Transcranial Magnetic Stimulation as an Intervention Tool to Recover from Language, Swallowing and Attentional Deficits after Stroke: A Systematic Review

Stroke, or cerebrovascular accident, represents the third leading cause of death and one of the most common sources of disability worldwide [1, 2]. Motor deficits after stroke have been given particular attention; however, other types of deficits are also relevant such as communication skills [3-5], swallowing abilities [6, 7] and attention [8, 9].

Aphasia, described as a total or partial loss of language functions, [3] is a very frequent disability among stroke patients and severely restricts communication and the ability to engage in social interactions [4, 10-12]. This syndrome arises from damage to the language dominant hemisphere – generally the left hemisphere in right-handed people [4, 13, 14]. In fact, aphasia is only associated to a lesion on the right hemisphere in 4% of the aphasia cases with poststroke patients [13]. Nonfluent aphasia originates problems in speech output such as interrupted speech, word omission or statements with limited syntactic complexity [15]. As a consequence of stroke, it affects about 38% of the patients [3-5] and becomes chronic in 10–18% of survivors [3, 5].

Dysphagia, characterized by a difficulty in swallowing, is also a common poststroke outcome, affecting up to 78% of patients [6, 7, 16]. Although the recovery is frequent within a few weeks, its extent varies considerably between subjects [6, 17]. A patient is more likely to develop dysphagia if the stroke affects the dominant hemisphere instead of the non-dominant one [6]. This condition raises the probability of death mostly owing to an increased risk of pulmonary complications [18] like pneumonia which, in turn, is associated with a third of stroke deaths [19].

Spatial neglect, that is, the inability to attend, reply, react or orient to stimuli located in the contralesional portion of space, [1, 2, 8, 9, 20-22] affects from 30 to 81% of patients in the acute phase [8, 9] and it is sustained in approximately a third of them [9]. It appears most often due to right hemispheric lesions of the middle cerebral artery, damaging the neural substrates of space representation and awareness [2, 8, 23]. Additionally, it can emerge associated to damage of other areas such as the parietal or frontal lobe, thalamus or basal ganglia [2]. This disability slows down the functional rehabilitation and increases the length of hospital stay [1, 22]. When poststroke patients become chronic, there is a decrease of the incidence of overt neglect and often only signs of visual extinction are observed, characterized by “an inability to detect a contralesional stimulus when an ipsilesional stimulus is simultaneously presented” [24].

The search for new methods able to increase the efficacy of recovery programs is crucial [17, 20]. It is well recognized that, following stroke, the balance of function between the hemispheres is disturbed and the affected hemisphere becomes more inhibited, while the hemisphere contralateral to the lesion shows an increased activity [3, 25, 26]. It is strongly believed that this imbalance of inter-hemispheric excitability limits considerably the recovery of function after stroke [25, 27]. Transcranial magnetic stimulation (TMS), a technique for the non-invasive brain stimulation, arises as a novel approach to rehabilitate motor and non-motor deficits in stroke patients, due to its ability to modulate brain plasticity [25, 26], either by increasing excitability (with high-frequency protocols or intermittent theta burst stimulation paradigms) or by decreasing it (low-frequency or continuous theta burst stimulation paradigms) [2, 5, 8, 11, 17, 25, 28].

In a previous work [29], we performed a systematic review to study the use of repetitive TMS on the rehabilitation of motor function, following stroke. Here, we assess the applicability of TMS to the rehabilitation of non-motor deficits such as post-stroke aphasia, dysphagia and neglect.

We identified articles on PubMed and ISI Web of Science, -using the search terms: (rTMS OR “repetitive transcranial magnetic stimulation”) AND (stroke OR “cerebrovascular accident” OR CVA) AND (rehab OR rehabilitation OR recover*). The last search was performed on September 12, 2017. We included studies that aimed to improve aphasia, dysphagia, neglect or visual extinction, in post-stroke patients, with repetitive TMS protocols. As for our preceding systematic review [29], we excluded (1) reviews; and studies (2) written in any language other than English; (3) in paediatrics; (4) performed in animals; (5) recruiting just healthy subjects; (6) including less than 5 participants; (7) not using repetitive TMS; (8) applying other stimulation techniques rather than TMS; (9) studying other disease or condition instead of stroke; (10) where the main goal was not to test the efficacy of repetitive TMS on the rehabilitation behavioural outcomes; and (11) that did not report explicitly the complete TMS protocol (including coil, stimulation area, number of sessions, frequency, intensity and pattern).

