Autonomic dysfunction in adults following mild traumatic brain injury: A systematic review

2.1Protocol and registration

The review was first registered in November 2019 (updated April 2020 and January 2021) with the National Institute for Health Research’s International Prospective Register of Systematic Reviews. The registered protocol (PROSPERO CRD42020157698) can be accessed at: https://www.crd.york.ac.uk/prospero/ display_record.php?RecordID = 157698. Institutional ethics approval was not required as no primary data collection occurred. The methods of this review followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Liberati et al., 2009).

2.2Information sources

Embase (1974 –January 21, 2021), Ovid MEDLINE (1946 –January 22, 2021), Cochrane Central Register of Clinical Trials (December 2020), PsycINFO (1803 –January Week 3, 2021), CINAHL (1973 –January 2021) and SPORTDiscus (1930 –January 2021) were searched. In databases that allowed, results were restricted to human studies. Search results were limited to English language studies. Filters were applied to select for adult and young adult populations as appropriate within the specific database searches. The final search was completed on January 24, 2021. No study authors were contacted to identify investigations beyond published work. Authors were encouraged to share additional references meeting inclusion criteria.

2.3Search

Development of the search strategy was done with assistance of a University of Calgary health sciences librarian. Search terms were related to 1) mTBI and 2) ANS. An example search strategy is provided for Embase in Table 1 and search strategies for additional databases are also available (Supplementary Tables 1-5). Records were exported directly from databases and imported into EndnoteX9 (EndNote Team, 2013). All results were exported from EndnoteX9 (EndNote Team, 2013) and imported into Covidence systematic review software where duplicates were removed.

2.4Eligibility criteria

Inclusion and exclusion criteria for study design, participants, outcome measures and comparators are provided in Table 2.

2.5Study selection

Following duplicate removal, all abstracts were screened by two independent reviewers based on eligibility criteria. Conflicts regarding inclusion/exclusion of studies were resolved by a third independent reviewer. Full text screening of included abstracts was completed by two independent reviewers and the same process was applied for conflict resolution.

2.6Data extraction

Data extraction and risk of bias (RoB) were performed by two reviewers independently for all studies meeting the inclusion criteria following full text screening. The following data was collected: participant characteristics (age, sex/gender), mTBI diagnosis and classification, control group, assessment timepoint(s), time since injury, physiologic measure with units and duration, test condition(s) and summarized results.

Each study was assessed for RoB by two independent reviewers using modified checklists evaluated and adapted to meet Scottish Intercollegiate Guidelines Network (SIGN) criteria (Healthcare Improvement Scotland, n.d.). A modified version of the SIGN Cohort checklist was used for cross-sectional and cohort study designs. A modified version of the SIGN Case-control checklist was used for case-control study design. The injury (mTBI) was considered as the “exposure” or “case” (for case-control studies) and metrics of ANS dysfunction as the “outcome”. The SIGN checklists contain 13–15 questions with four response categories: “yes”, “no”, “not applicable” and “can’t say”. In the cohort study checklist: A) Item 1.5 regarding drop outs was used to assess whether or not drop outs were reported; B) Item 1.11 was divided into two items, 1.11 and 1.12, which respectively assessed, “the outcome measure(s) is/are reliable” and “the outcome measure(s) is/are valid”; C) For item 1.13, “exposure level of prognostic factor is assessed more than once”, symptomatology/presence of persistent symptoms was considered the “exposure” and for studies with a single timepoint symptomatology had to be assessed once for the item to be rated as “yes”. In the case-control study checklist: the items, “the outcome measure(s) is/are reliable” (1.10) and “the outcome measure(s) is/are valid” (1.11) were added. Evaluation is based on reliability and validity of subject selection, assessment, confounding factors and statistical analysis. The overall assessment of the study was based on the ability to minimize RoB. Discrepancies were discussed and if a consensus was not achieved by two reviewers, conflicts were resolved by a third independent reviewer.

The corresponding authors for three included studies were contacted. The first researcher was contacted to provide supplementary tables referenced in the manuscript (Ding et al., 2020) (replied), the second researcher was contacted to clarify results (Huang et al., 2019) (did not reply) and a third author was contacted regarding result discrepancies within the manuscript (Kozlowski et al., 2013) (replied).

2.7Data synthesis

Due to heterogeneity of study methods, participant demographics, outcome measures and mTBI diagnostic criteria, data was not pooled for a meta-analysis. Demographics and results were summarized.

