Repetitive transcranial magnetic stimulation for motor function in stroke: a systematic review and meta-analysis of randomized controlled studies

Abstract

Objective

This study aimed to systematically evaluate the safety and effectiveness of repetitive transcranial magnetic stimulation (rTMS) in treating motor dysfunction in stroke patients.

Methods

A systematic search was conducted in five online databases, namely, Medline, EMBASE, the Cochrane Central Register of Controlled Trials (CENTRAL), CINAHL, and SPORTDiscus, from their inception to July 29, 2024. Studies meeting the predetermined inclusion criteria were included. The data were analyzed using RevMan 5.4.1 software and Stata 15.0. The subgroup analysis was conducted based on various disease stages and intervention frequencies. The overall effects were estimated using either the fixed effects model or the random effects model, with standardized mean differences (SMDs). The level of evidence was assessed using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) framework.

Results

A total of 70 studies encompassing 2951 stroke survivors were included. The results of the quantitative analysis revealed that the application of 1 Hz rTMS over the contralesional primary motor cortex (M1) significantly improved motor function during both the early stage (< 1 month) with moderate effect size (n = 443, SMD = 0.44, 95% CI 0.24 to 0.63, P < 0.00001, I² = 47%, fixed-effect model) and recovery period (1–6 months) with moderate effect size (n = 233, SMD = 0.61, 95% CI 0.34 to 0.87, P < 0.0001, I² = 33%, fixed-effect model). In the context of activities of daily living (ADLs), the application of 1 Hz rTMS over the contralesional M1 can lead to improvements in ADLs among individuals in the early stages of stroke with moderate effect size (n = 343, SMD = 0.67, 95% CI 0.44 to 0.89, I² = 79%, P < 0.00001, fixed-effect model). However, evidence to support that 1 Hz rTMS over contralesional M1 can improve motor dysfunction in the chronic phase of stroke (> 6 months) is insufficient.

Conclusion

Moderate- to high-quality evidence suggests that 1 Hz rTMS over the contralesional M1 may enhance motor function and independence in ADL during the early stages of stroke and the recovery period (within 6 months) with moderate effect. Nonetheless, as for the efficacy of 3, 5, 10, and 20 Hz rTMS in the treatment of motor dysfunction after stroke, it needs to be further determined. It is important to interpret these findings with caution in clinical practice due to the small sample sizes and low quality of the studies reviewed.

Systematic review registration

INPLASY, Registration number is INPLASY202360042. DOI number is https://doiorg.publicaciones.saludcastillayleon.es/10.37766/inplasy2023.6.0042.

Introduction

Stroke is a prominent contributor to mortality within the realm of noncommunicable diseases, and simultaneously serving as a primary instigator of incapacitation [1]. Approximately 34% of worldwide aggregate health care expenditures are allocated to stroke. Stroke affects individuals across various age groups, with over half of patients experiencing motor dysfunction. This impairment significantly impedes their ability to perform daily activities, consequently placing an increased burden on their families [1]. Approximately 80% of stroke patients experience motor dysfunction in either their upper or lower limbs following a stroke [2, 3]. Approximately 80% of individuals with mild paresis attain full upper limb function, whereas only 20% of those with severe paresis achieve the same level of function [2]. Only 50% of individuals who initially experienced upper limb paralysis exhibited partial motor recovery after a 6-month period [4, 5]. Approximately 70% of individuals with lower limb impairment are unable to walk independently soon following stroke, and only half of these individuals have independent walking function after rehabilitation [4, 5]. Upper and lower limb paresis stands out as a significant predictor of long-term functional recovery following a stroke [4, 5]. Consequently, numerous innovative therapeutic interventions have been suggested with the objective of augmenting motor recuperation [6]. Nevertheless, current medical interventions lack targeted therapies aim to restore function by repairing damaged tissue.

Noninvasive brain stimulation (NIBS) has been explored as a treatment for stroke populations [7,8,9]. There are several theoretical models that support the use of noninvasive brain stimulation techniques. Among them, the most prominent models are the interhemispheric competition model, the vicariation model, and the bimodal balance-recovery model. According to the interhemispheric competition model, a model of hemispheric interaction, mutual, balanced inhibition exists between the two hemispheres of the brain in healthy individuals [10, 11]. During stroke, damage to one hemisphere disrupts this balance: The affected hemisphere’s inhibition of the unaffected hemisphere weakens, and inhibition of the unaffected hemisphere consequently increases [12]. As a result, the affected hemisphere suffers from “double-disabled” and is therefore too inhibited [12]. In healthy individuals, interhemispheric inhibition (IHI) decreases before unilateral index finger movement begins, but reversibly shifts to promote movement [13, 14]. Stroke patients who have recovered relatively well can use the affected hand for movement of the index finger, and the lack of suppression-reversal promotion may interfere with the movement of the affected side, causing this mechanism to fail [13,14,15]. The vicariation model suggests that the activity of the unaffected hemisphere contributes to functional recovery after stroke, and this recombination pattern is called the compensatory model: That is, the activity of brain areas outside the lesion compensates for the function of the damaged brain area, including the activity of the contralateral undamaged hemisphere [16, 17]. The bimodal balance-recovery model introduces a new concept of structural reserve and defines it as the extent to which neural pathways and relays spared by the lesion contribute to recovery in an individual patient. The main idea of this model is that the degree of structural retention (e.g., retention of motor areas and the corticospinal tract) determines which of the interhemispheric competition models or the compensatory model is dominant. When degree of the structure retention is high, the interhemispheric competition model can better predict the degree of restoration. When the degree of structure retention is low, the compensatory model is dominant [15].

Transcranial magnetic stimulation (TMS), a form of NIBS, involves the application of an insulated coil to the scalp and targets a precise region of the brain. Upon the passage of an electrical current through this coil, the resulting magnetic signal can traverse the scalp and skull without any loss of intensity, thereby influencing brain metabolism and neural activity [7, 8]. This phenomenon induces the trans-synaptic activation of pyramidal cells, resulting in the generation of descending volleys within the pyramidal axons that project onto spinal motoneurons, commonly referred to as the corticospinal tract [18]. Motor neuron activation resulting from corticospinal projections induced by TMS can be measured as motor-evoked potentials (MEPs) using surface electromyography. The activation of motoneurons in response to corticospinal volleys elicited by TMS leads to contraction in a specific muscle, thereby evoking an MEP that is captured by electromyography (EMG) via surface electrodes placed over the muscle belly. The amplitude of the MEP’s peak-to-peak measurement is employed to estimate the excitability of the corticospinal tract [19, 20]. Substantial evidence suggests that the mechanism underlying the aftereffect of rTMS resembles to both long-term potentiation (LTP) and long-term depression (LTD) [19, 21]. Since the introduction of rTMS, a substantial body of research has been dedicated to investigating its impact on motor dysfunction following stroke, as evidenced by the growing number of scholarly publications over time. Currently, low-frequency rTMS (LF-rTMS, typically ≤ 1 Hz) is commonly administered over the unaffected cortex, whereas high-frequency rTMS (HF-rTMS, > 1 Hz) is typically administered over the affected cortex [22].

