Abstract
Background
The presence or absence of a motor evoked potential (MEP) in the post-stroke hemiparetic limb has been recommended by rehabilitation experts as a predictive biomarker which is ready for use in clinical trials. However, evidence remains limited for its prognostic value in the chronic stage.
Objective:
Determine if MEP status (MEP+ or MEP−) obtained within 1 week of starting treatment (baseline) predicts the magnitude of response to intervention in individuals with chronic, moderate–severe hemiparesis.
Methods
This is a retrospective analysis using data from a single-blind randomized controlled trial. Seventy-six individuals ≥6 months post-stroke with a baseline Fugl-Meyer Assessment of the Upper Extremity (FMUE) score of 23 to 40 underwent 30 hours of upper limb (UL) training over 6 weeks. Participants were stratified by baseline MEP status. The primary endpoint was change in FMUE score from baseline to post-test.
Results
Seventy-three participants provided FMUE scores and MEP status at baseline. Individuals who were MEP+ (n = 49) demonstrated a mean FMUE change score of 5.09 (standard deviation [SD] = 3.8) while MEP− (n = 24) individuals demonstrated a mean change score of 5.04 (SD = 4.0). There were no significant differences between the groups (mean difference = 0.05, P = .96, 95% confidence interval [−1.99, 2.09]).
Conclusions
Our results demonstrate that MEP status at the start of an intervention in the chronic stage does not predict recovery for people with moderate–severe UL impairments. This finding directly challenges recent expert recommendations to stratify trial groups by MEP status, suggesting that such stratification may not effectively reduce variability or predict treatment response at the chronic stage.
Clinical Trial Registration:
ClinicalTrials.gov, ID: NCT03517657.
Introduction
Rehabilitation of the post-stroke paretic upper limb (UL) represents a particularly complex and challenging area of research, characterized by neutral or failed results of several large-scale intervention trials over the last 2 decades.1-4 Three rigorous reviews attempting to explain these disappointing findings point primarily to flawed methodological designs.5-7 More specifically, they highlight that although stroke is a heterogeneous condition, clinical trials to date have largely taken approaches that fail to adequately account for the variability in patient responses to a given treatment. This leads to reduced statistical power, a decreased likelihood of detecting a treatment effect, and a reduced understanding of which patients stand to benefit from a particular therapeutic strategy.
To address methodological shortcomings and guide the development, execution, and reporting of stroke rehabilitation research for improved outcomes, a panel of experts has issued a series of papers outlining best practices. 6 Within these guidelines is the recommendation that clinical trials stratify groups based on their potential for neurobiological recovery using available biomarkers. 8 Amongst the biomarkers for motor recovery they deemed ready for use in clinical trials is the presence or absence of a motor evoked potential (MEP+/−) from the paretic UL, elicited using transcranial magnetic stimulation (TMS). The presence of a MEP in the affected limb is indicative of some amount of sparing of the corticospinal tract (CST), and a MEP+ response immediately following a stroke in the acute phase (1-7 days post-stroke) is associated with a better prognosis.9,10
Evidence supporting the MEP’s prognostic value comes primarily from studies in the acute phase which stratify participants by their baseline MEP status and subsequently follow their spontaneous recovery over time. Evidence for the utility of a MEP obtained in the chronic stage (defined as ≥6 months post-stroke) 6 as a predictor of response to treatment is more limited. This is illustrated by a systematic review of 71 studies on neurological biomarkers of post-stroke motor recovery which reported that 8 studies investigated spontaneous UL recovery by MEP status in the acute phase, but only 1 study tested the predictive value of a MEP after an UL intervention in the chronic phase.11,12 While other studies within the review claimed to investigate the prognostic value of MEP status in chronic populations, they included participants who were only 2, 13 3, 14 and 4 15 months post-stroke and therefore do not meet criteria for chronic as defined by the Stroke Recovery and Rehabilitation Roundtable (SRRR). 6
Within the review, the 1 study which included participants ≥6 months post-stroke demonstrated that MEP status was only helpful in making a prediction about treatment response when combined with other measures of spared function (fractional anisotropy of the internal capsule). 12 Beyond the review, a more recent study in chronic individuals by Powell et al 16 found no significant differences in outcomes when participants were analyzed by treatment condition and MEP status. 16 These results do not support the recommendation to stratify groups by MEP status in the chronic phase as a means to reduce variability and predict treatment response. However, the limited number of studies on this topic highlights the need for additional research to determine the utility of MEP stratification in chronic stroke rehabilitation trials.