2.8Data availability

Data sharing is not applicable to this article as no new data were created or analyzed in this study.

3.1Study selection

A total of 4,311 studies were identified, of which, 1,064 duplicates were removed. Two references were shared by a study author (Haider et al., 2020; Pyndiura et al., 2020). The title and abstract of 3,247 studies were screened with 3,092 not meeting inclusion criteria (see Table 2). Full texts of 155 studies were assessed for eligibility and 116 were excluded. The most common reason for exclusion was incorrect study design. Ultimately, 39 studies were included in the qualitative synthesis (see Fig. 1 for PRISMA flowchart).

Fig. 1

PRISMA flowchart.

3.2Study characteristics

Study characteristics, participant demographics and RoB grade are summarized in Table 3.

3.4Risk of Bias (RoB)

Results from the RoB assessment are reported in Table 4. No studies were graded as “high-quality”. Eleven studies (Ding et al., 2020; Howard et al., 2018; Hutchison et al., 2017; Johnson et al., 2020; Kozlowski et al., 2013; Leddy et al., 2010; Purkayastha, Williams et al., 2019; Solbakk et al., 2005; Truong & Ciuffreda, 2016; Wright et al., 2017, 2018) were of “acceptable” quality, with the remaining 28 (Abaji et al., 2016; Bishop et al., 2017; Brandt et al., 2020; Clausen et al., 2016; Dobney et al., 2018; Dobson et al., 2017; Gall et al., 2004; Haider et al., 2020; Hanna-Pladdy et al., 2001; Hilz et al., 2011, 2015, 2016, 2017, 2020; Huang et al., 2019; Johnson et al., 2018; La Fountaine et al., 2009, 2011, 2016, 2018; Liao et al., 2016; Mirow et al., 2016; Pyndiura et al., 2020; Russell et al., 2020; Senthinathan et al., 2017; Sung, Chen et al., 2016; Sung, Lee et al., 2016; Tan et al., 2009) graded as “unacceptable”. The most common biases were a lack of reporting how many people were approached to participate, not specifying whether the assessment of outcomes was blinded to exposure status, unreliable measures of exposure (not specifying mTBI diagnostic criteria), unreliable or invalid outcome measure and not addressing potential confounders.

Table 4

Risk of bias of included studies

Table 5

Reporting of study confounders

3.5Study results

A summary of HR and HRV results is presented in Supplementary Table 6. BP, BRS and alternative autonomic measure results are presented in Supplementary Table 7. Correlations between metrics of autonomic function and injury characteristics are reported in Supplementary Table 7.

4.1Proposed pathophysiology of ANS dysfunction following mTBI

The underlying pathophysiology of ANS dysfunction following mTBI remains unknown; however, pre-clinical and clinical work suggests this may result from either: a) diffuse axonal injury during the initial mTBI; b) reduced cerebral blood flow (CBF) post-mTBI or c) deconditioning following the injury. Higher brain centers, including the prefrontal cortex (anterior cingulate, insula, orbitofrontal and ventromedial cortex) and limbic structures (amygdala and hypothalamus) in addition to structures in the brainstem (periaqueductal gray matter, nucleus ambiguus, medulla) form the Central Autonomic Network (Benarroch, 1993). Therefore, depending on the mechanism of injury and whether there is diffuse axonal injury, centers responsible for control of HRV and other autonomic function may be affected.

CBF has been increasingly investigated following mTBI using a variety of methods including transcranial Doppler and arterial spin labeling (ASL) magnetic resonance imaging (MRI). Using ASL MRI, Meier et al. reported reduced CBF in the right dorsal midinsular cortex in collegiate athletes in the first week post-mTBI compared to controls (Meier et al., 2015). This difference did not persist at 1 month post-injury. Athletes who were slow to recover, characterized as > 14 days until clearance for return-to-play, had lower CBF in the right dorsal midinsular cortex at 1 month compared to those with a more rapid return-to-play (Meier et al., 2015). As noted previously, the insula is part of the Central Autonomic Network, playing an important role in cardiovascular ANS integration (Bernaroch, 1993). While preliminary studies suggest a link between ANS dysfunction and altered CBF, additional research is required.

