 2016), and depression (Sung, Lee, et al., 2016; Tessier et al., 2017). Two studies offered divergent findings: Robinson et al. (2008) suggested that people with depression had significantly lower SDNN than those without depression, although SDNN was not correlated with depression and anxiety (Robinson et al., 2008). In addition, Sung, Chen, et al. (2016) indicated that lower LF was correlated with higher anxiety, while higher LF/HF ratio was associated with higher depression (Table 1).

Cognitive functions.

Three studies addressed cognitive functions (Beer et al., 2017; Robinson et al., 2008; Tessier et al., 2017). Studies measured cognitive functions using the Montreal Cognitive Assessment (MoCA) (Beer et al., 2017; Tessier et al., 2017), and the Mini-Mental Status Exam (MMSE) (Robinson et al., 2008) (Table 3). One study demonstrated a positive relationship between HRV (including LF, HF, and RMSSD) and cognitive functions (Tessier et al., 2017). In contrast, two studies showed no relationship between HRV and cognitive functions (Beer et al., 2017; Robinson et al., 2008) (Table 1).

Behavioral functions.

Studies reported behavioral functions including functional independence and physical function. The modified Rankin Scale (mRS) (n = 4) (Graff et al., 2013; Szabo et al., 2018; Tang et al., 2020; Tsai et al., 2019) was the most frequently used measure of functional independence, followed by the Barthel Index (BI) (n = 3) (Bassi et al., 2007; Korpelainen et al., 1996; Robinson et al., 2008), and the Functional Independence Measure (FIM) (n = 2) (Arad et al., 2002; Katz-Leurer and Shochina, 2005) (Table 1). Two studies that used the FIM demonstrated positive correlations with HRV, including LF, HF, and SDNN (Arad et al., 2002; Katz-Leurer and Shochina, 2005). However, studies using the mRS noted divergent findings: Graff et al. (2013) indicated that lower LF/HF ratio was correlated with higher mRS scores (Graff et al., 2013). In Szabo et al.’s study (2018), patients with poor outcomes (mRS ≥ 4) had higher HFnu, lower LF/HF ratio, lower LFnu, and lower HF compared to those with good outcomes. Tang et al.’s (2020) and Tsai et al.’s (2019) studies indicated no significant differences in HRV between people with poor (mRS > 3) and good outcomes. Studies using the BI provided more consistent findings: Bassi et al. suggested that patients with lower functional independence (BI < 75) showed significantly lower SDNN and SDANN than those with higher functional independence (BI ≥ 75) (Bassi et al., 2007). Aligning with Bassi’s work, Korpelainen et al. demonstrated that higher HRV including SDNN, TP, VLF, and LF was correlated with better functional independence (Korpelainen et al., 1996). However, one study reported no correlations between SDNN and BI scores (Robinson et al., 2008).

The Motor Assessment Scale (MAS) (Katz-Leurer and Shochina, 2005), the Fugl-Meyer Assessment (FMA) (Sethi et al., 2016), and the 6-Minute Walk Test (6MWT) (Shapira-Vadler et al., 2015) were also used to investigate relationships between HRV indices and physical functions. All results suggested that higher HRV (i.e. SDNN, RMSSD, LF, and HF) were correlated with higher physical functions in ABI (Katz-Leurer and Shochina, 2005; Sethi et al., 2016; Shapira-Vadler et al., 2015) (Table 3).

3.2. Meta-Analyses

We included 16 studies in meta-analyses, of which 15 studies were included in the main analysis and seven for the secondary analysis. For the main analysis, we synthesized correlation coefficients according to the domains of neurological, emotional, and behavioral functions. We excluded cognitive functions from meta-analyses due to the limited number of studies. We performed meta-analyses of neurological functions for persons with acute ABI because the included studies only measured neurological functions in the acute stage. In contrast, we performed meta-analyses of emotional and behavioral functions separately with persons with acute and chronic ABI. In general, we included LF, HF, and LF/HF ratio in the frequency domain, and RMSSD and SDNN in the time domain.

3.2.1. Correlations Between HRV and Functional Outcomes

Neurological functions.

We included seven studies in meta-analyses (Chen et al., 2012; Hilz et al., 2011; Katz-Leurer et al., 2005; Korpelainen et al., 1996; Szabo et al., 2018; Tessier et al., 2017; Xu et al., 2016). Lower scores on measurements of neurological functions indicate better neurological function. Thus, negative correlation coefficients between neurological functions and HRV indices can be interpreted as: higher HRV is correlated with better neurological function (or lower scores on neurological measurements). Our results showed that LF and SDNN had significant medium negative correlations with neurological functions (r = −0.28 and −0.33, respectively, p < 0.05) (Figure 2). Other HRV indices, including HF, LF/HF ratio, and RMSSD, did not have significant correlations. We performed meta-regression with heterogeneities in meta-analyses with LF, HF, LF/HF ratio, and RMSSD. Results showed that sex could be potential moderating factors in correlations between RMSSD and neurological functions (Appendix B).

Figure 2. Correlations between HRV indices and neurological functions at the acute stage.

This figure includes forest plots of meta-analyses showing correlations between HRV indices and neurological functions after acquired brain injuries at the acute stage. We used Pearson’s correlation coefficients to quantify the effect sizes of correlations. We generated the plots according to the frequently used HRV indices in the included studies. Studies measuring neurological functions included the National Institutes of Health Stroke Scale and the Scandinavian Stroke Scale. Negative correlations in the plots indicate that higher HRV is correlated with better neurological function.

HRV: Heart rate variability, NIHSS: National Institutes of Health Stroke Scale, SSS: Scandinavian Stroke Scale. Note: Negative correlations indicate that higher HRV is related to better neurological function.

Emotional functions.

We included six studies in meta-analyses (Francis et al., 2016; Liao et al., 2016; Robinson et al., 2008; Sung, Chen, et al., 2016; Sung, Lee, et al., 2016; Tessier et al., 2017). Lower scores on measurements of emotional functions indicate better emotional function (i.e. fewer emotional symptoms). Thus, negative correlation coefficients between emotional functions and HRV indices can be interpreted as: higher HRV is correlated with better emotional function. In the chronic stage, our results showed that LF (r = −0.33), HF (r = −0.22), SDNN (r = −0.22), and RMSSD (r = −0.23) had small-to-medium negative correlations with emotional function (p < 0.05), while LF/HF ratio had no significant correlations (p > 0.05) (Figure 2). HF and RMSSD have long been used to estimate the vagally (parasympathetically) mediated HRV, which reflects the potential for fast parasympathetic modulation of autonomic control of the heart (Shaffer and Ginsberg, 2017). Our results suggest that higher vagally-mediated HRV is associated with better emotional function. Moreover, although higher LF and SDNN were correlated with better emotional function, these results do not provide clear information related to sympathetic activity due to controversies in interpreting LF and SDNN (Billman, 2013; Goldstein et al., 2011; Shaffer and Ginsberg, 2017). In contrast, we found no significant correlations in the acute stage (p > 0.05). We performed meta-regression with LF and LF/HF ratio by age, sex, HRV unit, and HRV recording duration but found no significant moderators (p > 0.05) (Figure 3).

Figure 3. Correlations between HRV indices and emotional functions.

This figure includes forest plots of meta-analyses showing correlations between HRV indices and emotional functions after acquired brain injuries. We used Pearson’s correlation coefficients to quantify the effect sizes of correlations. We generated the plots according to the frequently used HRV indices in the included studies. Studies measuring emotional functions included the Beck Anxiety Inventory, Toronto Alexithymia Scale, Hamilton Depression Rating Scale, and the Beck Depression Inventory. Out of the emotional assessments used in the included studies, studies frequently measured depression and anxiety. Negative correlations in the plots indicate that higher HRV is correlated with better emotional function.

