Effects of Spinal Cord Injury in Heart Rate Variability After Acute and Chronic Exercise: A Systematic Review

A literature search was performed on August 2016 in MEDLINE, PubMed, Embase, and Web of Science databases by 2 authors independently. MeSH terms used were “heart rate variability” [AND] “spinal cord injury” and “heart rate variability” [AND] “spinal cord injury” [AND] “exercise.” For Spanish publications and gray literature, SciELO and Google Scholar databases were searched for terms “variabilidad del ritmo cardíaco” “lesión medular” and “variabilidad de la frecuencia cardíaca” “lesión medular.” Search restrictions were used for gray literature according to the recommendations of Mahood et al.15 

Observational or experimental articles including linear HRV analysis on SCI individuals and one of the following reports were included: (1) comparison with able-bodied subjects, (2) comparison of active and inactive SCI subjects or inactive SCI individuals initiating an exercise intervention, (3) HRV acute response or recovery after exercise in SCI subjects. Additional inclusion criteria were articles in Spanish or English, published between 1998 and July 2016. A total of 279 articles were identified. After duplicates were removed, the titles and abstracts were screened independently by 2 reviewers (D.B, C.C.). Animal studies and nontraumatic SCI articles were excluded according to PRISMA methods.16 When there was disagreement in selection, a third reviewer (R.S.) resolved it. A total of 29 articles were fully reviewed, and 11 of them were discarded due to inclusion criteria. Eighteen articles were selected for the final review (Figure 1).

No quantitative analysis was performed. Risk of bias was assessed independently by 2 authors using the National Institutes of Health Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies (https://www.nhlbi.nih.gov/health-pro/guidelines/in-develop/cardiovascular-risk-reduction/tools/cohort) or the Quality Assessment Tool for Before-After (Pre-Post) Studies With No Control Group (https://www.nhlbi.nih.gov/health-pro/guidelines/in-develop/cardiovascular-risk-reduction/tools/before-after). Results of the assessment were presented with an adaptation of the Cochrane risk of bias summary figure17 for better understanding.

A total of 8 cross-sectional studies were included. Because the NIH assessment tool gives a lower score to cross-sectional studies compared to cohort studies, all included studies were identified as low quality. However, we decided to adjust the tool, removing items designed only for cohort studies (exposure assessed prior to outcome, sufficient timeframe to see an effect, and follow-up rate) due to the difficulty in performing studies other than cross-sectional for this outcome. Risk of bias classification is summarized in Figure 2.

Most studies found significant differences in rest LF values with a fair risk of bias, but there was a high level of variability within study subjects in regard to neurological level. Two studies used a 24-hour ECG recording,18,19 one study used a 4-hour recording,20 and 5 studies performed a short HRV recording of 15 minutes or less.21–25 Data are summarized in Table 1.

Most studies did not report differences in parasympathetic HRV values compared with able-bodied individuals; however, one study found lower values of HF in tetraplegics,22 whereas Krstacic et al found higher HF and time domain values in thoracic SCI patients.19 In summary, most studies with a fair risk of bias showed lower LF absolute power values with no major differences in HF and time domain HRV.

A total of 4 quasi-experimental and 3 cross-sectional studies were included for qualitative analysis. For analysis of quasi-experimental studies, the NIH assessment tool described above was used, but questions 11 and 12 were not included because of multiple outcome measures and group level interventions. Risk of bias assessment is summarized in Figure 3. Most studies were classified as fair risk of bias, mainly because sample size determination and blinding were not described. Nonetheless, the authors agreed that this did not invalidate the results. Only one cross-sectional study was classified as a fair risk of bias, however all of them were identified as low or very low quality for this outcome due to the intrinsic lower quality of observational studies and the feasibility of performing experimental or quasi-experimental studies for this outcome. Results of quasi-experimental studies are summarized in Table 2. Some studies did not find an association between regular physical activity and HRV adaptations in time18,24,26,27 and frequency domain,18,24,27,28 but other studies have found associations in nonlinear parameters of HRV24,26,27,29 and some frequency domain values.26,29,30 

In quasi-experimental articles, Ditor et al28 did not find changes on HRV in paraplegics after 4 months of training in a body weight–supported treadmill (BWST). The same group described a significant decrease in LF/HF ratio due to lower LF power after 6 months of BWST in tetraplegics following a similar protocol.31 Similarly, Millar et al27 compared HRV after BWST or tilt training protocol in a cross-over design. After 4 weeks of training, only BWST resulted in a significant increase in nonlinear HRV values and a trend to a decrease of LF and HF normalized units power. In contrast, Brizuela Costa et al29 described an increase in time and frequency domains in patients with tetraplegia after 8 weeks of training in arm ergometer at maximal tolerated intensity. However, an unclear definition of the intervention raises the risk of bias of this study.