Following mTBI, individuals may struggle with exercise intolerance (Kozlowski et al., 2013), characterized as the inability to return to one’s premorbid level of physical activity due to symptom exacerbation. This may result in reduced physical activity and subsequent deconditioning. It is understood that deconditioning (reduced exercise capacity) is associated with decreased parasympathetic (vagal) activity and thus an increased resting HR (Gourine & Ackland, 2019). For example, improvement in HR recovery, a measure of parasympathetic activity, observed following a 6 week training program is lost after only 4 weeks of detraining (Heffernan et al., 2007). Individuals with higher resting HR and lower percentage predicted peak oxygen-consumption have been shown to have slower HR recovery following exercise, a measure of an individual’s ability to recruit cardiac vagal tone (Machhada et al., 2017). Unfortunately, the link between physical activity/exercise intolerance and ANS function in individuals with mTBI has not been well characterized.

4.2Methodological confounders

Previous reviews of the ANS literature (Blake et al., 2016; Callaway & Kosofsky, 2019; Pertab et al., 2018; Purkayastha, Stokes et al., 2019) have failed to highlight the significant methodological confounders present in the included studies. Additionally, several of these were not systematic reviews, nor did they include participants will all mechanisms of injury. The current systematic review addressed these gaps in the literature by including a wide range of autonomic function outcomes, completing an RoB assessment, providing a comprehensive review of confounders for each study and outlining future directions for ANS research in the mTBI context. Advances in this field will be dependent on the methodological rigor of future research. Additional studies addressing confounders and with larger sample sizes will be necessary before these measures may be implemented clinically.

Providing a clear operational definition of the “exposure” (mTBI) is important for reproducibility and generalizability of the research. In the case of eighteen studies reliability and/or validity of exposure status could not be ascertained based on presented information. Future research should use accepted mTBI diagnostic criteria, such as the American Congress of Rehabilitation Medicine criteria (Mild Traumatic Brain Injury Committee of the Head Injury Interdisciplinary Special Interest Group, 1993) or the 5th Consensus statement on concussion in sport criteria (McCrory et al., 2017). Injury was self-reported with no additional detail or diagnostic criteria in three studies (Dobney et al., 2018; Hanna-Pladdy et al., 2001; Pyndiura et al., 2020). It is also important to consider the mTBI history of control groups. In eighteen studies (Bishop et al., 2017; Clausen et al., 2016; Gall et al., 2004; Hilz et al., 2011, 2015, 2016, 2017, 2020; Huang et al., 2019; Hutchison et al., 2017; Johnson et al., 2018, 2020; La Fountaine et al., 2016, 2018; Purkayastha, Williams et al., 2019; Russell et al., 2020; Senthinathan et al., 2017; Tan et al., 2009), controls either did have a reported lifetime history of mTBI or insufficient information on mTBI history was presented. Particularly when evaluating ANS function in the chronic stage of injury or in those with persistent symptoms, control groups should be free of lifetime mTBI history to confidently report between group differences.

Many studies failed to report confounders that could affect ANS function data, such as other medical conditions, medications, cardiorespiratory fitness and stage of menstrual cycle. Despite many of the cohorts being youth athletes, screening and recording of past medical history and medications affecting ANS function should be clearly reported. Unfortunately, this was not the case in many studies and limits the internal validity. This is especially important in older cohorts and those with persistent symptoms where polypharmacy may be more likely. Cardiorespiratory fitness should also be considered, especially in athletic populations. As mentioned previously, resting HR, and thus HRV (Monfredi et al., 2014), is affected by cardiorespiratory fitness (Machhada et al., 2017). Several studies (Clausen et al., 2016; Haider et al., 2020; Johnson et al., 2018) compared varsity athletes to recreational control athletes. In the absence of either a direct or indirect index of cardiorespiratory fitness completed by both groups, interpretation of study findings is limited. Those with higher cardiorespiratory fitness, and thus lower resting HR, would have higher time-domain HRV measures, compared to their peers with lower cardiorespiratory fitness, thus producing between group differences unrelated to mTBI history. Stage of the menstrual cycle is also known to affect HRV. A recent meta-analysis highlighted fluctuation in cardiac vagal activity across the menstrual cycle with a significant decrease from the follicular to luteal phase (Schmalenberger et al., 2019). Analysis comparing cycle sub-phases (i.e., menstrual, mid-to-late follicular, ovulatory, early-to-mid luteal, premenstrual) revealed a decrease in measures of cardiac vagal activity from menstrual to premenstrual sub-phases and mid-to-late follicular to premenstrual sub-phases, indicating that the change from follicular to luteal phases is likely driven by a decline in the premenstrual sub-phase (of the luteal phase) (Schmalenberger et al., 2019). The meta-analysis pooled multiple autonomic measures (HF, RMSSD, BRS) and separate analyses for each variable were not completed (Schmalenberger et al., 2019). Unfortunately, none of the included studies in our review reported if participants were eumenorrheic or menstrual cycle stage, which should be included in future studies.