HRV: Heart rate variability, BAI: Beck Anxiety Inventory, TAS: Toronto Alexithymia Scale, HDS: Hamilton Depression Rating Scale, BDI: Beck Depression Inventory. Note: Negative correlations indicate that higher HRV is related to better emotional function. Sung et al (2016a): Sung, Chen, et al (2016). Sung et al (2016b): Sung, Lee, et al (2016)

Behavioral functions.

We included five studies in meta-analyses (Arad et al., 2002; Katz-Leurer and Shochina, 2005; Korpelainen et al., 1996; Sethi et al., 2016; Shapira-Vadler et al., 2015). Higher scores on measurements of behavioral functions indicated better behavioral function. Thus, positive correlation coefficients can be interpreted as: higher HRV is related to higher behavioral function. In the acute stage, LF (r = 0.34), HF (r = 0.43), and SDNN (r = 0.43) had medium positive correlations with behavioral function (p <0.05), suggesting that general autonomic regulation, including parasympathetic and sympathetic activities as well as baroreflex activity, might influence behavioral function in the acute stage. However, this result should be cautiously interpreted because of controversies in interpreting LF and SDNN (Billman, 2013; Goldstein et al., 2011; Shaffer and Ginsberg, 2017). In the chronic stage, HF (r = 0.41), SDNN (r = 0.51), and RMSSD (r = 0.46) had medium-to-large positive correlations with behavioral functions (p < 0.05), may reflecting that higher parasympathetic activities may be related to better behavioral function (Figure 4). We performed no meta-regression of behavioral functions because no results showed heterogeneity (Figure 4).

Figure 4. Correlations between HRV indices and behavioral functions.

This figure includes forest plots of meta-analyses showing correlations between HRV indices and behavioral functions after acquired brain injuries. We used Pearson’s correlation coefficients to quantify the effect sizes of correlations. We generated the plots according to the frequently used HRV indices in the included studies. Studies measuring behavioral functions included the Functional Independence Measure, Motor Assessment Scale, Barthel Index, Lower Extremity of Fugl-Meyer Assessment, 6-Minute Walk Test, and Manual Muscle Testing. The Functional Independence Measure and Motor Assessment Scale were the two most common measures used in the included studies. Positive correlations in the plots indicate that higher HRV is correlated with better behavioral function.

HRV: Heart rate variability, FIM: Functional Independence Measure, MAS: Motor Assessment Scale, BI: Barthel Index, FM-LE: Lower Extremity of Fugl-Meyer Assessment, 6MWT: 6-Minute Walk Test, MMT: Manual Muscle Testing.

Note: Positive correlations indicate that higher HRV is related to better behavioral function.

3.2.2. Comparison of HRV Between Persons with ABI and Healthy Control

Seven studies were included in meta-analyses (Francis et al., 2016; Hilz et al., 2011; Hilz et al., 2017; Korpelainen et al., 1996; Liao et al., 2016; Tsai et al., 2019; Xu et al., 2016). We included LF, HF, LH/HF ratio, and SDNN. In general, persons with ABI showed significantly lower SDNN (SMD = −0.51, 95% CI = −0.93 to −0.09) compared to controls. Effect sizes of SDNN were medium. In subgroup analyses of groups of TBI and stroke, reduced HF (SMD = −0.50, 95% CI = −0.89 to −0.11) and reduced SDNN (SMD = −0.75, 95% CI = −1.19 to −0.32) were mainly observed in stroke samples, which were medium-to-large effect sizes (Figure 5). Considering that HF reflects parasympathetic activity that significantly contributes to SDNN (Shaffer and Ginsberg, 2017), the results may suggest reduced parasympathetic activities in people with stroke. We performed meta-regression with age, sex, ABI type, HRV unit, and HRV recording duration, but we found no significant moderators (Appendix B).

Figure 5. Comparisons of HRV indices between people with acquired brain injuries and healthy controls.

This figure includes forest plots of meta-analyses comparing differences in HRV indices between people with acquired brain injuries and healthy controls. We used means and standard deviations to quantify the effect sizes of differences between the two comparison groups. We generated the plots according to the frequently used HRV indices in the included studies. Higher effect sizes indicate higher HRV.

4. Discussion

Our main objective was to synthesize the effect sizes of correlation coefficients between HRV indices and functional outcomes to evaluate the use of HRV as a biomarker of functional outcomes in persons with ABI. Of the 22 studies included in the qualitative systematic review, we included 16 studies in meta-analyses. Through the qualitative review, we categorized functional outcomes into four domains: neurological, cognitive, emotional, and behavioral functions. The included studies frequently used LF, HF, and LF/HF ratio in the frequency domain and SDNN and RMSSD in the time domain of HRV indices. Thus, we performed separate meta-analyses with LF, HF, LF/HF ratio, SDNN, and RMSSD in neurological, emotional, and behavioral functions. We excluded the domain of cognitive functions due to a limited number of studies. The meta-analyses supported using HRV, especially LF, HF, SDNN, and RMSSD as a biomarker of functional outcomes after ABI because we found that these HRV indices were correlated with functional outcomes with small-to-large effects.

HRV is an indicator of cardiac ANS function as well as heart–brain interactions (Ernst, 2017a). Previous studies have reported that increases in HRV are generally related to better health outcomes (Ernst, 2017a; Kemp and Quintana, 2013). Our study added that higher LF and SDNN are likely associated with better neurological function in persons with acute ABI. These results align with those of previous qualitative reviews proposing that HRV indices can predict neurological severity after stroke (Lees et al., 2018; Yperzeele et al., 2015). This finding can be explained by the role of the ANS. For measuring neurological function, the included studies used the NIHSS and SSS, which consist of consciousness, eye movement, speech, facial paralysis, and orientation items (Gray et al., 2009). The ANS moderates these functions (McCorry, 2007). Therefore, the HF and SDNN indices of HRV are potential biomarkers of neurological function after ABI.

Researchers have studied the relationship between ANS and emotional deficits since the 19^th century, starting with the classical theory of William James and Carl Lange (Ellenbroek et al., 2019). Currently, evidence supports the use of HRV as a transdiagnostic biomarker of emotional deficits (Beauchaine and Thayer, 2015; Ellenbroek et al., 2019; Karri et al., 2017). Our results corroborate this evidence, indicating that reduced HRV including LF, HF, SDNN, and RMSSD is significantly associated with emotional dysfunction such as alexithymia, depression, and anxiety in persons with ABI. A previous meta-analysis of major depression supported our findings and demonstrated that persons with major depression have significantly lower LF, HF, SDNN, and RMSSD (Koch et al., 2019). Another study has also suggested that people with combined major depressive disorders and comorbid generalized anxiety disorder have the greatest reduction in HRV indices such as SDNN, RMSSD, and HF (Kemp et al., 2012). Although interpretations of LF and SDNN can be controversial (Shaffer and Ginsberg, 2017; Goldstein et al., 2011), the findings suggested that reduced parasympathetic activities, reflected in HF and RMSSD (i.e., the HRV indices indicative of vagally-mediated HRV) are associated with emotional dysfunction. Higher HRV, which is indicative of autonomic flexibility, is related to better psychological flexibility (Ellenbroek et al., 2019). Psychological flexibility refers to flexibility in changing one’s emotions in response to situational demands, which is essential to psychological health and adjustment (Kashdan and Rottenberg, 2010). In other words, reduced HRV is associated with lower psychological flexibility, which can make it more difficult for one to adapt to stressors (Ellenbroek et al., 2019) and can lead to emotional deficits such as depression and anxiety (Galea et al., 2018).

