PICO model

We used the PICO (Population, Intervention, Comparison, Outcome) model [27] to define the research question. Our research questions were as follows: “Is NIBS a feasible and safe tool in the neurorehabilitation of chronic stroke patients?” and “What are the optimal treatment approaches for stroke patients undergoing NIBS?”. To investigate these questions, we selected evidence with adults (>18 years) patients affected by chronic stroke (both ischaemic and haemorrhagic), as population; the intervention included all non-invasive neuromodulation techniques (e.g., TMS, rTMS, i-TBS, tDCS); the comparison included sham or placebo stimulation conducted in the control group, allowing for a comparative analysis of the effects of the active interventions; and outcomes included any motor improvements in the following functions: upper limb, lower limb (e.g., gait and balance), vocal and swallowing, shown by the patients and efficacy of treatment.

Search strategy and information sources

A systematic search, according to PRISMA guidelines [26], was conducted for all peer-reviewed articles published from January 2010 through September 2023, using the following databases: PubMed, Embase, Cochrane Database of Systematic Reviews, PEDro, RehabData, and Web of Science. The following terms were used: (“chronic stroke and neuromodulation”, and “neuroplasticity”) AND (“transcranial non-invasive stimulation” OR “NIBS”) AND/OR (“Transcranial magnetic stimulation “ OR “TMS”) AND/OR (“Transcranial direct current stimulation” or “tDCS”).

Eligibility criteria

The inclusion criteria were: (i) adult patients with chronic stroke; (ii) an applied approach to cognitive and motor rehabilitation; (iii) the English language; and (v) published in a peer-reviewed journal. We have excluded articles that describe theoretical models, including methodological approaches, algorithms, and basic technical descriptions, focusing instead on empirical and applied studies that present concrete and verifiable results. Additionally, we excluded: (i) animal studies; (ii) conference proceedings or reviews; and (iii) studies involving children; (iv) case reports and reviews. The list of articles was then refined for relevance, revised, and summarized, with the key topics identified from the summary based on the inclusion/exclusion criteria.

Selection and data collection processes

Two independent reviewers (M.G.M. and M.B.) screened titles, abstracts, and full texts, applying inclusion and exclusion criteria under blinded conditions to reduce selection and publication bias. Reviewers were blinded to the names, affiliations, and journals of the authors. Discrepancies were discussed and resolved with a third reviewer (R.S.C.). Inter-rater agreement was calculated using the kappa statistic, with κ > 0.61 indicating substantial concordance. [21]

Data items and data extraction

After the preliminary articles’ selection process, data from the included studies were extracted and organized into a summary table using Microsoft Excel (Version 2021). In accordance with PRISMA guidelines, rigorous procedures were employed for data extraction to ensure comprehensive coverage of relevant information. The extracted information included: assigned ID number, study title, year of publication or presentation and first author, study aims and design, study duration, recruitment method and setting, inclusion and exclusion criteria, presence of a control group, use of devices, informed consent, conflicts of interest and funding sources, type of intervention and control, number of participants, baseline characteristics, intervention setting, outcome measures and time-points for assessment, adverse events, results, and key conclusions. Once summarized, the data were uploaded into an online database (RYYAN) [28], where reviewers independently evaluated the relevance of each study based on titles, abstracts, and full texts. This meticulous approach ensured a thorough and reliable synthesis of the available evidence, enhancing the robustness of our review.

Assessing quality of included studies – Risk of bias

The risk of bias in controlled studies was assessed through a revised Cochrane risk of bias (RoB 2) [29], which consists of five domains: i) bias arising from the randomization process, ii) bias due to deviations from intended intervention, iii) bias due to missing outcome data, iv) bias in the measurement of the outcome, v) bias in the selection of the reported result. The screening and assessment were independently performed by two reviewers, who were blinded to each other’s evaluations during this stage to minimize potential bias. Once the initial assessments were completed, the blinding was removed. In cases where discrepancies between the reviewers’ assessments arose, a third reviewer mediated the discussion to reach a consensus. As a result from our analysis, the majority of the included studies (45 out of 58) were rated with a low risk of bias, as they reported robust methodologies (see S1 for the RoB analysis of the included studies).

