Repetitive transcranial magnetic stimulation of the contralesional dorsal premotor cortex for upper extremity motor improvement in severe stroke: a pilot randomized clinical trial

Abstract

Objective:

To compare the effects of repetitive transcranial magnetic stimulation targeting the contralesional dorsal premotor versus the ipsilesional primary motor cortex in chronic stroke survivors with severe hand paresis.

Design:

This assessor-blinded pilot randomized trial enrolled 16 participants ≥6 months post-stroke with severe hand paresis, defined by the absence of either 10° distal extension in wrist/fingers/thumb or motor evoked potentials in paretic muscles. Participants received 5Hz repetitive transcranial magnetic stimulation to the contralesional dorsal premotor or the ipsilesional primary motor cortex target, combined with upper extremity rehabilitation for 6-weeks. Outcomes included Upper Extremity Fugl-Meyer, Wolf Motor Function Test, Stroke Impact Scale-16, and inter-hemispheric inhibition.

Results:

Baseline characteristics were similar across groups, though the ipsilesional stimulation group had more participants with motor evoked potentials and ≥10° thumb extension (p=0.04). Both groups achieved similar gains on Fugl-Meyer, but wrist sub-score changes favored the ipsilesional stimulation group (p=0.011). No differences were observed for other outcomes.

Conclusion:

Stimulation of the contralesional dorsal premotor cortex was not superior to stimulation of the ipsilesional primary motor cortex for upper extremity rehabilitation outcomes in chronic stroke survivors with severe hand paresis, though both approaches yielded sustained motor improvements. Future trials should stratify by distal extension and physiology.

Introduction

Nearly two-thirds of stroke survivors have upper extremity (UE) paresis, and up to half have severe paresis.¹ Severe UE paresis is characterized by minimal voluntary movement, especially in the distal UE (wrist, thumb, fingers), restricting participation in functional activities. Rehabilitation options for this sub-population are limited given many therapies require at least some distal movement for participation in functional task-practice.

Repetitive transcranial magnetic stimulation (rTMS) is a widely studied brain stimulation technique to noninvasively prime motor cortical excitability and promote post-stroke rehabilitation outcomes.² Ipsilesional (stroke-affected) primary motor cortex (iM1) is the most common target based on its robust corticospinal tract (CST) connectivity for dexterous UE control. Typically, rTMS is delivered to increase the excitability of iM1, either directly, or indirectly by suppressing the excitability of contralesional (stroke-unaffected) M1—believed to have strong inter-hemispheric inhibition (IHI) on iM1. However, large randomized controlled trials (RCTs) have found no additional value of iM1 rTMS beyond rehabilitation alone.³ Effects are modest in individuals with severe UE paresis due to insufficient integrity of CST pathways.⁴ Alternative rTMS approaches are needed to promote rehabilitation outcomes in individuals with severe UE paresis.

Contralesional dorsal premotor cortex (cPMd) is a candidate target for severe stroke. CPMd is anatomically preserved following a typical unilateral stroke, and can widely engage bilateral networks via transcallosal connections.⁵ Temporary disruption of cPMd using TMS more strongly impairs arm movement in persons with severe weakness than temporary disruption of iM1.⁶ Conversely, temporary facilitation of cPMd using rTMS improves arm movement and reduces IHI more effectively than temporary facilitation of iM1 in persons lacking criterion CST viability (i.e., absent motor evoked potentials [MEPs)).⁷ Stronger cPMd-to-ipsilesional functional connectivity is positively associated with greater rehabilitation motor gains in persons with moderate-to-severe UE paresis, suggesting cPMd is functionally relevant for recovery in this sub-population.⁸

No study has evaluated the effects of facilitating cPMd on UE rehabilitation outcomes of chronic severe stroke survivors. We conducted a pilot RCT comparing cPMd facilitation rTMS to iM1 facilitation rTMS, each combined with 6-week UE rehabilitation, in chronic stroke survivors with severe hand paresis. We hypothesized that cPMd facilitation would lead to greater and lasting UE motor improvements and larger IHI reduction in participants with severe hand paresis. The trial was designed to estimate effect sizes for future efficacy studies.

Methods

Trial Design

This single-center, assessor-blinded, pilot RCT compared 2 parallel groups: cPMd facilitation rTMS plus rehabilitation versus iM1 facilitation rTMS plus rehabilitation (1:1 allocation). The Cleveland Clinic IRB approved the protocol. Full protocol is published.⁹ The trial was registered on clinicaltrials.gov (NCT03868410) and adheres to CONSORT guidelines.

