http://informahealthcare.com/bij ISSN: 0269-9052 (print), 1362-301X (electronic) Brain Inj, 2015; 29(3): 374–379 ! 2014 Informa UK Ltd. DOI: 10.3109/02699052.2014.969312
ORIGINAL ARTICLE
Association between the sensory-motor nervous system and the autonomic nervous system in neurorehabilitation patients with severe acquired brain injury Simon Tilma Vistisen1,2, Jim Jensen1, Jesper Fleischer3, & Jørgen Feldbæk Nielsen1 1
Hammel Neurorehabilitation Center and University Research Clinic, 2Research Centre for Emergency Medicine, Institute of Clinical Medicine, Aarhus University, Aarhus, Denmark, and 3Department of Endocrinology and Internal Medicine and the Medical Research Laboratories, Aarhus University Hospital, Aarhus, Denmark Abstract
Keywords
Introduction: The relation between motor and cognitive function and autonomic nervous system (ANS) function during neurorehabilitation following acquired brain injury (ABI) has only been investigated sporadically. In the present study, it was hypothesized that clinical measures in severely injured patients would relate to heart rate variability (HRV), a measure of autonomic function. Methods: HRV measurements were initially performed on 49 patients (enrolled in a previous study) and follow-up (4 28 days) HRV measurements were performed. Standard deviation of normal-to-normal intervals (SDNN) and low frequency (LF) were extracted and these HRV variables were related to the clinical measures, Early Functional Ability (EFA) and Functional Independence Measure (FIM). Associations between HRV and clinical measures were analysed on admission data (only EFA), at follow-up and for the longitudinal change in measures. Results: Follow-up HRV was extracted from 19 patients. SDNN and LF were significantly correlated (p50.05) to the EFA and FIM at follow-up, but not at admission. SDNN and LF changes were significantly correlated to EFA changes, but not FIM changes. Admission SDNN and LF were unable to provide prognostic information for the EFA and FIM at follow-up. Conclusion: HRV and its change during neurorehabilitation were associated to EFA and EFA changes over time. Further studies are required to clarify a number of limitations arising from this observational study.
Acquired brain injury, autonomic nervous system, heart rate variability, neurorehabilitation
Introduction Following acquired brain injury (ABI), improvements in cognitive and sensory-motor deficits are the primary target for rehabilitation interventions. One neglected aspect of neurorehabilitation in patients with ABI is the dysfunction observed in the autonomic nervous system (ANS). Unless obvious symptoms such as paroxysmal sympathetic hyperactivity are present [1–3], changes to the ANS are largely ignored. However, despite the lack of focus on ANS dysfunction, ANS symptoms also improve through the rehabilitation process [3,4] (similarly to cognitive and sensory motor deficits). Despite this, longitudinal development in ANS function and its relationship to cognitive and sensory-motor function during rehabilitation is not well investigated. The studies that have characterized/quantified ANS in patients admitted for neurorehabilitation [4,5] are limited and only one study [4] has presented longitudinal data comparing ANS measures with cognitive and sensory-motor outcome
Correspondence: Simon Tilma Vistisen, Voldbyvej 15, building 5, 1st floor, 8450 Hammel, Denmark. Tel: (+45) 2067 6868. Fax: (+45) 7841 9677. E-mail: [email protected]
History Received 9 December 2013 Revised 26 August 2014 Accepted 22 September 2014 Published online 28 October 2014
measures. In that study, Keren et al. [4] reported heart rate variability (HRV), as an estimate of the ANS function, in 20 patients with moderate-to-severe traumatic brain injury (TBI), and observed that HRV measures had a non-significant tendency to increase from admission to follow-up (38 days post-admission), indicating that autonomic regulation improves during rehabilitation. Despite this, changes in HRV were related to the time since the injury and not to changes in cognitive and sensory-motor scores [4]. Given the sparse data on autonomic function in patients with ABI, in particular those patients not diagnosed with TBI, the current study aims to investigate the relationship between HRV and clinical outcome measures in the sub-acute phase of patients with ABI (TBI, stroke, subarachnoid haemorrhage (SAH) and anoxia/hypoxia following cardiac arrest). It is hypothesized that (1) HRV measures are correlated to clinical outcome measures on admission and follow-up (4 28 days). (2) The change in HRV is correlated to the changes in clinical outcome measures. (3) Admission HRV predicts follow-up clinical outcome measures. (4) HRV increases (improves) from admission to follow-up.
Sensory-motor and autonomic nervous system
DOI: 10.3109/02699052.2014.969312
The secondary aim was to corroborate previous findings that the circadian pattern for patients with ABI was present at followup (HRV lower at 10 pm compared to 2 am and 6 am) [6].