Upper limb

The most prevalent risk of bias was in domain 1, related to randomization processes, in which only 16 [30–44] out of 26 studies were judged as having low risk. The remaining 10 studies were rated as having moderate [45,46] to serious [47] concerns regarding randomization procedures, primarily due to insufficient reporting of key methodological details, such as the method used to generate the random allocation sequence, the type of randomization, the allocation concealment mechanism, and whether allocation was blinded. One of these studies [47] was assessed with a serious concern, as no specific randomization procedure was described. Instead, participants were assigned to the experimental or control group in a seemingly random manner without the use of stratified randomization. This approach raises a potential risk of bias, particularly in terms of achieving balanced groups with respect to baseline functional severity. In contrast, the other 16 studies [30–44] with a low risk of bias reported information related to the type of procedure used and the accuracy of recruitment.

Despite this, 6 studies [33,37,48–51] presented biases in domain 2, because they did not report blinding or did not specify who was blinded. In domain 3, 9 studies [32,34,42,47–49,51–53] were judged to have some concerns regarding bias control, particularly due to a lack of transparency in handling missing data. One recurrent limitation was the insufficient adherence to standardized reporting frameworks, such as the CONSORT statement, which recommends practices like intention-to-treat analysis and detailed reporting of attrition. These omissions can undermine the interpretability and reproducibility of results. In 9 out of the selected articles [32,34,42,47–49,51–53], the absence of structured reporting limited confidence in the handling of missing data, raising concerns regarding the potential risk of bias in domain 3. Furthermore, domains 4 and 5 were generally rated as low risk of bias, except for two studies [51,54], which raised some concerns due to the limited number and inconsistency of outcome measures relative to their stated objectives (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. ROB2 Upper Limb.

https://doi.org/10.1371/journal.pone.0327583.g001

Lower limb and walking functions

The most prevalent risk of bias was in domain 2, because in two studies [55,56] it was not reported that blinding procedures were used, and in one study [56] there was no blinding. Furthermore, in domain 3, two studies [55,56] were judged to have some concerns regarding control of bias, and in one, the risk was judged to be high. Moreover, in domains 4 and 5, we consider the studies to have low bias, except for some concerns in three studies [56–58] for domain 4 and one [58] in domain 5, where outcome measures were few, and in some cases inconsistent with goals (Fig 2). For example, Tanaka et al. evaluated knee extension muscle force as the primary outcome and assessed hand-grip strength as a secondary measure that was quite inconsistent with the primary goal.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. ROB2 Lower limb and walking functions.

https://doi.org/10.1371/journal.pone.0327583.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. ROB2 of Robotic-assisted therapy combined with neuromodulation.

https://doi.org/10.1371/journal.pone.0327583.g003

Robotic-assisted therapy combined with neuromodulation

In Domain 1 (randomization process), 54 out of 58 studies were judged to be at low risk of bias. Three studies [59–61] raised some concerns due to the lack of description of the randomization procedure and, in one case [61], there was no information about allocation. One study [62] was rated as high risk due to the complete absence of details on randomization procedures and allocation methods. Regarding bias in Domain 2, Straudi et al., [63] did not specify the blinding procedures or clarify which individuals were blinded to the evaluation or treatment, raising concerns about potential performance and detection bias. In particular, Danzl et al. [60] did not implement specific strategies to address or mitigate potential bias related to missing outcome data.

A visual summary of risk of bias across all studies and domains is provided in Fig 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. ROB2 of Vocal and swallowing functions.

https://doi.org/10.1371/journal.pone.0327583.g004

Vocal and swallowing functions

Most studies in this category were judged to be at low risk of bias across the assessed domains. However, one study [64] was rated as high risk due to the absence of any bias control measures, limited outcome measures, and poor alignment between the outcomes assessed and the study objectives. Additionally, in Domain 1 (randomization process), two studies [65,66] did not report any randomization procedures, raising concerns about selection bias (Fig 4). Overall, conducting a robust quality assessment for this functional category was challenging due to the limited number of included studies. While several studies demonstrated a low risk of bias, others exhibited moderate to serious methodological difficulties, particularly in randomization, masking, outcome selection, and bias control, which reduce their reliability and should be interpreted with caution.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. PRISMA flow chart.