In behavioral functions, our results indicate that higher HRV indices (i.e. LF, HF, SDNN, and RMSSD) are associated with better behavioral functions including motor function and functional independence. The connection between HRV and behavioral functions has been under-examined compared to connections with cognitive and emotional functions. However, anatomical and physiological connections between the somatomotor system and the ANS may offer neurological underpinnings to explain this finding (Keizer and Kuypers, 1984; Kerman, 2008; Sequeira et al., 2000). Moreover, similar to emotional function, high psychological flexibility, manifested by higher HRV, can help persons with ABI overcome challenges caused by their ABI to gain more behavioral functional independence (Hardy and Segerstrom, 2017).

In the secondary analysis, we compared HRV indices between persons with ABI to healthy controls. Overall results indicated that persons with ABI have significantly lower SDNN than healthy controls. Moreover, in this study, significant differences in HRV between the comparison groups mainly come from the stroke sample, which showed that persons with stroke have significantly lower HF and SDNN than healthy controls. HF reflects parasympathetic activity that substantially contributes to SDNN (Shaffer and Ginsberg, 2017). Our results further support that persons with stroke have reduced parasympathetic activity, although the impairment of sympathetic activity remains unclear (Oka, 2017). However, this finding is inconsistent with a previous meta-analysis showing that persons with mild TBI had significantly lower SDNN (Galea et al., 2018). Differences between persons with stroke and TBI may result from lateralization and localization of brain injury (Yperzeele et al., 2015) because persons with stroke have relatively focal lesions compared to persons with TBI (Takahashi et al., 2015). This assertion may be inconclusive due to the small number of studies. Additional studies may resolve these conflicting findings.

The results of meta-regression suggested that sex can be factors moderating correlations between parasympathetic activities (RMSSD) and neurological functions. Previous studies have suggested that sex influence HRV because a decrease in HRV affect women more than men, even though the underlying mechanism is yet to be clarified (Fatisson et al., 2016). Interpretation of our meta-regression results should be cautious due to the small number of included studies.

4.1. Clinical and Research Implications

HRV has been considered a transdiagnostic biomarker, which can be commonly used in populations such as autism spectrum disorders (Cheng et al., 2020), dementia (da Silva et al., 2018), psychiatric illnesses (Beauchaine and Thayer, 2015; Jung et al., 2019), posttraumatic stress disorder (Beauchaine and Thayer, 2015; Jung et al., 2019; Robe et al., 2019), and spinal cord injury (Karri et al., 2017). To the best of our knowledge, this is the first study providing quantitative evidence to support using HRV in persons with ABI. Moreover, while studies of HRV have focused on psychological functions including cognitive and emotional functions, this study has extended opportunities to apply HRV to neurological and behavioral functions. Researchers and clinicians can objectively and precisely track ABI survivors’ functioning by using HRV biomarkers (Bayes-Genis et al., 2017; Bede and Pradat, 2019; Lees et al., 2018), which will be useful for identifying precision targets and developing individualized interventions (Bayes-Genis et al., 2018).

4.2. Limitations

Although our findings are valuable, we note several limitations. Most limitations result from the small number of included studies. Also, we were unable to conduct meta-analyses of cognitive functions because research on persons with ABI has disproportionately focused on motor and behavioral functions, while cognitive functions have received less attention (Shipley et al., 2020). Considering persons with ABI experience cognitive deficits and HRV is a critical biomarker, we suggest further studies to evaluate the relationships between cognitive functions and HRV in persons with ABI. Additionally, we could not include potential important moderators, such as lateralization and localization of brain injury (Yperzeele et al., 2015) and medications (Ellenbroek et al., 2019) in meta-regression analyses because the included studies did not measure and address these factors. Similarly, different resting recording conditions (e.g., position, stimulus, etc.) may be a source of heterogeneity. However, we could not address the resting recording conditions as some included studies did not report the conditions. Therefore, future research may evaluate these potential moderators of associations between HRV and functional outcomes after ABI.

4.3. Future Directions

We identified additional research gaps in this review. HRV is a primary biomarker for self-regulation (Holzman and Bridgett, 2017) and social engagement (Kemp and Quintana, 2013). Self-regulation refers to one’s ability to control their thoughts, emotions, and actions (Heatherton and Tice, 1994). Executive functions, closely linked to the prefrontal cortical area, are key components of self-regulation (Holzman and Bridgett, 2017). Because the prefrontal structures implicated in HRV influence self-regulation (Thayer and Lane, 2000), HRV can reflect self-regulation (Holzman and Bridgett, 2017; Thayer and Lane, 2000). Moreover, when people perceive threats, they have reduced parasympathetic activity, indicated by reduced HRV, triggering the fight-or-flight responses through sympathetic activity (Kemp and Quintana, 2013). This mechanism explains that HRV can be a biomarker of social withdrawal and social engagement (Kemp and Quintana, 2013). Although self-regulation and social engagement are instrumental to participation in daily activities and return to community living after ABI, no studies have examined the relationships between HRV and these functions in persons with ABI. In addition, despite that researchers measure nonlinear HRV parameters to overcome shortcomings of linear parameters (Constantinescu et al., 2018), only one study used nonlinear parameters. Thus, future studies may address the applicability and reflections of nonlinear parameters in persons with ABI. Last, HRV is increasingly being measured in real-world contexts via wearables and smartphone devices among the general population (Georgiou et al., 2018). However, the results of this study indicate that most studies of HRV and ABI are using ECG and devices in well-controlled settings. Future studies may investigate the feasibility and acceptability of obtaining real-world HRV data from wearables or other smart devices in persons with ABI. These digital data may become useful biomarkers of real-time, in vivo functioning after ABI and may facilitate an understanding of causal relationships between HRV and functional outcomes in real-world contexts.

5. Conclusions

We performed this study to quantitatively synthesize the effect sizes of correlation coefficients between functional outcomes and HRV after ABI. Our results suggest that, of the included HRV indices, LF, HF, SDNN, and RMSSD are candidate biomarkers for functional outcomes (i.e. neurological, emotional, and behavioral outcomes) after ABI. To move the field forward, we recommend research on nonlinear HRV indices, real-time HRV data using advanced technologies, and relationships between HRV and cognitive and social functions including self-regulation and social engagement after ABI.

Highlights.

• Electrocardiograms and wearable devices were used to measure heart rate variability

• Persons with acquired brain injury indicated decreased heart rate variability

• Heart rate variability indices were correlated with functional outcomes

• These indices can be used as biomarkers of functions after acquired brain injury

• These indices enable precise monitoring of functions after acquired brain injury

Appendix A. Search terms

Ovid Medline – 1,932 results on 01/31/20

Heart rate variability

exp Heart Rate/ph OR heart rate variability.mp. OR (HR adj3 variability).mp. OR (“Heart rate” adj3 (variability OR variance OR variation* OR fluctuat* OR oscillat*)).mp. OR hrv.mp. OR exp autonomic nervous system/ OR autonomic nervous system.mp. OR vagal.mp. oR vagus.mp. OR autonomic function.mp. OR autonomic dysfunction.mp. OR dysautonomia.mp. OR respiratory sinus arrhythmia.mp.