Methods Patients Consecutive patients included in a previous study [6] (comparing admission HRV measures with healthy controls and investigating circadian HRV pattern of patients) were reexamined at least 28 days from admission. The HRV measures obtained at follow-up were part of usual daily practice in the rehabilitation institution. The Danish Scientific Ethical Committee did not require informed patient consent because this study, according to the Danish Scientific Ethical Committee law, falls under the term quality assurance. Patients were included if they were aged 417 years and if they had no previous history of ABI. Patients were not observed by the investigators during the ECG recordings. Clinical outcome measures: Cognitive and sensory-motor clinical scores The clinical measurements collected at admission and follow up were the Early Functional Ability (EFA), Functional Independence Measure (FIM) and Ranchos Los Amigos Score (RLAS). The FIM and RLAS scores are widely used in rehabilitation institutions, but the use of EFA is less widespread. The EFA scale (developed by Heck et al. [7]) has been included as it is able to ‘bridge the gap’ between comatose patients (in which the coma recovery scale is most appropriate) and patients with greater function (in which the FIM is most appropriate [7]). As the FIM encounters a floor effect with lower functioning patients (with many patients scoring the lowest FIM score [8]) and the coma recovery scale encounters a ceiling effect (see Heck et al. [7]) the EFA scale, sensitive to small changes in function, is able to characterize patients with severe brain damage with the capacity to recover meaningful function. The EFA consists of 20 items, with each item scored from 1–5. Therefore, EFA scores range from 20–100, with a score of 100 indicating no loss of function. The FIM is an 18item sum score, with 13 items related to motor function and five items related to cognitive function, respectively, with each item scored from 1–7. Therefore, FIM scores range from 18–126, with a score of 126 indicating complete independence. RLAS is an 8-point score indicating consciousness level, behavioural and cognitive deficits. A score of ‘8’ indicates a purposeful and appropriate level of cognitive functioning. Other measures such as Glasgow Coma Scale (GCS) and Post-Traumatic Amnesia (PTA) duration (usually used to estimate initial injury severity in patients with TBI) were not available and those measures are not the most useful measures to indicate the injury severity in these heterogenic neurorehabilitation patients; rather based on the FIM and EFA measures, the patients’ injury severity is considered at the time of neurorehabilitation [8]. ECG recordings, processing and R spike detection The acquisition and processing of the admission ECG data has been described previously [6]. Briefly, the ECG data were
375
sampled using a standard clinical monitoring (S/5 monitor, General Electrics, Datex-Ohmeda Division, Instrumentarium Corp., UK). The data were pre-processed offline and R spike detection was implemented with a precision of 1 millisecond using a customized automatic algorithm developed in Matlab (Mathworks Inc., Natick). Follow-up ECG data were acquired by the Actiheart acquisition unit (CamNtech, Cambridge, UK), which is able to identify R spikes and determine RR intervals with a precision of 1 millisecond. HRV variables and extraction Overall, there exist three HRV quantification methods: Time domain analysis methods, frequency analysis methods and non-linear analysis methods [9]. Different HRV measures are thought to reflect parasympathetic tone, sympathetic tone and the balance between the two systems or the total variation [9]. In line with a previous study [6], the most appropriate HRV variables are considered to be Standard Deviation of Normalto-Normal heart beats (SDNN; time domain) and the Low Frequency component (LF) derived from the frequency method (0.04–0.15 Hz, extracted based on the fast Fourier transformed spectrum). HRV values were extracted according to suggested standards [9] (From 5-minute RR interval time series). Measurements were performed at 6 pm, 10 pm, 2 am and 6 am (the following day) as patients were scheduled to rest at these time points. HRV analysis was performed in Kubios-HRV software (University of Eastern Finland, Kuopio, Finland). RR time series (the tachogram) were artefact corrected using the correction algorithm in the Kubios-HRV software (correction setting set at ‘very strong’ correction). Five minute windows with a stable heart rate (adequate artefact correction, no ‘paroxysmal bursts’, no upward or downward trend in the RR interval) were chosen and visually inspected by one of the authors (STV). Statistics The demographic and basic clinical data are reported as the median, interquartile range (IQR) and range. HRV was summarized for each patient defined as the geometric mean from the available HRV values at the four time points. The EFA scores on admission and follow-up and FIM scores at follow-up were compared to corresponding LF and SDNN admission and follow-up scores (admission FIM scores were not compared to admission HRV as 75% of the patients had admission FIM scores 5 21). Additionally, the changes in HRV between admission and follow-up (DLF and DSDNN, both absolute and relative) were compared to the corresponding changes in the EFA and FIM (DEFA and DFIM). Finally, admission HRV data were compared with follow-up EFA, follow-up FIM and DEFA and DFIM to investigate the prognostic value of HRV. The relationship between HRV measures and clinical measures were analysed using linear least squares regression when appropriate. The regression models were evaluated by inspecting QQ plots and a plot of the residuals. HRV variables were log transformed before comparing HRV to clinical scores. Spearman rank correlation analysis was performed if regression analysis assumptions were violated.
376
S. T. Vistisen et al.
One-tailed paired t-tests were used to compare admission LF and SDNN with follow-up LF and SDNN. Changes in cardiopulmonary variables (heart rate, respiratory frequency, diastolic blood pressure and systolic blood pressure) between admission and follow-up were analysed using a two-tailed paired t-test. As a secondary analysis, this study also analysed heart rate’s (HR) relation to clinical measures using the same statistical approach explained for HRV variables. On admission, through the four time points, some patients had variable HRV whilst others had regular HRV. Due to this, a secondary analysis was performed and investigated the relationship between the variation in HRV (standard deviation of the available HRV variables for the four time points) and clinical scores. One patient was excluded from this analysis as the HRV was only recorded at one of the four time points. Regarding half-circadian rhythm analysis at follow-up, a one-tailed paired t-test was used to compare LF and SDNN at 10 pm with 2 am and 6 am (as previously shown for admission data). The data were log-transformed as following inspection of QQ plots. Statistics were performed in Stata (StataCorp LP, College Station) or Matlab (MathWorks Inc., Natick); p50.05 was considered significant.
Results Of 49 patients with admission HRV data, 19 patients were evaluated for HRV at follow-up (see flow chart in Figure 1). Admission EFA scores of patients included and lost to follow-up were inspected for possible selection bias (see Electronic Supplement, Figure S1). Patient characteristics, including the clinical measures at admission and follow-up, are shown in Table I Patient follow-up occurred 35 days (median) after admission, interquartile range ¼ 31–39 days. For two patients, follow-up occurred after more than 45 days (63 and 97 days
Brain Inj, 2015; 29(3): 374–379
post-admission). The median time lag between collection of the clinical scores and ECG data at follow-up was 1 day (ECG obtained 1 day after clinical score); interquartile range ¼ 4–0 days. Two patients had an absolute time lag of more than 4 days (8 and 10 days). Correlation data and corresponding statistics are shown in Figures 2 and 3, as well as the Supplemental Material. Based on analysis of the residuals, approximately half of the least squares regression analyses were deemed inappropriate. Therefore, Spearman rank correlation was used for all analyses and corresponding discussion. Despite this, Pearson regression analyses have also been reported. On admission, EFA scores were not statistically significantly correlated to LF or SDNN (pSDNN ¼ 0.27 and pLF ¼ 0.23, see Figure 2). At follow-up, EFA and FIM were significantly correlated to SDNN and LF (pSDNN, EFA ¼ 0.005 and pLF, EFA ¼ 0.009; pSDNN, FIM ¼ 0.027 and pLF, FIM ¼ 0.026; see Figure 3 and Supplemental Figure S2). Relative and absolute DSDNN and relative DLF (but not absolute DLF) were significantly correlated to DEFA (pabsolute DSDNN, DEFA ¼ 0.023 and prelative DSDNN, DEFA ¼ 0.041; pabsolute DLF, DEFA ¼ 0.10 and prelative DLF, DEFA ¼ 0.037. DFIM was not correlated to DSDNN and DLF (pabsolute DSDNN, DFIM ¼ 0.07 and prelative DSDNN, DFIM ¼ 0.14; pabsolute DLF, DFIM ¼ 0.16 and prelative DLF, DFIM ¼ 0.095, see Supplemental Figures S3 and S4). Heart rate was not related to EFA at follow-up, but was significantly correlated to EFA on admission (p ¼ 0.01). Additionally, HR changes were related to changes in EFA (p ¼ 0.005; see Figure S5). On admission, the standard deviation of the four (in four patients, only three) HRV measures were negatively Table I. Patient characteristics. n Male gender (n ¼ 19) 10 ABI diagnosis TBI 5 SAH 5 Stroke 5 Anoxic 4 Diabetes Type 1 0 Type 2 0 Cardiac drug changes 6 between admission and follow-up (n) Ca+ antagonists added 1 Beta blockers added 2 Anti-hypertensive 1 drug added Ca+ antagonists withdrawn 1 ACE inhibitors withdrawn 1 Intrathecal baclofen pump changes Added 1 Age 55 Days since injury, 37 admission 2 BMI (kg m ), 22.9 admission EFA, admission 46 FIM, admission 18 RLAS, admission 3 EFA, follow-up 63 FIM, follow-up 23 RLAS, follow-up 5
Figure 1. Flow chart of included patients with admission HRV data.
