Effects of Repetitive Transcranial Magnetic Stimulation on Motor Functions in Patients With Stroke
Background and Purpose—The purpose of this study was to perform a meta-analysis of studies that investigated the effects of repetitive transcranial magnetic stimulation (rTMS) on upper limb motor function in patients with stroke.
Methods—We searched for randomized controlled trials published between January 1990 and October 2011 in PubMed, Medline, Cochrane, and CINAHL using the following key words: stroke, cerebrovascular accident, and repetitive transcranial magnetic stimulation. The mean effect size and a 95% CI were estimated for the motor outcome and motor threshold using fixed and random effect models.
Results—Eighteen of the 34 candidate articles were included in this analysis. The selected studies involved a total of 392 patients. A significant effect size of 0.55 was found for motor outcome (95% CI, 0.37–0.72). Further subgroup analyses demonstrated more prominent effects for subcortical stroke (mean effect size, 0.73; 95% CI, 0.44–1.02) or studies applying low-frequency rTMS (mean effect size, 0.69; 95% CI, 0.42–0.95). Only 4 patients of the 18 articles included in this analysis reported adverse effects from rTMS.
Conclusions—rTMS has a positive effect on motor recovery in patients with stroke, especially for those with subcortical stroke. Low-frequency rTMS over the unaffected hemisphere may be more beneficial than high-frequency rTMS over the affected hemisphere. Recent limited data suggest that intermittent theta-burst stimulation over the affected hemisphere might be a useful intervention. Further well-designed studies in a larger population are required to better elucidate the differential roles of various rTMS protocols in stroke treatment.
- neuronal plasticity
- repetitive transcranial magnetic stimulation (rTMS)
- motor function
Stroke is the leading cause of death and the principle cause of long-term neurological disability worldwide.1–3 Approximately two thirds of patients with stroke have profoundly impaired motor function.4 Reduced upper limb function leads to significant disability that affects daily living and increases the burden on these patients and their families.5,6 However, various patterns of neural reorganization occur after stroke, and functional recovery is associated with neural plastic changes in the brain.7,8 Novel strategies that enhance beneficial plastic changes and improve recovery are emerging.9
According to the concept of interhemispheric competition, the equilibrium of cortical excitability within the 2 hemispheres changes after stroke.10–12 Cortical excitability and the representation area of the affected muscles are decreased in the affected hemisphere,13–15 whereas excitability in the unaffected hemisphere is enhanced. The abnormally increased interhemispheric inhibition driven from the unaffected to the affected hemisphere16–18 is associated with motor impairment.11,19 Rebalancing the cortical excitability between the hemispheres has been associated with a better overall prognosis.14 Based on this model, 2 therapeutic strategies to restore interhemispheric balance exist: enhancement of the cortical excitability of the affected hemisphere or inhibition of the cortical excitability of the unaffected hemisphere.
Repetitive transcranial magnetic stimulation (rTMS) is a painless, noninvasive method that modulates cortical excitability. High-frequency rTMS facilitates cortical excitability,20,21 whereas low-frequency rTMS decreases the cortical excitability of the stimulated hemisphere22,23 and increases the cortical excitability of the nonstimulated hemisphere.24–26 Theta burst stimulation (TBS), which is another form of rTMS, uses low-intensity stimulation to modulate cortical excitability. Intermittent pattern TBS (iTBS) enhances cortical excitability, and continuous TBS (cTBS) suppresses cortical excitability.27 Therefore, rTMS can be used to increase and decrease the cortical excitability of the affected and unaffected hemispheres, respectively, and may facilitate motor function after stroke.