However, the findings of prior studies investigating the efficacy of rTMS for addressing motor impairment following stroke have yielded inconclusive and conflicting results. The initial Cochrane review did not provide support for the utilization of rTMS in stroke rehabilitation [23]. Conversely, another systematic review and meta-analysis conducted during the same time frame concluded that rTMS has a favorable impact on motor recovery among stroke patients [24]. The ongoing debate surrounding the impact of rTMS on motor function following stroke persists. A recent systematic review, published in 2022, highlighted the potential of rTMS therapy to enhance the outcomes of poststroke sequelae, specifically by ameliorating upper limb function, hand dexterity, and muscle tonicity [25]. However, a subsequent systematic review conducted in the same year demonstrated that NIBS, which includes rTMS, was unable to effectively alter the Fugl-Meyer Assessment Upper Limb (FMA-UL) subscale score among individuals with chronic stroke [26]. Furthermore, significant variability exists in the findings of contemporary research regarding the optimal timing of rTMS intervention. A systematic review has suggested that the efficacy of excitatory rTMS in promoting upper limb motor function recovery is limited to the initial 3 months following a stroke [27]. Interestingly, another study revealed that the effects of real and sham rTMS did not significantly differ within a 3-month period following a stroke [28]. According to a recent systematic review, rTMS significantly improved in upper limb and fine motor function in the short term (0–1 month) and medium term (2–5 months). However, no significant improvement was observed in long-term upper limb function (6 months or more) [25].

In addition, the rapid expansion of publications and ongoing research have exacerbated the challenge of remaining current with empirical contributions and extensive research trajectories. As a result, synthesizing definitive evidence from previous studies is increasingly intricate. Within this framework, reviews play a crucial role in aggregating literature findings, enhancing the knowledge base, identifying theoretical, practical, and methodological gaps, and refining research agendas. However, the issue of ensuring reliability and objectivity in systematic reviews remains a significant challenge that necessitates resolution. To enhance the reliability and transparency of systematic reviews, numerous ongoing studies advocate for the integration of both quantitative and qualitative methodologies. These approaches aim to improve systematic reviews by systematically reorganizing and synthesizing the findings from existing literature [29]. Thus, this study carried out systematic review and meta-analysis to assess the effectiveness and safety of rTMS in addressing motor dysfunction following stroke which is imperative, with the aim of offering valuable insights for the healthcare decisions, patients, and their families.

Methods

Protocol and registration

The protocol for this systematic review was registered with the International Platform of Registered Systematic Review and Meta-analysis Protocols (INPLASY, Registration number is INPLASY202360042. DOI number is https://doiorg.publicaciones.saludcastillayleon.es/10.37766/inplasy2023.6.0042). The entire study was conducted in strict adherence to the guidelines outlined in the Cochrane Handbook for Systematic Reviews of Interventions.

Retrieval strategy

A systematic search was conducted in five online electronic databases, namely, EMBASE (via embase.com), Medline (via PubMed), the Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library, CINAHL, and SPORTDiscus (via EBSCO host), from their inception to September 27, 2023, without any language limitations. Following the completion of the data analysis, the literature was updated until July 29, 2024. Additionally, other research reports were identified by examining the reference lists of the articles. The search strategy employed for Medline via PubMed is presented in Table 1. The search strategy employed for additional databases can be found online in the Supplementary Information (SI).

Table 1 Search strategy of Medline (via PubMed)

Inclusion and exclusion criteria

The inclusion and exclusion criteria for the studies were defined as follows:

• (1) Population

  Studies that exclusively examined adult stroke patients, without any limitations on the type, severity, location, or progression of the disease, were included. Studies that included multiple conditions, such as stroke combined with Parkinson’s disease or spinal cord injury, were excluded from consideration.

• (2) Intervention

  The intervention utilized in this study involved the application of simple rTMS, specifically low-frequency rTMS or high-frequency rTMS. Patterned rTMS techniques, such as theta burst stimulation (θTBS), continuous TBS (cTBS), intermittent TBS (iTBS), paired pulse stimulation (PPS), quadripulse stimulation (QPS), and paired associative stimulation (PAS), were not included. Importantly, the intervention consisted of a single mode of rTMS and did not involve the combination of two or more different rTMS techniques (e.g., LF-rTMS combined with HF-rTMS). Only studies that utilized rTMS, rTMS combined with rehabilitation therapy, or rTMS combined with conventional medication therapy were included in the experimental group. Studies that combined rTMS with electrical stimulation (e.g., transcranial direct current stimulation (tDCS) or transcutaneous electrical nerve stimulation (TENS)) or rTMS with specific drugs (excluding routine medications for stroke) were excluded. The intervention parameters, including time, coil position, frequency, intensity, and duration, were not restricted.

• (3) Comparison

  Control groups included sham-rTMS, sham plus rehabilitation, rehabilitation, or no treatment. Studies that did not have appropriate control groups or sham intervention groups were excluded from the analysis [30]. Studies that utilized patterned rTMS as an intervention for the sham group were excluded from the analysis [31].

• (4) Outcome

  The primary outcome of this study was motor function, with no limitations on the instruments used for measurement (e.g., FMA-UL, Fugl-Meyer Assessment for lower limb (FMA-LL), and grip strength). The secondary outcomes included activities of daily living (ADLs), adverse events, and dropout, with no restrictions on the measurement tools employed.

• (5) Study

  In terms of study design, only randomized controlled trials (RCTs) were included in this study. In the case of cross-over studies, only the initial segment that adhered to the study protocol was considered. The minimum requirement for total participants was set to 6. The inclusion criteria were restricted to publications in the English language.

Study selection and data extraction

Acquisition, screening, and data extraction were performed between March 2023 and September 2023, and this process was updated to July 29 2024. All acquired documentation was carried out using Endnote X7. Following the removal of duplicate references, two authors (XGL and TW) independently assessed the title and abstract of the literature reports to identify potentially eligible studies. The final selection of included studies was based on reading the full text, in accordance with the preestablished inclusion criteria.