https://doi.org/10.1371/journal.pone.0327583.g005

Synthesis of evidence

Electronic searches identified 1.132 papers. We assessed and included 75 studies, according to their pertinence and relevance to the topic. After reading the full text of the selected publication and applying the predefined inclusion criteria, we excluded 17 articles, and 58 articles were included in our qualitative analysis (see Fig 5) (see S2 and S3 for all excluded and included manuscripts). We divided our results in four sections as follows: (1) upper limb, (2) lower limb, (3) robotic assisted therapy combined with neuromodulation, and (4) vocal and swallowing functions.. The summary presented below reflects the relative distribution of studies included across functional domains: 26 studies [30–54,67] focused on upper limb rehabilitation, 13 studies [55,56,58,68–77] on Lower limb and walking functions rehabilitation, 13 [60–63,11,78–85] on Robotic-assisted therapy combined with neuromodulation, and 6 [59,64–66,86,87] on Vocal and swallowing functions. Therefore, the level of detail varies proportionally to the available evidence within each category, to ensure accuracy without over-interpreting limited data.

Upper limb

Our search revealed 26 [30–54, 67] articles focusing on the use of NIBS, including i-TBS, rTMS, and tDCS for the rehabilitation treatment of upper limb functions in 684 chronic post-stroke patients. In particular, we found 25 RCTs [30,32,34–54, 67], 1 pilot study [33], and 1 randomised cross-over study [31]. The sample size ranged from 6 to 80 patients. In particular, the intervention was carried out through i-TBS, rTMS, and tDCS. In detail, 5 RCTs [32,33,43,46,67] investigated the effects of i-TBS in chronic stroke patients. Each study addressed the brain stimulation on the ipsilateral or contralateral motor cortex according to inhibitory or excitatory stimulation, after conventional physiotherapy. In addition, Kuzu et al. [32] examined the effects of i-TBS and low-frequency rTMS combined with physiotherapy. The authors [32] suggested that both techniques were effective in improving upper limb functions, but benefits are limited for spasticity. Both the real rTMS and real TBS groups showed significant improvements in total upper extremity Fugl-Meyer (FM) scores both post-treatment (p = 0.017 for both groups) and at the 4-week follow-up (p = 0.18 for rTMS and p = 0.018 for cTBS). The real rTMS group exhibited significant improvements in spasticity scores, measured with Modified Ashworth Scale (MAS), for the elbow flexor, pronator, wrist flexor, and finger flexor muscle groups post-treatment (p = 0.025, p = 0.025, p = 0.039, p = 0.038, respectively) and at the 4-week follow-up (p = 0.025, p = 0.025, p = 0.039, p = 0.038, respectively). The real TBS group showed significant improvements in MAS scores for the elbow flexor and wrist flexor muscle groups post-treatment (p = 0.025, p = 0.034, respectively), and for the wrist flexor at the 4-week follow-up (p = 0.024).

On the other hand, 9 studies [31,38,39,41,45,46,53,54] (8 RCTs and one randomized cross-over study) investigated the effects of rTMS on the paretic upper limb. In 5 studies [31,35,38,39,41] out of 9 addressed rTMS on the contra lesional/unaffected primary motor cortex, and in 2 of these [35,39] authors associated conventional physiotherapy with rTMS. One study Aşkın A, Tosun A, Demirdal ÜS. Effects of low-frequency repetitive transcranial magnetic stimulation on upper extremity motor recovery and functional outcomes in chronic stroke patients: A randomized controlled trial. Somatosens Mot Res. 2017;34(2):102–107. https://doi.org/10.1080/08990220.2017.1316254 administered low-frequency rTMS (LF-rTMS) stimulating contra lesional primary motor cortex, in association with conventional physiotherapy and reach-to-grasp training. Revill et al. [45] used a particular rTMS technique called Hebbian, which has been used to induce long term potentiation in postsynaptic pyramidal tract neurons. Interestingly, these authors found that significant correlations occurred between improvements in hand function and changes in motor cortex (M1) activity for both hemispheres (p < 0.05). Participants who showed greater improvement in hand function also had greater increases in task-related M1 activity (p < 0.05). These findings were particularly strong in the treatment group with the Hebbian-based training compared to the sham therapy, suggesting that this approach led to both functional and neural improvements. In addition, Takeuchi et al. [53] combined LF-rTMS with anodal tDCS to prevent deterioration of coordination of bimanual movements.