Our study aimed to clarify the extent to which stratification by MEP status is predictive of response to treatment by examining data from a randomized clinical trial focused on restoring movement in the moderate–severely impaired UL in a chronic population. Understanding this relationship will inform the design of future UL rehabilitation trials to ultimately optimize outcomes for individuals in the chronic phase of stroke. The randomized clinical trial was one of the largest chronic stroke rehabilitation clinical trials to date (n = 76), and we demonstrated a significant treatment effect across all study participants. Thus, we are well-positioned to examine this relationship. We conducted a retrospective analysis to assess whether the presence of a MEP in the paretic UL obtained at baseline testing (within 1 week prior to the start of the intervention) can predict the magnitude of response to treatment.
Methods
Clinical Trial Overview
The current study analyzed data from a randomized, single-blind, clinical trial for rehabilitation of the moderate–severely impaired UL in patients with chronic stroke. 17 The trial took place at a university and rehabilitation hospital in Chicago, IL in the United States. Study procedures were approved by the Northwestern University Institutional Review Board. The study was done according to the Declaration of Helsinki and written informed consent was obtained from all participants at the time of enrollment. This trial is registered at ClinicalTrials.gov (NCT03517657) and is closed to participants.
Seventy-six participants were recruited to the study. Participants were deemed eligible if they met the following criteria: (1) at least 18 years old, (2) unilateral stroke ≥6 months prior to enrollment, (3) moderate–severe impairment in the paretic UL defined as 23 to 40 on the Fugl Meyer Assessment for the Upper Extremity (FMUE), and (4) ≤3 on the Modified Ashworth Scale. Participants were excluded for: (1) contraindications to TMS, (2) Botox in the affected arm or hand 6 months prior to enrollment, (3) concomitant neurological diagnosis, and (4) a score on the mini mental cognitive screen of <21.
Intervention
Participants were stratified by high (30-40) and low scores (23-29) on the FMUE and subsequently randomized to receive either bilateral motor priming (BUMP) of the UL or control priming (CP and sham transcutaneous electrical stimulation) prior to an UL task specific training (TST) protocol. Briefly, BUMP is a repetitive sensorimotor technique consisting of mirror symmetric, bilateral, wrist flexion, and extension movements using a device with a mechanical linkage that ensures the unaffected and affected hands move in exact symmetry. The intervention consisted of 30 hours of treatment (priming + TST) delivered in 15 days over 6 weeks. At each visit, the participant completed 15 minutes of their assigned priming method followed by 45 minutes of TST. They had a 30-minute break and then repeated this schedule. Additional details regarding the intervention can be found in the published protocol paper. 17
Outcome Measures
Behavioral Assessment
Two days of testing were scheduled for each assessment timepoint including: baseline, post-intervention, and follow up, 8 weeks after treatment ended. One day was allocated for behavioral testing and a second day was designated for TMS. The primary outcome measure was change on the FMUE from baseline to post-test. The FMUE is an impairment scale with established interrater and intrarater reliability that addresses both synergistic and isolated movements of the UL. Research has estimated the minimum clinically important difference (MCID) of FMUE change scores to be within the range of 4.25 to 7.25 points for people with moderate and chronic stroke. 18 We also report the percentage of individuals in each group who achieved a clinically meaningful change in FMUE score at post-test compared to baseline. A clinically meaningful change is defined as an improvement of 6 points or more, based on previous research. 18
TMS Assessment
Corticomotor excitability was measured using TMS. Muscle activity was recorded from the extensor carpi radialis (ECR) of the paretic forearm with surface electromyography (EMG). The ECR was selected because it is frequently impaired following a stroke and its recovery is critical to improved hand function. 19 Magnetic stimuli were delivered using the MagStim 200 and a focal figure-of-8 coil (wing diameter 9 cm). A hotspot for the ECR was determined by incrementally repositioning the coil over M1 of the hemisphere contralateral to the paretic limb (ie, the ipsilesional hemisphere) and stimulating beginning at 50% of maximal stimulator output (MSO) to identify a facilitatory EMG response at rest. 20 To detect the MEP, the stimulator intensity was incrementally increased as high as 100% MSO if the participant could tolerate it. If a MEP response was not detectable at rest, participants were asked to extend their affected wrist. Participants for whom a resting or active MEP could be elicited at or below 100% MSO at the baseline TMS appointment were labeled MEP+; those who did not have a resting or active MEP at intensities up to 100% MSO at baseline were labeled MEP−.