Several trials have investigated the effect of rTMS on upper limb motor function in patients with stroke. High-frequency rTMS over the primary motor cortex (M1) in the affected hemisphere could improve motor learning performance in patients with chronic stroke28 and have a positive, long-term effect on motor recovery in acute29 and subacute30 patients with stroke. Ameli et al31 suggested that high-frequency rTMS has favorable effects on subcortical rather than patients with cortical stroke, and iTBS over the affected hemisphere was reported to improve motor behavior.32,33 Low-frequency rTMS to the unaffected hemisphere also has beneficial effects on hand dexterity,34 pinch acceleration,35,36 grip force,37 reaction time,38,39 and finger tapping.40 Some of these studies also measured the impact of rTMS on cortical excitability.28,29,35,36,38,41 However, other reports did not show measurable therapeutic effects of rTMS on motor function after stroke.42–44 Because of the inconsistent findings and methodological discrepancies across these trials, there is a lack of consensus regarding the effect of rTMS on motor recovery in patients with stroke.
The primary goal of the present study was to evaluate the effects of rTMS on upper limb motor recovery and cortical excitability in patients with stroke by systematically reviewing the available data. The secondary goals were to determine whether high- or low-frequency rTMS leads to enhanced motor recovery and to determine factors that may contribute to better motor outcomes. We also examined possible adverse effects of using rTMS in patients with stroke.
Computerized searches were performed in PubMed, Medline, Cochrane, and CINAHL to identify relevant studies. The search terms were stroke/cerebrovascular accident and repetitive transcranial magnetic stimulation/rTMS. Manual searches of the reference lists of the retrieved articles and pertinent reviews were also conducted. The searches were limited to human studies that were written in English and published between January 1990 and October 2011. Studies were included when the following criteria were met: (1) the patients were diagnosed with a stroke; (2) the patients were adults; (3) ≥5 patients were recruited; (4) the focus was placed on rTMS effects on upper limb motor function in patients with stroke; (5) the outcome measures were reported with continuous scales that evaluated the motor function of the affected hand; and (6) the study was a randomized controlled trial. Two reviewers independently searched and evaluated the literature for the inclusion of studies based on their titles and abstracts. We reviewed the full text of articles that appeared to be relevant.
To evaluate the methodological quality of the included studies, we used a modified checklist derived from a quality screening form that was revised by Moher et al.45 The quality of each study was assessed according to the following criteria: (1) random allocation; (2) blinding procedure; (3) dropout number; (4) description of baseline demographic data; (5) statistical comparison between the experiment and control groups; (6) reports of point estimates and measures of variability; and (7) description of adverse events. Randomization was recorded as 1 if the patients were randomly allocated into groups. The blinding procedure rating ranged from 0 to 2, in which 0 represented a nondescribed or nonblinded procedure and 1 and 2 indicated single-blind and double-blind procedures, respectively. Dropouts were recorded as the number of participants who withdrew from the study. Descriptions of baseline demographic data and statistical comparisons as well as point estimates and measures of variability were denoted as 1 if provided. Adverse effects were recorded as the number and type of adverse event.
The relevant data from each study were independently extracted by 2 reviewers using a standard data recording form that included study design, number of subjects, mean age, stroke duration, treatment protocol (ie, rTMS frequency, intensity, number of pulses, and additional interventions), dropout number, information regarding study quality (see previously), outcome measures, and mean differences and SDs of the change scores. The mean and SD of the change scores relative to the baseline or the posttreatment mean and SD were recorded for each outcome measure in the experimental and control groups. Various motor function assessment outcome measures were used across the studies, and some assessed multiple measures. For the purposes of this meta-analysis, the measure used to assess each study was the explicitly declared primary outcome. If the primary outcome was not clearly defined, the first outcome that was reported with a mean and SD in the results section was used. For cortical excitability, the motor threshold (MT) data in the affected hemisphere were extracted. The SD was calculated from the SEM in 1 study,31 estimated from a plot in another study,44 and calculated with 95% CIs provided for the MT data by Fregni et al.38 A few studies did not report the mean and SD for outcome measures,29,30,32,33,35,36 and the data were estimated from the figures. Six articles contributed >1 trial because they applied 2 rTMS protocols with different stimulus frequencies29,41,46 or different TBS patterns,32,33 or they divided patients into 2 subgroups.31
To determine the factors that may have influenced motor recovery, planned subgroup analyses were conducted by grouping the studies based on stimulus frequency (high versus low), poststroke duration (acute [<2 weeks] versus subacute [2 weeks to 6 months] versus chronic [>6 months]), and lesion site (nonspecified versus subcortical stroke).