Participants

Individuals aged 18-90 years with severe UE paresis following ischemic/hemorrhagic stroke occurring ≥6 months prior were eligible. Severe paresis was defined by either (1) failure to achieve ≥10° voluntary extension at wrist/thumb/two-digits (constraint induced movement therapy or CIMT criterion), or (2) absence of MEPs in relaxed/resting paretic extensor digitorum communis (EDC) muscle, which indicates insufficient ipsilesional CST viability. Participants were required to have enough proximal control to bring the weak arm from side to their lap for therapy.

Exclusion criteria were brainstem/cerebellar stroke; bilateral or multiple lesions involving sensorimotor pathways; severe cognitive impairment (failure of 2/3 word recall at 30 min and/or inability to follow 3-step commands); severe spasticity/contracture (Modified Ashworth Scale ≥3) in wrist/hand; severe sensory loss or hemineglect; contraindications to TMS/magnetic resonance imaging (MRI) (e.g., head metal, seizures, pacemaker); and recent UE therapy or botulinum toxin (<2 months).

Recruitment occurred via electronic medical record review, referrals, support groups, and phone inquiries. Participants underwent phone screening before an in-person eligibility testing visit, at which the neurologist (K.U.) confirmed diagnosis, and study team established severity eligibility using TMS and CIMT movement testing.

Randomization and Blinding

Covariate-adaptive randomization was utilized to minimize group imbalances on variables relevant to UE recovery: time since stroke (<2 vs. ≥2 years), lesion location (subcortical vs. cortical/mixed), active wrist extension (≥0° vs. <0° from neutral), and dominance of the paretic limb. A custom MATLAB script was utilized (The MathWorks, Inc., Natick, MA, USA). Study staff not involved in assessment/therapy generated group assignments. Only the rTMS operator was unblinded to group assignment. Outcome assessors and treating therapists were blinded. Participants knew which side of the head was targeted with rTMS (affected vs unaffected), but were not aware of the hypothesis and were instructed not to disclose details to blinded staff.

Motor Outcomes Assessments

Assessments occurred at baseline (PRE), end-of-treatment at 6-weeks (POST), and 3-month follow-up (FU). The primary endpoint was change in UE Fugl-Meyer (UEFM) from PRE to POST. UEFM is a 33-item impairment scale (0-66) with established reliability/validity.¹⁰ Secondary outcomes were Wolf Motor Function Test (WMFT) and Stroke Impact Scale (SIS-16). WMFT evaluates activity limitations across 15 timed performance tasks and includes two strength tasks. The time is converted to performance rate, i.e., repetitions per minute and averaged across items, a more meaningful measure for severe stroke.¹¹ WMFT functional ability was also measured on 0-5 scale, where 0 denotes “does not attempt with the weaker arm” and 5 denotes “weaker arm participates, normal movement”. SIS-16 is a reliable and valid index of patient perceived participation restrictions specific to physical functioning. The measure consists of 16 items across 4 physical domains (strength, hand function, mobility, and activities of daily living).¹²

TMS and Neurophysiology Assessments

Single-pulse TMS was used to evaluate CST excitability and IHI at PRE and POST. A monophasic pulse stimulator with a figure-of-eight coil was used to deliver TMS (Magstim 200², The Magstim Company, UK). Stereotactic neuro-navigation guided by individual anatomic MRI (whenever scans were available) ensured precision and consistency in targeting (Brainsight, Rogue Research Inc., Montreal, Quebec, Canada).

Participants sat comfortably with hands resting on a flat surface, elbow flexed at 90°, forearms pronated, and shoulders slightly flexed/abducted. Surface electromyography (EMG) was recorded from paretic EDC, the primary muscle for finger extension/opening, at 4kHz sampling rate with 10Hz high-pass filter and analyzed offline using LabChart 8 (PowerLab 4/25T, AD Instruments, Colorado Springs, CO).

Before TMS, participants completed isometric maximum voluntary contraction (MVC) testing of EDC muscle. They were required to perform (or attempt) maximum isometric finger extension at metacarpophalangeal joints against manual resistance. Three trials were completed for paretic and non-paretic EDC (3-5s trials at ~30s intervals). Root-mean-square EMG was used to define target contraction for subsequent TMS experiments.