Data Analyses
The clinical trial examined between-group (ie, BUMP and CP) differences in FMUE change scores from baseline to follow-up and found that: (1) there were no statistically significant between-group differences in FMUE change scores; (2) both groups improved to a similar extent, exceeding the MCID for the FMUE; (3) the greatest magnitude of change occurred from baseline to post-intervention; and (4) there were no significant changes in FMUE scores from post-intervention follow-up for either group. Given that the most substantial behavioral changes occurred between baseline and post-intervention for both groups, we focused our current analysis on this period to examine how the behavioral changes relate to MEP status (MEP+ or MEP−) at baseline.
Subject characteristics, including age, sex, date and type of stroke, and UL impairment level (FMUE score) were collected at the baseline evaluation and summarized using appropriate descriptive statistics (mean ± standard deviation [SD], count/%). A Shapiro–Wilk test was applied to residuals to assess the normality of the data distribution. Two-sample t-tests were used to determine differences in baseline demographics including age, time since stroke and UL impairment across MEP status groups.
We assessed the pairwise relationship between MEP status (+/−) and change in FMUE total score at post-test from baseline using 2-sample t-tests and point-biserial correlation. We further evaluated the bivariate relationship between MEP status and clinically meaningful response (≥6 point improvement at post-test compared to baseline FMUE) using Fisher’s exact test. Further, to assess the relevance of MEP status in analytic models used to estimate treatment effects from the trial, separate generalized linear models (GLM) were fit for change in FMUE (identity link) and presence of a clinically meaningful improvement in FMUE (logit link). For each outcome, we fit 2 models: 1 including study arm (BUMP vs CP) and another including both study arm and MEP status. We report all relevant point estimates and standard errors (SEs) for GLM; unless otherwise noted, all tests were 2-sided and assume a significance level of α = .05. To handle a small amount of missing post-treatment FMUE scores (5/73 participants), we leveraged multiple imputation (MI) with 40 imputations using predictive mean matching that included baseline FMUE, treatment arm, and MEP status as predictors. Complete case analyses were also performed. There were no differences in results, so we report the planned intention-to-treat results with MI here. This method is consistent with that of several other large-scale stroke intervention trials,2,21,22 as well as how we report the primary endpoint of our clinical trial. 23 All analyses were performed using R version 4.3.2.
Results
Seventy-three participants completed baseline behavioral and TMS assessments and were included in this analysis. Of these participants, N = 49 exhibited a MEP+ response and N = 24 exhibited a MEP− response at baseline. MEP status was almost perfectly balanced across study arm (BUMP = 25/37 MEP+; CP = 24/36 MEP+). Baseline characteristics and demographics are shown in Table 1. All variables met the assumption of normality based on the Shapiro–Wilk test (P < .05). There were no statistically significant differences in age or time since stroke between MEP+ and MEP− groups. There was a difference in impairment level at baseline between groups. The mean baseline FMUE score for the MEP+ group was 31.45 (SD = 5.05) and the mean score for the MEP− group was 26.88 (SD = 4.67) with a between group point difference of 4.57 (P < .001).
Participant Characteristics.
Values are frequency counts and means (standard deviation).