Analyses were performed with Comprehensive Meta Analysis 2.0 software (Biostat Inc, Englewood, NJ). The standardized effect sizes and 95% CIs were calculated to test the results of the different trials. The effect sizes were calculated based on the differences between posttreatment evaluations,30,31,33,37,38,40,46 mean changes between pre- and posttreatments,39,43 or changes relative to the baseline32,34–36,42 in the experimental and control groups divided by the pooled SD. The posttreatment data were based on the first assessment performed after the intervention period, except for trials that only provided posttreatment data that were collected 2 weeks42 or 3 months41 after the intervention. Because effect sizes may be influenced by sample sizes and effects may be overestimated in studies with low numbers of patients, a weighting factor was applied that gave more weight to studies with larger samples. Finally, the mean effect sizes were obtained after combining the weighted effect size of each study. Absolute effect sizes that ranged from 0.2 to 0.49 were considered to be small,47 and a value of 0.5 was likely to be clinically meaningful.48
The heterogeneity across each effect size was evaluated with Q-statistics49 and the I2 index,50 which is useful for assessing consistency between trials.51 When significant heterogeneity was found by Q-statistics or when I2 was >50%, the random effect model was applied. By contrast, the fixed effect model was used when I2 was <50%. Begg and Mazumdar rank correlation and Egger regression test were used to assess possible publication bias. The statistical significance level was set at 0.05.
The searches yielded 706 citations. After exclusion based on title and abstract, 34 potentially relevant articles were obtained and evaluated by 2 independent reviewers, and 18 articles that met our inclusion criteria were identified.28–44,46 The other 16 articles were excluded for the following reasons: they were review articles or case reports, they addressed topics other than motor outcome, or they did not include a control condition.
Table 1 shows the main characteristics of the studies included in our meta-analysis. In total, 392 patients with stroke were included, and 370 were re-evaluated postintervention. Three studies recruited patients in the acute phase,29,34,41 3 studies focused on the subacute phase,30,37,40 and 7 other studies followed patients with chronic stroke.28,32,33,35,36,38,42 Regarding lesion sites, 6 trials only recruited patients with subcortical stroke,32,34–37,40 whereas the other studies recruited both patients with cortical stroke and those with subcortical stroke.28–30,33,38,39,41–44,46 Ameli et al31 subgrouped patients according to the lesion site (cortical versus subcortical stroke).
The studies used different interventions. Eight studies applied 1 Hz rTMS over the unaffected hemisphere,34–40,44 5 delivered high-frequency rTMS over the affected hemisphere,28–31,42 and 2 used both low- and high-frequency stimulations.41,46 Pomeroy and coworkers43 applied 1 Hz rTMS over the affected hemisphere. Two trials applied iTBS and cTBS to affected and unaffected hemispheres, respectively.32,33 Several methods of sham rTMS were applied. Six trials used a sham coil,32,34,38,39,42,43 8 studies changed the coil angle,28–30,33,35,36,41,46 3 studies delivered the stimulation over the vertex,31,37,40 and 1 study used 0 stimulation output.44 Additional treatments were applied in 7 studies.28,30,32,36,42–44
Various outcome measures were used in the selected articles. Kinematic motion analyses were performed in 5 studies,32,35,36,40,43 hand grip studies were reported in 2 trials,29,37 and the Wolf Motor Function Test was used in 2 articles.42,44 Other reported outcome measures included finger tapping,31,46 the Purdue Peg Board Test,39 reaction time,33,38 movement accuracy,28 the Nine Hole Peg Test,34 keyboard tapping,41 and the motricity index.30
Table 2 shows the quality assessment results for the included studies. Random allocation was achieved in all of the trials. Most of the studies were double-blind29,30,34–36,38,39,41,42,44,46 or single-blind.28,30,43 Only 4 articles did not report a blinding procedure.31,33,37,40 All but 1 study34 described baseline demographic data. Four studies had dropouts.28,29,39,43 Statistical comparisons were performed in all the articles; however, 8 of the studies did not provide point estimates or the variability of outcome measures.28–30,32,33,35,36,44
Thirteen of the 18 studies reported adverse effects. Only 1 trial found adverse events, including 2 patients with headaches, 1 patient with increased anxiety, and 1 patient with increased fatigue.38
Table 3 summarizes the outcome measures; the number of patients in the postintervention evaluation; and the mean, SD, and effect size of each study. Totally 23 effect sizes were obtained from the 17 articles involving 362 patients with stroke. The meta-analysis of motor outcome showed a statistically significant mean effect size of 0.55 (95% CI, 0.37–0.72; P<0.01). We did not find heterogeneity across the studies (Q=17.52; I2=0.00; P>0.05; Figure 1).