TMS was delivered to 14-25 scalp sites (1cm² resolution) while participants maintained slight contraction of EDC (20%±5% MVC). Active contraction increases likelihood of eliciting MEPs in paretic muscles. While participants may lack resting MEPs per inclusion criterion, they could still possess active MEPs in contracted paretic EDC. The site which elicited active MEPs ≥100μV (peak-to-peak) larger than background EMG (over 140ms) at the lowest TMS intensity in ≥6/10 trials was called motor hotspot.¹³ The lowest TMS intensity used was called active motor threshold (AMT), expressed as %maximal stimulator output (%MSO). AMT is a measure of corticospinal excitability with excellent reliability in paretic EDC post-stroke (ICC>0.94).¹⁴ (Note a similar TMS thresholding procedure was performed at eligibility visit to determine absence of criterion resting MEPs— ≥50μV amplitude across ≥6/10 trials in relaxed paretic EDC—for selection criterion of severe loss to ipsilesional CST).

After AMT, we tested IHI using the ipsilateral silent period (ISP) method. Supra-threshold TMS was delivered to contralesional motor hotspot (100% MSO, 21-50 trials), while participants maintained submaximal (50%±5% MVC) contraction of paretic EDC. Supra-threshold TMS suppresses ongoing ipsilateral EMG activity, hence the term ipsilateral silence. EMG data was downsampled to 2kHz for analysis. ISP onset was defined as the time point (in ms) when the averaged rectified EMG dropped below 1.77×MCD (mean consecutive difference) of pre-stimulus EMG (over 150ms period) for at least five consecutive data points. ISP offset was defined as when at least half the data points returned above this level in a 5ms period.¹⁵ ISP duration was calculated as the time between onset and offset. ISP depth, i.e., degree of EMG suppression, was calculated as %reduction in EMG during ISP relative to pre-stimulus EMG (Equation 1). ISP onset, duration and depth serve as reliable measures of IHI from the targeted motor cortex. Depth exhibits good-to-excellent between- and within-session reliability (ICC = 0.83 and 0.96, respectively),¹⁴ while onset and duration have moderate-to-good reliability (ICC = 0.68 and 0.71, respectively).¹⁶

$$ISPDepth=\frac{MeanAverageRectifiedEM{G}_{pre-stimulus}-MeanAverageRectifiedEM{G}_{ISP}}{MeanAverageRectifiedEM{G}_{pre-stimulus}}\times 100\%$$

Imaging Assessments

MRI images were collected on a Siemens Prisma 3T scanner equipped with a standard 20-channel head-neck coil. T1-weighted inversion recovery turboflash (MPRAGE) scans were acquired with the following parameters: axial, 256mm×256mm×120mm field of view (FOV), 1.0mm×1.0mm×1.2mm voxels, echo time (TE)/repetition time (TR)/inversion time (TI) = 1.71ms/1900ms/900ms.

Lesions were classified as subcortical or cortical/mixed. Manual tracing was performed at each slice to calculate lesion volume (by J.L. and X.L.). High angular resolution diffusion imaging (HARDI) data was collected for diffusion tensor imaging (DTI) with the following parameters: axial, 128mm×128mm×52mm FOV, 2.0mm×2.0mm×2.0mm voxels, TE/TR = 66ms/6200ms, 71 diffusion-weighting gradients with b=1000s/mm² and 8 T2-weighted image volumes.

T1-weighted and diffusion images (DWIs) were skull-stripped using the Brain Extraction Tool¹⁷ in FSL (FMRIB Software Library, Oxford).¹⁸ DWI data was corrected for motion artifacts and eddy current distortions with eddy_correct, followed by generation of whole-brain maps of fractional anisotropy (FA) with dtifit. FA is a unitless measure of water diffusion with values ranging 0 to 1, where higher values indicate higher direction-specific diffusion and greater white matter integrity. FA is a valid and reliable measure of microstructural white matter tract integrity in stroke. It can be acquired reliably even in the presence of severe CST damage—making it ideal for use in this pilot RCT.¹⁹

FA values were extracted from the posterior limb of the internal capsule (PLIC) region of interest (ROI) using an objective, validated, template-based approach.²⁰ The MNI152 T1 template was aligned to native T1 space, and the resulting transformation was used to map the right and left PLIC labels from the Johns Hopkins University DTI-based white matter atlas.²¹ Registered PLIC labels were manually adjusted if they overlapped with basal ganglia or ventricles. The FA maps were also transformed into native T1 space and visually inspected to ensure accuracy. Mean FA values were extracted from the contralesional and ipsilesional PLIC ROIs. FA asymmetry was calculated as:

$$F{A}_{asymmetry}=\frac{F{A}_{contralesional}-F{A}_{ipsilesional}}{F{A}_{contralesional}+F{A}_{ipsilesional}}$$

FA asymmetry ranges between −1.0 and +1.0, where 0 indicates symmetrical FA on both sides, while positive values indicate reduced FA in ipsilesional PLIC.