Acquired Brain injury - stroke, traumatic brain injury, brain tumor, encephalitis

Exp stroke/ OR stroke.mp. OR (cerebrovascular adj1 (accident OR injury OR insult OR insufficiency OR failure OR arrest OR lesion)).mp. OR cerebrum vascular accident.mp. OR brain vascular accident.mp. OR apoplexy.mp. OR apoplexia.mp. OR ischemic cerebral attack.mp. OR brain infarction.mp OR brain stem infarction.mp.

OR

Exp brain injuries/ OR exp traumatic brain injuries/ OR traumatic brain injury.mp. OR traumatic brain injuries.mp. OR ((brain or cerebral OR encephalopathy) adj2 (lesion OR concussion OR contusion OR trauma*)).mp. OR (traumatic adj1 encephalopath*).mp.

OR

Exp basal ganglia cerebrovascular disease/ OR Exp brain ischemia/ OR Exp carotid artery disease/ OR Exp cerebral small vessel diseases/ OR Exp cerebrovascular trauma/ OR Exp intracranial arterial diseases/ OR Exp intracranial arteriovenous malformations/ OR exp “Intracranial embolism and thrombosis”/ OR Exp intracranial hemorrhages/ OR Exp periventricular leukomalacia/ OR Exp sneddon syndrome/ OR exp susac syndrome/ OR Exp vascular headaches/ OR Exp central nervous system vasculitis/ OR Exp intracranial vasospasm/ OR exp chronic brain injury/ OR exp persistent vegetative state/

OR

Exp brain neoplasms/ OR ((brain or intracranial) adj1 (metastasis OR tumor or tumour or cancer* or neoplasm*)).mp.

OR

Exp encephalitis/ OR encephalitis.mp. OR meningoencephalitis.mp. OR cerebromenengitis.mp.

functional outcomes

Exp motor activity/ OR Exp motor skills/ OR Exp exercise tolerance/ OR Exp depressive disorder/ OR Exp psychiatric status rating scales/ OR Exp psychomotor performance/ OR (motor adj1 (function OR assessment OR impairment or performance)).mp. OR ((outcome* OR recovery OR rehabilit* OR performance or activity) adj1 (predict* OR cognitive OR emotional OR function* OR neurological OR physical OR social)).mp. OR ((emotional or social) adj4 function*).mp. OR gait.mp. OR balance.mp. OR fitness.mp. OR physical performance.mp. OR range of motion.mp. OR ((executive OR extremity) adj1 function*).mp. or dexterity.mp. OR (motor adj1 skill*).mp. OR walking.mp. OR ((walking Or language) adj1 test).mp. OR attention.mp. OR processing speed.mp. OR ((spatial OR visual) adj1 ability).mp. OR memory.mp. OR depression.mp. OR anxiety.mp. OR disability.mp. OR independence.mp. OR participation.mp. OR (emotional adj1 disorder*).mp. OR National Institutes of Health Stroke Scale.mp. OR level of consciousness.mp. OR responsiveness.mp. OR commands.mp. OR horizontal eye moevement.mp. OR visual field test.mp. OR facial palsy.mp. OR motor arm.mp. OR motor leg.mp. OR limb ataxia.mp. OR sensory.mp. OR language.mp. or speech.mp. or extinction.mp. or inattention.mp. OR speech disturbance.mp. OR weakness.mp. OR modified rankin scale.mp. OR barthel scale.mp. OR glasgow outcome scale.mp. OR ((glasgow OR stroke OR balance OR lawton OR rankin) adj2 scale).mp. OR neurological scale.mp. or hess scale.mp. OR subarachnoid hemorrhage scale.mp. OR berg balance scale.mp. OR quality of life.mp. or ssqol.mp. OR barthel index.mp. OR functional independence.mp. OR sf36.mp. or sf12.mp. OR sf 36.mp. or sf 12.mp. or community integration questionnaire.mp. or arm test.mp. or blessed dementia.mp. OR dementia scale.mp. or orpington.mp. or concentration test.mp. OR exp Brain Injuries/rh OR exp Brain Neoplasms/rh or exp Encephalitis/rh OR hrqol.mp.

NOT ((Exp Animals/ NOT (Exp Animals/ AND Exp Humans/)) or rabbit.ti. or rabbits.ti. or rat.ti. or rats.ti. or cattle.ti. or bovine.ti. or mice.ti. or mouse.ti. or ovine.ti. or sheep.ti. or goat.ti. or dog.ti.)

NOT (Exp Child/ NOT (Exp Child/ AND Exp adult/))

Embase – 1842 results on 01/31/20

Heart rate variability

‘heart rate variability’/exp OR (HR near/3 variability):ti,ab,kw,de OR (‘Heart rate’ near/3 (variability OR variance OR variation* OR fluctuat* OR oscillat*)):ti,ab,kw,de OR hrv:ti,ab,kw,de OR ‘autonomic nervous system’:ti,ab OR vagal:ti,ab,kw oR vagus:ti,ab,kw OR ‘autonomic function’:ti,ab,kw,de OR ‘autonomic dysfunction’:ti,ab,kw,de OR dysautonomia:ti,ab,kw,de OR ‘respiratory sinus arrhythmia’:ti,ab,kw OR ‘aortic nerve’/exp

Acquired Brain injury - stroke, traumatic brain injury, brain tumor, encephalitis

‘cerebrovascular accident’/exp OR stroke:ti,ab,kw,de OR (cerebrovascular near/1 (accident OR injury OR insult OR insufficiency OR failure OR arrest OR lesion)):ti,ab,kw,de OR ‘cerebrum vascular accident’:ti,ab,kw,de OR ‘brain vascular accident’:ti,ab,kw,de OR apoplexy:ti,ab,kw,de OR apoplexia:ti,ab,kw,de OR ‘ischemic cerebral attack’:ti,ab,kw,de OR ‘brain infarction’:ti,ab,kw,de OR ‘brain stem infarction’:ti,ab,kw,de

OR

‘brain injury’/exp OR ‘traumatic brain injury’/exp OR ‘acquired brain injury’:ti,ab,kw,de OR ‘traumatic brain injury’:ti,ab,kw,de OR ‘traumatic brain injuries’:ti,ab,kw,de OR ((brain or cerebral OR encephalopathy) near/2 (lesion OR concussion OR contusion OR trauma*)):ti,ab,kw,de OR (traumatic near/1 encephalopath*):ti,ab,kw,de

OR

‘cerebrovascular disease’/exp OR

‘brain tumor’/exp OR ((brain or intracranial) near/1 (metastasis OR tumor or tumour or cancer* or

neoplasm*)):ti,ab,kw,de

OR

‘encephalitis’/exp OR encephalitis:ti,ab,kw,de OR meningoencephalitis:ti,ab,kw,de OR