The data extraction process was conducted by a single researcher (DL) using a preestablished table, which was subsequently reviewed by another researcher (ZLM). The extracted data encompassed various components, including but not limited to the following: (1) Fundamental details (such as authors, title, publication date, number of participants, sex, age, duration, stroke location, and severity). (2) Methodological quality indicators (such as the randomization method, allocation concealment, operator blinding, presence of a sham stimulation group, statistician blinding, and intention-to-treat analysis). (3) The interventions included various parameters, such as the placement and shape of the coil, as well as the intensity, frequency, duration, and follow-up period. (4) The outcomes of interest included both primary and secondary outcomes, which were assessed using specific evaluation tools. Additionally, follow-up assessments, drop-out rates, and adverse events were also considered. In cases where raw data were not provided, data extraction software (Getdata 2.2) was used to obtain relevant information from statistical charts, such as line charts or bar charts (SHM and LL). To ensure the precision of data extraction, data were exclusively employed for analysis when the consistency of independently extracted data by two authors surpasses 95%. Any missing data were acquired through communication with the original author via email. If no mean was reported, the median was considered as the mean. If no standard deviation was reported, then the standard deviation was calculated from the confidence interval or interquartile range. Data were managed and transformed using Microsoft Excel. Data were analyzed using RevMan 5.4 and Stata (version 15.0). In instances of incongruity during the literature screening and data extraction process, consensus was reached after consulting with a third author possessing substantial expertise (LJL).

Research quality assessment

The methodological quality of the included studies was assessed by two independent reviewers (ZCY and LL) using the “Risk of Bias 2” table. This table evaluated six domains, namely, selection bias, performance bias, detection bias, attrition bias, reporting bias, and other biases. Each domain consisted of one or more entries. Selection bias included random sequence generation and allocation concealment, whereas performance bias involved the blinding of participants, rTMS intervention operators, and rehabilitation therapists. Detection bias encompasses the blinding of outcome assessments and data analysts, whereas attrition bias encompasses incomplete outcome data, reporting bias, and selective reporting. To evaluate the extent of bias, each entry in every study was categorized as low risk, high risk, or unclear. In instances of incongruity during process of risk assessment, consensus was reached after consulting with a third author possessing substantial expertise (LJL).

Data analysis

Overall effects

The Cochrane Review Manager (RevMan 5.4) and Stata 15.0 software programs were used for data analysis. To evaluate the overall effects, standardized mean differences (SMDs) or mean differences (MDs) with 95% confidence intervals (95% CIs) were employed for continuous data. According to Cochrane guidelines and literature [32, 33], the SMD should be used to evaluate overall effects in specific scenarios to mitigate the impact of inter-study variability on results. These scenarios include as follows: firstly, when different measurement units are used to assess the same outcome; secondly, when diverse measurement tools and methodologies are applied to the same outcome; and thirdly, when there is significant variability or a doubling in outcome values across different studies. In situations not characterized by these conditions, the MD is recommended for data synthesis. In this study, although a consistent measurement tool was used for assessing the primary outcome of motor function (encompassing both upper and lower limbs), the reported values across studies demonstrated considerable variability. Consequently, the SMD was employed as the index for data synthesis. Furthermore, for outcome measures such as motor function assessed via the Modified Ashworth Scale (MAS) or Ashworth Scale (AS), activities of daily living (ADLs) (via Modified Barthel Index(MBI) or Barthel Index (BI)), and neurological function (via National Institute of Health stroke scale (NHISS), resting motor threshold (RMT), or motor evoked potentials(MEPs)), the use of differing measurement units or tools necessitated the selection of SMD for these outcomes as well. According to the Cochrane criteria, an absolute SMD value of approximately 0.2 is indicative of a small effect, around 0.5 represents a moderate effect, and a value of 0.8 or higher indicates a large effect size. The pooled effect was estimated using either the fixed-effect model or random-effect model, depending on the results of chi-square test and Higgins I² value. Statistical heterogeneity was assessed using the chi-square test and Higgins I² value. A significant level of statistical heterogeneity was observed: if a chi-square test results P < 0.1 and I² > 75%, in which case a random effects model is used; otherwise, a fixed effects model is used.

Subgroup analysis

Subgroup analysis was conducted to identify clinical heterogeneity based on the varying frequencies of rTMS (1 Hz, 3 Hz, 5 Hz, 10 Hz, 20 Hz) and different disease periods (< 1 month, 1–6 months, > 6 months). The duration of early rehabilitation was defined as less than 1 month from stroke onset, the recovery period was considered to be between 1 and 6 months, and chronic rehabilitation was defined as more than 6 months, following the guidelines [34]. In studies encompassing patients with stroke episodes spanning multiple time periods as defined by our study, we calculated the mean stroke time or percentage of the population in each period to ascertain the specific time period under consideration.

Sensitivity analysis

The robustness of results was assessed using sensitivity analysis. In this study, a comprehensive sensitivity analysis was conducted at each decision node, encompassing literature search, literature inclusion, data extraction, and data analysis, to evaluate the impact of these decisions on the study results. Detailed analytical procedures and corresponding processing methods are documented in Supplementary Information (SI) Table S1. At the statistical method selection stage, either a random effects model or fixed effects model was applied to evaluate the overall effect. Furthermore, Stata software (Version 15.0) was utilized to sequentially exclude studies one by one and subsequently re-perform the meta-analysis to evaluate the robustness of the results. A significant alteration in the pooled effect size following the exclusion of a specific study suggested its considerable impact on the overall findings.

Publication bias

In cases where the reported outcomes consisted of more than five studies or involved a participant pool exceeding 250 individuals, Egger’s test was conducted to investigate the presence of publication bias.

Grading the quality of evidence

The evidence level was evaluated via the GRADEprofile software version 3.6, in accordance with the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) framework. The assessment considered five factors for downgrading the evidence (risk of bias, inconsistency, indirectness, imprecision, publication bias) and three factors for upgrading the evidence (large effect, plausible confounding that could alter the effect, dose‒response gradient). Based on this evaluation, the evidence quality was categorized as high, moderate, low, or very low.

Results

Study identification

A comprehensive search of online databases yielded a total of 7006 articles. Among these, 3069 records were identified as duplicate reports, while 3937 records were deemed irrelevant based on abstracts and titles by two reviewers. Subsequently, 176 full texts were carefully screened, resulting in the exclusion of 111 records that did not meet the inclusion criteria. Additionally, five records were included in this study through reference screening. A total of 70 RCTs meeting the predetermined inclusion criteria were ultimately included in this systematic review. Not all studies included in this research reported the necessary outcome indicators for quantitative analysis, specifically FMA-UL, FMA-LL, MAS/AS, MBI/BI, NIHSS, and RMT. To address this gap, the original authors of these studies were contacted via email to inquire about the availability of the relevant data. By the time of revision, responses were received from nine authors, all of whom indicated that they had not collected the specified outcome indicators. Ten authors did not respond. As a result, the 19 studies lacking these indicators were included exclusively in the qualitative analysis, while the remaining 51 papers were incorporated into both qualitative and quantitative analyses. The complete procedure of the literature screening is depicted in Fig. 1.