Interventions

CPMd and iM1 facilitation involved high-frequency (5Hz) rTMS delivered to respective target immediately before therapy, twice per week for 6 weeks (total 12 sessions). RTMS followed similar procedures as single-pulse TMS but this time an air-cooled figure-of-eight coil was used with a biphasic stimulator (Magstim Rapid², Whitland, Wales, UK). Target iM1 was the ipsilesional motor hotspot for paretic EDC; if active MEPs could not be found then mirror location of contralesional hotspot was used as proxy target. CPMd target was defined 2cm anterior and 1cm medial to the contralesional motor hotspot.²² Target location was consistently monitored using stereotactic neuro-navigation.

Stimulation intensity was 90% AMT or 72% MSO, whichever was lower to limit discomfort and risk of seizure.²³ Forty-two trains were delivered at 5Hz (10s, 50 pulses/train) with 20.3-second inter-train intervals, totaling 2100 pulses (~22 min). RTMS parameters and dose followed established safety guidelines,²³ balancing across considerations of maximizing effect (dose/pulses), participant comfort (intensity), and practicability (session duration).

Immediately after rTMS, participants underwent 60-minutes of therapist-guided task practice emphasizing high-repetition, goal-directed movements of the paretic UE. Unilateral practice was adopted from previous studies of iM1 rTMS—which paired stimulation with unilateral paretic therapy—to test whether cPMd is a more effective target than iM1.³ Additionally, neuroscientific studies have indicated cPMd’s positive contributions to paretic UE when there is insufficient ipsilesional substrate.^5-7 Therefore, tasks included paretic limb reaching in multiple directions (in/out of synergy as tolerated) with grasp/manipulation/release when possible. When participants could not complete full tasks, preparatory exercises built requisite components—passive/active-assisted range, stretching, part-task training. Tasks were repeated across sessions and “shaping” was followed for progression, involving adjustment of distance, height, load, time, and rate, with goal of difficulty level 7/10. On average, 4 tasks were completed per session with a range of 3-9 tasks. The average number of repetitions per session was 112 (SD 58). At the end of the 6-week program, therapists provided a personalized home program to support continued UE activity during follow-up.

Statistical Analysis

Sample size planning used a prior case series: between-group ΔUEFM=4.7 and pooled SD=2.71 implied n=6 per arm (power 80%). Allowing for 20% attrition, target N=16 was selected to support preliminary efficacy estimation, consistent with pilot neuromodulation trials.^24,25

Analyses used R 4.3.1. The primary analysis followed intent-to-treat (ITT) with last-observation-carried-forward for missing data. A sensitivity ITT analysis without imputation and a per-protocol analysis (participants completing all sessions and assessments) assessed robustness. The primary endpoint was change in UEFM over 6weeks. Linear mixed-effects models (lme4) were fitted using fixed effects of GROUP (cPMd, iM1), TIME (PRE, POST, FU), and GROUP×TIME, with random intercepts for participants. ANOVA tests were performed on the model, and post-hoc contrasts (emmeans) for within- and between-group differences (with Holm-Bonferroni corrections) were performed if there is significant GROUP×TIME interaction. UEFM sub-scores (proximal 36, wrist 10, hand 14), WMFT (paretic and non-paretic performance rates, functional ability) and SIS-16 were analyzed using the same method.

Response was defined a priori as ≥3-point UEFM gain—minimal detectable change (MDC) for the measure (minimal clinically important difference for UEFM is lacking in chronic severe stroke).²⁶ Exploratory associations between baseline characteristics and change scores were examined using t-tests and Wilcoxon rank sum tests for categorical variables, and bivariate correlations for continuous variables. Neurophysiology outcomes (i.e., AMT, ISP) were analyzed per-protocol due to expected missingness (e.g., absent active MEPs even at 100% MSO) using linear mixed-effects models analogous to motor outcomes.

Statement

During the preparation of this work the author(s) used ChatGPT 5 to improve grammar, readability, conciseness, and language. After using this tool, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication.