The remaining twelve studies investigated the effects of tDCS on upper limb functions in chronic stroke. Half of them administered bi- or mono-hemispheric anodal and cathodal tDCS on the primary motor cortex. The other two studies combined tDCS with physiotherapy [45] and with a resistance training program [37] to improve upper limb functions.

Rocha et al. [39] and Bolognini et al. [47] administered cathodal tDCS in addition to constraint-induced movement therapy (CIMT).

Moreover, Chen et al. [33] suggested that tDCS-iTBS combined physical therapy can be a promising approach to enhance therapeutic benefits of upper limb treatment. In particular, these authors, found significant improvements in functional hand motor skills, measured with Jebsen-Taylor hand function test (JTT) (p = 0. 016), as well as in upper limb coordination, measured with finger-to-nose test (FNT) (p = 0. 037). Also, Menezes et al. [52] applied tDCS in combination with repetitive peripheral nerve sensory stimulation (RPSS) could enhance the effects of functional electric stimulation on active range of movement of paretic wrist, (see Table 1). They found that the real NIBS group had greater improvements in the JTT (p = 0. 016) and FNT (p = 0. 037) scores than the sham NIBS group.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Description of studies regarding the use of NIBS on upper limb outcomes.

https://doi.org/10.1371/journal.pone.0327583.t001

Lower limb function

In addition to upper limb rehabilitation, another key area investigated was lower limb function. We found 13 articles [55,56,58,68–77] dealing with the use of NIBS to promote balance and gait recovery in 267 chronic post-stroke patients. Specifically, our research found 8 RCTs [55,56,69–71,73,75,77], 3 clinical trials [58,68,72], 2 pilot studies [74,76]. The sample size of the selected RCTs ranges from 10 to 48 patients. The studies included investigated different types of NIBS, including i-TBS, rTMS and tDCS (see Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Description of studies regarding the use of NIBS on gait and balance outcomes.

https://doi.org/10.1371/journal.pone.0327583.t002

Firstly, the i-TBS was used in four studies [70,72–74] (two RCTs and two pilot studies) to promote gait and balance recovery stimulating cerebellar [70,73] and bilateral cerebral motor cortex [72]. Each study coupled the brain stimulation with conventional physiotherapy. Physiotherapy sessions were arranged immediately after the i-TBS stimulation. Additionally, patients enrolled in the studies [70,72,73] presented moderate motor deficits and were randomly assigned to i-TBS or sham groups.

The rTMS and TMS were performed in three studies [75–77] (2 RCTs and 1 pilot study) to improve lower limb functions, walking ability and spasticity. Specifically, only Chieffo et al. [76] administered rTMS using the H-coil which is designed to reach deep brain regions. Study comparisons were made with sham controls in all three studies and Wang et al. [77] performed a task-oriented gait training after brain stimulation.

Lastly, tDCS was used in 7 studies [55,56,58,68,69,71,72] (5 RCTs and 2 controlled clinical trials) to improve walking, balance abilities and muscle strength in lower limb. Combined approach was used by da Cunha et al. [68], who administered tDCS, footdrop stimulation and gait training simultaneously, while Madhavan et al administered tDCS with treadmill training. Specifically, da Cunha et al., [68] found statistically significant time effects were found in clinical tests to evaluate gait and balance functions, such as the Timed-Up-Go (TUG) (p < 0.001), while for spasticity, a main effect of time was seen for plantar flexors and knee extensors on the MAS (P < .05), with reductions in resistance in both groups over time. In addition, walking distance improved significantly over time (p = 0.001) in both groups, with an average increase of 76 meters, indicating clinical relevance. Moreover, Madhavan et al., [72] found that he tDCS plus ankle motor tracking group showed significantly greater increases in corticomotor excitability, compared to other groups (e.g., high-intensity speed-based treadmill training) (p = 0.02), suggesting enhanced neuroplastic effects. Zandvliet et al. [55] stimulated cerebellar cortex using tDCS during medio-lateral postural tracking task on force platform to improve balance performance in post-stroke patients. In particular, they found that after the tDCS stimulation, several centre of pressure (CoP) measurements were significant improved, including CoP amplitude (p = 0.02), CoP variability (p = 0.01), CoP range and velocity (p = 0.01, p = 0.02). Tanaka et al. [58] and Ojardias et al. [71], used anodal tDCS stimulation to improve respectively knee extension strength (p < 0.01) and walking abilities tested with 6-minutes walking test (6MWT) (p = 0.038), compared to the sham therapy (see Table 2).