As previously described, 3 studies evaluated differences in rest HRV between physically active and inactive SCI subjects in a cross-sectional design. Higher parasympathetic-associated HRV values were found in active SCI compared with sedentary SCI and neurological intact subjects,26,30 but the lower intrinsic quality of observational studies does not allow the generalization of this finding.

A total of 7 quasi-experimental articles were included. Five studies evaluated HRV after a single session of aerobic exercise and 2 of them after an isometric exercise. Risk of bias is summarized in Figure 4. No sample size justification and lack of blinding are the main biases of included studies. However, the authors agreed that, although interpreted with caution, studies could be considered as a fair risk of bias for internal validity.

Included articles evaluating HRV acute response to aerobic exercise are summarized in Table 3. Wecht et al32 analyzed HRV recovery in active and inactive individuals with paraplegia with lesions below T6 after a maximal effort test in arm ergometer. Successive measurements during a 90-minute period post workout showed a significant decrease in HF, LF, and LF/HF ratio values immediately after exercise, as previously reported.4,21,25,33 

HRV response to other exercise modalities has also been described. Takahashi et al34 reported a decrease in LF and HF immediately after static arm exercise in patients with tetraplegia, with a recovery to normal values in 90 seconds. In another study,23 a decrease in LF with no alterations in HF was reported 5 minutes after hand grip exercise. However, definitive conclusions cannot be made due to heterogeneity and limited number of studies.

HRV is a noninvasive tool that can be used to evaluate the autonomic tone in SCI. The objective of this review was to analyze the effects of SCI on HRV compared with able-bodied subjects and determine how HRV is modified with acute and chronic exercise in SCI subjects. In the reviewed articles, lower rest LF values were observed in SCI subjects compared with neurological intact controls independently of recording time. Unfortunately, these findings could not be interpreted as an alteration in autonomic balance because of difficulties interpreting LF and inconsistency in other HRV parameters. Despite some authors who suggested that LF could be used as a marker of sympathetic tone, it is now believed to reflect sympathetic and parasympathetic tone in neurologically intact and SCI subjects.14,34 Also, the lack of homogeneity in neurological level and completeness impedes identification of these differences as part of alterations in sympathetic tone, other comorbidities, or as an adaptation to a sedentary behavior. Indeed, the physical inactivity hypothesis could be viable considering that HRV values are similar between active SCI and able-bodied subjects in observational studies.26,30 However, these findings are not validated by higher quality studies. If lower HRV values in individuals with paraplegia below T6 are a consequence of low physical activity, it could be expected that chronic training could improve autonomic tone. Furthermore, if improvement in HRV values occurs in SCI subjects with an interruption of sympathetic pathways, exercise could be an important therapeutic tool independently of previous autonomic injury. Indeed, while no differences in HRV have been reported between injuries above T6 and those below T6,24 alterations in sympathetic innervation in high SCI subjects should be considered for future studies with larger sample size.

HRV adaptations to regular exercise in able-bodied subjects have been well defined,35,36 however these adaptations are less understood in SCI individuals due to few quasi-experimental studies.18,24,26–30 Most articles had a small sample size and lack of standardization in exercise intensity and duration, which affected their external validity. Some studies found a trend to an improvement in HF values post training, but exercise intensity was limited by spasticity and the training protocol did not meet the SCI recommendations for physical activity.37 If these articles are compared,27–29,31 it could be interpreted as indicating that a higher intensity29 or a longer period of training in protocols of low intensity31 is required to obtain positive cardiovascular adaptations in subjects with an altered sympathetic innervation. Furthermore, time and frequency domains could have low sensitivity in early phases, and nonlinear parameters might be more adequate at short term.24,26,27 Positive adaptations were observed even in high SCI, which reinforces the idea of maintained exercise as a feasible option to decrease cardiovascular risk in persons with tetraplegia with sympathetic compromise. Future training protocols with controlled exercise variables and standardized characteristics of SCI subjects are encouraged to obtain better conclusions.