Results

A total of 217 individuals were contacted for phone screening (Figure 1). Of them, 179 did not advance to eligibility assessment (99 did not meet eligibility, 31 were not interested, 29 could not be reached, 17 had schedule conflicts, and 3 were deceased). Thirty-eight provided informed consent and underwent eligibility assessment to confirm severity eligibility. Twenty of 38 candidates failed to meet eligibility—12 had CIMT distal extension, 6 had resting MEPs, 1 did not have enough arm movement, 1 had a history of seizure—and two declined to participate. Overall, 16 participants (42%) were randomized between December 2019 and March 2023. The study was on hold during the COVID-19 pandemic. There was attrition of 2 participants (12.5%): one withdrew from the iM1 group after the 8^th treatment session for unrelated medical reasons; another from the cPMd group could not undergo POST assessment due to COVID-19 shutdown. A total of 14 participants were included in the per-protocol analysis.

Figure 1.

CONSORT flow diagram. ITT: intent to treat.

Clinical and demographic details are presented in Table 1. The sample was on average (SD) 64.2 (11.7) years old and 46.4 (37.8) months post-stroke. UEFM score was 22.8 (7.8), range 13-41 (inclusive). Participants met severity criteria, defined by the absence of resting MEPs (12/16) or the absence of CIMT distal extension (12/16), or both (8/16). Groups were balanced at baseline on all pre-defined characteristics, including covariate-adaptive randomization variables (all p>0.08) and eligibility criteria related to severity: resting MEP status (p=1.0) and composite CIMT criterion (p=0.08). However, notable differences emerged within these over-arching categories. While balanced on binary time post-stroke (<2 vs. ≥2 years), the iM1 group tended to be less chronic than the cPMd group (26 (12.3) vs. 37.5 (53) months post-stroke, p=0.06). While well-matched on composite CIMT criterion, the iM1 group had more participants with ≥10° thumb extension (7/8 vs. 2/8, p=0.04). While balanced on resting MEPs, more participants in the iM1 group had active MEPs (6/8 vs. 1/8, p=0.04).

Table 1.

Participant Baseline Demographic and Clinical Data.

ID	Group	Age	Sex	Stroke Type	Months Since Stroke	Dominant Paresis	Cortical Stroke	Lesion Volume (mm3)	FA Asym	Resting MEP	Active MEP	CIMT	CIMT Wrist	CIMT Thumb	CIMT Fingers	UEFM
A003	cPMd	60	Female	Hemorrhagic	29	No	No	65	0.17	No	No	No	No	No	No	19
A093	cPMd	76	Female	Ischemic	27	Yes	No	536	0.18	No	No	No	No	Yes	No	17
A117	cPMd	68	Female	Ischemic	29	Yes	Yes	22344	0.09	No	No	No	No	No	No	17
A167	cPMd	60	Male	Ischemic	109	Yes	No	-	-	Yes	No	No	No	Yes	Yes	24
A203	cPMd	73	Female	Ischemic	41	Yes	No	721	0.09	No	No	No	No	No	No	17
A221	cPMd	59	Male	Hemorrhagic	153	No	No	331	0.12	Yes	Yes	No	Yes	No	No	26
A222	cPMd	75	Female	Ischemic	34	No	No	520	0.11	No	No	No	Yes	No	No	21
A339	cPMd	34	Male	Ischemic	73	Yes	Yes	114301	0.10	No	No	No	Yes	No	No	34
Mean (SD) or Median (IQR)		63.1 (13.7)	5F/3M	6I/2H	37.5 (53.0) †	5Y/3N	2Y/6N	536.4 (11107)	0.12 (0.04)	2Y/6N	1Y/7N*	0Y/8N†	3Y/5N	2Y/6N*	1Y/7N	21.9 (6.0)
A034	iM1	61	Male	Ischemic	69	Yes	No	76	0.08	No	Yes	Yes	Yes	Yes	Yes	41
A095	iM1	53	Male	Ischemic	46	No	Yes	89360	0.20	No	No	No	No	No	No	13
A145	iM1	66	Female	Ischemic	15	Yes	No	1106	0.12	Yes	Yes	No	No	Yes	Yes	20
A152	iM1	59	Female	Hemorrhagic	29	No	No	1848	0.26	No	No	No	No	Yes	No	16
A155	iM1	60	Female	Ischemic	29	Yes	No	-	-	No	Yes	Yes	Yes	Yes	Yes	28
A169	iM1	81	Female	Ischemic	23	Yes	No	22	0.11	Yes	Yes	No	No	Yes	No	16
A198	iM1	80	Female	Ischemic	13	Yes	No	487	0.08	No	Yes	Yes	Yes	Yes	Yes	33
A206	iM1	62	Female	Ischemic	23	Yes	Yes	-	-	No	Yes	Yes	Yes	Yes	Yes	23
Mean (SD) or Median (IQR)		65.3 (10.1)	6F/2M	7I/1H	26.0 (12.3) †	6Y/2N	2Y/6N	796.8 (1484)	0.14 (0.07)	2Y/6N	6Y/2N*	4Y/4N†	4Y/4N	7Y/1N*	5Y/3N	23.8 (9.6)