cerebromenengitis:ti,ab,kw,de

functional outcomes

‘mini mental state examination’/exp OR ‘motor performance’/exp OR ‘psychological rating scale’/exp OR ‘beck depression inventory’/exp OR ‘beck anxiety inventory’/exp OR ‘hospital anxiety and depression scale’/exp OR ‘visual analog scale’/exp OR ‘national institutes of health stroke scale’/exp OR ‘activity of daily living assessment’/exp OR ‘exercise tolerance’/exp OR ‘physical activity’/exp OR ‘depression’/exp OR ‘injury scale’/exp OR ‘motor activity’/exp OR ‘quality of life assessment’/exp OR ‘activity of daily living assessment’/exp OR ‘functional status assessment’/exp OR ‘social support assessment’/exp OR ((psychomotor OR task OR motor OR cognitive) near/1 (function OR assessment OR impairment or performance)):ti,ab,kw,de OR ((outcome* OR recovery OR rehabilit* OR performance or activity) near/1 (predict* OR cognitive OR emotional OR function* OR neurological OR physical OR social)):ti,ab,kw,de OR (predict* near/1 value):ti,ab,kw,de OR ((emotional or social) near/4 function*):ti,ab,kw,de OR gait:ti,ab,kw,de OR balance:ti,ab,kw,de OR fitness:ti,ab,kw,de OR ‘physical performance’:ti,ab,kw,de OR ‘range of motion’:ti,ab,kw,de OR ((executive OR extremity) near/1 function*):ti,ab,kw,de or dexterity:ti,ab,kw,de OR (motor near/1 skill*):ti,ab,kw,de OR walking:ti,ab,kw,de OR ((walking Or language) near/1 test):ti,ab,kw,de OR ‘processing speed’:ti,ab,kw,de OR ((spatial OR visual) near/1 ability):ti,ab,kw,de OR memory:ti,ab,kw,de OR depression:ti,ab,kw,de OR anxiety:ti,ab,kw,de OR disability:ti,ab,kw,de OR independence:ti,ab,kw,de OR participation:ti,ab,kw,de OR (emotional near/1 disorder*):ti,ab,kw,de OR ‘National Institutes of Health Stroke Scale’:ti,ab,kw,de OR ‘level of consciousness’:ti,ab,kw,de OR responsiveness:ti,ab,kw,de OR commands:ti,ab,kw,de OR ‘horizontal eye movement’:ti,ab,kw,de OR ‘visual field test’:ti,ab,kw,de OR ‘facial palsy’:ti,ab,kw,de OR ‘motor arm’:ti,ab,kw,de OR ‘motor leg’:ti,ab,kw,de OR ‘limb ataxia’:ti,ab,kw,de OR sensory:ti,ab,kw,de OR language:ti,ab,kw,de or speech:ti,ab,kw,de or extinction:ti,ab,kw,de or inattention:ti,ab,kw,de OR ‘speech disturbance’:ti,ab,kw,de OR ‘weakness’:ti,ab,kw,de OR ‘modified rankin scale’:ti,ab,kw,de OR ‘barthel scale’:ti,ab,kw,de OR ‘glasgow outcome scale’:ti,ab,kw,de OR ((glasgow OR stroke OR balance OR lawton OR rankin) near/2 scale):ti,ab,kw,de OR ‘neurological scale’:ti,ab,kw,de or ‘hess scale’:ti,ab,kw,de OR ‘subarachnoid hemorrhage scale’:ti,ab,kw,de OR ‘berg balance scale’:ti,ab,kw,de OR ‘quality of life’:ti,ab,kw,de or ssqol:ti,ab,kw,de OR ‘barthel index’:ti,ab,kw,de OR ‘functional independence’:ti,ab,kw,de OR sf36:ti,ab,kw,de or sf12:ti,ab,kw,de OR ‘sf 36’:ti,ab,kw,de or ‘sf 12’:ti,ab,kw,de or ‘community integration questionnaire’:ti,ab,kw,de or ‘arm test’:ti,ab,kw,de or ‘blessed dementia’:ti,ab,kw,de OR ‘dementia scale’:ti,ab,kw,de or orpington:ti,ab,kw,de or ‘concentration test’:ti,ab,kw,de OR hrqol:ti,ab,kw,de

NOT (‘Animal’/exp NOT (‘Animal’/exp AND ‘Human’/exp)) or rabbit:ti or rabbits:ti or rat:ti or rats:ti or cattle:ti or bovine:ti or mice:ti or mouse:ti or ovine:ti or sheep:ti or goat:ti or dog:ti

NOT (‘Child’/exp NOT (‘Child’/exp AND ‘adult’/exp))

TRUE SEARCH:

(‘heart rate variability’/exp OR ((hr NEAR/3 variability):ti,ab,kw,de) OR ((‘heart rate’ NEAR/3 (variability OR variance OR variation* OR fluctuat* OR oscillat*)):ti,ab,kw,de)) AND (‘cerebrovascular accident’/exp OR stroke:ti,ab,kw,de OR ((cerebrovascular NEAR/1 (accident OR injury OR insult OR insufficiency OR failure OR arrest OR lesion)):ti,ab,kw,de) OR ‘cerebrum vascular accident’:ti,ab,kw,de OR ‘brain vascular accident’:ti,ab,kw,de OR apoplexy:ti,ab,kw,de OR apoplexia:ti,ab,kw,de OR ‘ischemic cerebral attack’:ti,ab,kw,de OR ‘brain infarction’:ti,ab,kw,de OR ‘brain stem infarction’:ti,ab,kw,de OR ‘brain injury’/exp OR ‘traumatic brain injury’/exp OR ‘acquired brain injury’:ti,ab,kw,de OR ‘traumatic brain injury’:ti,ab,kw,de OR ‘traumatic brain injuries’:ti,ab,kw,de OR (((brain OR cerebral OR encephalopathy) NEAR/2 (lesion OR concussion OR contusion OR trauma*)):ti,ab,kw,de) OR ((traumatic NEAR/1 encephalopath*):ti,ab,kw,de) OR ‘cerebrovascular disease’/exp OR ‘brain tumor’/exp OR (((brain OR intracranial) NEAR/1 (metastasis OR tumor OR tumour OR cancer* OR neoplasm*)):ti,ab,kw,de) OR ‘encephalitis’/exp OR encephalitis:ti,ab,kw,de OR meningoencephalitis:ti,ab,kw,de OR cerebromenengitis:ti,ab,kw,de) AND (‘heart rate variability’/exp OR ((hr NEAR/3 variability):ti,ab,kw,de) OR ((‘heart rate’ NEAR/3 (variability OR variance OR variation* OR fluctuat* OR oscillat*)):ti,ab,kw,de) OR hrv:ti,ab,kw,de OR ‘autonomic nervous system’:ti,ab OR vagal:ti,ab,kw OR vagus:ti,ab,kw OR ‘autonomic function’:ti,ab,kw,de OR ‘autonomic dysfunction’:ti,ab,kw,de OR dysautonomia:ti,ab,kw,de OR ‘respiratory sinus arrhythmia’:ti,ab,kw OR ‘aortic nerve’/exp) NOT (‘animal’/exp NOT (‘animal’/exp AND ‘human’/exp) OR rabbit:ti OR rabbits:ti OR rat:ti OR rats:ti OR cattle:ti OR bovine:ti OR mice:ti OR mouse:ti OR ovine:ti OR sheep:ti OR goat:ti OR dog:ti) NOT (‘child’/exp NOT (‘child’/exp AND ‘adult’/exp))

Scopus – 975 results on 01/31/20

Heart rate variability

(TITLE-ABS-KEY (HR w/3 variability)) OR (TITLE-ABS ( “autonomic nervous system”)) OR (TITLE-ABS (vagal)),kw oR (TITLE-ABS (vagus)),kw OR (TITLE-ABS ( “autonomic function”)) OR (TITLE-ABS ( “autonomic dysfunction”)) OR (TITLE-ABS (dysautonomia)) OR (TITLE-ABS ( “respiratory sinus arrhythmia”)) OR (TITLE-ABS-KEY (“Heart rate” w/3 (variability OR variance OR variation* OR fluctuat* OR oscillat*)))

Acquired Brain injury - stroke, traumatic brain injury, brain tumor, encephalitis