Fig. 1

Flow diagram of the study selection process

Study characteristics

(1) Population

The pertinent details of the studies included in this analysis are presented in Table 2 and Table 3. The aggregate number of participants encompassed 2951 individuals, with participant counts varying between 9 and 240. Within the experimental group, the number of participants ranged from 4 to 132, whereas the control group included participants ranging from 5 to 160. The female proportion ranged from 13.33 to 78.38%, and the mean age of participants ranged from 49.7 to 79 years. The study included patients with illness durations ranging from 6 h to more than 12 months, with 5 studies failing to report the type of stroke [35,36,37,38,39], 30 studies including only ischemic stroke patients [28, 40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68], and 35 studies including both ischemic and hemorrhagic stroke patients [39, 69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102]. Among the included studies, 19 studies did not provide information on the severity of the patients [36, 37, 39,40,41, 43, 45, 49, 54,55,56,57, 60, 61, 63, 65, 68, 86, 90], while the remaining studies reported the severity of the included patients. Of these, seventeen studies included patients with mild to moderate or to severe impairment [28, 38, 42, 45,46,47,48, 58, 59, 69, 77, 81,82,83, 95, 98, 99], and various assessment methods, such as independent walking of 10 m [80, 102], or walk 10 m with assistance [80], minimum 10° of volitional flexion and extension of fingers and wrist in the affected limb [67], the Brunnstrom scale [28, 51, 62, 71, 77, 97, 99,100,101], Fugl-Meyer Assessment (FMA) [58, 74, 91, 99, 103], functional ambulation classification (FAC) [72, 79, 96], National Institute of Health stroke scale (NHISS) [46, 64, 92, 98], Modified Rankin Scale (MRS) [35], Timed Up and Go Test (TUGT) [78], Movement-related cortical potential (MRCP) [58], the Medical Research Council (MRC) Scale [59], Glasgow Coma Scale (GCS) [87], Chedoke–McMaster [44], Modified Ashworth Scale (MAS) [62, 70, 71, 88], Lovett scale [52], and Bruininks–Oseretsky Test (BOT) [94], were employed to determine the functional status of the included patients in other studies.

Table 2 Characteristics of included studies in this systematic review

Table 3 Characteristics of included studies in this systematic review (continue)

(2) Intervention

In the experimental group, eight studies used transcranial magnetic stimulation alone [41, 44, 48, 58, 72, 80, 85, 88]. Other studies integrated a combination of rTMS with diverse rehabilitation training approaches, including physical therapy (PT), occupational therapy (OT), pinching tasks, and constraint-induced therapy, augmented reality gait adaptive training, and augmented reality gait adaptive training. The types of TMS coils used in these studies included 8-figure collars, H-type coils [80], double-cone coils. One trial did not report the type of coil used [44]. The frequencies used in rTMS included 0.1 Hz, 0.5 Hz, 1 Hz, 3 Hz, 5 Hz, 10 Hz, and 20 Hz. The coil was positioned over the primary motor cortex (M1) in 58 studies, over the dorsal premotor cortex (PMd) (located 3 cm anterior to the M1 cortex) in one study [89], over the surface of the scape of the left dorsolateral prefrontal cortex (DLPFC) in one study [102], 2 cm below the inion and 2 cm lateral to the midline on the cerebellar hemisphere ipsilateral to the ataxic side, with the handle pointing superiorly, targeting the posterior cerebellar lobe in one study [57], and over the dorsal premotor (the site 2.5 cm anterior to the motor “hot spot” area of the first dorsal interosseous (FDI) muscle) in one study [88]. Specifically, in a total of 55 studies employing 1 Hz rTMS, the stimulation site was the contralateral M1. Four studies utilized 3 Hz rTMS, with two studies stimulating the contralateral M1 and two studies stimulating the ipsilesional M1. Furthermore, eight studies employed 5 Hz transcranial magnetic stimulation, with five studies stimulating the ipsilesional M1, two studies stimulating the ipsilesional M1, and one study stimulated the DLPFC. Ten studies utilized 10 Hz rTMS, with three studies stimulated the contralateral M1, five studies stimulated the ipsilesional M1, one study stimulated the bilateral leg motor areas, and one study stimulated the FDI. The three studies employing 20 Hz rTMS stimulated the ipsilesional M1. The intensity of stimulation varied between 200 and 7500 pulses. The times of intervention sessions ranged from twice per week to six times per week, whereas the total number of intervention sessions ranged from one to 40 sessions. Two studies employed the active motor threshold (AMT) to determine intervention intensity [45, 78], whereas other studies used the resting motor threshold (RMT). The intervention intensity used in the two AMT studies was set at 120% AMT and 90% AMT. The intervention intensity in the other studies varied from 80 to 120% of RMT. The duration of each intervention varied between 2 and 30 min. Thirty-three studies conducted interventions without any subsequent follow-up [35, 38,39,40, 45, 46, 51, 58,59,60,61,62, 65, 67, 71, 73, 74, 79, 84, 86, 87, 89,90,91, 93,94,95,96,97, 99, 101,102,103]. The remaining studies had follow-up periods ranging from 7 days to 18 months.

(3) Comparison

Among the 68 studies included, nine studies compared rTMS with sham rTMS [41, 48, 58, 72, 80, 85, 88, 94, 95], whereas sixteen studies compares rTMS with rehabilitation [37, 39, 40, 45, 51, 52, 62, 67, 71, 74, 92, 93, 96, 97, 99, 103]. Additionally, other studies compared rTMS with a combination of sham and rehabilitation. The rehabilitation interventions included physical therapy (PT), occupational therapy (OT), motor task practice, hand motor training, pinching task, and conventional rehabilitation, augmented reality gait adaptive training, and augmented reality gait adaptive training.

(4) Outcome

Motor function was assessed using a variety of measures, including the FMA, FMA-UL, FMA-LL, MAS, Box and Block test (BBT), Brunnstrom Recovery Stages (BRS), range of motion (ROM), ten-meter walking test (10MT), six-min walking test (6MWT), Jebsen-Taylor Hand Function Test (JHFT), RMT, grip strength, Pegboard Tasks, Ankle dorsiflexion, Hip flexion, Berg Balance Scale (BBS), Action Research Arm Test (ARAT), Postural Assessment Scale for Stroke (PASS), balance subscale of the Tinetti Performance Oriented Mobility (POMA-b), Wolf Motor Function Test (WMFT), TUGT, and pinch strength. The assessment of ADLs involved the utilization of the Modified Barthel Index (MBI) and Barthel Index (BI). Additionally, the evaluation of quality of life involved the use of various instruments, such as NHISS, Stroke Impact Scale (SIS), Stroke Specific Quality of Life Scale (SSQOL), and Euro-QoL Five Dimensions Questionnaire (EQ-5D).

