Utility of Transcranial Magnetic Stimulation and Diffusion Tensor Imaging for Prediction of Upper-Limb Motor Recovery in Acute Ischemic Stroke Patients.

Background: The recovery of the upper-limb (UL) motor function after ischemic stroke (IS) remains a major scientific, clinical, and patient concern and it is hard to predict alone from the clinical symptoms. Objective: To determine the accuracy of the prediction of the recovery of UL motor function in patients with acute ischemic middle cerebral artery (MCA) stroke using individual clinical, transcranial magnetic stimulation (TMS) or diffusion tensor imaging (DTI) parameters or their combination. Methods and Material: The first-ever acute ischemic MCA stroke patients within 7 days of the stroke onset who had an obvious UL motor deficit underwent TMS for the presence of motor-evoked potential (MEP) and DTI to evaluate the integrity of corticospinal tracts. Multivariate logistic regression analysis was done to test for the accuracy of the prediction of the recovery of UL motor function.
Results: Twenty-nine acute ischemic MCA stroke patients (21 males and 8 females) with a mean age of 51.45 ± 14.26 years were recruited. Model-I included clinical scales (Fugl-Meyer Assessment [FMA] + Motricity Index [MI]) + TMS (MEP) + DTI (fractional anisotropy [FA]) were found to be the most accurate predictive model, with the overall predictive ability (93.3%; 95% confidence interval [CI]: 0.87-0.99) and sensitivity: 94.9% (95% CI: 0.87-1.0) and specificity: 95.8% (95% CI: 0.89-1.0); respectively.
Conclusion: The accuracy of UL motor recovery can be predicted through the clinical battery and their elements as well as TMS (MEP) and DTI (FA) parameters. Further, well-designed prospective studies are needed to confirm our findings. Copyright:

INTRODUCTION

The prediction of the upper-limb (UL) motor recovery in post-stroke patients with UL weakness is extremely vital when making an overall treatment plan. UL motor recovery following an ischemic stroke (IS) is highly variable and hard to predict only through the clinical symptoms. The clinical scales including the Fugl-Meyer Assessment (FMA), Action Research Arm Test (ARAT), and Motricity Index (MI); the neuroimaging tools such as functional magnetic resonance imaging (fMRI), diffusion tensor imaging (DTI), and neurophysiological techniques like transcranial magnetic stimulation (TMS) parameters have been reported to be used to measure the extent of the damage to the motor system and predict the subsequent UL recovery of the function after the IS onset, but these techniques are not yet used routinely due to lack of enough evidence.[123456789]

TMS and diffusion tensor tractography (DTT) were performed in the early stage (7–28 days) of a stroke by Kwon et al. (2011)[10] in 53 patients with complete motor weakness of the affected hand. TMS showed higher positive predictability and DTT showed higher negative predictability. Kwon et al. (2011)[10] concluded that a combined study of TMS and DTT appeared to be more advantageous in the prediction of negative motor outcomes than did every single study. Single TMS appeared to be more advantageous in the prediction of positive motor outcomes. In another published study of forty-three patients with first MCA stroke using TMS and DTT by Kim et al. (2016)[11] showed a better motor recovery and ambulatory function than the other groups at the 4-week follow-up in the patients with the presence of both measurable MEPs and a preserved corticospinal tracts (CST). To date, all the published studies had measured the baseline parameters beyond 2 weeks or more of the stroke onset and have not utilized these parameters in a combination. A limited number of studies had measured UL recovery only through the clinical scales including finger extension and shoulder abduction in the acute phase (within 1 week) of IS but TMS and DTI have not been performed within the acute phase of stroke onset.[47]

Therefore, there is an extreme need for investigating these measures either alone or in combinations for predicting the UL motor recovery during the acute phase of IS onset. The present study was performed to ascertain the accuracy of the prediction of the UL motor recovery in the acute ischemic middle cerebral artery (MCA) stroke using individual clinical, DTI, or TMS parameters or their combinations.