Data from each study was extracted and the most relevant information was added to a data extraction sheet. This included experimental design, number of participants, clinical characteristics of the patients, therapies subjects were undergoing besides the TMS, description of the stimulation protocol, outcome measures and main results.

The search we performed in PubMed and ISI Web of Science retrieved a total of 745 records, as can be observed in Figure 1 (designed according to the PRISMA statement requirements [30]). From these records, 299 were duplicates, remaining 446 for the initial screening based on titles and abstracts. After excluding 272 articles that did not fit the defined criteria, we accessed a total of 174 full-text articles. At the end of this procedure, we included in our qualitative synthesis 38 articles, from which 21 focused on aphasia recovery, 8 on dysphagia, 8 on neglect and 1 on visual extinction rehabilitation. These articles were all published between 2009 and 2017 and included a total of 827 adult patients.

We provide relevant information regarding the stimulation protocols and the reported results on online -supplementary material Tables S1.1, S1.2 and S1.3 (for all online suppl. material, see www.karger.com/doi/10.1159/000494213), for aphasia, dysphagia and neglect or visual extinction, respectively. Moreover, on online supplementary Tables S2.1–S2.3, we present clinical characterization of the patients that were included in the studies we revised.

In this section, we point out the main findings reported by the authors, concerning the stimulation protocols and their efficacy as a rehabilitative intervention.

For aphasia rehabilitation, we considered a total of 21 articles, published between 2011 and 2017. The majority of the studies on aphasia recovery focused on chronic stroke patients; 62% of the included research recruited patients that had had the stroke event more than 1 year before. The stable baseline condition, in the chronic phase, favours a more objective assessment of the TMS effects. However, Kindler et al. [3] observed that the best responders to TBS were those with a shorter interval poststroke. The most part of the analysed studies (20 out of 21) studied the unaffected (right) hemisphere in the inferior frontal gyrus as a stimulation target, the majority in pars triangularis (Brodmann area 45). Medina et al. [15] found the right pars triangularis to be the optimal site of stimulation in 9 of the 10 patients and Naeser et al. [14] claimed that suppression of right pars triangularis, but not of pars opercularis, improved naming in aphasia.

Studies included in this review reported improvements in aphasia recovery with the application of rTMS, mainly in picture naming. Rubi-Fessen et al. [10] also noted that the literature pointed out the most pronounced improvements for picture naming, when rehabilitating the language function. However, the authors obtained significant improvements caused by stimulation on auditory and written comprehension, writing and reading and on functional communication too [10]. Additionally, Barwood et al. [31, 32] also described improved spontaneous speech and auditory comprehension. In this way, besides the improvements in expressive language, rTMS can improve receptive language performance as well [31]. Some authors reported an association between changes in brain activity induced by rTMS and language improvement [4, 33, 34], while others did not [12]. Barwood et al. [32], Chieffo et al. [11] and Seniów et al. [5] observed that the patients that obtained larger improvements were those with global aphasia and more severe deficits, suggesting that this tool can modulate language performance even in individuals with quite significant lesions. On the contrary, Naeser et al. [14] observed less improvement in patients with more severe impairment.

Barwood et al. [32] postulated a need to understand the variability of patients’ response to rTMS and to explain why there are different responses to treatment. Tsai et al. [35] found no correlation between the effects of the treatment and either subject’s age, education, National Institute of Health Stroke Scale score, aphasia type and severity, time since stroke onset or baseline Functional Independent Measurement. Interestingly, they provided evidence that hyper-excitability of the contralesional hemisphere and the absence of diabetes mellitus comorbidity are related to a better response to rTMS treatment for aphasia recovery [35].

Chieffo et al. [11] described improvements with 10 Hz-rTMS that were significantly larger than with 1 Hz-rTMS, both applied in the right hemisphere. Heiss et al. [13] recruited both right-handed patients with left-hemispheric stroke and left-handed subjects that experienced a right-hemispheric infarct that led to aphasia. They -obtained promising results with low-frequency rTMS (LF-rTMS) and speech and language therapy in the right-handed patients, observing an activation shift to the dominant hemisphere with a concomitant significant recovery in language function. Interestingly, left-handed patients also revealed some improvement in language although a significant shift of activity to the affected hemisphere was not observed [13]. Findings from functional magnetic resonance imaging described greater activation in the right hemisphere of stroke patients that were recovering from aphasia, when compared to healthy subjects, which suggests that, although it is reported that LF-rTMS applied to the right hemisphere can be effective in improving language, it can be deleterious in those patients [36]. Indeed, it was argued that the characteristics of a left-hemispheric lesion, namely, its location and size, can influence the contribution of the right hemisphere to the improvement of language function [36]. Hara et al. [36] used functional near-infrared spectroscopy to localize the hemisphere that was activated for language function, and, this way, group subjects. Patients with a stronger activation for the language functioning on the left hemisphere received a low-frequency protocol applied to the right hemisphere, whereas a high-frequency protocol was applied to the right hemisphere of those subjects with a stronger activation in this hemisphere. Both groups showed improvements in aphasia at comparable levels [36].