Multiple studies did not specify if reporting sex/gender or interchanged the two terms and no studies explicitly reported both sex and gender. Sex refers to a set of biological attributes (primarily associated with physical and physiological features including chromosomes, gene expression, hormone levels and function, and reproductive/sexual anatomy), whereas gender refers to the socially constructed roles, behaviours, expressions and identities of girls, women, boys, men, and gender diverse people (Canadian Institutes of Health Research, 2020). Future research should consider reporting both sex and gender. There is a lack of evaluation of both sex and gender differences in autonomic function following mTBI.

Length of recording periods are generally divided as ultra-short-term (<3 minutes), short-term (5-20 minutes) or 24 hour recordings. It is inappropriate to compare HRV metrics when collected from epochs of different lengths, as longer recordings are associated with increased HRV (Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology, 1996). Unfortunately, one study (Tan et al., 2009) did compare data from short-term recordings (mTBI participants) to 24 hour control data. Ultra-short-term recordings for analysis of HRV remain controversial, yet these were used in the majority of included studies (Abaji et al., 2016; Brandt et al., 2020; Haider et al., 2020; Hilz et al., 2011, 2015, 2016, 2017; Huang et al., 2019; Johnson et al., 2018; La Fountaine 2009, 2011; Russell et al., 2020). Recent research suggests that recordings of at least 4 minutes should be used to approximate short-term data (Burma et al., 2021). Using 1 minute recordings, metrics can vary by ≥30% when compared to a 5 minute recording (Burma et al., 2021). Based on these findings, analysis using epochs < 4 minutes were rated as unreliable/invalid in the RoB assessment.

Recording methods should also be carefully considered. While higher-end HR monitors may obtain fairly close values to an electrocardiogram (ECG) at rest and during exertion (pNN50 being an exception) (Vanderlei et al., 2008), ECG remains the gold-standard for measurement of HRV. For specific variables, such as QTVI, a 12-lead ECG should be used in place of a 3-lead ECG (which was used in an included study (La Fountaine et al., 2011)) as electrode placement can cause variation in QT time of 25-30 ms (Salvi et al., 2012). Respiratory rate is also known to affect certain HRV variables, specifically HF. Respiratory rate was measured in nine studies (Hilz et al., 2011, 2015, 2016, 2017, 2020; Huang et al., 2019; La Fountaine et al., 2018; Purkayastha, Williams et al., 2019; Solbakk et al., 2005). HF is commonly referred to as the respiratory frequency as it includes frequencies between 0.15–0.4 Hz, corresponding to physiologic events occurring 9–24 times per minute. At rest, respirations occur ∼12–20 times per minute, thus falling into the HF range (Singh et al., 2018); however, altered breathing faster than the normal range, potentially due to test anxiety, would elevate this rate above the HF band and thus skew interpretation of results. Therefore, future studies investigating HRV following mTBI should ensure there is a steady and consistent rate and depth of breathing for all participants at all time points. Additional considerations should be taken for certain test conditions. For example, studies using face cooling (Haider et al., 2020; Johnson et al., 2018) or cold pressor test (Johnson et al., 2020) conditions should address whether participants (in the case of the included studies, athletes) use ice baths in their training regimen as this may affect response.