Methodological quality of the studies

The risk of bias summary for the included studies is presented in Fig. 2. Among the included studies, twenty-eight trials did not provide information on randomization allocation [36, 41, 43, 45, 48, 55, 58, 60,61,62,63, 65, 66, 72, 74, 76, 78,79,80,81, 84, 85, 87, 91, 95, 98, 100, 103]. As a result, these studies were evaluated as having an “unclear” risk of bias concerning the “Random Sequence Generation (Selection Bias)” criterion. In contrast, forty-one studies reported employing various randomization methods, such as sealed envelopes [28, 35, 46, 47, 50, 54, 56, 57, 82, 83, 89, 99], blank-coded magnetic cards [80], random number computerized generator [37, 39, 40, 42, 44, 49, 51, 53, 59, 67, 69, 73, 75, 77, 88, 90, 92,93,94, 96, 97, 101], website [70], tables of random number [68, 71, 102], and simple random sampling method [38, 38]. These studies were identified as having a “low” risk of bias in the “Random sequence generation (selection bias)” category. One trial, which assigned patients based on the date of admission, was identified as having a high risk of bias [86]. A total of twenty-three studies provided information regarding the concealment of allocation, employing methods such as sealed envelopes [28, 35,36,37, 43, 46, 47, 49, 50, 54, 56, 57, 64, 73, 82, 88, 89, 97, 99], locked cabinet [42], and independent secretary [76]. Therefore, these studies were assessed as having a “low” risk of bias in the domain of “Allocation concealment (selection bias)”. Conversely, other studies were identified as having unclear risks in this area. Thirty-two studies reported the blinding of participants [36, 39, 42,43,44, 47, 49, 50, 55,56,57, 59, 63, 64, 69, 70, 73,74,75,76, 78, 79, 82, 85,86,87,88, 92, 95, 98], indicating a low risk of bias in the domain of “Blinding of participants (performance bias)”. Furthermore, five trials implemented rTMS operator blindness [39, 57, 73, 80, 90], demonstrating a low risk of bias in the category of “Blinding of rTMS intervention operators (performance bias)”. Additionally, twenty-two studies implemented the blinding of rehabilitation therapists [36, 42, 43, 47, 49, 57, 64, 69,70,71, 75,76,77,78,79, 82,83,84, 86, 89, 92, 98], demonstrating a low risk of bias in the area of “Blinding of rehabilitation therapists (performance bias)”. However, eight studies focused solely on comparing the efficacy difference between rTMS and sham interventions, without conducting any rehabilitation intervention [41, 48, 58, 72, 80, 85, 88, 95]. Consequently, these studies were categorized as not applicable in terms of “blinding of rehabilitation therapists”, resulting in a blank box in the risk of bias summary. A total of forty-four studies provided explicit documentation of the blinding of the outcome assessmenter [28, 35, 36, 39, 40, 42,43,44, 47,48,49,50, 53,54,55,56,57,58, 62,63,64, 67, 69,70,71, 74,75,76,77,78,79, 81,82,83, 86,87,88, 90, 92, 95,96,97,98, 102], indicating a low risk of bias in the category of “Blinding of outcome assessment (detection bias)”. In contrast, only nineteen studies reported the blinding of the data analyst [36, 40, 43, 47, 59, 63, 65,66,67, 70, 75, 76, 79, 81, 88, 90, 97, 98, 100], demonstrating a low risk of bias in the item of “Blinding of data analysts (detection bias)” category. Regarding incomplete outcome data, one trial reported that thirteen participants were excluded from the analysis, primarily because of incomplete data sets [35], which presents a high risk of bias in the “Incomplete outcome data (attrition bias)” category. Furthermore, nineteen studies were considered as having a high risk for other biases, due to factors such as small sample size, poor randomization and lack of blinding [38, 40,41,42, 45, 46, 48, 59, 60, 62, 68, 71, 72, 76, 78, 80, 86, 98, 99].

Fig. 2

Risk of bias graph: review authors’ judgements about each risk of bias item presented as percentages across all included studies

Effects of intervention

Motor function of the upper limb

Twenty-two studies, involving a total of 950 stroke patients, examined the effects of 1 Hz rTMS over the contralesional M1 on the motor function of the upper limb in stroke survivors measured using FMA-UL. The findings of these studies revealed that 1 Hz rTMS over the contralesional M1 was associated with improvements in motor function among stroke patients in the early stage (< 1 month) with a moderate effect size (study number: 10, n = 443, SMD = 0.44, 95% CI 0.24 to 0.63, P < 0.00001, I² = 47%, fixed-effect model) as well as the recovery phase (1–6 months) with a moderate effect size (study number: 8, n = 233, SMD = 0.61, 95% CI 0.34 to 0.87, P < 0.00001, I² = 33%, fixed-effect model). However, the available evidence does not support the notion that 1 Hz rTMS can enhance motor function in individuals with chronic stroke (> 6 months) with a small effect size (study number: 4, n = 274, SMD = 0.13, 95% CI − 0.11 to 0.36, P = 0.28, I² = 0%, the fixed-effect model) by increasing FMA-UL scores (as shown in Fig. 3A and the SI Table S2). Moreover, six trials investigated the effects of 10 Hz rTMS on stroke patients. The overall results showed that 10 Hz rTMS over the ipsilesional M1 can improve the motor function of upper limbers among stroke patients in the early stage (< 1 month) with a large effect size (study number: 3, n = 229, SMD = 0.75, 95% CI 0.49 to 1.02, P < 0. 00001, I² = 0%, fixed-effect model, Fig. 3B and SI Table S2) using the FMA-UL score. The included studies employed various assessment tools to evaluate upper limb motor function, such as the grip strength of limb, JTHFT, the Nine-Hole Peg Test (NHPT), the Purdue Pegboard test (PPT), WMFT, BBT, and the Action Research Arm Test (ARAT).

Fig. 3

A Meta-analysis of 1 Hz rTMS over contralesional M1 for stroke on FMA-UL. B Meta-analysis of 10 Hz rTMS over ipsilesional M1 for stroke on FMA-UL. C Meta-analysis of 1 Hz rTMS over contralesional M1 for stroke on FMA-LL

Motor function of the lower limb

Eight trials were conducted to investigate the impact of 1 Hz rTMS over the contralesional M1 on the motor function of the lower limb in stroke survivors, as measured by FMA-LL. The findings of these trials revealed that 1 Hz rTMS has the potential to increase the FMA-LL score in both early-stage stroke patients (< 1 month) with a moderate effect size (n = 166, SMD = 0.67, 95% CI 0.35 to 0.99, P < 0.0001, I² = 60%, fixed-effect model, Fig. 3C and SI Table S2) and recovery phase patients (1–6 months) with a moderate effect size (n = 223, SMD = 0.59, 95% CI 0.32 to 0.86, P < 0.0001, I² = 0%, fixed-effect model, Fig. 3C and SI Table S2). The included studies employed additional measures to evaluate lower limb motor function, including the10MWT, the 6MWT and the Brunnstrom-Low test, with small samples.