IQR, Interquartile range.
Median
IQR
38–69 31–44
23–78 21–110
20.7–24.6 41–56 18–20 3–5 43–86 18–57 3–6
16.4–30.3 26–71 18–37 2–6 31–95 18–82 2–7
Range
Sensory-motor and autonomic nervous system
DOI: 10.3109/02699052.2014.969312
logSDNN and EFA score on admission
logLF and EFA score on admission
(B)
100
100
90
90
80
80
70
70
EFA score
EFA score
(A)
60
50
60
50
40
40
30
30
20
r =0.3465, p
Pearson
Rho =0.1621, p
Spearman
1
1.5
2 2.5 3 logSDNN (ms))
20
=0.1209
r =0.3716, p
Pearson
=0.2657 3.5
Rho =0.176, p
=0.08
Spearman
4
377
1
2
3
4 5 logLF (ms2))
=0.2265 6
7
Figure 2. Scatter plots of admission EFA score and admission HRV variables.
correlated to the EFA score (pSDNN ¼ 0.009 and pLF ¼ 0.02; see Figure S6), but follow-up standard deviation of LF and SDNN were not related to follow-up EFA and FIM scores (p-value range ¼ 0.13–0.89). No admission HRV variables were able to predict the clinical measurements at follow-up or their change from admission to follow-up, i.e. admission HRV measures were not correlated to follow-up clinical scores or changes in clinical scores. SDNN (p ¼ 0.020) and LF (p ¼ 0.038) increased from admission to follow-up (SDNN on admission: 11.4 milliseconds, Confidence Interval (CI) ¼ [8.4; 15.4] milliseconds vs. SDNN at follow-up: 15.9 milliseconds, CI ¼ [12.6; 20.0] milliseconds; LF on admission: 22.4 ms2, CI ¼ [11.7; 43.0] ms2 vs. LF at follow-up: 42.9 ms2, CI ¼ [23.6; 77.9] ms2). Cardiopulmonary variables either significantly decreased or had a tendency to decrease, see Table II Half-circadian rhythm: SDNN at 10 pm was lower than SDNN at 2 am (p ¼ 0.017), but not SDNN at 6 am (p ¼ 0.28). LF at 10 pm was not significantly different from LF 2 am (p ¼ 0.069) or LF at 6 am (p ¼ 0.19).
Discussion The current study aimed at quantifying cardiac autonomic regulation and its longitudinal changes in patients with heterogenic ABI diagnoses admitted for neurorehabilitation.
In particular, this is unknown for other diagnoses than TBI. The study demonstrates that the investigated HRV variables (SDNN and LF) at follow-up were significantly correlated to the EFA and FIM; however, admission HRV was not significantly correlated to admission EFA. In addition, changes in HRV variables from admission to follow-up were generally significantly correlated to corresponding changes in the EFA score but not FIM score. Furthermore, HRV increased from admission to follow-up and the analysed cardiopulmonary variables significantly decreased or tended to decrease. HR was related to EFA on admission and HR changes were also associated with EFA changes. Therefore, the current study indicates a reduction in sympathetic tone from admission to follow-up and an improvement in the autonomic regulation of HR. Additionally, the data generally indicate an association between measures of the autonomic nervous system and clinical measurements of cognitive and somato-sensory function (EFA and FIM). The patients in the current study were severely brain injured (75% of patients had FIM 521 on admission) and likely encountered the floor effect for the FIM. Therefore it was not surprising that the EFA score (being more sensitive when evaluating these patients [7,8,10]) demonstrated the best relationship to HRV variables. Follow-up scatter plots (Supplemental Figure S2) demonstrated that nine of 19 patients had FIM scores 20 throughout the study period and,
S. T. Vistisen et al. logSDNN and EFA score at follow−up
(A)
EFA score
Brain Inj, 2015; 29(3): 374–379
logLF and EFA score at follow−up
(B)
100
100
90
90
80
80
70
70
EFA score
378
60
50
60
50
40
40
30
30
20
r =0.642, p
Pearson
1.5
r =0.6458, p
Pearson
Rho =0.619, p
=0.0047
2 2.5 3 logSDNN (ms)
3.5
Spearman
10 1
20
=0.0417
Rho =0.5795, p
=0.0394
Spearman
4
10 1
2
3
4 5 logLF (ms2)
=0.0093 6
7
Figure 3. Scatter plots of follow-up EFA score and follow-up HRV variables.