( TITLE-ABS-KEY ( stroke ) ) OR ( TITLE-ABS-KEY ( cerebrovascular W/1 ( accident OR injury OR insult OR insufficiency OR failure OR arrest OR lesion ) ) ) OR ( TITLE-ABS-KEY ( “cerebrum vascular accident” ) ) OR ( TITLE-ABS-KEY ( “brain vascular accident” ) ) OR ( TITLE-ABS-KEY ( apoplexy ) ) OR ( TITLE-ABS-KEY ( apoplexia ) ) OR ( TITLE-ABS-KEY ( “ischemic cerebral attack” ) ) OR ( TITLE-ABS-KEY ( “brain infarction” ) ) OR ( TITLE-ABS-KEY ( “brain stem infarction” ) ) OR ( TITLE-ABS-KEY ( “acquired brain injury” ) ) OR ( TITLE-ABS-KEY ( “traumatic brain injury” ) ) OR ( TITLE-ABS-KEY ( “traumatic brain injuries” ) ) OR ( TITLE-ABS-KEY ( ( brain OR cerebral OR encephalopathy ) W/2 ( lesion OR concussion OR contusion OR trauma* ) ) ) OR ( TITLE-ABS-KEY ( traumatic W/1 encephalopath* ) ) OR ( TITLE-ABS-KEY ( ( brain OR intracranial ) W/1 ( metastasis OR tumor OR tumour OR cancer* OR neoplasm* ) ) ) OR ( TITLE-ABS-KEY ( encephalitis ) ) OR ( TITLE-ABS-KEY ( meningoencephalitis ) ) OR ( TITLE-ABS-KEY ( cerebromenengitis ) )

functional outcomes

(TITLE-ABS-KEY ((psychomotor OR task OR motor OR cognitive) w/1 (function OR assessment OR impairment or performance))) OR (TITLE-ABS-KEY ( (outcome* OR recovery OR rehabilit* OR performance or activity) w/1 (predict* OR cognitive OR emotional OR function* OR neurological OR physical OR social))) OR (TITLE-ABS-KEY ( predict* w/1 value)) OR (TITLE-ABS-KEY ( gait )) OR (TITLE-ABS-KEY ( balance )) OR (TITLE-ABS-KEY ( fitness )) OR (TITLE-ABS-KEY ( “physical performance” )) OR (TITLE-ABS-KEY ( “range of motion” )) OR (TITLE-ABS-KEY (function* w/1 (social OR emotional OR executive OR extremity))) or (TITLE-ABS-KEY ( dexterity )) OR (TITLE-ABS-KEY (motor w/1 skill*)) OR (TITLE-ABS-KEY ( walking )) OR (TITLE-ABS-KEY (test w/1 (walking Or language)) OR (TITLE-ABS-KEY ( “processing speed” )) OR (TITLE-ABS-KEY ( ability w/1 (spatial OR visual)) OR (TITLE-ABS-KEY ( memory )) OR (TITLE-ABS-KEY ( depression )) OR (TITLE-ABS-KEY ( anxiety )) OR (TITLE-ABS-KEY ( disability )) OR (TITLE-ABS-KEY ( independence )) OR (TITLE-ABS-KEY ( participation )) OR (TITLE-ABS-KEY ( emotional w/1 disorder* )) OR (TITLE-ABS-KEY ( “National Institutes of Health Stroke Scale” )) OR (TITLE-ABS-KEY ( “level of consciousness” )) OR (TITLE-ABS-KEY ( responsiveness )) OR (TITLE-ABS-KEY ( commands )) OR (TITLE-ABS-KEY ( “horizontal eye movement” )) OR (TITLE-ABS-KEY ( “visual field test” )) OR (TITLE-ABS-KEY ( “facial palsy” )) OR (TITLE-ABS-KEY ( “motor arm” )) OR (TITLE-ABS-KEY ( “motor leg” )) OR (TITLE-ABS-KEY ( “limb ataxia” )) or (TITLE-ABS-KEY ( inattention )) OR (TITLE-ABS-KEY ( “speech disturbance” )) OR (TITLE-ABS-KEY ( “weakness” )) OR (TITLE-ABS-KEY ( “modified rankin scale” )) OR (TITLE-ABS-KEY ( “barthel scale” )) OR (TITLE-ABS-KEY ( “glasgow outcome scale” )) OR (TITLE-ABS-KEY (scale w/2 (glasgow OR stroke OR balance OR lawton OR rankin))) OR (TITLE-ABS-KEY ( “neurological scale” )) or (TITLE-ABS-KEY ( “hess scale” )) OR (TITLE-ABS-KEY ( “subarachnoid hemorrhage scale” )) OR (TITLE-ABS-KEY ( “berg balance scale” )) OR (TITLE-ABS-KEY ( “quality of life” )) or (TITLE-ABS-KEY ( ssqol )) OR (TITLE-ABS-KEY ( “barthel index” )) OR (TITLE-ABS-KEY ( “functional independence” )) OR (TITLE-ABS-KEY ( sf36 )) or (TITLE-ABS-KEY ( sf12 )) OR (TITLE-ABS-KEY ( “sf 36” )) or (TITLE-ABS-KEY ( “sf 12” )) or (TITLE-ABS-KEY ( “community integration questionnaire” )) or (TITLE-ABS-KEY ( “arm test” )) or (TITLE-ABS-KEY ( “blessed dementia” )) OR (TITLE-ABS-KEY ( “dementia scale” )) or (TITLE-ABS-KEY ( orpington )) or (TITLE-ABS-KEY ( “concentration test” )) OR (TITLE-ABS-KEY ( hrqol ))

NOT ((Exp Animals/ NOT (Exp Animals/ AND Exp Humans/)) or rabbit.ti. or rabbits.ti. or rat.ti. or rats.ti. or cattle.ti. or bovine.ti. or mice.ti. or mouse.ti. or ovine.ti. or sheep.ti. or goat.ti. or dog.ti.)

NOT (Exp Child/ NOT (Exp Child/ AND Exp adult/))

TRUE:

( ( TITLE-ABS-KEY ( hr W/3 variability ) ) OR (TITLE-ABS ( “autonomic nervous system”)) OR (TITLE-ABS (vagal)) oR (TITLE-ABS (vagus)) OR (TITLE-ABS ( “autonomic function”)) OR (TITLE-ABS ( “autonomic dysfunction”)) OR (TITLE-ABS (dysautonomia)) OR (TITLE-ABS ( “respiratory sinus arrhythmia”)) OR ( TITLE-ABS-KEY ( “Heart rate” W/3 ( variability OR variance OR variation* OR fluctuat* OR oscillat* ) ) ) ) AND ( ( TITLE-ABS-KEY ( stroke ) ) OR ( TITLE-ABS-KEY ( cerebrovascular W/1 ( accident OR injury OR insult OR insufficiency OR failure OR arrest OR lesion ) ) ) OR ( TITLE-ABS-KEY ( “cerebrum vascular accident” ) ) OR ( TITLE-ABS-KEY ( “brain vascular accident” ) ) OR ( TITLE-ABS-KEY ( apoplexy ) ) OR ( TITLE-ABS-KEY ( apoplexia ) ) OR ( TITLE-ABS-KEY ( “ischemic cerebral attack” ) ) OR ( TITLE-ABS-KEY ( “brain infarction” ) ) OR ( TITLE-ABS-KEY ( “brain stem infarction” ) ) OR ( TITLE-ABS-KEY ( “acquired brain injury” ) ) OR ( TITLE-ABS-KEY ( “traumatic brain injury” ) ) OR ( TITLE-ABS-KEY ( “traumatic brain injuries” ) ) OR ( TITLE-ABS-KEY ( ( brain OR cerebral OR encephalopathy ) W/2 ( lesion OR concussion OR contusion OR trauma* ) ) ) OR ( TITLE-ABS-KEY ( traumatic W/1 encephalopath* ) ) OR ( TITLE-ABS-KEY ( ( brain OR intracranial ) W/1 ( metastasis OR tumor OR tumour OR cancer* OR neoplasm* ) ) ) OR ( TITLE-ABS-KEY ( encephalitis ) ) OR ( TITLE-ABS-KEY ( meningoencephalitis ) ) OR ( TITLE-ABS-KEY ( cerebromenengitis ) ) ) AND ( ( TITLE-ABS-KEY ( ( psychomotor OR task OR motor OR cognitive ) W/1 ( function OR assessment OR impairment OR performance ) ) ) OR ( TITLE-ABS-KEY ( ( outcome* OR recovery OR rehabilit* OR performance OR activity ) W/1 ( predict* OR cognitive OR emotional OR function* OR neurological OR physical OR social ) ) ) OR ( TITLE-ABS-KEY ( predict* W/1 value ) ) OR ( TITLE-ABS-KEY ( gait ) ) OR ( TITLE-ABS-KEY ( balance ) ) OR ( TITLE-ABS-KEY ( fitness ) ) OR ( TITLE-ABS-KEY ( “physical performance” ) ) OR ( TITLE-ABS-KEY ( “range of motion” ) ) OR ( TITLE-ABS-KEY ( function* W/1 ( social OR emotional OR executive OR extremity ) ) ) OR ( TITLE-ABS-KEY ( dexterity ) ) OR ( TITLE-ABS-KEY ( motor W/1 skill* ) ) OR ( TITLE-ABS-KEY ( walking ) ) OR ( TITLE-ABS-KEY ( test W/1 ( walking OR language ) ) ) OR ( TITLE-ABS-KEY ( “processing speed” ) ) OR ( TITLE-ABS-KEY ( ability W/1 ( spatial OR visual ) ) ) OR ( TITLE-ABS-KEY ( memory ) ) OR ( TITLE-ABS-KEY ( depression ) ) OR ( TITLE-ABS-KEY ( anxiety ) ) OR ( TITLE-ABS-KEY ( disability ) ) OR ( TITLE-ABS-KEY ( independence ) ) OR ( TITLE-ABS-KEY ( participation ) ) OR ( TITLE-ABS-KEY ( emotional W/1 disorder* ) ) OR ( TITLE-ABS-KEY ( “National Institutes of Health Stroke Scale” ) ) OR ( TITLE-ABS-KEY ( “level of consciousness” ) ) OR ( TITLE-ABS-KEY ( responsiveness ) ) OR ( TITLE-ABS-KEY ( commands ) ) OR ( TITLE-ABS-KEY ( “horizontal eye movement” ) ) OR ( TITLE-ABS-KEY ( “visual field test” ) ) OR ( TITLE-ABS-KEY ( “facial palsy” ) ) OR ( TITLE-ABS-KEY ( “motor arm” ) ) OR ( TITLE-ABS-KEY ( “motor leg” ) ) OR ( TITLE-ABS-KEY ( “limb ataxia” ) ) OR ( TITLE-ABS-KEY ( inattention ) ) OR ( TITLE-ABS-KEY ( “speech disturbance” ) ) OR ( TITLE-ABS-KEY ( “weakness” ) ) OR ( TITLE-ABS-KEY ( “modified rankin scale” ) ) OR ( TITLE-ABS-KEY ( “barthel scale” ) ) OR ( TITLE-ABS-KEY ( “glasgow outcome scale” ) ) OR ( TITLE-ABS-KEY ( scale W/2 ( glasgow OR stroke OR balance OR lawton OR rankin ) ) ) OR ( TITLE-ABS-KEY ( “neurological scale” ) ) OR ( TITLE-ABS-KEY ( “hess scale” ) ) OR ( TITLE-ABS-KEY ( “subarachnoid hemorrhage scale” ) ) OR ( TITLE-ABS-KEY ( “berg balance scale” ) ) OR ( TITLE-ABS-KEY ( “quality of life” ) ) OR ( TITLE-ABS-KEY ( ssqol ) ) OR ( TITLE-ABS-KEY ( “barthel index” ) ) OR ( TITLE-ABS-KEY ( “functional independence” ) ) OR ( TITLE-ABS-KEY ( sf36 ) ) OR ( TITLE-ABS-KEY ( sf12 ) ) OR ( TITLE-ABS-KEY ( “sf 36” ) ) OR ( TITLE-ABS-KEY ( “sf 12” ) ) OR ( TITLE-ABS-KEY ( “community integration questionnaire” ) ) OR ( TITLE-ABS-KEY ( “arm test” ) ) OR ( TITLE-ABS-KEY ( “blessed dementia” ) ) OR ( TITLE-ABS-KEY ( “dementia scale” ) ) OR ( TITLE-ABS-KEY ( orpington ) ) OR ( TITLE-ABS-KEY ( “concentration test” ) ) OR ( TITLE-ABS-KEY ( hrqol ) ) ) AND ( LIMIT-TO ( DOCTYPE, “ar” ) OR LIMIT-TO ( DOCTYPE, “re” ) ) AND ( LIMIT-TO ( EXACTKEYWORD, “Human” ) OR LIMIT-TO ( EXACTKEYWORD, “Humans” ) OR LIMIT-TO ( EXACTKEYWORD, “Adult” ) )

Cochrane Central - 180 results on 01/31/20

Heart rate variability

(HR near/3 variability):ti,ab,kw OR (“Heart rate” near/3 (variability OR variance OR variation* OR fluctuat* OR oscillat*)):ti,ab,kw OR “heart rate variability”:ti,ab,kw OR “autonomic nervous system”:ti,ab OR vagal:ti,ab oR vagus:ti,ab OR “autonomic function”:ti,ab OR “autonomic dysfunction”:ti,ab OR dysautonomia:ti,ab OR “respiratory sinus arrhythmia”:ti,ab

Acquired Brain injury - stroke, traumatic brain injury, brain tumor, encephalitis

stroke:ti,ab,kw OR (cerebrovascular near/1 (accident OR injury OR insult OR insufficiency OR failure OR arrest OR lesion)):ti,ab,kw OR “cerebrum vascular accident”:ti,ab,kw OR “brain vascular accident”:ti,ab,kw OR apoplexy:ti,ab,kw OR apoplexia:ti,ab,kw OR “ischemic cerebral attack”:ti,ab,kw OR “brain infarction”:ti,ab,kw OR “brain stem infarction”:ti,ab,kw OR “acquired brain injury”:ti,ab,kw OR “traumatic brain injury”:ti,ab,kw OR “traumatic brain injuries”:ti,ab,kw OR ((brain or cerebral OR encephalopathy) near/2 (lesion OR concussion OR contusion OR trauma*)):ti,ab,kw OR (traumatic near/1 encephalopath*):ti,ab,kw

OR ((brain or intracranial) near/1 (metastasis OR tumor or tumour or cancer* or neoplasm*)):ti,ab,kw