Small sample size is another limitation of the included research given there is large between-subject variability for many HRV variables (Nunan et al., 2010). Of the included studies, only twelve had a sample size ≥25 in both groups (mTBI, controls) (Brandt et al., 2020; Dobney et al., 2018; Hanna-Pladdy et al., 2011; Hilz et al., 2016; Howard et al., 2018; Hutchison et al., 2017; Liao et al., 2016; Pyndiura et al., 2020l; Russell et al., 2020; Sung, Chen et al., 2016; Sung, Lee et al., 2016; Truong & Ciuffreda, 2016). The n = 25 sample size as been used as a cut-off for small sample size papers in other systematic reviews of this population (Iverson et al., 2020). Notably, only three of the studies (Howard et al., 2018; Hutchison et al., 2017; Truong & Ciuffreda, 2016) with a larger sample (≥25 in both groups) were rated as “acceptable” on the risk of bisk assessment. Ability to detect between-group differences may be affected by these small sample sizes with potentially great inter-subject variability. A lack of evaluation for confounding/modification by factors that may have affected test results in addition to the use of unreliable/invalid autonomic function outcome measures (i.e., ultra-short-term recordings) are reflected in the RoB assessment where 72% (28 out of 39) of studies (Abaji et al., 2016; Bishop et al., 2017; Brandt et al., 2020; Clausen et al., 2016; Dobney et al., 2018; Dobson et al., 2017; Gall et al., 2004; Haider et al., 2020; Hanna-Pladdy et al., 2001; Hilz et al., 2011, 2015, 2016, 2017, 2020; Huang et al., 2019; Johnson et al., 2018; La Fountaine et al., 2009, 2011, 2016, 2018; Liao et al., 2016; Mirow et al., 2016; Pyndiura et al., 2020; Russell et al., 2020; Senthinathan et al., 2017; Sung, Chen et al., 2016; Sung, Lee et al., 2016; Tan et al., 2009) were rated as “unacceptable”.

4.3Participant symptomatology

Whether a degree of ANS dysfunction persists in some or all cases following symptom resolution remains a question of great clinical interest. The relationship between specific post-concussive symptoms and autonomic dysfunction also requires further study. With return to sport protocols currently based on symptom exacerbation with increasing exertion/physical contact (McCrory et al., 2017), assessment of specific autonomic parameters could offer an objective measure of recovery. Preliminary data suggests certain autonomic measures may be altered in asymptomatic individuals in the post-acute stage of injury (Abaji et al., 2016; Haider et al., 2020), although these studies are limited by methodological confounders (see Table 5). In general, symptomatology is poorly characterized in the included literature, with sixteen studies (Abaji et al., 2016; Dobson et al., 2017; Gall et al., 2004; Hilz et al. 2011, 2015, 2016, 2017; Johnson et al., 2020; La Fountaine et al., 2009, 2016, 2018; Pyndiura et al., 2020; Senthinathan et al., 2017; Solbakk et al., 2005; Tan et al., 2009; Truong & Ciuffreda, 2016) not reporting results of a specific symptom questionnaire nor a general symptom battery. Unfortunately, several studies noted a symptom checklist had been completed, but results were not reported in the manuscript (Gall et al., 2004; Johnson et al., 2020; La Fountaine et al., 2009; Senthinathan et al., 2017; Tan et al., 2009). Additionally, the heterogeneity of questionnaires does not allow for easy comparison between studies and standardization should be considered. There is a great need for a study with the primary aim of evaluating change in multiple post-concussive symptoms and autonomic metrics across recovery timepoints in a large cohort. Finally, study of ANS dysfunction in cohorts with persistent post-concussive symptoms is limited. While several studies did complete assessments in the chronic stage of injury (>3 months post), symptoms were poorly characterized in many cases (Hilz et al., 2011, 2015, 2016, 2017; Pyndiura et al., 2020; Solbakk et al., 2005). Research in cohorts with persistent post-concussive symptoms, diagnosed using standardized criteria, is necessary to understand potential underlying physiologic mechanisms contributing to persistent symptoms, such as exercise intolerance, dizziness and headache.

4.4Limitations

There are several limitations based on the methods of this review. Only published work was included in the review, thus the findings are subject to publication bias. Additionally, only English language papers were included. Our results and conclusions are therefore subject to language-of-publication bias. The majority of studies included in this review had a high RoB due to methodological shortcomings. As pre-clinical studies were not included, this review does not address translational research on this topic. It is understood that different critical appraisal tools uniquely assess the methodological quality of a publication and different RoB tools have been shown to lead to different conclusions (ratings of study quality) (Losilla et al., 2018). The SIGN checklist was chosen for this review as it is a comprehensive, easy to use RoB assessment with checklists tailored to different study designs (i.e., cohort, case-control studies). The heterogeneity of methods, outcomes and samples precluded a quantitative review or meta-analysis.

4.5Future directions

There is great opportunity for the use of ANS function parameters, such as HRV, to understand recovery following mTBI and potentially guide return to play; however, the significant limitations of the presented literature limit its clinical applicability. We propose the following recommendations to build upon the current literature and ensure methodological rigor in the field.