CIMT: constraint-induced movement therapy, “Yes” for meeting criteria for CIMT, i.e., ≥10° extension in wrist, thumb and two other digits, “No” for not meeting criteria for CIMT; FA Asym: fractional anisotropy asymmetry; IQR: interquartile range; MEP: motor evoked potential; SD: standard deviation; UEFM: upper extremity Fugl-Meyer.

^*Between-group p<0.05.

^†Between-group p<0.1.

Table 2 presents motor outcomes across time for each group. For the primary outcome (UEFM), the null hypothesis was not rejected (p=0.38, Figure 2). From PRE to POST, the estimated difference in UEFM (95% confidence interval [CI]) was 2.6 (0, 5.2) for both cPMd and iM1 groups. Four of 8 participants in the cPMd group (50%) and 5 of 8 participants in the iM1 group (62.5%) had ≥3-point (MDC) UEFM gain at primary endpoint (Fisher’s exact test, p=1). UEFM improvements were partially maintained through 3-month FU, with PRE-to-FU gain of 1.8 (−0.8, 4.3) and 3.3 (0.7, 5.8) in the cPMd and iM1 groups.

Table 2.

Motor Outcomes by GROUP and TIME (mean±SD).

	iM1 Group			cPMd Group
	PRE	POST	FU	PRE	POST	FU	GROUP x TIME p-value	GROUP p-value	TIME p-value
UEFM	23.8±9.6	26.4±10.4	27.0±12.5	21.9±6.0	24.5±6.8	23.6±6.3	0.380	0.598	<0.001
UEFM Wrist	1.9±1.8	2.9±2.0	2.9±2.3	1.9±1.6	2.1±2.0	1.6±1.8	0.011	0.486	0.011
UEFM Hand	3.9±2.8	4.5±3.7	4.8±4.0	1.9±1.4	2.6±2.1	2.5±1.9	0.849	0.152	0.060
WMFT Rate (P)	14.1±16.1	17.1±18.9	16.8±17.9	9.4±5.8	10.4±5.4	11.3±6.7	0.249	0.405	0.001
WMFT Rate (NP)	47.2±8.1	50.3±10.1	49.9±8.3	49.9±8.0	52.4±8.9	53.7±9.5	0.534	0.520	0.001
WMFT FA	1.9±0.9	2.1±1.1	2.2±1.2	1.6±0.3	1.7±0.4	1.8±0.4	0.315	0.382	0.001
SIS-16	58.9±5.9	60.3±8.6	61.6±8.6	61.9±11.0	66.1±9.4	64.3±7.9	0.267	0.374	0.025

FA: functional ability; NP: non-paretic; P: paretic; SIS-16: Stroke Impact Scale-16; UEFM: upper extremity Fugl-Meyer; WMFT: Wolf Motor Function Test.

Figure 2.

A) UEFM scores across pre-test, post-test and 3-month follow-up time points for iM1 and cPMd rTMS groups. Error bars show standard error of the mean. # indicates significant contrasts between timepoints, collapsed across groups. B) UEFM scores across time points for individual subjects, by group.

Analysis of UEFM sub-scores revealed a significant GROUP x TIME interaction for UEFM wrist (p=0.011). Post hoc contrasts revealed significant effects favoring the iM1 group: the estimated difference from PRE-to-POST in UEFM wrist sub-score was 1.0 (0.2, 1.8) for the iM1 group (p=0.009) and 0.3 (−0.6, 1.1) for the cPMd group (p=1). These gains remained stable through FU, post hoc contrasts showed PRE-to-FU gains being 1.0 (0.2, 1.8) in the iM1 group (p=0.009) and 0.25 (−1.1, 0.6) in the cPMd group (p=1). There were no between-group differences on any of the other UEFM sub-score changes (all p>0.7). Exploration of baseline characteristics revealed that UEFM wrist sub-score gains at 6-week were larger among participants with active MEPs than those without (p=0.043). As a reminder, 6 of 8 participants in the iM1 group had active MEPs at baseline, compared to only 1 of 8 in the cPMd group.