METHODS

This was a prospective cohort study in which the patients with the first-ever MCA IS within the acute phase (7 days of stroke onset) were recruited after their proper informed consent. The ethical clearance was obtained by the local Institutional Ethical Committee. The inclusion criteria were: (1) the first-ever acute ischemic MCA stroke, (2) hemiplegic motor weakness, and (3) no serious medical complications such as pneumonia. The exclusion criteria were: (1) a previous history of stroke, (2) patients with any implanted metal (e.g. artificial pacemaker), (3) refusal to DTI or TMS.

All the assessors were blinded to the clinical assessment, MEP, and DTI observations during the entire study. The follow-up assessment for the stroke recovery was done by a blinded third investigator at the third month post-stroke. TMS (MEP) were recorded from both the abductor pollicis brevis (APB) and biceps brachii muscle (BB) muscles in a relaxed state, and both hemispheres were examined. The diffusion tensor calculation and tractography were performed by using the DTIStudio Software version 3.03 (. Three regions of interest were drawn on a two-dimensional FA color map: medulla, middle anterior pons, and posterior limb of the internal capsule (PLIC). The demographic and comorbidities details such as age, sex, hypertension, diabetes mellitus, dyslipidemia, and cardiac disease were captured from the medical records. The recovery outcome was based on the ARAT scale; good recovery was to be considered as ARAT >10 and poor recovery ARAT ≤9. The study methods and protocols are reported in the Supplementary Appendix-1.

Sample size calculation

It was estimated that out of the 200 stroke patients, approximately 80 patients (40%) meeting the eligibility criteria would recover in 3 months. One variable per 10 patients as a prediction in the logistic regression model was considered for the sample size calculation.

Statistical analysis

Descriptive statistics were examined using a mean/median, range depending on the distribution of the data. Bivariate analysis was conducted using the Chi-square test for categorical data and t-test for continuous data. Multivariable analysis was done using logistic regression with recovery (Yes = Recovery; ARAT >10; No = Recovery; ARAT ≤9 as dependent variable[12] and (1) sex, (2) hemisphere affected, (3) mean diffusivity (MD), (4) fractional anisotropy (FA), (5) the presence or absence of motor cortex damage MEPs in the affected target muscle, (6) the days between the stroke and first assessment, (7) consciousness after stroke onset (National Institute of Health Stroke Scale [NIHSS], –42 points), (8) Barthel Index [BI], 0–100 points), (9) MI-Arm score and (11) FMA-Arm score as independent variables. The prognostic test properties were expressed as sensitivity and specificity, positive predictive value (PPV), and negative predictive value (NPV). The P value < 0.05 was considered to be statistically significant. All the statistical analysis was performed using the STATA 13.0 software.

RESULTS

Twenty-nine acute ischemic MCA stroke patients (21 males and 8 females with a mean age of 51.45 ± 14.26 years) were recruited after the screening of 386 stroke patients. Figure 1 represents the study flow diagram for the inclusion, exclusion, investigations, and follow-up details conducted in the study. The clinical assessment and TMS were done in 29 patients but only 24 patients had data on DTI. We had complete data (clinical assessment and TMS) for 24 acute MCA IS patients at baseline and 3 months of follow-up. The mean duration of the stroke onset was 5.24 ± 1.39 days. The time interval between the DTI and TMS assessments in each patient was between 03 and 36 h. All the investigations were performed within 7 days of the stroke onset with a deviation of ± 02 days.

Figure 1

Flow chart for inclusion in the study.

Table 1 shows the patients’ demographics and clinical data. Seventeen (58.6%) patients had an infarction in the left while the remaining 12 (41.4%) had an infarction in the right hemisphere. Five patients died within 3 months after the stroke because of a recurrent stroke or other comorbidities. The data were tested for normal distribution and accordingly non-parametric Mann–Whitney U test for continuous variables as well as the Chi-square test were applied for categorical variables and were finally used for the baseline and clinical characteristics analysis. Based on the logistic regression analysis, 6 out of 11 variables were significantly associated to the chance of the return of the dexterity on the ARAT scale acquired 3 months after the stroke. The highest odds ratio (OR) for the total scores of FMA (Arm), TMS (MEP) present, and DTI (FA) parameters was observed. The specificity and sensitivity of the TMS (MEP) (82.3 and 64.2%) were higher than that of DTI (FA) (71.4 and 20.0%) [Table 2].