The exploratory study from Medina et al. [15] showed less promising results. Although they observed an improvement on discourse productivity, in the other 3 measures of fluency they tested, namely, sentence productivity, grammatical accuracy and lexical selection, they did not find significant improvement in addition to those observed with sham stimulation. Waldowski et al. [37], in turn, conducted a randomized controlled trial where they made some reservations about the effect of low-frequency rTMS stating that, even though this therapy can be favourable for the rehabilitation of patients with a lesion including the anterior part of language area, it cannot be assumed to be an effective method for all poststroke aphasic patients and that its efficacy should be confirmed. Seniów et al. [5] also observed in their randomized controlled study that this approach was not effective for all aphasics and that the response could depend on individual differences and factors such as lesion site and extent. Interestingly, as previously reported by Waldowsky et al. [37], they also described here a superior, although modest, improvement on those patients with a lesion affecting the frontal part of the language area [5]. According to Rubi-Fessen et al. [10] 1 possible explanation for the lack of efficacy is the method Seniów et al. [5] used to place the coil over the inferior frontal gyrus (10–20 method), which is said to be not as accurate as the surface distance measurement method adopted by the former.

We included 8 studies focused on dysphagia recovery, published from 2009 until 2017. Out of the 8 studies, only two of these recruited patients in the chronic phase of the stroke. Cheng et al. [16] pointed out a need for more studies confirming the efficacy of TMS in improving chronic dysphagia. Moreover, applying rTMS in the early phase can be a source of bias when interpreting the intervention results, since patients frequently recover their swallowing abilities in this stage by natural mechanisms [18]. rTMS was applied to the oesophageal cortical representation area [6, 19], pharyngeal motor cortex [18], suprahyoid muscle cortical area [7], mylohyoid cortical area [17, 38, 39], tongue cortical area [16] or abductor pollicis brevis cortical area [7]. A great difference between protocols was observed. Studies applied low-frequency rTMS to suppress the nonlesioned hemisphere [17, 38], or high-frequency to facilitate the excitability of either the unaffected [18], the affected [6, 7, 16, 17] or even both hemispheres simultaneously [19]. Park et al. [39] set up a treatment group where patients received high-frequency rTMS (HF-rTMS) over mylohyoid cortical area of the affected hemisphere followed by the same protocol over the unaffected hemisphere and a group receiving HF-rTMS to the affected hemisphere and sham stimulation on the unaffected hemisphere. Stimulation of either hemisphere is hypothetically valid since the swallowing musculature has its representation on both hemispheres and is innervated bilaterally [17].

All but one [16] of the included studies demonstrated qualitatively good results in improving dysphagia, and were able to describe that patients recovered swallowing ability into different extents. Du et al. [17] applied 1 Hz stimulation to the contralesional hemisphere or 3 Hz to the lesioned hemisphere over the mylohyoid cortical area. The authors reported 1 Hz significantly decreasing the cortical excitability of the unaffected hemisphere and, simultaneously, enhancing significantly the excitability of the affected side. On the other hand, using 3 Hz induced a significant increase of the affected hemisphere’s excitability, but regarding the unaffected hemisphere it only caused a modest change [17]. However, according to Khedr et al. [6], the recovery after delivery of 3 Hz-rTMS to the affected hemisphere was related with an upregulation of excitability in the corticobulbar projections from both hemispheres. Therefore, stimulation of the swallowing cortical representations of one hemisphere conducted to an increase in excitability in both hemispheres [6]. Oropharyngeal dysphagia is thought to be related with a smaller pharyngeal representation on the unaffected hemisphere that grows with return of swallowing [18]. Thus, Park et al. [18] stimulated the pharyngeal motor cortex in the unaffected hemisphere with 5 Hz-rTMS to increase its excitability and potentially enhance recovery from dysphagia. They reported significant improvement in the function of the pharyngeal phase but not in the oral phase [18].