There was no between-group difference on change in any of the WMFT measures (all p>0.24). The estimated difference in WMFT paretic rate was 1.0 (−1.5, 3.5) repetitions per minute in the cPMd group and 3.0 (0.5, 5.5) in the iM1 group from PRE to POST. Improvements were maintained through FU with gains of 1.9 (−0.6, 4.4) in the cPMd group and 2.7 (0.2, 5.2) in the iM1 group (Figure 3A). Estimated differences in WMFT functional ability scores for both PRE-to-POST and PRE-to-FU comparisons were 0.1 (−0.1, 0.3) for the cPMd group and 0.2 (0, 0.4) for the iM1 group (Figure 3E).

Figure 3.

WMFT performance rate of the paretic side (A), non-paretic side (C), and functional ability scores (E) across pre-test, post-test and 3-month follow-up time points for iM1 and cPMd rTMS groups. Error bars show standard error of the mean. # indicates significant contrasts between timepoints, collapsed across groups. Respective individual subject data are shown in B, D, and F.

Change in SIS-16 scores also did not differ across groups (p=0.27). From PRE to POST, the estimated difference in SIS-16 was 4.3 (−0.3, 8.8) for the cPMd group and 1.4 (−3.2, 5.9) for the iM1 group with PRE-to-FU changes of 2.4 (−2.2, 6.9) and 2.8 (−1.8, 7.3), respectively.

Sensitivity analysis performed without imputation for missing data revealed similar results for UEFM, WMFT paretic rate, WMFT non-paretic rate, WMFT functional ability, and SIS-16. Per-protocol analysis (n=14) also revealed the same results for all motor outcomes.

TMS analysis revealed no difference between groups on PRE-to-POST change in AMT or incidence of active MEPs. There was a between-group difference in the change on contralesional ISP onset (p=0.034), albeit post hoc contrasts were not significant (p>0.34) (Figure 4). The estimated difference in ISP onset was −3.1ms (−8.8ms, 2.6ms) in the cPMd group—indicating shortened onset latency, versus 3.5ms (−2.2ms, 9.2ms) in the iM1 group—indicating delayed onset latency. Change in no other ISP measure differed across groups (all p>0.6).

Figure 4.

A) Contralesional ISP onset across pre-test and post-test time points for iM1 and cPMd rTMS groups. Error bars show standard error of the mean. B) Contralesional ISP onset across time points for individual subjects, by group.

There were no serious adverse events. Eleven adverse events were recorded in total across four participants—two from each group. These included headaches (n=2), neck pain (n=2), scalp irritation (n=2), dizziness (n=2), fall at home (n=1), high blood pressure upon arrival (n=1), and dental pain (n=1). Fidelity of rTMS targeting (cPMd, iM1) was measured at each visit (total 181 out of 188 visits). An error was made in <1% of visits, with rTMS given to cM1 instead of cPMd.

Discussion

This pilot RCT is the first to compare cPMd versus iM1 facilitation rTMS, each combined with identical UE rehabilitation, in chronic severe stroke. Both groups improved after 6 weeks of treatment and largely maintained gains at 3 months. We found no additional advantage of targeting one region over another in terms of motor impairment, function, or participant-perceived physical functioning. Proportion of responders (UEFM gain ≥3) was also not different between groups (50% and 62.5%). While the study found no overall advantage of facilitating cPMd over iM1 in chronic stroke participants with severe UE paresis, between-group differences were likely mitigated by the presence of key clinical and physiologic characteristics favoring response in the iM1 group (i.e., more participants with active MEPs and criterion thumb extension at baseline), indicating availability of some residual ipsilesional substrate. Conceivably, it is less about the target and more about the conditions favoring response. Future efficacy trials should stratify by distal extension and active MEP status to rigorously evaluate the effects of targeted rTMS.

CPMd stimulation was no more effective than iM1 in improving UE rehabilitation outcomes among stroke survivors with severe paresis. This challenges the prevailing theory that the intact hemisphere has a more favorable role in cases with severe injury. The bimodal-balance recovery hypothesis postulates that intact contralesional motor cortices can contribute to ipsilateral paretic UE movement when there is insufficient contralateral substrate, such as via reduction in IHI or unmasked alternate (ipsilateral) connections.^19,27 CPMd is well positioned to make ipsilateral contributions based on its rich transcallosal and ipsilateral connectivity.⁵ Cross-sectional and single-session crossover studies in severely impaired survivors have shown that cPMd disruption impairs, and cPMd facilitation transiently improves, paretic arm movement, often more than experimental manipulation of iM1.^6,14,22 However, our multi-session RCT did not demonstrate superiority of facilitating cPMd over iM1 in rehabilitation of participants with severe impairment.