OR encephalitis:ti,ab,kw OR meningoencephalitis:ti,ab,kw OR cerebromenengitis:ti,ab,kw

functional outcomes

((psychomotor OR task OR motor OR cognitive) near/1 (function OR assessment OR impairment or performance)):ti,ab,kw OR ((outcome* OR recovery OR rehabilit* OR performance or activity) near/1 (predict* OR cognitive OR emotional OR function* OR neurological OR physical OR social)):ti,ab,kw OR (predict* near/1 value):ti,ab,kw OR ((emotional or social) near/4 function*):ti,ab,kw OR gait:ti,ab,kw OR balance:ti,ab,kw OR fitness:ti,ab,kw OR “physical performance”:ti,ab,kw OR “range of motion”:ti,ab,kw OR ((executive OR extremity) near/1 function*):ti,ab,kw or dexterity:ti,ab,kw OR (motor near/1 skill*):ti,ab,kw OR walking:ti,ab,kw OR ((walking Or language) near/1 test):ti,ab,kw OR “processing speed”:ti,ab,kw OR ((spatial OR visual) near/1 ability):ti,ab,kw OR memory:ti,ab,kw OR depression:ti,ab,kw OR anxiety:ti,ab,kw OR disability:ti,ab,kw OR independence:ti,ab,kw OR participation:ti,ab,kw OR (emotional near/1 disorder*):ti,ab,kw OR “National Institutes of Health Stroke Scale”:ti,ab,kw OR “level of consciousness”:ti,ab,kw OR responsiveness:ti,ab,kw OR commands:ti,ab,kw OR “horizontal eye movement”:ti,ab,kw OR “visual field test”:ti,ab,kw OR “facial palsy”:ti,ab,kw OR “motor arm”:ti,ab,kw OR “motor leg”:ti,ab,kw OR “limb ataxia”:ti,ab,kw OR sensory:ti,ab,kw OR language:ti,ab,kw or speech:ti,ab,kw or extinction:ti,ab,kw or inattention:ti,ab,kw OR “speech disturbance”:ti,ab,kw OR “weakness”:ti,ab,kw OR “modified rankin scale”:ti,ab,kw OR “barthel scale”:ti,ab,kw OR “glasgow outcome scale”:ti,ab,kw OR ((glasgow OR stroke OR balance OR lawton OR rankin) near/2 scale):ti,ab,kw OR “neurological scale”:ti,ab,kw or “hess scale”:ti,ab,kw OR “subarachnoid hemorrhage scale”:ti,ab,kw OR “berg balance scale”:ti,ab,kw OR “quality of life”:ti,ab,kw or ssqol:ti,ab,kw OR “barthel index”:ti,ab,kw OR “functional independence”:ti,ab,kw OR sf36:ti,ab,kw or sf12:ti,ab,kw OR “sf 36”:ti,ab,kw or “sf 12”:ti,ab,kw or “community integration questionnaire”:ti,ab,kw or “arm test”:ti,ab,kw or “blessed dementia”:ti,ab,kw OR “dementia scale”:ti,ab,kw or orpington:ti,ab,kw or “concentration test”:ti,ab,kw OR hrqol:ti,ab,kw

Clinicaltrials.gov

01/31/20 3 results

( brain injury OR stroke OR traumatic brain injury OR brain neoplasms OR encephalitis ) AND and heart rate variability and AND EXPAND[Concept] ( “functional outcome” OR “functional outcomes” ) AND AREA[OverallStatus] EXPAND[Term] COVER[FullMatch] “Completed”

Appendix B. Results of meta-regression analyses

Model	Moderator	Estimate	p value	95% CI		Q	R2 (%)
				Lower	Upper
Correlations between HRV indices and neurological functions at acute stages
LF	Age	−0.031	0.170	−0.096	0.033	2.177	100.00
	Male (%)	0.022	0.089	−0.008	0.052	1.188	100.00
	Stroke type	−0.234	0.436	−1.276	0.808	4.701	69.85
	Unit (ln)	0.234	0.436	−0.808	1.276	4.701	69.85
	Measurement	0.036	0.932	−1.580	1.652	6.998	0.28
	Recording	0.036	0.932	−1.580	1.652	6.998	0.28
HF	Age	−0.053	0.057	−0.114	0.007	0.093	100.00
	Male (%)	0.033	0.196	−0.101	0.168	1.076	100.00
	Hemorrhage	−0.209	0.775	−7.420	7.001	10.282	16.04
	Unit (ln)	0.209	0.775	−7.002	7.420	10.282	16.04
LF/HF ratio	Age	0.016	0.743	−0.448	0.479	5.275	31.58
	Male (%)	0.001	0.998	−0.329	0.329	6.266	0.00
	Stroke type	−0.354	0.432	−3.979	3.272	2.471	100.00
	Unit (ln)	0.354	0.432	−3.272	3.979	2.471	100.00
RMSSD	Age	−0.019	0.523	−0.271	0.234	4.066	78.35
	Male (%)*	0.028	0.001	0.028	0.029	0.000	100.00
	Unit (ln)	0.391	0.368	−2.846	3.628	2.241	100.00
	Recording	0.010	0.987	−5.933	5.953	7.515	0.01
Correlations between HRV indices and emotional functions at chronic stages
LF	Age	0.026	0.384	−0.201	0.252	1.421	100.00
	Male (%)	0.004	0.701	−0.099	0.108	3.502	99.96
	ABI type	−0.336	0.153	−1.381	0.708	0.248	100.00
	Unit (ln)	0.217	0.666	−4.560	4.994	3.529	15.62
LF/HF ratio	Age	−0.014	0.831	−0.657	0.630	7.174	25.53
	Male (%)	−0.010	0.420	−0.109	0.089	3.214	99.99
	ABI type	−0.108	0.874	−6.960	6.743	8.542	0.00
	Unit (ln)	0.603	0.174	−1.544	2.751	0.623	100.00
	Recording	−0.603	0.174	−2.751	1.544	0.623	100.00
Comparison of HRV measures between ABIs and healthy controls
LF	Age	0.016	0.613	−0.073	0.104	14.944	16.07
	Male (%)	0.014	0.439	−0.037	0.065	12.903	34.26
	Chronic ABI	0.367	0.554	−1.389	2.122	13.804	24.15
	Unit (ln)	0.813	0.252	−1.017	2.643	10.177	61.95
	Recording	−0.930	0.158	−2.511	0.652	10.389	64.67
LF/HF ratio	Age	0.027	0.117	−0.017	0.071	1.219	100.00
	Male (%)	0.010	0.272	−0.019	0.040	2.593	100.00
	Chronic ABI	−0.327	0.403	−1.663	1.010	3.555	100.00
	Unit (ln)	0.701	0.057	−0.049	1.452	0.607	100.00
	Recording	−0.133	0.805	−2.161	1.896	5.438	0.00
SDNN	Age	−0.011	0.571	−0.066	0.044	11.742	17.19
	Male (%)	0.024	0.308	−0.039	0.088	8.554	60.38
	Chronic ABI	0.653	0.148	−0.419	1.725	6.103	81.68
	Unit (ln)	0.701	0.057	−0.049	1.452	0.607	100.00
	Recording	−0.336	0.513	−1.780	1.108	11.922	15.51

^*p < 0.05

LF: Low frequency, HF: high frequency, ABI: acquired brain injury.

Age and male (%) are continuous variables. Stroke type (0 = ischemic stroke, 1 = hemorrhagic stroke), Unit (0 = Raw, 1 = Normalized), Measurement (0 = NIH Stroke Scale, 1 = Scandinavian Stroke Scale), Recording (0 = Short-term recording, 1 = Long-term recording), ABI type (0 = stroke, 1=traumatic brain injury), Chronic ABI (0 = Acute ABI, 1 = Chronic ABI).