Motor function using MAS/AS

MAS/AS was used to assess motor function in six studies. The results did not support that 1 Hz rTMS over the contralesional M1 could improve the MAS score for stroke less than 1 month with a small effect size (n = 58, SMD = 0.03, 95% CI − 0.49 to 0.54, P = 0.39, I² = 0%, the fixed-effect model, SI Table S2), for the stroke with a small effect size (1–6 months) (n = 107, SMD = − 0.15, 95% CI − 0.54 to 0.23, P = 0.25, I² = 28%, the fixed-effect model, SI Table S2), and for stroke more than 6 months with a moderate effect size (n = 21, SMD = − 0.49, 95% CI − 1.36 to 0.39, P = 0.27, the fixed-effect model, SI Table S2).

Activities of daily living (ADLs)

Fifteen studies were conducted to investigate the effects of 1 Hz rTMS over the contralesional M1 on ADLs in individuals with stroke using the MBI or BI. The results indicated that 1 Hz rTMS over the contralesional M1 was found to significantly improve ADLs in individuals with stroke within 1 month of onset with a moderate effect size (n = 343, SMD = 0.67, 95% CI 0.44 to 0.89, I² = 79%, P < 0.00001, fixed-effect model, Fig. 4A and SI Table S2). However, no significant improvement in ADL was observed for individuals with stroke beyond 1 month (Fig. 4A and SI Table S2). Additionally, five studies examined the effect of 10 Hz rTMS on stroke and the results demonstrated that 10 Hz rTMS over the ipsilesional M1 can improve ADLs in individuals with stroke within 1 month of stroke onset with a large effect size (n = 160, SMD = 1.98, 95% CI 1.60 to 2.36, P < 0. 00001, fixed-effect model, Fig. 4B and SI Table S2) and with the relatively small sample size. However, evidence suggesting that 3 Hz or 5 Hz rTMS can improve the ADLs for stroke patients between the 1–6 months is a lacking (SI Table S2).

Fig. 4

A Meta-analysis of 1 Hz rTMS over contralesional M1 for stroke on MBI. B Meta-analysis of 10 Hz rTMS (contralesional M1 or ipsilesional M1) for stroke on MBI

Neurological function

There is no evidence to suggest that 1 Hz rTMS over contralesional M1 can improve the NHISS score for the stroke-related neurological impairment with a small effect (P = 0.18, P = 0.61 and P = 0.73, respectively, as shown in Fig. 5A and SI Table S2). However, four studies examining the effects of 3 Hz rTMS on stroke patients within 1 month of onset indicated that 3 Hz can improve the NHISS score with a large effect (n = 154, SMD = − 0.73, 95% CI − 1.06 to − 0.41, I² = 0%, P < 0.0001, fixed-effect model, as shown in Fig. 5B and SI Table S2).

Fig. 5

A Meta-analysis of 1 Hz rTMS over contralesional M1 for stroke on NHISS. B Meta-analysis of 3 Hz rTMS over ipsilesional M1 for stroke on NHISS

Using RMT of unaffected hemisphere, 1 Hz rTMS over the contralesional M1 has a supportive effect on the recovery of unaffected hemispheres in stroke patients within 1 month of onset with a moderate effect (n = 123, SMD = 0.67, 95% CI 0.30 to1.03, I² = 0%, P = 0.0003, fixed-effect model, Fig. 6A and SI Table S1). The results suggest that 1 Hz rTMS over the contralesional M1 can improve the amplitude of contralesional motor evoked potentials (MEPs) in individuals with stroke within 1 month of onset with a moderate effect (n = 167, SMD = − 0.58, 95% CI − 0.90 to − 0.27, I² = 0%, P = 0.0003, fixed-effect model, Fig. 6B and SI Table S2). Furthermore, the study revealed that 1 Hz rTMS improve the amplitude of ipsilesional MEPs in individuals with stroke within 1 month of onset with a large effect (n = 121, SMD = 1.09, 95% CI 0.70 to 1.48, I² = 50%, P < 0.0001, fixed-effect model, Fig. 6C and SI Table S2).

Fig. 6

A Meta-analysis of 1 Hz rTMS over contralesional M1 for stroke on rMT of unaffected hemisphere. B Meta-analysis of 1 Hz over contralesional M1 for stroke on amplitude of contralesional MEPs. (C) Meta-analysis of 1 Hz over contralesional M1 for stroke of Latency of contralesional MEPs

Drop out

Information for the drop-offs in each study is shown in Table 3. A total of twenty-seven studies carried out intent-to-treat analyze [28, 39, 44, 48, 51, 52, 54,55,56,57, 64, 67, 74,75,76, 83, 84, 86, 87, 90,91,92, 94, 99, 100, 102, 103]. Eight studies did not report information of on drop-off [45,46,47, 65, 66, 69, 71, 85]. Twenty-seven studies reported that all patients completed protocols [39,40,41, 51, 52, 55, 58, 60, 67, 78, 81, 82, 84,85,86, 88,89,90, 92, 94, 98, 99, 102, 103].

Adverse events

Twenty-two studies did not report information on adverse events [39,40,41, 45, 46, 50,51,52, 61,62,63, 69, 71, 72, 78, 81, 83, 95, 98, 99, 102, 104], whereas twenty-three studies reported that no patients experienced any adverse reactions [49, 54, 56,57,58, 65,66,67, 70, 74, 76, 81,82,83, 85,86,87,88,89, 94, 100, 101, 103]. However, 21 studies reported various adverse events, such as epileptic seizures, headaches, anxiety, tiredness, pain, spasm, myalgia, paresthesia, and drowsiness [28, 35,36,37, 42,43,44, 47, 48, 53, 59, 60, 64, 73, 75, 77,78,79,80, 90, 96].

Sensitivity analysis

The detailed analysis process and the treatment taken are shown in the SI Table S1. The random-effects model was employed instead of the fixed-effects model for subgroups exhibiting substantial heterogeneity across studies. The robustness of the findings was evaluated via a sensitivity analysis, in which each included study was systematically excluded, and the effect sizes were recalculated to detect any variations. This approach also enables the identification of studies that exert significant influence on the overall results and assists in examining clinical heterogeneity. The analysis demonstrated that the effect sizes remained largely consistent upon the exclusion of individual studies, thereby indicating the stability of the study’s findings (Fig. 7A–D and SI Table S3 to Table S6).

