Introduction

Approximately one third of all people with acute stroke have aphasia, the vast majority due to left hemisphere (LH) lesions [1, 2]. Improvements in language are fastest in the first few months after stroke and gradually slow down over time [3]. Although recovery is highly variable, on average survivors achieve 70 % of the maximum possible recovery on common aphasia tests 90 days after stroke [4]. The only widely accepted treatment for post-stroke aphasia is speech-language therapy, which improves outcomes in some aspects of language and functional communication [5]. However, the effects of speech-language therapy on overall aphasia outcomes are relatively modest, and about two thirds of people with aphasia at stroke onset who survive to follow up continue to have chronic language deficits 18 months later [6]. Living with aphasia reduces participation in life activities, independence, and mood and increases the cost of care [7–10]. Clearly, new treatments are needed to improve outcomes for people living with aphasia.

Aphasia specialists have long sought biologically based interventions to augment recovery. A number of medications have been tested for aphasia, based on the theory that increasing neurotransmitter availability in partially disrupted pathways may mitigate deficits related to these disruptions. Correcting these neurotransmitter disruptions could enhance language abilities on a day-to-day basis or potentially facilitate relearning of language skills during speech-language therapy, much as stimulants that improve attention in a child with attention deficit disorder might improve learning at school. The results for some medications have been encouraging [See 11 for review]. In general, these studies have been small, and larger, more definitive trials are needed.

A more ambitious and potentially more impactful goal of biologically based treatments is to alter the process of brain reorganization that underlies aphasia recovery. After the initial stabilization of blood flow and resolution of brain swelling in the first weeks after a stroke, recovery relies on reorganization of brain networks, which occurs both spontaneously and in response to behavioral training [12], such as speech-language therapy. As in animal models of sensorimotor stroke, the primary changes after a stroke causing aphasia include recruitment of perilesional tissue adjacent to the stroke in the LH and recruitment of homotopic (mirror image) right hemisphere (RH) sites (Fig. 1) [13•]. Engagement of preserved language areas of the LH and recruitment of nearby perilesional tissue is widely thought to support recovery of language functions [14–22]. The role of the RH in recovery is less clear [23••]. Optimizing brain reorganization to achieve a maximally efficient language network could theoretically yield significant gains in aphasia recovery. This could potentially be accomplished in various ways, for instance enhancing plasticity in the months following stroke, coaxing the brain’s language network into a more efficient organization, or by restoring plasticity long after stroke. Based on these goals, a number of small studies have recently examined whether noninvasive electrical or magnetic brain stimulation can improve aphasia recovery. These techniques are used to excite or inhibit particular areas of the brain and are being tested for use on a wide range of neurologic and psychiatric conditions [24]. There has been a great deal of excitement about these techniques amongst aphasia researchers, clinicians, patients, and families. Indeed, results from early studies have been encouraging, although certainly not definitive.

Fig. 1
figure 1

Reorganization of language networks in post-stroke aphasia. Results are shown from a meta-analysis of functional neuroimaging studies of people with chronic post-stroke aphasia and matched control subjects (Adapted from [13•]). The top row shows normal brain activity in the left-lateralized language network in control subjects. The bottom row shows the reorganized brain activity during the same language tasks in people with aphasia. Results are rendered onto a template brain so the locations of stroke damage are not shown. Three patterns are notable: (1) preserved activity in native LH language areas when viable tissue remains in these areas, (2) shifts in activity within the LH to so called “perilesional” locations, and (3) activation of a “RH language network” that mirrors the native LH language network activated by controls

Full size image

Most brain stimulation studies on aphasia to date have aimed either to enhance activity in brain areas thought to support good recovery from aphasia or, more commonly, to suppress activity in brain areas thought to interfere with recovery. These studies have proceeded, however, even as neuroscientists continue to debate basic questions about how language networks reorganize after stroke. There is a striking lack of consensus in the literature regarding key aspects of the neuroscientific theories guiding the use of brain stimulation for aphasia, particularly the role of the RH in recovery [23••]. In addition, the results of brain stimulation trials are often used as evidence supporting particular theories of aphasia recovery, despite a lack of neurobiological data to confirm the mechanisms of action of these methods. Without a better understanding of the brain basis of aphasia recovery and the mechanisms by which brain stimulation improves outcomes, the field risks developing suboptimal brain stimulation treatments based on erroneous assumptions and then reinforcing those assumptions based on weekly positive clinical results. A more thorough understanding of the brain basis of aphasia recovery and of the neurobiological effects of brain stimulation techniques will increase our chances of developing new interventions that have a meaningful clinical impact on outcomes for people with aphasia.

Below, I will describe the two main forms of noninvasive brain stimulation currently being investigated for use in post-stroke aphasia, and the current state of the evidence supporting their use. I will then describe the controversy surrounding the neurobiological theories guiding the most common approaches to treatment and suggest ways to improve the chances of turning these promising investigational techniques into meaningful clinical interventions.