Effects observed in cross-sectional and single-session experimental studies may not consistently generalize to favorable outcomes in multi-session clinical studies. Cross-sectional and single-session crossover studies offer rigorous control for within-subject variability and often rely on laboratory-based measures of rapid response (e.g., reaching response, cortical excitability, neuromuscular coherence). Efficacy RCTs involving multiple treatment sessions, on the other hand, must demonstrate lasting impact on clinical measures beyond inter-individual variability. Therefore, while experimental studies are useful to investigate feasibility and potential mechanisms of a novel treatment under controlled conditions, efficacy designs are better suited to informing the potential long-term impact of a treatment.

While the study found no advantage of cPMd over iM1 facilitation in chronic severe stroke, between-group differences were likely mitigated by the presence of characteristics favoring response in the iM1 treatment arm. The iM1 group had more participants with active MEPs and thumb extension at baseline and tended to be less chronic (~2 vs. >3 years post-stroke) than the cPMd group. Survivors less than 2 years post-stroke have higher likelihood of achieving clinically relevant dexterity gains from hand therapies than those that are more chronic.²⁸ Similarly, criterion distal extension and MEP+ status signal the availability of critical ipsilesional CST substrate.²⁹ Therefore, it may be less about the target and more about the conditions that favor generating a response. The small sample size likely has affected the ability to effectively minimize all potential group imbalances at baseline and study significant baseline characteristics as covariates. Future efficacy trials stratified by distal extension and active MEP status will be better suited to investigate effects of targeted treatments.

The null findings of this study present key opportunities for developing effective targeted treatments. First, future studies should investigate whether target individualization is effective—iM1 for patients with residual distal extension and CST availability (e.g., active MEPs), and cPMd for those without, where intact contralesional networks could be more consequential. We have recently proposed one such preliminary framework for stratification, called Stratification Algorithm for rTMS in stroke or START.¹⁴ START uses clinical motor impairment (UEFM) score and imaging of white matter CST integrity (FA asymmetry) to guide treatment selection. Developed based on a single-session cross-over experimental study, START revealed that patients with better UEFM scores (>42) are more likely to improve with rTMS facilitating iM1 than cPMd, whereas those with UEFM scores ≤42 are more likely to improve with rTMS facilitating cPMd than iM1 (91.67% and 83.33% accuracy, respectively). UEFM cut-offs revealed in START were validated by another group using a multi-session treatment design.³⁰ Second, the comparison of two competing rTMS approaches in this study cannot rule out that gains observed in both groups were from rehabilitation alone. Exceeding gains of rehabilitation is a much harder standard to meet and would require larger studies in future with stratification. Third, the importance of the choice of task-practice cannot be overstated. In this pilot study, only the iM1 group received task practice directly linked to the stimulation target (contralateral M1 target and paretic UE practice), which may have had higher synergistic effects than in the cPMd group that received non-corresponding task practice (contralesional PMd target and ipsilateral paretic UE practice). Future studies should explore whether effects of cPMd stimulation in persons with severe hand impairment are potentiated with task practice of the corresponding limb, i.e., intact-hand mediated paretic UE therapy or bilateral therapies. The dose may also limit effects in severe paresis. In this study, brief 22-min rTMS sessions were combined with 12, 1-hr blocks of task-oriented therapy, but the ceiling for additive neurostimulation in severe stroke survivors may be higher.

In summary, cPMd facilitation rTMS was not better than iM1 facilitation rTMS in promoting UE rehabilitation outcomes in chronic stroke survivors with severe hand paresis, though both approaches yielded sustained gains. Future trials should use distal extension and active MEP status for stratification, test higher doses, and consider bimanual training to capitalize on preserved resources in severe stroke.

What is Known/What is New.

What is Known

• Repetitive transcranial magnetic stimulation (rTMS) is a noninvasive technique to increase activity of motor cortices for post-stroke upper extremity recovery.

• Experimental studies reveal contralesional (intact) dorsal premotor cortex (cPMd) is a better target than ipsilesional (damaged) primary motor cortex (iM1) for severe upper extremity paresis.

What is New

• Pilot trial demonstrates cPMd rTMS yields similar motor improvements as iM1 rTMS, both combined with rehabilitation, in persons with severe paresis.

• Wrist sub-score gains favor the iM1 group, which includes more participants with distal extension and viable corticospinal pathways.

• Findings inform patient selection for future efficacy trials of targeted treatments.

Data Availability statement

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.