Table II. Summary of cardiopulmonary clinical variables from admission to follow-up. Variable 1
HR (min ) Respiratory frequency (min1) Diastolic blood pressure (mmHg) Systolic blood pressure (mmHg)
Admission
Follow-up
94 20 86 139
81 18 80 127
(23) (6.3) (14) (23)
(16) (7.7) (10) (18)
p Value 0.008 0.08 0.08 0.06
Data are presented as mean (standard deviation).
therefore, the EFA is sensitive to detect functional progression and in turn relationships with HRV improvements in these patients [8]. This could be the reason Keren et al. [4] demonstrated no relationship between HRV variables and FIM. In accordance with the previous findings, it was found that the HRV variable, SDNN, at follow-up was statistically significantly lower at 10 pm than at 2 am. LF had a nonsignificant tendency to reduce from at 10 pm than at 2 am (p ¼ 0.069). There was a trend when comparing HRV variables at 10 pm with HRV variables at 6 am and, although non-significant, the trend was consistent with a previous finding. This is likely due to the loss of patients to follow-up. To summarize, the half circadian pattern analyses of patients on admission and follow-up are comparable to healthy
subjects, although HRV variables are generally lower in patients with ABI than in healthy subjects [6]. There was a significant relationship between admission EFA score and the standard deviation of four time measurements of the HRV variables on admission. This was not observed during follow-up and the admission relationship is merely regarded as a hypothesis generating finding. Limitations There are a number of limitations to the current study. First, the follow-up study population had high loss to follow-up, which may lead to selection bias. Data from 19 of 49 patients were included. Eight of the ‘lost to follow-up’ group were discharged prior to scheduled follow-up. Patients discharged within a few weeks of admission are usually discharged as they (by clinical judgement) are either in a too poor condition and have little potential for rehabilitation (and admitted to e.g. a nursing home) or in a too good condition (and typically admitted to less intensive rehabilitation). This tendency might be present in the data (Supplemental Figure S1) and, although speculative, these patients could develop differently in the outcome measures assessed than the included patients. It is unlikely that the other main reasons for exclusion (follow-up registration not initiated (6), logistic reasons (7) and tachogram quality inadequate (5)) were a potential source
Sensory-motor and autonomic nervous system
DOI: 10.3109/02699052.2014.969312
for selection bias, neither theoretically nor indicated by admission EFA scores. Second, it was assumed that patients were resting during their scheduled resting periods; however, the patients were not observed during their ECG recordings over these times. Third, patients treated with cardiac drugs have been included, some of whom had changes in drug administration during the study. Drugs affecting the cardiovascular system like -blockers have previously been shown to be associated with improved parasympatic modulation and reduced heart rate [11]. In spite of this possible beneficial intervention, no difference was found in HRV parameters between patients receiving cardiac drugs compared to those not receiving cardiac drugs [6]. Fourth, given the observational nature of the study, different ECG equipment was used for detection of RR intervals on admission and follow-up. This is mainly an issue regarding interpretation of the statistically significant results that SDNN and LF increase from admission to follow-up. The RR interval detection method on admission offers a precision of 1 millisecond. This is identical to the reported detection from the manufacturer of the system used at follow-up. Nonetheless, the two technologies have not been validated against each other (by e.g. Bland-Altman analysis) and the conclusion that SDNN and LF increases in the rehabilitation period should be treated with caution. Despite this the data are consistent with the trends observed by Keren et al. [4] for all four investigated HRV variables in that study. Also, the observational nature of the study gave rise to variability between the timing of HRV recordings and the clinical evaluation of patients. Yet, it is speculated that this impacts only slightly if at all the discussion of the present data. Fifth, the analyses in the current study are based on data derived from visually inspected tachograms, which reduces external validity. Still, it is considered a pre-requisite to visually inspect tachograms from patients with ABI to extract reliable HRV data. Finally, in healthy subjects, HRV variables depend on the underlying HR and age and it has been suggested to correct for HR and age [11]. Given the low number of patients, the residual problems with linear regression (where a correction would typically be performed) and the unknown impact of HR and age on HRV for patients with ABI, this study only reported the HR in relation to EFA (in Supplemental Material). Given these reported HR data, the associations between clinical scores and HRV might be partly explained by HR itself and not HRV alone. Nonetheless, HR and HRV indicate that sympathetic activity and its development are related to clinical scores and their development. In conclusion, the autonomic nervous system and its function, as represented by HRV, appear to be related to the clinically evaluated sensory-motor and cognitive function of
379
neurorehabilitation patients and that change in HRV is related to changes in the clinical measures. This observational study offers insights to the autonomic nervous function, an area rarely addressed in neurorehabilitation. However, as the study is observational and has a number of limitations, the results need to be confirmed in more controlled settings on a larger set of patients.
Acknowledgements The authors thank Peter W. Stubbs for proofreading this manuscript.
Declaration of interest The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.
References 1. Baguley IJ. Autonomic complications following central nervous system injury. Seminars in Neurology 2008;28:716–725. 2. Baguley IJ, Heriseanu RE, Cameron ID, Nott MT, Slewa-Younan S. A critical review of the pathophysiology of dysautonomia following traumatic brain injury. Neurocritical Care 2008;8:293–300. 3. Perkes I, Baguley IJ, Nott MT, Menon DK. A review of paroxysmal sympathetic hyperactivity after acquired brain injury. Annals of Neurology 2010;68:126–135. 4. Keren O, Yupatov S, Radai MM, Elad-Yarum R, Faraggi D, Abboud S, Ring H, Groswasser Z. Heart rate variability (HRV) of patients with traumatic brain injury (TBI) during the post-insult sub-acute period. Brain Injury 2005;19:605–611. 5. King ML, Lichtman SW, Seliger G, Ehert FA, Steinberg JS. Heartrate variability in chronic traumatic brain injury. Brain Injury 1997; 11:445–453. 6. Vistisen ST, Hansen TK, Jensen J, Nielsen JF, Fleischer J. Heart rate variability in neurorehabilitation patients with severe acquired brain injury. Brain Injury 2014;28:196–202. 7. Heck G, Steiger-Ba¨chler G, Schmidt T. Early functional abilities (EFA) – eine skala zur evaluation von behandlungsverla¨ufen in der neurologischen fru¨hrehabilitation. Neurology & Rehabilitation 2000;6:125–133. 8. Stubbs PW, Pallesen H, Pedersen AR, Nielsen JF. Using EFA and FIM rating scales could provide a more complete assessment of patients with acquired brain injury. Disability and Rehabilitation 2014. DOI: 10.3109/09638288.2014.904935. 9. [Anonymous]. Heart rate variability. Standards of measurement, physiological interpretation, and clinical use. Task force of the European Society of Cardiology and The North American Society of Pacing and Electrophysiology. European Heart Journal 1996;17: 354–381. 10. Alvsaker K, Walther SM, Kleffelgard I, Mongs M, Draegebo RA, Keller A. Inter-rater reliability of the early functional abilities scale. Journal of Rehabilitation Medicine: Official Journal of the UEMS European Board of Physical and Rehabilitation Medicine 2011;43: 892–899. 11. Tsuji H, Venditti Jr FJ, Manders ES, Evans JC, Larson MG, Feldman CL, Levy D. Determinants of heart rate variability. Journal of the American College of Cardiology 1996;28:1539–1546.