Fig. 7

A Sensitivity analysis of 1 Hz rTMS over contralesional M1 for stroke (< 1 month) on FMA-UL. B Sensitivity analysis of 1 Hz rTMS over contralesional M1 for stroke (1–6 months) on FMA-UL. C Sensitivity analysis of 1 Hz rTMS contralesional M1 for stroke (< 1 month) on MBI/BI. D Sensitivity analysis of 1 Hz rTMS contralesional M1 for stroke (1–6 months) on MBI/BI

Publication bias

Despite the Cochrane Handbook’s recommendation to apply publication bias for outcomes reported in more than 10 studies, Egger’s test was still conducted for 1 Hz rTMS on MBI/BI due to the inclusion of more than 5 studies and a total participant count exceeding 250. The findings suggest that there is no evidence of publication bias in relation to the FMA-UL at a frequency of 1 Hz for stroke patients within the first month of onset (P = 0.94), as well as for stroke patients between 1 and 6 months post onset (P = 0.831). Similarly, the same lack of publication bias was observed for the MBI at a frequency of 1 Hz for stroke patients within the first month of onset (P = 0. 508), and for stroke patients between 1 to 6 months post onset (P = 0.418).

The quality of evidence

The level of evidence supporting the use of 1 Hz rTMS over the contralesional M1 for stroke (within 1 month) using FMA-UL and MBI was increased as a result of a substantial sample size exceeding 300 participants. However, for other outcomes, the level of evidence was decreased due to imprecision resulting from a small sample size and publication bias. Consequently, the level of evidence was downgraded. Finally, the quality of evidence for the effect of 1 Hz rTMS over the contralesional M1 on motor function (FMA-UL) in stroke patients within 1 month was determined to be “high quality”. The quality of evidence for other outcomes, such as 1 Hz rTMS over the contralesional M1 for stroke between 1 and 6 months on FMA-UL and 1H z rTMS over the contralesional M1 for stroke patients within 1 month on MBI/BI, were downgraded to “moderate quality”, whereas for other outcomes, not mentioned above, the quality of evidence were also downgraded to “low” or “very low quality” (Table 4).

Table 4 The level of GRADE for outcomes

Discussion

This research adhered to the rigorous methodological guidelines outlined in the Cochrane intervention Handbook. As reported in the previous literature [105], the scope of this study was limited to English literature but considers all health-related areas. The databases were selected cross-disciplinary research, which can provide a wide range of opinions. The process of study selection comprises a systematic search of literature sources followed by three iterative stages of screening and filtering. In the initial stage, duplicate articles are excluded, and relevant articles are collected using Endnote software. The second stage involves screening titles and abstracts to exclude articles that fall outside the scope of the research domain. Finally, the third stage entails a thorough full-text review to exclude articles that do not meet the specified domain and criteria. And additional records identified through reference of full text. All iterations adhere to consistent eligibility criteria and are subsequently subjected to a process of screening and review. During the data extraction process, the initial data were independently extracted by one researcher and subsequently verified by another. For the studies that did report the raw data, data extraction software (Getdata 2.2) was used to obtain relevant information from statistical charts by two well trained authors. Any missing data were acquired through communication with the original author via email. The above process ensures the completeness, accuracy and reliability of the data, which are core elements of high-quality data. In the context of healthcare decision-making, the availability of accurate and reliable predictions is of paramount importance, underscoring the crucial role of data quality [106]. For this reason, our findings can aid healthcare practitioners in making educated decisions.

Prior to commencing the study, a well-structured study protocol was developed, reviewed by experts, and made available online, thereby ensuring the reliability and validity of the findings. The process of literature screening, data extraction, and data analysis was independently conducted by two authors, with a consensus reached through the involvement of a third author to ensure internal consistency. Following the completion of data analysis, the literature was updated until July 29, 2024, to ensure the timeliness of the results. This review specifically included studies that focused on the use of simple rTMS either alone or in combination with rehabilitation, excluding studies involving complex-pattern rTMS. This approach aimed to establish a clear association between rTMS and its impact on motor function and ADLs. The subgroup analysis of different frequencies of rTMS and durations of stroke explained the clinical heterogeneity and methodological heterogeneity.

The study included a total of seventy RCTs that involved 2951 patients across various stages of stroke, including early-stage stroke, recovery-phase stroke, and chronic stroke. The existing evidence indicates that the efficacy of different frequencies of rTMS in ameliorating motor dysfunction among stroke patients varies inconsistently across different stages. The results derived from the FMA-UL suggest that rTMS at frequencies of 1, 10, and 20 Hz may exert a beneficial effect on upper limb motor function in patients with early-stage stroke (< 1 month). Specifically, the SMD analysis suggests that 1 Hz rTMS has a moderate effect on enhancing upper limb motor function, whereas 10 and 20 Hz rTMS demonstrate a large effect size. Nonetheless, these conclusions should be interpreted with caution in clinical practice due to limitations such as the small sample size, the low quality of the included studies, and other confounding factors. The moderate effect size observed for 1 Hz rTMS over the contralesional M1 suggests its potential to positively influence upper limb motor function recovery during the stroke recovery phase. However, the results of grip strength assessments did not indicate that 1 Hz over the contralesional M1 or 10 Hz rTMS over ipsilesional M1 provides positive benefits for individuals in the early stages of stroke. In terms of ADLs, the overall results suggest that 1 Hz over the contralesional M1, 3 Hz over the ipsilesional M1, 10 Hz over the ipsilesional M1, and 20 Hz over the ipsilesional M1 rTMS may offer potential benefits for early-stage stroke patients. Specifically, the SMD result suggests that 1 Hz stimulation over the contralesional M1 exerts a moderate effect on improving upper limb ADLs, whereas 10 and 20 Hz rTMS over the ipsilesional M1 exhibit a large effect size. However, these findings should be approached with caution in clinical practice due to limitations such as small sample sizes, the low quality of the included studies, and other confounding variables. However, the current evidence does not support the notion that rTMS can improve neurological function scores as measured by the NHISS. The primary outcomes, namely FMA-UL, FMA-LL, grip strength, and MAS/AS, exhibited varying levels of evidence. Specifically, the evidence supporting the efficacy of 1 Hz rTMS over the contralesional M1 for early-stage stroke motor function was upgraded to a “high-quality” rating. Conversely, the evidence pertaining to other outcomes was downgraded to “moderate-quality”, “low-quality”, and “very low-quality” ratings. This decision was influenced by the presence of substantial clinical heterogeneity and limited sample sizes.