Table 1

Baseline clinical characteristics of acute ischemic MCA stroke

Clinical Characteristics	Acute Ischemic MCA Stroke (N=29)
Sex (Male/Female); n	21/08
Age (Years); Mean±SD	51.45±14.26
Systolic BP; Mean±SD	139.8±73.2
Diastolic BP; Mean±SD	87.5±34.9
Risk Factors
Diabetes mellitus; n (%)	4 (13.7)
Hypertension; n (%)	7 (24.1)
Dyslipidemia; n (%)	2 (7)
Smoking; n (%)	6 (20.6)
Alcohol intake; n (%)	4 (13.7)
Hemisphere of stroke (Left/Right); n	17/12
Subcortical/Cortical/Cortical+Subcortical stroke; n	18/7/4
Stroke onset (Days); Mean±SD	5.24±1.39
Length of hospital stay (Days); Mean±SD	14.62±8.25
Clinical assessment (Days) from onset of stroke [n=29]	5.67±2.67
TMS sssessment (Days) from onset of stroke [n=29]	6.59±2.12
DTI assessment (Days) [n=24] from onset of stroke	4.32±3.4
NIHSS Score (0-42); median (range) at admission	11.5 (7.3-20)
BI Score (0-100) at admission; Mean±SD	50.20±25.02
MI-Arm (0-100) at admission; Mean±SD	58.5±21.9
FMA-Arm (0-52) at admission; Mean±SD	29.5±17.3
ARAT total score (0-57) at admission; Mean±SD	32.5±11.6
mRS at discharge; Mean±SD	3.58±1.13

Table 2

UL motor recovery between baseline and 3 months based on the ARAT scale

Group	Total	ARAT ≥10	ARAT ≤9
DTI (n=24)	Number	Good UL recovery	Poor UL recovery
DTT (+)	18	10	08
DTT (−)	06	04	02
	Value	95% CI
Sensitivity	71.43%	41.90-91.61%
Specificity	20.00%	2.52-55.61%
Positive likelihood ratio	0.89	0.57-1.41
Negative likelihood ratio	1.43	0.32-6.34
Positive predictive value	55.56%	44.26-66.30%
Negative predictive value	33.33%	10.12-68.95%
Accuracy	50.00%	29.12-70.88%
TMS (n=29)
MEP (+)	17	12	05
MEP (−)	12	03	09
	Value	95% CI
Sensitivity	80%	51.9-96%
Specificity	64.3%	35.1-87.2%
Positive likelihood ratio	2.24	1.10-4.82
Negative likelihood ratio	0.31	0.11-0.92
Positive predictive value	70.59%	53.2-83.5%
Negative predictive value	75.00%	50.3-89.8%
Accuracy	72.41%	52.8-87.2%

The findings from the multivariate logistic regression for UL motor outcome at 3 months post-stroke suggest that the MI (Arm) score, FMA (Arm) score, the duration between the stroke onset and first assessment, TMS (MEP) presence, DTI (FA) parameters are strong predictors for UL recovery after acute stroke [Table 3]. For developing models, we included 24 acute ischemic MCA stroke patients in our final analysis as the data for all the three components (clinical, TMS, DTI) during the first week of the stroke onset and the survived patients during 3 months of follow-up was available for 24 patients. Five predictive models were made including the following variables. Model-I includes clinical scales [FMA-Arm score + MI-Arm score] + TMS (MEP) + DTI (FA); Model-II: Clinical scales [FMA-Arm score + MI-Arm score] + TMS (MEP); Model-III: Clinical scales [FMA-Arm score + MI-Arm score]+ DTI (FA); Model-IV: TMS (MEP) + DTI (FA); and Model-V: Clinical scales [FMA-Arm score + MI-Arm score]. Among the five models, Model-I including clinical scales (FMA + MI) + TMS (MEP) + DTI (FA) was found to be the most accurate predictive model, with an overall predictive ability of 93.3% (95% confidence interval [95% CI] 0.87–0.99) and sensitivity 94.9% (95% CI 0.87–1.0) and specificity 95.8% (95% CI 0.89–1.0), respectively [Table 4].