The rank correlation (τ=0.19; P>0.05) and regression test (intercept [b]=0.52; P>0.05) indicated negative results for publication bias. Kim et al28 have reported enhancement effect of high-frequency rTMS on movement accuracy and speed. However, they did not merge the data from the first and second session to provide appropriate mean and SD values for determining the effect size.
The impact of each trial on the mean effect size was also evaluated by performing a meta-analysis that excluded 1 trial at a time. Figure 2 shows the results of removing individual trials. The statistically significant effect size (>0.5) was obtained even when any 1 trial was eliminated.
Stimulus Frequency: High Frequency Versus Low Frequency
Because TBS is another form of rTMS, the subgroup analysis for stimulus frequency (high versus low) was only based on studies that applied high- or low-frequency rTMS. The analysis revealed a mean effect size of 0.69 (95% CI, 0.42–0.95; P<0.001) for patients who received low-frequency rTMS (Q=9.82; I2=0.00; P>0.05). The mean effect size for patients who received high-frequency rTMS was 0.41 (95% CI, 0.14–0.68; P<0.01) without heterogeneity (Q=4.4; I2=0.00; P>0.05; online-only Data (Supplement Table I).
Poststroke Duration: Acute Versus Subacute Versus Chronic
The subgrouped mean effect size for acute stroke was 0.79 (95% CI, 0.42–1.16; P<0.001) without heterogeneity (Q=0.54; I2=0.00; P>0.05). The mean effect size for subacute stroke was 0.63 (95% CI, 0.18–1.08; P<0.01) without heterogeneity (Q=1.27; I2=0.00; P>0.05). The mean effect size for chronic stroke was 0.66 (95% CI, 0.31–1.00; P<0.001) without heterogeneity (Q=4.82; I2=0.00; P>0.05; online-only Data Supplement Table I).
Lesion Site: Nonspecified Versus Subcortical Stroke
The subgrouped mean effect size for nonspecified lesion sites was 0.45 (95% CI, 0.23–0.67; P<0.001) without heterogeneity (Q=12.63; I2=0.00; P>0.05). The mean effect size for subcortical lesions was 0.73 (95% CI, 0.44–1.02; P<0.001); no heterogeneity was detected (Q=2.61; I2=0.00; P>0.05; online-only Data Supplement Table I).
The effect of rTMS on cortical excitability was evaluated based on MT data from the affected hemisphere. The meta-analysis for MT showed a mean effect size of 0.30 (95% CI, −0.09 to 0.68; P>0.05) without heterogeneity (Q=3.20; I2=0.00; P>0.05; Figure 3).
The present study provides evidence to support the efficacy of rTMS on motor recovery in patients with stroke. No statistical evidence was found for publication bias or heterogeneity, and the results remained significant after any 1 of the trials was removed.