Robotic assisted therapy combined with neuromodulation

Beyond domain-specific functional targets, some studies have explored the synergistic effects of combining NIBS with other rehabilitation modalities. In particular, we identified 13 RCTs [60–63,11, 78–85] dealing with robot assisted therapy (RAT) combined with neuromodulation in 397 chronic post-stroke patients. In 5 RCTs [61,63,78–80] RAT was focused on the treatment of upper limb, while in the remaining 8 [60,62,11,81–85] RCTs RAT was focused on gait training.

In the 5 RCTs [78,79], sample size ranged between 18 and 82 chronic post-stroke patients. The intervention regarded the use of RAT for upper limbs in combination with tDCS to improve manual dexterity. Specifically, in 3 studies [63,79,80] compared tDCS with sham stimulations plus RAT, except for Ochi et al. [61], which compared anodal and cathodal tDCS using a cross-over study design. The RAT was carried out through different robotic devices including exoskeleton like Armeo Power II [79], and end-effectors such as MIT Manus [80], REO therapy system [63] and Bi-Manu-Track robotic arm trainer (Reha-Stim, Berlin) [61]. In contrast, Zhang et al. [78] used two types of robotic devices such as one exoskeleton (HandRehab) and one end-effector (Armodus) combined with i-TBS to improve upper limb functioning. Specifically, they found a significant time-by-group interaction in FM scale scores (p = 0.011), with both priming and nonpriming iTBS outperforming sham stimulation. In patients with higher upper limb function, priming iTBS led to significantly greater FM improvements than both nonpriming iTBS (p = 0.025) and sham (p = 0.029).

The remaining 8 RCTs [60,62,11,81–85] were focused on the administration of RAT for gait function in chronic post-stroke patients. In details, sample size ranged between 8 and 45 chronic stroke patients. Interventions were carried out through tDCS in seven studies [60–63,79–81,11, 83–85] and rTMS in one study [82] in addition to robotic devices which included exoskeletons like Lokomat [60], Walbot_S [84], and end-effectors such as GE-O System [83,11,85] and Gait Trainer [62], except for Wong which used treadmill device (see Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Description of studies regarding combined approach through NIBS and robotic-assisted training (RAT).

https://doi.org/10.1371/journal.pone.0327583.t003

Vocal and swallowing functions

In addition to limb-related motor functions, a smaller subset of studies investigated the use of NIBS for orofacial rehabilitation. We found 6 articles [59,64–66,86,87] focusing on the use of NIBS to promote swallowing skills, speech-related and vocal functions among post-stroke dysarthric patients. Specifically, our search revealed 5 RCTs [59,64–66,86] and 1 pilot study [87]. The sample size of the selected RCTs ranges from 4 to 26 patients. The studies included two types of NIBS (rTMS and tDCS).

rTMS was performed in four studies [65,66,86,87] (3 RCTs and one pilot study) to improve swallowing functions and dysphagia. Five studies used comparisons with sham controls [59,64–66,86]. Only Michou et al. [87] performed single pulse rTMS over the pharyngeal motor cortex and an assessment of pharyngeal electromyographic responses. Furthermore, they divided patients into three types of stimulation: electrical pharyngeal impaction (PES), paired associative stimulation (PAS), or rTMS. Instead, Chen et al., in an RCT and another pilot study [65,66], performed high frequency rTMS at 5 Hz applied to the tongue region of the motor cortex. Finally, only Ünlüer et al. [86] performed conventional dysphagia rehabilitation in both groups (sham and experimental), and then the experimental group also received 1 Hz rTMS in the unaffected hemisphere.