Supplementary material available online Supplementary Figures 1–6
Copyright of Brain Injury is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
ORIGINAL ARTICLE
Association between the sensory-motor nervous system and the autonomic nervous system in neurorehabilitation patients with severe acquired brain injury Simon Tilma Vistisen1,2, Jim Jensen1, Jesper Fleischer3, & Jørgen Feldbæk Nielsen1 1
Hammel Neurorehabilitation Center and University Research Clinic, 2Research Centre for Emergency Medicine, Institute of Clinical Medicine, Aarhus University, Aarhus, Denmark, and 3Department of Endocrinology and Internal Medicine and the Medical Research Laboratories, Aarhus University Hospital, Aarhus, Denmark Abstract
Keywords
Introduction: The relation between motor and cognitive function and autonomic nervous system (ANS) function during neurorehabilitation following acquired brain injury (ABI) has only been investigated sporadically. In the present study, it was hypothesized that clinical measures in severely injured patients would relate to heart rate variability (HRV), a measure of autonomic function. Methods: HRV measurements were initially performed on 49 patients (enrolled in a previous study) and follow-up (4 28 days) HRV measurements were performed. Standard deviation of normal-to-normal intervals (SDNN) and low frequency (LF) were extracted and these HRV variables were related to the clinical measures, Early Functional Ability (EFA) and Functional Independence Measure (FIM). Associations between HRV and clinical measures were analysed on admission data (only EFA), at follow-up and for the longitudinal change in measures. Results: Follow-up HRV was extracted from 19 patients. SDNN and LF were significantly correlated (p50.05) to the EFA and FIM at follow-up, but not at admission. SDNN and LF changes were significantly correlated to EFA changes, but not FIM changes. Admission SDNN and LF were unable to provide prognostic information for the EFA and FIM at follow-up. Conclusion: HRV and its change during neurorehabilitation were associated to EFA and EFA changes over time. Further studies are required to clarify a number of limitations arising from this observational study.
Acquired brain injury, autonomic nervous system, heart rate variability, neurorehabilitation
Introduction Following acquired brain injury (ABI), improvements in cognitive and sensory-motor deficits are the primary target for rehabilitation interventions. One neglected aspect of neurorehabilitation in patients with ABI is the dysfunction observed in the autonomic nervous system (ANS). Unless obvious symptoms such as paroxysmal sympathetic hyperactivity are present [1–3], changes to the ANS are largely ignored. However, despite the lack of focus on ANS dysfunction, ANS symptoms also improve through the rehabilitation process [3,4] (similarly to cognitive and sensory motor deficits). Despite this, longitudinal development in ANS function and its relationship to cognitive and sensory-motor function during rehabilitation is not well investigated. The studies that have characterized/quantified ANS in patients admitted for neurorehabilitation [4,5] are limited and only one study [4] has presented longitudinal data comparing ANS measures with cognitive and sensory-motor outcome
Correspondence: Simon Tilma Vistisen, Voldbyvej 15, building 5, 1st floor, 8450 Hammel, Denmark. Tel: (+45) 2067 6868. Fax: (+45) 7841 9677. E-mail: [email protected]
History Received 9 December 2013 Revised 26 August 2014 Accepted 22 September 2014 Published online 28 October 2014
measures. In that study, Keren et al. [4] reported heart rate variability (HRV), as an estimate of the ANS function, in 20 patients with moderate-to-severe traumatic brain injury (TBI), and observed that HRV measures had a non-significant tendency to increase from admission to follow-up (38 days post-admission), indicating that autonomic regulation improves during rehabilitation. Despite this, changes in HRV were related to the time since the injury and not to changes in cognitive and sensory-motor scores [4]. Given the sparse data on autonomic function in patients with ABI, in particular those patients not diagnosed with TBI, the current study aims to investigate the relationship between HRV and clinical outcome measures in the sub-acute phase of patients with ABI (TBI, stroke, subarachnoid haemorrhage (SAH) and anoxia/hypoxia following cardiac arrest). It is hypothesized that (1) HRV measures are correlated to clinical outcome measures on admission and follow-up (4 28 days). (2) The change in HRV is correlated to the changes in clinical outcome measures. (3) Admission HRV predicts follow-up clinical outcome measures. (4) HRV increases (improves) from admission to follow-up.
Sensory-motor and autonomic nervous system
DOI: 10.3109/02699052.2014.969312
The secondary aim was to corroborate previous findings that the circadian pattern for patients with ABI was present at followup (HRV lower at 10 pm compared to 2 am and 6 am) [6].