The results of this meta-analysis suggest that rTMS might be helpful in improving upper limb motor function. Among the 362 patients with stroke of the studies analyzed in the present review, the significant mean effect size was 0.55, which is considered to be clinically significant.48 To determine factors that might influence the motor outcome, a subgroup analysis was performed. The mean effect size of low-frequency rTMS (mean effect size, 0.69) was larger than that of high-frequency rTMS (mean effect size, 0.41), which indicates that the motor improvement was more pronounced in studies that applied low-frequency rTMS to the unaffected hemisphere. The results support the concept that interhemispheric competition is altered after stroke,16,17 and this imbalance could be remedied by reducing the cortical excitability in the unaffected hemisphere.35,46 A previous study demonstrated a stronger effect of low-frequency rTMS (1 Hz) on motor recovery and reported decreased excitability of the intact hemisphere as well as concurrent enhanced excitability of the stroke hemisphere.41 Regardless, the present findings show a positive effect on motor recovery by low- or high-frequency rTMS.
A subgroup analysis for the effects of cTBS versus iTBS was not performed because the TBS data are limited. Talelli et al33 found that cTBS suppressed excitability in the unaffected hemisphere but did not improve paretic hand motor function; however, iTBS enhanced motor behavior. Ackerley et al reported that the excitability of the affected hemisphere was enhanced by iTBS but decreased by cTBS and that motor function deteriorated after cTBS. Although the subgroup analysis indicated a greater beneficial effect of low-frequency rTMS compared with high-frequency rTMS, the TBS studies revealed that iTBS may be more helpful for motor recovery than cTBS. The minor effects of cTBS on motor improvement might be the result of the low intensity of stimulation and subsequent inability to induce responses in the transcallosal connections.32 Further investigation is necessary to elucidate the underlying mechanism.
The poststroke duration subgroup analysis revealed a clinically significant effect for each stage of stroke. However, these results must be viewed with caution because only a few studies exclusively recruited patients according to stroke duration. The lesion site subgroup analysis showed a relatively larger effect size (mean effect size, 0.73) for patients with subcortical stroke when compared with patients who had nonspecified lesion sites (mean effect size, 0.45), indicating that patients with subcortical stroke benefit more from rTMS. Ameli et al investigated changes in neural activity in response to facilitatory rTMS over the ipsilesional M1 and found that patients with subcortical stroke showed reduced neural activity in the unaffected hemisphere that was associated with motor recovery. In contrast, patients with additional cortical stroke showed increased neural activity in the frontal and parietal motor areas that might counteract the effect of rTMS.31
The mean effect size for MT did not reach statistical significance. However, only a few trials evaluated MT (N=6), which makes it difficult to reach a definite conclusion regarding the effect of rTMS on MT in patients with stroke.
The effect sizes were calculated based on measurements that were recorded immediately after the intervention period in most of the studies.28,30–41,43,44,46 Only 7 studies followed up motor outcomes with durations that ranged from 2 weeks to 1 year.29,30,36,38,41,42,46 For these reasons, more research is required to determine whether the effects of rTMS are transient or sustained. All of the included studies were randomized controlled trials, and blinding procedures were described in most of them. Therefore, the results obtained in the present meta-analysis may be valid and clinically meaningful. However, a few studies did not provide point estimations or the variability of the outcome measures29,30,32,33,35,36,44; these data were estimated from their figures. The influence of nonprecise data on the mean effect size cannot be fully excluded.
Safety is an important consideration because rTMS could induce potential adverse effects such as headaches and seizures. Thus, we analyzed adverse effects in the present meta-analysis. Twelve of the studies reported that all patients tolerated the intervention well without any adverse events28–30,32,33,35,36,41–44,46; only 1 study reported 4 subjects with mild, benign side effects of headaches (n=2), anxiety (n=1), and fatigue (n=1).38 Although rTMS is generally thought to be safe in patients with stroke, investigators should follow safety guidelines52,53 and monitor the potential risk of poststroke seizure associated with rTMS.