Finally, tDCS was used in 2 RCTs studies [59,64] to improve physiological, vocal, and swallowing function. Both studies used tDCS over the pharyngeal motor cortex and used experimental and sham groups. Only Ahn et al. [64] used conventional swallowing therapy in combination with tDCS (see Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Description of the studies regarding the use of NIBS on vocal and swallowing outcomes.

https://doi.org/10.1371/journal.pone.0327583.t004

tDCS

We found that the most used NIBS technique was the tDCS for restoring upper and lower limb outcomes. Relevant benefits provided by tDCS were registered for lower limb functions [68,69,73]. In particular, contra-lesional cerebellar tDCS showed promising results on balance recovery [55], whereas anodal tDCS over M1 promoted gait recovery [58,71]. However, the potential mechanisms underlying the improvement of gait performance induced by tDCS applied to M1-LL remain mainly speculative. An fMRI study on healthy subjects showed that four sessions of anodal tDCS applied to the lower limb area of the primary motor cortex (M1-LL) increased activation across a broad bilateral sensorimotor network, including the anterior cingulate gyrus, supplementary motor area (SMA), and somatosensory cortices [105]. This transient enhancement of sensorimotor network activity may underlie improvements in gait endurance and potentially contribute to increased gait speed.

This is further enhanced when the tDCS is combined with robotic gait training (RAGT), as suggested by other authors [60,11,84,85]. A recent systematic review [23] highlighted that tDCS applied to the leg area of the motor cortex in the affected hemisphere, or cerebellar tDCS targeting the ipsilesional or contralesional cerebellum in combination with RAGT, can lead to meaningful improvements in walking function. These improvements are especially evident in walking ability and capacity, as measured by tools such as the 6MWT.

The administration of tDCS in addition to RAGT suggests that this combined approach may promote spinal mechanisms of locomotion, such as spinal reflexes, central pattern generators and lower extremity electromyographic activity [23,56,106]. This approach could explain the rationale for the use of this combined approach. However, upper limb RAT combined with tDCS did not provide additional benefits to the only RAT, in three studies [61,79,80]. Specifically, Morone et al. [79] employed dual tDCS, combining inhibition of the unaffected hemisphere with excitation of the affected hemisphere. This approach is particularly recommended for patients with moderate to severe stroke, who may benefit from enhancing the compensatory role of the unaffected hemisphere. In contrast, for patients with milder stroke, single site tDCS targeting the affected hemisphere appears to be more effective in improving motor function [96].

Moreover, the effects of tDCS for the upper limbs included both sensory [107] and motor improvements [30,33,34,37,39,44,47,50,51]. Some authors [39,47] combined CIMT with brain stimulation, obtaining gains in grip strength and hand dexterity. Indeed, Rocha et al. [39] suggested that anodal tDCS seems to have a greater impact on upper limb motor outcomes than cathodal tDCS.

On the other hand, positive effects of tDCS on vocal and speech functions were also found, as suggested by Wong et al. [59]. The authors applied a combined intervention of speech therapy and tDCS to enhance outcomes in dysarthric speech. Similarly, Ahn et al. [64] reported positive effects of tDCS paired with conventional therapy in improving swallowing function in patients with dysphagia.

i-TBS

i-TBS demonstrated its efficacy on the recovery of gait and balance functions, especially when the stimulation was applied to cerebellar lobes [70]. Indeed, Koch et al. [73] demonstrated that i-TBS was effective in improving not only gait and balance functions, but also ankle stability of the paretic lower limb. However, Lin et al. [74] concluded that applying i-TBS to bilateral lower extremity (LE) M1 was not effective in enhancing lower limb functions. The authors suggested that bilateral stimulation may be appropriate when motor evoked potentials (MEPs) from the affected hemisphere are unrecordable, as it ensures that the induced current reaches both motor cortices. However, when MEPs from the affected hemisphere are present, unilateral stimulation targeting the affected LE M1 may be more effective.