Table 3

Multivariate logistic regression for upper-limb motor outcome at 3 months post-stroke

Predictors	Odds Ratio (OR)	95% CI	P
Demographics (n = 29)
Sex (M/F)	0.59	0.34-1.02	0.060
Hemisphere affected (L/R)	0.87	0.38-2.01	0.75
Days between stroke onset and first assessment	0.94	0.89-0.98	0.009
Clinical Scales [n = 29]
NIHSS	1.96	1.16-3.31	0.012
MI-Arm score	24.80	7.96-77.30	<0.001
BI at admission	1.39	1.18-1.64	<0.001
FMA-Arm score	22.71	7.63-67.63	<0.001
TMS [n = 29]
MEP present	28.33	9.18-87.46	<0.001
DTI [n = 24]
FA	7.00	2.70-18.72	<0.001
MD	0.71	0.42-1.19	0.191

Table 4

Predictive properties of the models for predicting upper-limb motor outcome at 3 months post-stroke (n=24)

Recovery=No; ARAT (Arm) ≤ 9, Yes; ARAT (Arm) ≥ 10
	Model-I (95% CI)	Model-II (95% CI)	Model-III (95% CI)	Model-IV (95% CI)	Model-V (95% CI)
Sensitivity %	94.9 (0.87-1.0)	89.3 (0.81- 0.94)	63.1 (0.43-0.69)	64.7 (0.37-0.76)	74.9 (0.56-0.81)
Specificity %	95.8 (0.89 -1.0)	86.4 (0.76-0.95)	54.3 (0.41-0.58)	59.3 (0.39-0.85)	73.4 (0.67-0.85)
PPV%	92.7 (0.76-0.96)	85.2 (0.63-0.83)	55.2 (0.44-0.67)	65.2 (0.43-0.72)	69.1 (0.52-0.78)
NPV %	93.3 (0.87-0.99)	82.7 (0.54-0.88)	54.1 (0.59-0.72)	62.7 (0.54-0.89)	74.3 (0.77-0.91)
Overall %	94.4 (0.76-0.95)	83.7 (0.65-0.90)	51.5 (0.45-0.67)	53.7 (0.65-0.70)	68.7 (0.64-0.83)

Model-I: Clinical scales [FMA-Arm score+MI-Arm score] + TMS (MEP) + DTI (FA); Model-II: Clinical scales [FMA-Arm score+MI-Arm score] + TMS (MEP); Model-III: Clinical scales [FMA-Arm score+MI-Arm score] + DTI (FA) Model-IV: TMS (MEP) + DTI (FA); Model-V: Clinical scales [FMA-Arm score+MI-Arm score]

DISCUSSION

In the present study, we examined the prognostic value of clinical scales, TMS, and DTI for UL motor outcome at the acute phase within 7 days of the IS onset. Our results specify that the patients who showed MEP response in TMS and had a preserved CST integrity in DTT had a better chance of UL motor recovery than those patients without these characteristics. The FA values of the DTI parameter indicate CST integrity of white matter lesions and the decreased FA values indicate interrupted integrity of the neural tract and correlate with the motor deficits. On comparing these two predictive tools, TMS reveals a higher PPV than DTI and DTI has a higher NPV than TMS. In contrast, the presence or absence of MEPs, as well as the disruption of the CST tract detected at an early stage of the stroke onset, can predict the motor outcome by their unique characteristics.

The previously published studies had confirmed the correlation of the neuromotor outcome with the conventional neuroimaging variables such as the extent of infarction, infarct size, and volume.[14151617] The individual utility of DTI or TMS has been reported by numerous studies;[131415161718192021222324252627] however, only two studies to date have been performed to investigate the combined utility of TMS and DTI for the prediction of motor outcome during the acute phase of intracerebral hemorrhage and corona radiata infarct.[1020] More recent studies have investigated the correlation between a decrease in the FA value measured through DTI and the motor function in the patients with subacute or chronic stages of IS.[282930] Two recent meta-analyses conducted by Kumar P et al. 3132] reported that DTI-based FA is a strong predictor of UL motor recovery after subacute ischemic as well as hemorrhagic stroke.