FMUE change scores had almost no bivariate relationship with MEP status: the point-biserial correlation was r = .01 [−0.23, 0.24] and there was no statistically significant difference in FMUE change scores between the MEP+ group (mean change score = 5.09, SD = 3.8) and the MEP− group (mean change score = 5.05, SD = 4.0), with a between-group differences of .05 (P = .96). See Figure 1. Based on frequency counts, 45% of the MEP+ group and 47% of the MEP− achieved a clinically meaningful improvement of ≥6 points on the FMUE from baseline to post-test (P = .34).
Individual and mean imputed change scores on the FMUE for groups that were MEP+ or MEP− at baseline are shown. The mean imputed change score is marked with a bolded line and the box is representative of the standard deviation.
GLM revealed that MEP status at baseline included no additional information in the estimation of treatment effects (Table 2). MEP status was not a statistically significant predictor of either outcome in GLM with coefficients very near zero. Moreover, it did not appear to explain any variation in outcomes as SEs of treatment effects increased after inclusion of MEP status, indicating that its inclusion as a predictor hindered the efficiency of effect estimates.
Linear Models Report Point Estimate (SE) and P-value.
Abbreviations: SE, standard error, FMUE, Fugl-Meyer Assessment of the Upper Extremity.
Discussion
The results of our study showed that individuals with moderate–severe chronic stroke respond equally to an UL rehabilitation intervention regardless of whether they are classified as MEP+ or MEP− prior to treatment. Both groups demonstrated a mean improvement that surpassed the MCID for the FMUE. Additionally, both groups had similar proportions of participants achieving an improvement of 6 or more points. These findings do not support the guidelines issued by the SRRR in their 2017 consensus paper which stated “We recommend that future studies of UL interventions determine whether patients are MEP+ or MEP− for the purposes of stratification.” 8 Our findings challenge the current consensus on the utility of MEP status as a stratification factor in chronic stroke recovery, particularly for those with moderate–severe hemiparesis. Alongside more recent evidence, 16 they indicate that such stratification may not provide the anticipated predictive value for recovery in the chronic phase. MEP status was not marginally associated with change in FMUE scores, nor did it improve the efficiency of treatment effect estimates in this trial. These insights are important for the design of future chronic stroke rehabilitation trials.
In combination with clinical measures, MEP status in the acute stage can be a valuable prognostic biomarker for estimating spontaneous recovery potential. 24 However, its utility in the chronic stage to predict treatment response has, thus far, remained an open question. Examining the evidence cited by the SRRR to support the MEP as a biomarker of recovery potential in chronic stroke reveals that only one of the studies includes a set of patients who all meet criteria as chronic (≥ 6 months from stroke), as defined by the same group.6,8 This study, conducted by Stinear et al., investigated recovery outcomes for 17 chronic stroke survivors following a 30-day UL motor training program. 12 They found that MEP status alone is insufficient for predicting UL functional recovery. To have predictive value for individuals who are MEP− at baseline, MEP status needed to be paired with diffusion tensor imaging (DTI) to evaluate CST structural integrity. Although the SRRR acknowledges there is increased power for effect detection when MEP status is combined with other measures of spared connectivity, they stop short of recommending using these measures together in their guidelines. If MEP status on its own does not meaningfully predict recovery in the chronic phase, its use as a stratification factor may result in analyses that are less powerful or precise and introduce unnecessary complexity in study design without improving the accuracy of outcome predictions.