Moreover, the i-TBS was also administered to improve upper limb functions. According to findings of the included studies, it seems to be effective in reducing upper limb spasticity [36] and in modulating corticospinal excitability [40] when it is applied alone. However, Sung et al. [42] and Di Lazzaro et al. [43] administered iTBS joined with rTMS, suggesting that this kind of stimulation is safe and promotes motor recovery in chronic stroke patients. Zhang et al. [78] investigated the effects of both priming and non-priming iTBS combined with RAT for upper limb rehabilitation. They found that priming iTBS, specifically, applying continuous TBS before iTBS, was more effective in patients with relatively preserved upper limb function compared to non-priming approaches. The authors hypothesized that priming iTBS may enhance motor learning by increasing the responsiveness of the ipsilesional supplementary motor area to therapeutic inputs, such as mirror visual feedback. However, Talelli et al. [67] reported no additional benefit of TBS for improving upper limb and hand function. They emphasized the importance of patient selection, noting that attempting to promote ipsilesional reorganization may be unrealistic in individuals with severe brain damage.

rTMS

Lastly, rTMS was particularly used to improve upper limb functions [31,32,35,38,41–43,45,53]. Some authors [32,35,41,48] suggested its use to reduce spasticity by increasing the unaffected hemisphere’s excitability and decreasing spinal excitability. Similarly, Rastgoo et al. [75] found improvements in lower limb muscle spasticity but using low frequency rTMS. However, Etoh et al. [49] did not register any changes in spasticity after rTMS, despite finding gains in motor functions. According to the authors’ conclusion [49], this finding is linked to the limited ability of the MAS to detect small changes in spasticity, as it primarily identifies only more significant improvements. Despite being the most widely used clinical scale for assessing spasticity, the MAS may not capture subtle clinical changes [108]. The authors hypothesise that improvements in spasticity are more evident in patients with a severe MAS score.

Moreover, we found some evidence [75–77] about rTMS in the recovery of the lower limb in chronic post-stroke patients. According to literature [75–77,82], rTMS promoted functional recovery of lower limb and gait functions. Specifically, Wang et al. [77,82] found that rTMS alone [77] and/or combined with treadmill training [82] had positive effects on gait speed and spatial symmetry.

On the other hand, rTMS was used to improve swallowing functions [65,66,86,87], reducing risks of aspiration in patients with chronic post-stroke. However, only Cheng et al. [66] found positive effects of excitatory rTMS on swallowing post-stroke symptoms. Other authors [65,86] did not found any improvement in dysphagic symptoms, which may be related to their study design and severity of stroke. The rTMS for dysphagic symptoms should be applied on the unaffected hemisphere, considering also that patients with severe stroke may have smaller improvements than moderate and mild stroke.

Comparative analysis of the different NIBS techniques

Among the various NIBS techniques used in chronic stroke rehabilitation, tDCS was the most widely applied, particularly for enhancing both upper [30,33,34,37,39,40,44,48,50–53] lower limb [55,56,58,60,62,68,69,71,81,11, 83–85] motor functions. tDCS showed significant benefits for lower limb recovery, with contra-lesional cerebellar stimulation improving balance [55] and anodal stimulation over the M1 promoting gait function [23]. These effects are further enhanced when it is combined with RAGT [11,60,62,81,83,84,85]. Additionally, tDCS has been linked to improvements in upper limb strength and dexterity, particularly when paired with constraint-induced movement therapy [47], and it showed promise for treating speech and swallowing disorders [59,64,86]. The advantages of tDCS include its simplicity, low cost, portability, and ease of integration with physical therapies, making it a widely accessible option for post-stroke rehabilitation. However, a significant limitation of current tDCS protocols is their reliance on fixed stimulation parameters, such as electrode placement, stimulation intensity, and polarity (anodal vs. cathodal). These factors may not account for individual differences in neurophysiological responses. Future research could benefit from a more tailored approach, adjusting tDCS to preserve critical neural pathways, such as the CST, and guiding it with MEPs. This individualised approach, where both stimulation parameters and electrode sites are customised to each patient’s unique brain activity, could enhance tDCS efficacy and lead to more consistent improvements in motor function. Furthermore, in the future, neuro-navigation and real-time EEG monitoring of NIBS effects may help reduce variability in outcomes, offering a more precise and personalized treatment method.