The neural basis of swallowing has been related to multiple cortical and subcortical areas [7]. Lee et al. [7] divided their patients into 2 groups. In the first group, they used HF-rTMS to stimulate a specific dysphagia-related target, namely, the cortical region representing the suprahyoid muscle of the affected side, and described improvements of the swallowing function. In the other group, they used the same parameters to stimulate the cortical area representing the abductor pollicis brevis muscle of the affected side and also observed improvements in swallowing. Possible explanations for these findings are that they are due to the natural recovery or to stimulation of the interconnected site. In fact, once the white matter is interconnected, the stimulation of M1 could have triggered the stimulation of swallowing-related regions. Nevertheless, the authors reported that stimulation of the cortical region representing the suprahyoid muscle was more effective for dysphagia rehabilitation than stimulation of the cortical area representing the abductor pollicis brevis [7].

Park et al. [39] compared the efficacy of the stimulation of both hemispheres to the stimulation of the affected hemisphere, both at 10 Hz, and observed a considerable superior improvement with bilateral stimulation. Actually, unilateral stimulation was not significantly more effective than sham stimulation in their work, contrarily to what was reported in the literature, which the authors justified with the variability on subjects’ characteristics and the small sample size.

Cheng et al. [16] conducted a double-blind, randomized, controlled study where they performed 5 Hz rTMS applied to the tongue cortical area of chronic patients’ affected hemisphere and failed to observe significant effects of the treatment of the swallowing function. They identified various possible explanations for their negative results, including the stimulation protocol, which might not be optimized; the outcome measure that was used to assess swallowing function, that could lack sensitivity; the low severity of the deficits, and the absence of an additional therapy such as tongue or swallowing exercises.

We studied 8 publications focusing on neglect rehabilitation and 1 dealing with visual extinction. These were published between 2009 and 2016. The literature applied cTBS [1, 20] or LF-rTMS [8, 9, 21, 22] to the unaffected hemisphere over posterior parietal cortex or HF-rTMS to the stroke affected hemisphere, [8] as an adjuvant to conventional therapy. Moreover, Cha and Kim [23] applied LF-rTMS over P3, based on the International 10/20 system, and Yang et al. [2] stimulated the contralateral posterior parietal cortex either with cTBS, LF-rTMS or HF-rTMS.

For neglect rehabilitation, 88% of the studies included in this review admitted patients within the first 6 months, considering the mean time post-stroke. Kim et al. [21] recruited subjects that had the stroke more than 1 year before entering the study.

All studies reported some improvement. However, although Lim et al. [9] observed improvements on their pilot study in the line bisection test, indicating a potential enhancement of recovery in patients with neglect, the Albert test did not show significant differences between the results obtained with or without stimulation. The authors pointed out a need for a prospective randomized, sham-controlled study to evaluate the efficacy of stimulation on hemispatial neglect.

Kim et al. [8] reported that HF significantly improved neglect more than LF-rTMS, which suggests future studies should evaluate the delivery of HF-rTMS to the affected hemisphere. Yang et al. [2] compared different protocols of stimulation, using low-frequencies (1 Hz), high-frequencies (10 Hz) or cTBS and obtained the greater effectiveness with the continuous TBS, demonstrating more clear-cut results. Actually, as confirmed by diffusion tensor imaging, continuous theta burst stimulation enhanced connections of white matter’s tract considerably, which suggests recovery at the structural level [2].

Kim et al. [21] found significantly larger improvements with 10 sessions of LF-rTMS than with a single session. Furthermore, the treatment effects were superior for allocentric compared to egocentric neglect, which was explained by the fact that allocentric neglect patients had wider brain lesions and, thereafter, more severe baseline symptoms [21]. Fu et al. [20] provided initial evidence that increasing number of training runs per day and of stimulation days might enhance and lengthen cTBS efficacy on improvement of visuospatial neglect.

The effects of treatment were evaluated by different outcome measures. Inconsistent changes across outcome measures were noticed, which suggests that different forms of spatial neglect might show distinct test measure responses [8, 9]. Yang et al. [2] observed different responses to the selected tests; patients showed significant greater improvements after intervention in star cancellation test, in comparison to line bisection test. The authors also assumed that this difference could have been originated by the functional heterogeneity of patients [2]. Fu et al. [20] suggested that applying a combination of multiple neglect tests could be more sensitive than using only a test to detect and define the presence of visuospatial neglect.

Agosta et al. [24] applied 1 Hz-rTMS over the left parietal cortex to reduce visual extinction due to right parietal damage and observed improvements in sustained attention only in the left visual field.

When searching for the application of repetitive TMS protocols to the rehabilitation of aphasia, dysphagia and neglect on stroke patients, it stands out the relatively smaller number of records retrieved, in comparison to the works that deal with motor function rehabilitation [29]. This observation stresses the need for more studies evaluating the efficacy of this therapeutic intervention on these disabilities. Moreover, a great number of the studies included in this review recruited a small number of participants.