Methods Patients Consecutive patients included in a previous study [6] (comparing admission HRV measures with healthy controls and investigating circadian HRV pattern of patients) were reexamined at least 28 days from admission. The HRV measures obtained at follow-up were part of usual daily practice in the rehabilitation institution. The Danish Scientific Ethical Committee did not require informed patient consent because this study, according to the Danish Scientific Ethical Committee law, falls under the term quality assurance. Patients were included if they were aged 417 years and if they had no previous history of ABI. Patients were not observed by the investigators during the ECG recordings. Clinical outcome measures: Cognitive and sensory-motor clinical scores The clinical measurements collected at admission and follow up were the Early Functional Ability (EFA), Functional Independence Measure (FIM) and Ranchos Los Amigos Score (RLAS). The FIM and RLAS scores are widely used in rehabilitation institutions, but the use of EFA is less widespread. The EFA scale (developed by Heck et al. [7]) has been included as it is able to ‘bridge the gap’ between comatose patients (in which the coma recovery scale is most appropriate) and patients with greater function (in which the FIM is most appropriate [7]). As the FIM encounters a floor effect with lower functioning patients (with many patients scoring the lowest FIM score [8]) and the coma recovery scale encounters a ceiling effect (see Heck et al. [7]) the EFA scale, sensitive to small changes in function, is able to characterize patients with severe brain damage with the capacity to recover meaningful function. The EFA consists of 20 items, with each item scored from 1–5. Therefore, EFA scores range from 20–100, with a score of 100 indicating no loss of function. The FIM is an 18item sum score, with 13 items related to motor function and five items related to cognitive function, respectively, with each item scored from 1–7. Therefore, FIM scores range from 18–126, with a score of 126 indicating complete independence. RLAS is an 8-point score indicating consciousness level, behavioural and cognitive deficits. A score of ‘8’ indicates a purposeful and appropriate level of cognitive functioning. Other measures such as Glasgow Coma Scale (GCS) and Post-Traumatic Amnesia (PTA) duration (usually used to estimate initial injury severity in patients with TBI) were not available and those measures are not the most useful measures to indicate the injury severity in these heterogenic neurorehabilitation patients; rather based on the FIM and EFA measures, the patients’ injury severity is considered at the time of neurorehabilitation [8]. ECG recordings, processing and R spike detection The acquisition and processing of the admission ECG data has been described previously [6]. Briefly, the ECG data were
375
sampled using a standard clinical monitoring (S/5 monitor, General Electrics, Datex-Ohmeda Division, Instrumentarium Corp., UK). The data were pre-processed offline and R spike detection was implemented with a precision of 1 millisecond using a customized automatic algorithm developed in Matlab (Mathworks Inc., Natick). Follow-up ECG data were acquired by the Actiheart acquisition unit (CamNtech, Cambridge, UK), which is able to identify R spikes and determine RR intervals with a precision of 1 millisecond. HRV variables and extraction Overall, there exist three HRV quantification methods: Time domain analysis methods, frequency analysis methods and non-linear analysis methods [9]. Different HRV measures are thought to reflect parasympathetic tone, sympathetic tone and the balance between the two systems or the total variation [9]. In line with a previous study [6], the most appropriate HRV variables are considered to be Standard Deviation of Normalto-Normal heart beats (SDNN; time domain) and the Low Frequency component (LF) derived from the frequency method (0.04–0.15 Hz, extracted based on the fast Fourier transformed spectrum). HRV values were extracted according to suggested standards [9] (From 5-minute RR interval time series). Measurements were performed at 6 pm, 10 pm, 2 am and 6 am (the following day) as patients were scheduled to rest at these time points. HRV analysis was performed in Kubios-HRV software (University of Eastern Finland, Kuopio, Finland). RR time series (the tachogram) were artefact corrected using the correction algorithm in the Kubios-HRV software (correction setting set at ‘very strong’ correction). Five minute windows with a stable heart rate (adequate artefact correction, no ‘paroxysmal bursts’, no upward or downward trend in the RR interval) were chosen and visually inspected by one of the authors (STV). Statistics The demographic and basic clinical data are reported as the median, interquartile range (IQR) and range. HRV was summarized for each patient defined as the geometric mean from the available HRV values at the four time points. The EFA scores on admission and follow-up and FIM scores at follow-up were compared to corresponding LF and SDNN admission and follow-up scores (admission FIM scores were not compared to admission HRV as 75% of the patients had admission FIM scores 5 21). Additionally, the changes in HRV between admission and follow-up (DLF and DSDNN, both absolute and relative) were compared to the corresponding changes in the EFA and FIM (DEFA and DFIM). Finally, admission HRV data were compared with follow-up EFA, follow-up FIM and DEFA and DFIM to investigate the prognostic value of HRV. The relationship between HRV measures and clinical measures were analysed using linear least squares regression when appropriate. The regression models were evaluated by inspecting QQ plots and a plot of the residuals. HRV variables were log transformed before comparing HRV to clinical scores. Spearman rank correlation analysis was performed if regression analysis assumptions were violated.
376
S. T. Vistisen et al.
One-tailed paired t-tests were used to compare admission LF and SDNN with follow-up LF and SDNN. Changes in cardiopulmonary variables (heart rate, respiratory frequency, diastolic blood pressure and systolic blood pressure) between admission and follow-up were analysed using a two-tailed paired t-test. As a secondary analysis, this study also analysed heart rate’s (HR) relation to clinical measures using the same statistical approach explained for HRV variables. On admission, through the four time points, some patients had variable HRV whilst others had regular HRV. Due to this, a secondary analysis was performed and investigated the relationship between the variation in HRV (standard deviation of the available HRV variables for the four time points) and clinical scores. One patient was excluded from this analysis as the HRV was only recorded at one of the four time points. Regarding half-circadian rhythm analysis at follow-up, a one-tailed paired t-test was used to compare LF and SDNN at 10 pm with 2 am and 6 am (as previously shown for admission data). The data were log-transformed as following inspection of QQ plots. Statistics were performed in Stata (StataCorp LP, College Station) or Matlab (MathWorks Inc., Natick); p50.05 was considered significant.
Results Of 49 patients with admission HRV data, 19 patients were evaluated for HRV at follow-up (see flow chart in Figure 1). Admission EFA scores of patients included and lost to follow-up were inspected for possible selection bias (see Electronic Supplement, Figure S1). Patient characteristics, including the clinical measures at admission and follow-up, are shown in Table I Patient follow-up occurred 35 days (median) after admission, interquartile range ¼ 31–39 days. For two patients, follow-up occurred after more than 45 days (63 and 97 days
Brain Inj, 2015; 29(3): 374–379
post-admission). The median time lag between collection of the clinical scores and ECG data at follow-up was 1 day (ECG obtained 1 day after clinical score); interquartile range ¼ 4–0 days. Two patients had an absolute time lag of more than 4 days (8 and 10 days). Correlation data and corresponding statistics are shown in Figures 2 and 3, as well as the Supplemental Material. Based on analysis of the residuals, approximately half of the least squares regression analyses were deemed inappropriate. Therefore, Spearman rank correlation was used for all analyses and corresponding discussion. Despite this, Pearson regression analyses have also been reported. On admission, EFA scores were not statistically significantly correlated to LF or SDNN (pSDNN ¼ 0.27 and pLF ¼ 0.23, see Figure 2). At follow-up, EFA and FIM were significantly correlated to SDNN and LF (pSDNN, EFA ¼ 0.005 and pLF, EFA ¼ 0.009; pSDNN, FIM ¼ 0.027 and pLF, FIM ¼ 0.026; see Figure 3 and Supplemental Figure S2). Relative and absolute DSDNN and relative DLF (but not absolute DLF) were significantly correlated to DEFA (pabsolute DSDNN, DEFA ¼ 0.023 and prelative DSDNN, DEFA ¼ 0.041; pabsolute DLF, DEFA ¼ 0.10 and prelative DLF, DEFA ¼ 0.037. DFIM was not correlated to DSDNN and DLF (pabsolute DSDNN, DFIM ¼ 0.07 and prelative DSDNN, DFIM ¼ 0.14; pabsolute DLF, DFIM ¼ 0.16 and prelative DLF, DFIM ¼ 0.095, see Supplemental Figures S3 and S4). Heart rate was not related to EFA at follow-up, but was significantly correlated to EFA on admission (p ¼ 0.01). Additionally, HR changes were related to changes in EFA (p ¼ 0.005; see Figure S5). On admission, the standard deviation of the four (in four patients, only three) HRV measures were negatively Table I. Patient characteristics. n Male gender (n ¼ 19) 10 ABI diagnosis TBI 5 SAH 5 Stroke 5 Anoxic 4 Diabetes Type 1 0 Type 2 0 Cardiac drug changes 6 between admission and follow-up (n) Ca+ antagonists added 1 Beta blockers added 2 Anti-hypertensive 1 drug added Ca+ antagonists withdrawn 1 ACE inhibitors withdrawn 1 Intrathecal baclofen pump changes Added 1 Age 55 Days since injury, 37 admission 2 BMI (kg m ), 22.9 admission EFA, admission 46 FIM, admission 18 RLAS, admission 3 EFA, follow-up 63 FIM, follow-up 23 RLAS, follow-up 5
Figure 1. Flow chart of included patients with admission HRV data.