Moreover, iTBS has demonstrated promise in recovering gait, balance, and lower limb stability [70,72–74], especially when applied over the cerebellar lobes. It also reduces upper limb spasticity [36] and modulates corticospinal excitability, with more notable outcomes when used with priming protocols or in combination with other NIBS approaches like rTMS [42,43]. However, iTBS requires highly accurate targeting and appears to be more effective in patients with preserved motor function, limiting its use in more severe cases.

rTMS is primarily used to support upper limb motor recovery and reduce spasticity, often through low-frequency stimulation of the unaffected hemisphere [31,32,38,40–43,45,49,53,54]. It has also shown potential benefits in lower limb function and gait symmetry [75–77], particularly when combined with treadmill training [82]. However, the role of rTMS in swallowing rehabilitation remains inconclusive due to mixed study results [65,66,86,87]. rTMS is valued for its precise cortical targeting and well-established protocols, but it is more expensive, less accessible, and less portable than other NIBS techniques. Additionally, its efficacy may vary depending on stroke severity and stimulation parameters.

Although overall results showed clinical benefits across all NIBS modalities, differences in underlying mechanisms may partly explain the variability in treatment response. tDCS modulates cortical excitability through subthreshold changes in neuronal membrane potentials, leading to long-term neuroplastic effects when combined with task-specific training [12,13]. In contrast, rTMS induces direct neuronal depolarization via magnetic pulses, allowing for more focal and frequency-dependent modulation of cortical circuits [1,3,4]. These mechanistic distinctions may explain the stronger or more consistent effects reported with rTMS in some motor outcomes [2,5]. However, individual variability, stimulation parameters, and integration with rehabilitation protocols also contribute to treatment efficacy [13–15].

In summary, while each NIBS technique offers distinct benefits, their effectiveness depends on patient-specific factors such as stroke severity, residual function, and therapy goals (see Table 5) [39,45,55,73].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Summary of advantages and disadvantages of each NIBS techniques, according to the literature.

https://doi.org/10.1371/journal.pone.0327583.t005

The best outcomes are often achieved through individualized, combined approaches that leverage the strengths of each technique. The choice of intervention should be tailored to the patient’s clinical profile and the targeted functional outcomes, with the understanding that a personalised, multi-modal approach may yield the most significant benefits in post-stroke rehabilitation.

Strengths and limitations

This review presents a comprehensive and qualitative synthesis of the current evidence on the effectiveness of NIBS in the field of chronic stroke. It highlights the potential of NIBS interventions to promote motor functional recovery, in addition to conventional or robotic-assisted therapy, fostering neuroplastic changes. The strengths of this review lie in its analysis of diverse applications of NIBS, providing insights into its therapeutic possibilities. Additionally, the inclusion of robotic devices in addition to neuromodulation therapy reflects the growing emphasis on personalised treatment approaches.

However, several limitations must be acknowledged. This review is based solely on a qualitative analysis, which, while providing valuable descriptive insights, does not allow for statistical evaluation of effect sizes or direct comparisons between studies. As a result, our review provided a comprehensive qualitative synthesis of the available evidence, offering valuable insights into the field of NIBS applications for specific motor deficits in individuals with chronic stroke, identifying key implications for clinical practice and considerations for future investigation.

Despite the large number of selected studies, many included trials were conducted with small sample sizes and heterogeneous methodologies, limiting the generalizability of the findings. Furthermore, the lack of standardised NIBS protocols complicates comparisons across studies. These limitations pertain to the quality and design of the included studies and should be considered when interpreting the results.

Limitations of the review process itself must also be acknowledged. Although rigorous methods were applied, including blinded screening and inter-rater agreement assessment, the possibility of publication bias and language bias cannot be fully excluded, as only studies published in English and peer-reviewed journals were considered.

Several questions remain open, such as the optimal NIBS dosage, the most effective timing of administration (e.g., before or after rehabilitation therapy), and the duration of its retained effects. Future research should prioritise longitudinal clinical trials to assess whether the benefits of NIBS are sustained over extended periods. Given the persistent nature of motor deficits, it is essential to determine if NIBS-induced motor recovery remains effective over months or years, thus providing stronger evidence for its integration into long-term rehabilitation management strategies.