At first glance, it seems logical to expect that the presence of a MEP from the paretic UL prior to intervention would be a good predictor of treatment response. Although the precise neurophysiological mechanisms underlying MEPs induced by TMS are not fully understood, previous studies have established MEPs as a valuable method for assessing changes in CST excitability.25,26 It is possible that the presence or absence of a MEP could serve as a physiological marker of the excitability of pre-synaptic inputs and post-synaptic components of CST neurons activated by TMS. A decrease in excitability may reflect a decrease in the overall structural integrity of the CST, in which the ability of the motor cortex to transmit signals through the spinal cord and to the muscles is disrupted. Presumably, this disruption would have an aberrant effect on the ability to perform voluntary movements and limit recovery potential. However, a direct causal relationship between MEP status and voluntary motor recovery in the chronic stage has yet to be demonstrated. 27 Further, whether greater CST structural integrity leads to increased recruitment of CST fibers and a correspondingly larger or more consistent MEP response has not yet been established. Small studies conducted in the chronic population have produced contradictory results; while some found a correlation between measures of CST structural integrity using MRI and CST excitability assessed via the MEP, others did not.28-31 It is probable that some degree of CST preservation is critical for treatment-induced UL recovery to occur in the chronic stage. However, if the ipsilesional MEP (CST excitability) cannot sufficiently capture the degree of preserved descending connectivity (CST structural integrity) and the presumed capacity for recovery along the CST, then its predictive value as a marker of treatment response is significantly diminished. Future studies in a large cohort of individuals with chronic stroke, which incorporate both CST structural and excitability measures alongside behavioral recovery outcomes may clarify this relationship and lead to a better understanding of the pathways that drive recovery in this population.
To avoid potential pitfalls in trial design and to sufficiently account for the heterogeneity of patients’ response to treatment, viable biomarkers for motor recovery potential in the chronic phase are desirable. While biomarkers that predict functional recovery could transform treatment strategies by guiding individualized and targeted interventions, those treatment strategies must be based on robust and validated biomarkers to truly maximize patient outcomes and improve quality of life. However, given the substantial cost and effort associated with incorporating biomarkers into study design, the decision to implement them should be based on the best available evidence. On this basis, we assert that the recommendation by the SRRR to stratify chronic participants by MEP status in clinical trials lacks empirical support. Our results, along with those of others, 16 suggest that stratification by MEP status alone fails to provide predictive power regarding who will respond to an intervention for the moderate–severely impaired UL in the chronic phase post-stroke. Therefore, reliance on MEP status as a biomarker for stratification is not only unwarranted but could potentially lead to further inefficiencies in analyses and trial designs.
Conclusions
Our results suggest that baseline ipsilesional MEP status in individuals with chronic, moderate-to-severe post-stroke UL motor impairments does not reliably predict treatment response. Most rehabilitation intervention trials are conducted in the chronic phase. As such, the recommendation to stratify by MEP status is likely to have significant implications for the design of future UL rehabilitation research studies and the outcomes for the participants involved. Moving forward, it is crucial for future researchers to reconsider using MEP status alone to stratify participants. Future research should focus on identifying appropriate biomarkers relevant specifically for moderate–severely impaired individuals in the chronic phase of recovery.
Footnotes
Acknowledgements
We thank participants for their contribution to this research project. We would like to thank the National Institutes of Health, Northwestern University, University of Illinois at Chicago, Shirley Ryan AbilityLab, and University of Chicago for their support, grants, and funding of ongoing clinical trials for advancement of clinical practice and education.
Author Contributions
Erin C. King: Conceptualization; Investigation; Project administration; Writing - original draft; Writing - review & editing. Michael Trevarrow: Formal analysis; Investigation; Writing - original draft; Writing - review & editing. Sebastian Urday: Investigation; Writing - review & editing. Jacob M. Schauer: Formal analysis; Methodology; Writing - review & editing. Daniel M Corcos: Conceptualization; Investigation; Supervision; Writing - original draft; Writing - review & editing. Mary Ellen Stoykov: Conceptualization; Funding acquisition; Investigation; Methodology; Writing - review & editing.
Data Availability
Data collected for the study, including deidentified individual participant data, study protocol, and statistical analysis plan will be available upon the approval of the proposal by the principal investigator. Requests should be sent to the corresponding author.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research is funded by National Institutes of Health (1RO1HD091492). ECK is supported by 1F31HD111318-01.
Ethical Considerations and Participant Consent
Study procedures received ethical approval by the Northwestern University Institutional Review Board. The study was done according to the Declaration of Helsinki and written informed consent was obtained from all participants at the time of enrollment by a study team member.
Consent for Publication
Not applicable.