IQR, Interquartile range.
Median
IQR
38–69 31–44
23–78 21–110
20.7–24.6 41–56 18–20 3–5 43–86 18–57 3–6
16.4–30.3 26–71 18–37 2–6 31–95 18–82 2–7
Range
Sensory-motor and autonomic nervous system
DOI: 10.3109/02699052.2014.969312
logSDNN and EFA score on admission
logLF and EFA score on admission
(B)
100
100
90
90
80
80
70
70
EFA score
EFA score
(A)
60
50
60
50
40
40
30
30
20
r =0.3465, p
Pearson
Rho =0.1621, p
Spearman
1
1.5
2 2.5 3 logSDNN (ms))
20
=0.1209
r =0.3716, p
Pearson
=0.2657 3.5
Rho =0.176, p
=0.08
Spearman
4
377
1
2
3
4 5 logLF (ms2))
=0.2265 6
7
Figure 2. Scatter plots of admission EFA score and admission HRV variables.
correlated to the EFA score (pSDNN ¼ 0.009 and pLF ¼ 0.02; see Figure S6), but follow-up standard deviation of LF and SDNN were not related to follow-up EFA and FIM scores (p-value range ¼ 0.13–0.89). No admission HRV variables were able to predict the clinical measurements at follow-up or their change from admission to follow-up, i.e. admission HRV measures were not correlated to follow-up clinical scores or changes in clinical scores. SDNN (p ¼ 0.020) and LF (p ¼ 0.038) increased from admission to follow-up (SDNN on admission: 11.4 milliseconds, Confidence Interval (CI) ¼ [8.4; 15.4] milliseconds vs. SDNN at follow-up: 15.9 milliseconds, CI ¼ [12.6; 20.0] milliseconds; LF on admission: 22.4 ms2, CI ¼ [11.7; 43.0] ms2 vs. LF at follow-up: 42.9 ms2, CI ¼ [23.6; 77.9] ms2). Cardiopulmonary variables either significantly decreased or had a tendency to decrease, see Table II Half-circadian rhythm: SDNN at 10 pm was lower than SDNN at 2 am (p ¼ 0.017), but not SDNN at 6 am (p ¼ 0.28). LF at 10 pm was not significantly different from LF 2 am (p ¼ 0.069) or LF at 6 am (p ¼ 0.19).
Discussion The current study aimed at quantifying cardiac autonomic regulation and its longitudinal changes in patients with heterogenic ABI diagnoses admitted for neurorehabilitation.
In particular, this is unknown for other diagnoses than TBI. The study demonstrates that the investigated HRV variables (SDNN and LF) at follow-up were significantly correlated to the EFA and FIM; however, admission HRV was not significantly correlated to admission EFA. In addition, changes in HRV variables from admission to follow-up were generally significantly correlated to corresponding changes in the EFA score but not FIM score. Furthermore, HRV increased from admission to follow-up and the analysed cardiopulmonary variables significantly decreased or tended to decrease. HR was related to EFA on admission and HR changes were also associated with EFA changes. Therefore, the current study indicates a reduction in sympathetic tone from admission to follow-up and an improvement in the autonomic regulation of HR. Additionally, the data generally indicate an association between measures of the autonomic nervous system and clinical measurements of cognitive and somato-sensory function (EFA and FIM). The patients in the current study were severely brain injured (75% of patients had FIM 521 on admission) and likely encountered the floor effect for the FIM. Therefore it was not surprising that the EFA score (being more sensitive when evaluating these patients [7,8,10]) demonstrated the best relationship to HRV variables. Follow-up scatter plots (Supplemental Figure S2) demonstrated that nine of 19 patients had FIM scores 20 throughout the study period and,
S. T. Vistisen et al. logSDNN and EFA score at follow−up
(A)
EFA score
Brain Inj, 2015; 29(3): 374–379
logLF and EFA score at follow−up
(B)
100
100
90
90
80
80
70
70
EFA score
378
60
50
60
50
40
40
30
30
20
r =0.642, p
Pearson
1.5
r =0.6458, p
Pearson
Rho =0.619, p
=0.0047
2 2.5 3 logSDNN (ms)
3.5
Spearman
10 1
20
=0.0417
Rho =0.5795, p
=0.0394
Spearman
4
10 1
2
3
4 5 logLF (ms2)
=0.0093 6
7
Figure 3. Scatter plots of follow-up EFA score and follow-up HRV variables.
Table II. Summary of cardiopulmonary clinical variables from admission to follow-up. Variable 1
HR (min ) Respiratory frequency (min1) Diastolic blood pressure (mmHg) Systolic blood pressure (mmHg)
Admission
Follow-up
94 20 86 139
81 18 80 127
(23) (6.3) (14) (23)
(16) (7.7) (10) (18)
p Value 0.008 0.08 0.08 0.06
Data are presented as mean (standard deviation).