The influence of the recruitment strategy of the participants must be focus of particular attention, when considering potential sources of bias. We observed that in both dysphagia and neglect rehabilitation, in most of the studies, patients were recruited in the early phase after stroke. This may represent an important source of bias due to the natural recovery mechanisms that frequently occur at this stage [18]. On the other hand, most studies on aphasia focused their attention on the recovery of chronic patients, although shorter times after stroke are believed to be optimal to achieve the greatest modulation of plasticity [3]. The selection of the best moment to intervene, and the time dependence of the intervention should be further studied in larger cohorts of patients, including sham-stimulation groups.

Common to the studies focusing the recovery of aphasia and neglect on this review was the overall preference by applying inhibitory protocols, either low frequencies or continuous theta burst stimulation. Supporting this approach, Yang et al. [2] found cTBS over the contralesional hemisphere to be more effective than low-frequency or high-frequency TMS on the rehabilitation of spatial neglect. Yet, Chieffo et al. [11] and Kim et al. [8] both reported superior improvements with high frequencies applied to the affected hemisphere in opposition to low frequency stimulation of the contralateral hemisphere in the rehabilitation of aphasia and neglect, respectively. Moreover, Hara et al. [36] stated that applying LF-rTMS to the right hemisphere in aphasia rehabilitation is not suitable for all patients and can even be detrimental on those patients with stronger activation for language on the right hemisphere. In their study, they obtained good results with an HF protocol over the right hemisphere [36]. On the other hand, in the most part of the studies dealing with dysphagia, researchers chose to apply excitatory protocols, either high frequencies or intermittent theta burst stimulation.

The stimulation area was more consensual across studies. For aphasia rehabilitation, most part of the authors chose to stimulate the inferior frontal gyrus, while for neglect the posterior parietal cortex was selected. Dysphagia, in turn, presented a wider inter-study variability both on the stimulation area and on the design of the protocols.

The existing variability limits the conclusions that can be drawn through the observation of the included literature results. In fact, the lack of consistency across studies regarding the selection of the participants, of the protocols and even of the outcome measures that are used to evaluate the efficacy of intervention render definite conclusions about the best protocol not yet possible, because criteria for a metanalysis are not fullfilled. Future studies are mandatory to define the optimal TMS parameters, number of sessions and suitable outcome measures, since protocol standardization would be crucial to enable the ultimate evaluation of the efficacy of this technique as a rehabilitative intervention.

As we had already observed in our previous work [29], another critical point is the inter-subject variability and patient stratification. We believe that authors should concentrate their efforts on understanding the different responses to protocols, select those patients that could benefit the most from TMS and define biomarkers that could act as predictors of greater efficacy.

Albeit there are several outstanding issues, almost all works included in this systematic review supported the use of TMS on stroke rehabilitation, presenting positive results on the improvement of communication, swallowing and attentional deficits, without the occurrence of serious adverse effects.

The application of repetitive TMS protocols to the recovery of a stroke event has been receiving increasing attention in the last years. Considering the novelty of this approach, there are still major issues that need to be investigated, being the most prominent the definition of the parameters of stimulation that bring out the best results. Still, the potential of this method is undeniable. The large majority of the included studies supported the use of TMS for this purpose, reporting statistically significant improvements in aphasia, dysphagia and neglect. Larger clinical trials are needed to validate the efficacy of this technique for stroke rehabilitation.

The authors have no ethical conflicts to disclose.

The authors have no conflicts of interest to declare.

This work was supported by BIGDATIMAGE (CENTRO-01–0145-FEDER-000016 “From computational modelling and clinical research to the development of neuroimaging big data platforms for discovery of novel biomarker”), EU H2020-MSCA-IF-2015, 708492 TMS_ATT, and by Fundação Luso-Americana para o Desenvolvimento (Prémio FLAD Life Sciences 2020 – “Linking inhibition from molecular to systems and cognitive levels: a preclinical and clinical approach in autism spectrum disorders and neurofibromatosis”). The sponsors had no involvement in study design, in the collection, analysis and interpretation of data, article preparation or in the decision to submit the article for publication.

M.C.-B. gave his substantial contribution to the conception of the work. A.D., M.P., and M.C.-B. contributed to the study design. A.D. collected, analysed and interpreted data, with the substantial support from I.C.D., M.P., and M.C.-B. on data analysis and from I.C.D. and M.C.-B. on interpretation. A.D. wrote the draft and I.C.D., M.P., and M.C.-B. revised it critically for important intellectual content. All authors approved the final version to be published and agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.