The autonomic nervous system has a key role in cardiac functioning and hemodynamics during acute exercise, helping in the delivery of oxygen and glucose for muscular contraction.38 During exercise, a decrease in parasympathetic tone and an increase in sympathetic response is maintained even during recovery due to mechanisms in which mechano- and baroreceptors are involved.39 In able-bodied subjects, HRV changes immediately after exercise have been described,40–42 however few studies and heterogeneous protocols of exercise in the SCI population make it difficult to extrapolate results to clinical practice. Indeed, only one recent study evaluated HRV response after continuous exercise at 50% and 80% of heart rate reserve,25 which is more related to clinical practice. In the articles reviewed, HRV values decreased immediately after exercise,4,21,25,32,33 but recovery kinetics are unclear. It has been described that a rise in endocrine catecholamines is involved in the decrease of HRV during exercise43 and is proportional to exercise intensity,44 which can explain the sympathetic response to exercise in persons with tetraplegia. Regarding recovery kinetics, Wecht et al32 reported that 90 minutes post exercise, only active SCI subjects return to normal HRV values compared to inactive SCI subjects. The authors attributed these findings to a faster recovery of vagal tone and a better autonomic control post workout with chronic exercise. Preservation of vagal tone and influence of parasympathetic response in exercise is observed in a nonpublished work33 in which able-bodied and active tetraplegics with sympathetic alterations performed an incremental aerobic exercise test. In addition to a decrease in HRV values immediately after exercise, recovery of parasympathetic-related HRV parameters showed similar values in both groups. This finding could be interpreted as an inhibition and recovery of parasympathetic activity independent of sympathetic compromise, as observed by other authors.34 Even though the interest of aerobic exercise response in this population is increasing, it could be prudent to evaluate continuous protocols at standardized intensities to obtain a better picture in recovery.

The biggest limitation of this systematic review is the vast heterogeneity of included articles regarding characteristics of subjects and exercise protocols. Completeness and lesion level in SCI directly affects autonomic innervation, however actual classification guidelines do not fully differentiate autonomic completeness.2 Even if future studies homogenize by the current American Spinal Injury Association Impairment Scale (AIS), it cannot be guaranteed that autonomic innervation follows a similar standardization.

Other limitations are the lack of experimental studies in the literature and the low sample size in quasi-experimental studies. In long-term studies, it is hard to obtain an adequate sample size of subjects with similar characteristics who complete the follow-up; this could be even more difficult if SCI participants require longer periods of training. In addition, studies involving exercise adaptations or acute response to exercise lack standardization of the protocols used. Even if intensity and duration of exercise is standardized, the HRV changes could be different depending on whether an arm ergometer or a BWST is used, so results of current evidence must be interpreted in the correct context.

HRV values in SCI subjects show differences with able-bodied subjects, specifically in LF, with a low risk of bias. However, it is not possible to attribute these differences to anatomic, functional, or behavior characteristics due to the high heterogeneity of included subjects. In the SCI population, there is a nonsignificant trend toward an improvement in HF values after a training period with a fair risk of bias. Because of small sample sizes and different exercise parameters used among studies, future research is needed with larger sample sizes and with adherence to SCI exercise recommendations. Regarding acute response of HRV to exercise in SCI subjects, all HRV values decrease immediately after exercise with a low risk of bias. Nevertheless, the role of intensity and duration of exercise in this response is unclear. In addition, studies in HRV recovery kinetics to exercise are lacking. Therefore, more experimental studies are needed to evaluate how exercise affects autonomic response and recovery with the purpose of obtaining a tool for monitoring intensity and rest periods during training and rehabilitation. Future studies must standardize AIS to minimize confounding variables in a highly heterogeneous population.

We thank Loreto Vergara, Sandra Mahecha, Natalia Gattini, and Erwin Buckel for comments and advice that improved this article.

The authors report no conflicts of interest.