therefore, the EFA is sensitive to detect functional progression and in turn relationships with HRV improvements in these patients [8]. This could be the reason Keren et al. [4] demonstrated no relationship between HRV variables and FIM. In accordance with the previous findings, it was found that the HRV variable, SDNN, at follow-up was statistically significantly lower at 10 pm than at 2 am. LF had a nonsignificant tendency to reduce from at 10 pm than at 2 am (p ¼ 0.069). There was a trend when comparing HRV variables at 10 pm with HRV variables at 6 am and, although non-significant, the trend was consistent with a previous finding. This is likely due to the loss of patients to follow-up. To summarize, the half circadian pattern analyses of patients on admission and follow-up are comparable to healthy
subjects, although HRV variables are generally lower in patients with ABI than in healthy subjects [6]. There was a significant relationship between admission EFA score and the standard deviation of four time measurements of the HRV variables on admission. This was not observed during follow-up and the admission relationship is merely regarded as a hypothesis generating finding. Limitations There are a number of limitations to the current study. First, the follow-up study population had high loss to follow-up, which may lead to selection bias. Data from 19 of 49 patients were included. Eight of the ‘lost to follow-up’ group were discharged prior to scheduled follow-up. Patients discharged within a few weeks of admission are usually discharged as they (by clinical judgement) are either in a too poor condition and have little potential for rehabilitation (and admitted to e.g. a nursing home) or in a too good condition (and typically admitted to less intensive rehabilitation). This tendency might be present in the data (Supplemental Figure S1) and, although speculative, these patients could develop differently in the outcome measures assessed than the included patients. It is unlikely that the other main reasons for exclusion (follow-up registration not initiated (6), logistic reasons (7) and tachogram quality inadequate (5)) were a potential source
Sensory-motor and autonomic nervous system
DOI: 10.3109/02699052.2014.969312
for selection bias, neither theoretically nor indicated by admission EFA scores. Second, it was assumed that patients were resting during their scheduled resting periods; however, the patients were not observed during their ECG recordings over these times. Third, patients treated with cardiac drugs have been included, some of whom had changes in drug administration during the study. Drugs affecting the cardiovascular system like -blockers have previously been shown to be associated with improved parasympatic modulation and reduced heart rate [11]. In spite of this possible beneficial intervention, no difference was found in HRV parameters between patients receiving cardiac drugs compared to those not receiving cardiac drugs [6]. Fourth, given the observational nature of the study, different ECG equipment was used for detection of RR intervals on admission and follow-up. This is mainly an issue regarding interpretation of the statistically significant results that SDNN and LF increase from admission to follow-up. The RR interval detection method on admission offers a precision of 1 millisecond. This is identical to the reported detection from the manufacturer of the system used at follow-up. Nonetheless, the two technologies have not been validated against each other (by e.g. Bland-Altman analysis) and the conclusion that SDNN and LF increases in the rehabilitation period should be treated with caution. Despite this the data are consistent with the trends observed by Keren et al. [4] for all four investigated HRV variables in that study. Also, the observational nature of the study gave rise to variability between the timing of HRV recordings and the clinical evaluation of patients. Yet, it is speculated that this impacts only slightly if at all the discussion of the present data. Fifth, the analyses in the current study are based on data derived from visually inspected tachograms, which reduces external validity. Still, it is considered a pre-requisite to visually inspect tachograms from patients with ABI to extract reliable HRV data. Finally, in healthy subjects, HRV variables depend on the underlying HR and age and it has been suggested to correct for HR and age [11]. Given the low number of patients, the residual problems with linear regression (where a correction would typically be performed) and the unknown impact of HR and age on HRV for patients with ABI, this study only reported the HR in relation to EFA (in Supplemental Material). Given these reported HR data, the associations between clinical scores and HRV might be partly explained by HR itself and not HRV alone. Nonetheless, HR and HRV indicate that sympathetic activity and its development are related to clinical scores and their development. In conclusion, the autonomic nervous system and its function, as represented by HRV, appear to be related to the clinically evaluated sensory-motor and cognitive function of
379
neurorehabilitation patients and that change in HRV is related to changes in the clinical measures. This observational study offers insights to the autonomic nervous function, an area rarely addressed in neurorehabilitation. However, as the study is observational and has a number of limitations, the results need to be confirmed in more controlled settings on a larger set of patients.
Acknowledgements The authors thank Peter W. Stubbs for proofreading this manuscript.
Declaration of interest The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.
References 1. Baguley IJ. Autonomic complications following central nervous system injury. Seminars in Neurology 2008;28:716–725. 2. Baguley IJ, Heriseanu RE, Cameron ID, Nott MT, Slewa-Younan S. A critical review of the pathophysiology of dysautonomia following traumatic brain injury. Neurocritical Care 2008;8:293–300. 3. Perkes I, Baguley IJ, Nott MT, Menon DK. A review of paroxysmal sympathetic hyperactivity after acquired brain injury. Annals of Neurology 2010;68:126–135. 4. Keren O, Yupatov S, Radai MM, Elad-Yarum R, Faraggi D, Abboud S, Ring H, Groswasser Z. Heart rate variability (HRV) of patients with traumatic brain injury (TBI) during the post-insult sub-acute period. Brain Injury 2005;19:605–611. 5. King ML, Lichtman SW, Seliger G, Ehert FA, Steinberg JS. Heartrate variability in chronic traumatic brain injury. Brain Injury 1997; 11:445–453. 6. Vistisen ST, Hansen TK, Jensen J, Nielsen JF, Fleischer J. Heart rate variability in neurorehabilitation patients with severe acquired brain injury. Brain Injury 2014;28:196–202. 7. Heck G, Steiger-Ba¨chler G, Schmidt T. Early functional abilities (EFA) – eine skala zur evaluation von behandlungsverla¨ufen in der neurologischen fru¨hrehabilitation. Neurology & Rehabilitation 2000;6:125–133. 8. Stubbs PW, Pallesen H, Pedersen AR, Nielsen JF. Using EFA and FIM rating scales could provide a more complete assessment of patients with acquired brain injury. Disability and Rehabilitation 2014. DOI: 10.3109/09638288.2014.904935. 9. [Anonymous]. Heart rate variability. Standards of measurement, physiological interpretation, and clinical use. Task force of the European Society of Cardiology and The North American Society of Pacing and Electrophysiology. European Heart Journal 1996;17: 354–381. 10. Alvsaker K, Walther SM, Kleffelgard I, Mongs M, Draegebo RA, Keller A. Inter-rater reliability of the early functional abilities scale. Journal of Rehabilitation Medicine: Official Journal of the UEMS European Board of Physical and Rehabilitation Medicine 2011;43: 892–899. 11. Tsuji H, Venditti Jr FJ, Manders ES, Evans JC, Larson MG, Feldman CL, Levy D. Determinants of heart rate variability. Journal of the American College of Cardiology 1996;28:1539–1546.
Supplementary material available online Supplementary Figures 1–6
Copyright of Brain Injury is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