In addition to safety concerns, some crucial issues should be taken into account: (1) Would the positive therapeutic effect be preserved when applying the milder stimulus intensity that might reduce the risk of potential complications related to rTMS? (2) Is motor function improvement accompanied by changes in neural activity? (3) Is it possible that some of the beneficial effects of rTMS might be the result of potential antidepressive actions of rTMS?54
Some limitations exist in the present study. First, methodological variations existed between the selected studies. The outcome measures, experimental designs (eg, randomization, parallel design, and crossover design), patient inclusion criteria, and rTMS protocols (eg, stimulus frequency, duration, intensity, and sham condition methods) all varied. Second, the applied additional treatments or motor function training28,30,32,36,42–44 in some of these studies represent potential confounders. Finally, we may have missed relevant studies that were published in languages other than English.
This meta-analysis study suggests a clinically positive effect of rTMS on motor recovery in the affected upper limb of patients with stroke. Low-frequency rTMS over the unaffected hemisphere is more effective than high-frequency rTMS over the affected hemisphere, which is compatible with the concept of interhemispheric inhibition. Compared with patients with cortical stroke, patients with subcortical stroke may benefit more from rTMS. Moreover, recent limited data suggest that iTBS over the stroke hemisphere might be a useful intervention. Further well-designed studies are necessary to determine the effect duration and the plasticity change of cortical excitability after individual rTMS protocols.
Sources of Funding
The study was supported in part by research grants from Taipei Veterans General Hospital (V97ER3-006; VGHUST97-P6-24; V97C1-034; V98C1-095; V98ER3-002; VGH-S4--98-001; VGH-ER3-99-006; V99C1-156; V100C-146) and from the National Science Council (NSC-95-2314-B-010-030-MY3; NSC-96-2628-B-010-030-MY3; NSC-98-2321-B-010-007; NSC-99-2321-B-010-004; NSC-99-2628-B-010--011-MY3); and a grant from the Ministry of Education (Aim for the Top University Plan), Taipei, Taiwan.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.111.649756/-/DC1.
- Received January 1, 2012.
- Revision received February 27, 2012.
- Accepted March 13, 2012.
- © 2012 American Heart Association, Inc.
- Kolominsky-Rabas PL,
- Weber M,
- Gefeller O,
- Neundoerfer B,
- Heuschmann PU
- Cicinelli P,
- Pasqualetti P,
- Zaccagnini M,
- Traversa R,
- Oliveri M,
- Rossini PM
- Shimizu T,
- Hosaki A,
- Hino T,
- Sato M,
- Komori T,
- Hirai S,
- et al
- Classen J,
- Schnitzler A,
- Binkofski F,
- Werhahn KJ,
- Kim YS,
- Kessler KR,
- et al
- Kobayashi M,
- Hutchinson S,
- Theoret H,
- Schlaug G,
- Pascual-Leone A
- Rossini PM,
- Rossi S
- Kim YH,
- You SH,
- Ko MH,
- Park JW,
- Lee KH,
- Jang SH,
- et al
- Ackerley SJ,
- Stinear CM,
- Barber PA,
- Byblow WD
- Takeuchi N,
- Chuma T,
- Matsuo Y,
- Watanabe I,
- Ikoma K
- Takeuchi N,
- Tada T,
- Toshima M,
- Chuma T,
- Matsuo Y,
- Ikoma K
- Fregni F,
- Boggio PS,
- Valle AC,
- Rocha RR,
- Duarte J,
- Ferreira MJ,
- et al
- Mansur CG,
- Fregni F,
- Boggio PS,
- Riberto M,
- Gallucci-Neto J,
- Santos CM,
- et al
- Nowak DA,
- Grefkes C,
- Dafotakis M,
- Eickhoff S,
- Kust J,
- Karbe H,
- et al
- Pomeroy VM,
- Cloud G,
- Tallis RC,
- Donaldson C,
- Nayak V,
- Miller S
- Higgins JP,
- Thompson SG,
- Deeks JJ,
- Altman DG
- Rossini PM,
- Johnston CS