Despite successfully demonstrating the benefits of rTMS for stroke patients, this study has notable limitations that warrant attention. Importantly, significant clinical heterogeneity was observed across studies, which stemmed from variations in disease periods, injury sites, intervention timing, intervention type, frequency, control measures, site of application, duration, and assessment instruments. The second concern pertains to the inclusion of patients in certain studies who were unaware of rTMS and were exposed to simulated sounds or counterfeit coils, albeit not in an entirely identical manner. While the sensation experienced on the scalp may be similar similarities, it cannot be considered identical to that of the rTMS group. The third point of consideration pertains to the challenging nature of implementing rTMS operator blinding, thereby categorizing operator blindness as a high-risk factor. Furthermore, the fourth aspect to contemplate involves the limited sample size, absence of intention-to-treat analysis reporting, and methodological deficiencies, all of which contributed to a reduction in the quality of our evidence. The final and paramount inquiry that warrants our attention pertains to the limited ability to ascertain the safety of implementing rTMS due to inadequate and inconsistent documentation of adverse events, such as epileptic seizures, headaches, anxiety, fatigue, pain, spasms, myalgia, paresthesia, and drowsiness, which have only been partially reported in certain studies. Consequently, it is imperative to exercise caution when evaluating the robustness of the evidence.

Future research directions

This study significantly contributes to our comprehension of the impact of rTMS on the motor function of stroke patients. Furthermore, it offers valuable insights and recommendations for future investigations. The recovery of motor function in stroke patients is intricately linked to the specific location and type of stroke. However, the present studies encompassed the utilization of comparable interventions in stroke patients with diverse site injuries (cortical, subcortical, or others) and distinct types (ischemic and hemorrhagic stroke). Importantly, the baseline motor function following a stroke serves as a significant indicator of eventual functional recovery post-stroke. The initial motor function status of stroke patients should be taken into account during the assessment of treatment effectiveness. For example, ensuring that groups are matched for initial motor function impairment can enhance the assessment of the effectiveness of rTMS. Hence, patient stratification becomes imperative when formulating treatment plans for future studies. Trials evaluating rTMS focused on the motor domain should carefully consider selecting endpoint measures that are tailored to assess motor impairment, capacity, and performance. In early-phase exploratory trials, the use of measures at all three levels could provide valuable insights, whereas later-phase trials may benefit from prioritizing measures of capacity and performance that effectively capture the impact of the intervention on patients’ daily activities and participation. Broad measures of independence or disability, such as the modified Rankin scale, may lack the necessary sensitivity to serve as primary endpoint measures. Therefore, carefully selecting domain-specific endpoint measures that align with the intervention objectives, the patient’s stage of stroke recovery, and the stroke phase is crucial. Moreover, previous studies on rTMS have focused primarily on factors such as suitability, patient comfort, availability of stimulation devices, and impact on corticospinal excitability, without considering the underlying mechanism of action. Therefore, an optimal rTMS regimen should be contingent upon the patient’s functional preservation, the specific type of stroke (subcortical or cortical; ischemic or hemorrhagic), and the stage of stroke (early stage, convalescence, or chronic period). The use of rTMS exhibits considerable inter-individual variability. This variability is often attributed to the stimulation parameters employed to enhance corticospinal excitability, because the protocol’s impact on a diverse population of cortical neurons can result in an inhibitory effect. The clinical heterogeneity observed in rTMS trials for stroke primarily stems from variations in stimulus programs, such as the number of sessions administered and the duration of concurrent physical therapy. These variations have constrained the outcomes of rTMS trials for stroke. Intervention parameters, such as location of the coil, intensity, cycle treatment, and stroke type, increased the variability of the study. Few studies have examined the relationships between stimulation parameters and neurophysiological assessments of the motor system, as well as the impact of the interventions on excitation connectivity and clinical outcomes. Consequently, the selection of rTMS intervention parameters should be guided by test findings, such as functional magnetic resonance imaging (fMRI), magnetic resonance spectroscopy, near infrared spectroscopy (NIR), EEG, and other detection methods. These tools provide a comprehensive understanding of the specific mode of operation employed by patients and the natural activities occurring during rTMS intervention. This understanding is crucial for the successful implementation of personalized rTMS programs. Additionally, the preceding study failed to establish a definitive correlation between the efficacy of rTMS and the degree of stimulation, thereby highlighting the significance of investigating this crucial aspect in subsequent research endeavors. Notably, during the acute phase, MEPs are frequently not elicited on the affected side even when the stimulation intensity is at its maximum [15]. In patients who retain MEPs, it was observed that the threshold for movement on the affected side is typically elevated, whereas the threshold for MEPs is reduced compared with that on to the unaffected or healthy side [107, 108]. Within the initial months, MEPs may resurface and progressively amplify, whereas the motion threshold tends to diminish [104, 109]. Numerous scholars have proposed that the evaluation of early corticospinal tract integrity and the enhancement of corticospinal tract integrity through rTMS during the early recovery phase are linked to enduring functional outcomes [110]. Hence, acknowledging the potential influence of the integrity of the spinal cord cortex on the effectiveness of TMS is imperative and warrants further investigation in subsequent research endeavors. Moreover, the issue of spatial and temporal sensitivity has long been a prominent consideration in the realm of non-invasive interventions for stroke treatment. Precise spatial sensitivity cannot be quantified with absolute precision because of the variability in factors such as the number of coils in the device and the magnitude of the stimulus current, which determine the precise positioning and depth of the stimulus. Notably, circular and figure-8 coils yield distinct magnetic field ranges and stimulation points. Therefore, forthcoming studies should allocate greater consideration to this particular facet.

Conclusion

The application of rTMS in stroke patients, especially in improving the motor function of patients, brings hope to patients. However, the current research findings have introduced considerable ambiguity for health policymakers, clinicians, patients, and their families. This study adhered to a meticulously structured protocol, thereby ensuring the reliability and validity of the findings. After rigorous literature screening, data analysis, and evaluation of evidence quality, the results of this study showed that 1 Hz rTMS over the contralesional M1 may enhance motor function during the early stages of stroke and the recovery period and 1 Hz rTMS over the contralesional M1 can improve ADLs for stroke patients (1–6 months) of the event. Despite the inability to locate certain sources of “grey” literature, their inclusion is unlikely to have significantly impacted our results. It is crucial to interpret these findings with caution in clinical practice due to the small sample sizes and the low quality of the studies reviewed. The efficacy of 3, 5, 10, and 20 Hz rTMS in the treatment of motor dysfunction following stroke requires further investigation. To be specific, meticulously planned, multi-center, large-scale trials are needed to assess the safety and clinical efficacy of 3, 5, 10, and 20 Hz rTMS in individuals afflicted by stroke.