Presently, we investigated the clinical, TMS, and DTI measured for the accuracy of prediction in the acute phase (within 7 days of the stroke onset). Although we screened 386 stroke patients, only 29 acute ischemic MCA stroke patients were recruited due to our stringent inclusion criteria of evaluation. The reason behind the exclusion of the screened cases include (a) most of the patients were unstable as they had both upper and lower limb weaknesses, (b) hemicraniectomy done in some of the MCA stroke patients, (c) the patients mostly reached All India Institute of Medical Sciences (AIIMS), New Delhias a referral center and that time they were at a subacute and chronic phase of the stroke, (d) denied providing consent, (e) stroke other than MCA territory, and (f) hemorrhagic stroke. The enrolment of a small number of subjects, lack of bounded time of investigations, and resources deficit in terms of funding were found to be the major limitations of the study.

Up to 80% of the stroke survivors have UL impairment early after the stroke, and a few demonstrate complete functional recovery at 6 months post-stroke.[33343536] The UL rehabilitation trials designed to improve the recovery rates have been largely unsuccessful. As a result, the burden of the UL impairment after a stroke remains high.[37] Therefore, understanding how to improve the potential for the recovery of the UL function remains a major scientific, clinical, and patient priority. There is a growing interest in using biomarkers to predict motor recovery and outcomes after a stroke. The Predict REcovery Potential (PREP2) algorithm proposed by Connell et al. (2021), Smith et al. 2019, and Stinear et al. (2017)[383940] combines the clinical assessment with biomarkers in an algorithm, to predict the UL functional outcomes for individual patients. Active and theoretically underpinned implementation strategies are needed to ensure that the biomarkers are successfully used in clinical practice for predicting motor outcomes after a stroke, and should be considered in parallel with biomarker development. The past work has suggested that the brain biomarkers may help advance our understanding of the recovery phenotypes.[41] Routine clinical scans (often CT, a few MRIs) describe the lesion location as defined by a neurologist: cortical, subcortical, mixed). Unfortunately, this approach to analyze routine scans is not helpful to identify the meaningful UL recovery in stroke patients. Therefore, there is a need for longitudinal studies that include measures that are potentially more sensitive to change than clinical measures and the lesion location which may include biomarkers such as the impact of the stroke on the primary motor pathways (corticospinal tract) using TMS to index MEP status (present or absent),[41] as well as more nuanced interpretations of the lesion location (e.g. internal capsule impacted yes/no)[42] that account for the primary motor pathway impact. There is an ongoing study focused on people with severe paresis that is focused on identifying the brain biomarkers of recovery.[43] Together, the advancements in the biomarkers may improve the prediction of recovery, patient selection, and individualized intervention.

Further research regarding the re-perfusion, MRI signal changes, imaging of Wallerian degeneration in the descending cortico-spinal tracts, may be other imaging methods which may be implemented in further research. To recruit a large number of subjects, a long duration of studies with extensive follow-up is required in order to reach a definite conclusive finding. Future prospective cohort studies are recommended to put a spotlight on the perceptive mechanisms that can define the decisive time frame of motor recovery after a stroke.

CONCLUSION

The accuracy of UL motor recovery can be predicted through a clinical battery and their elements as well as TMS (MEP) and DTI (FA) parameters. Further, well-designed prospective studies are needed to confirm our findings.

Declaration of patient consent

The authors certify that they have obtained all appropriate patient consent forms. In the form the patient(s) has/have given his/her/their consent for his/her/their images and other clinical information to be reported in the journal. The patients understand that their names and initials will not be published and due efforts will be made to conceal their identity, but anonymity cannot be guaranteed.

Financial support and sponsorship

Indian Council of Medical Research, New Delhi, India [Ref. No-: 5/4-5/Neuro/74/2012-NCD-I].

Conflicts of interest

There are no conflicts of interest.