Noninvasive Brain Stimulation Methods Used for Aphasia

The two most common noninvasive brain stimulation techniques are repetitive transcranial magnetic stimulation (rTMS) and transcranial direct current stimulation (tDCS) (Table 1). rTMS involves passing a very brief high-current electrical pulse through an insulated coil of wire held over the scalp. The pulse produces a rapidly changing magnetic field that induces electrical current in underlying brain tissue, causing neuronal firing. rTMS, as commonly used for aphasia, is thought to affect a small area of brain tissue directly under the coil (approximately 1 cm3), although downstream effects are expected in areas connected to the site of stimulation [25, 26]. Low-frequency rTMS (typically 1 Hz) reduces excitability of the stimulated cortical site, whereas high-frequency rTMS (>5 Hz) increases excitability. These effects persist minutes to hours after rTMS is stopped, and daily repeated sessions can induce more durable effects although the biological basis of these long-lasting effects is less clear. This “off-line” effect is the basis for therapeutic uses of TMS in a variety of neurological and psychiatric conditions [27].

Table 1 Comparison of practical aspects of brain stimulation methods used in treatment studies of post-stroke aphasia
Full size table

In contrast to rTMS, which induces electrical currents in the brain using magnetic fields, tDCS directly applies a low level of constant electrical current to the scalp using electrodes [28]. In ex vivo studies, this direct electrical current slightly depolarizes or hyperpolarizes neurons, making them more or less likely to fire [29]. In aphasia treatment studies, the current is typically applied for 10–30 min, often in combination with speech-language therapy. Unlike the highly focal effect of TMS, tDCS has an anatomically broader effect. Electrical field modeling suggests that entire lobes of the brain may be impacted by typical tDCS methods, which use large saline-soaked sponges as electrodes [30]. Recently, “high definition tDCS” methods have been developed using multiple smaller electrodes to focus the electrical current [31•] although the effects are still thought to be less localized than typical rTMS methods. Like rTMS, tDCS can induce localized excitation or inhibition of neuronal populations that can last for minutes to hours after a short session [28, 32]. The polarity of the tDCS electrode is widely thought to determine whether the effect on the underlying brain tissue is excitatory or inhibitory. The anode is generally thought to induce excitation, whereas the cathode induces inhibition. This rubric, however, has been brought into question by electrical field modeling studies demonstrating large areas of current flow between the electrodes [31•] in sometimes unexpected distributions especially in people with brain lesions [33]. Further, effects of cathodal stimulation on cognitive tasks in healthy subjects have been inconsistent [34]. Despite these issues, most tDCS studies in clinical populations assume that activity in the brain area under the anode will be facilitated by stimulation, whereas the area under the cathode will be inhibited. Both rTMS and tDCS can enhance learning during motor or language training in healthy subjects [35–38]. These effects have generated a great deal of hope that rTMS or tDCS might improve language and communication outcomes for people with post-stroke aphasia.

Evidence to Date for TMS and tDCS Treatments of Aphasia

Naeser and colleagues provided the first evidence that noninvasive brain stimulation might improve aphasia, applying low frequency rTMS to the pars triangularis of the right inferior frontal gyrus (IFG) in a small group of people with chronic nonfluent aphasia [39, 40]. These initial open-label studies suggested that this type of stimulation, aimed at suppressing the RH homolog to Broca’s area, improved picture naming for up to 8 months after 10 daily sessions of treatment applied over 2 weeks.

Since the initial reports of Naeser and colleagues, over 20 other studies have tested rTMS treatments for post-stroke aphasia, with or without adjunct speech-language therapy. Collectively, these studies have included over 200 patients during either the subacute or chronic phases of recovery. Although a few have used other stimulation strategies (e.g., [41]), most rTMS studies on aphasia have used low frequency stimulation to the RH homolog of Broca’s area, as in the original Naeser and colleagues’ work. A recent meta-analysis of seven such trials including 160 patients found significant positive effects on naming, repetition, writing, comprehension, and global language impairment, with standardized mean differences of 0.32 to 0.70 on individual language domains, and 1.26 on global impairment [42••]. Studies that have examined durability of effects suggest that language improvements may last months after a single 2-week course of treatment in chronic patients [39, 43, 44].

tDCS studies on aphasia have emerged more recently and have been somewhat more varied in their design. Studies have targeted the LH [45], or the RH [46], or both [47]. Anodal and cathodal stimulation have variously been applied to each hemisphere in both fluent and nonfluent patients [48•]. Compared to rTMS, relatively fewer tDCS studies have been published to date, and the variability in the methods used makes it difficult to determine if effects on aphasia are reliable. A recent Cochrane review of six tDCS studies including a total of 66 patients concluded there were no reliable effects on picture naming [49•]. It is perhaps too early for this kind of quantitative review on tDCS, and future meta-analyses including more homogeneous groups of studies may yield different results. One recent meta-analysis combined six rTMS and three tDCS studies that aimed to suppress RH activity and found a significant positive effect of these methods on picture naming with a standardized mean difference of 0.52 [50•].

A common criticism of brain stimulation treatment studies for aphasia has been the focus on specific language tasks, particularly picture naming, as the outcome measures, rather than more ecologically valid measures of functional communication [49•]. Although it is hard to draw any firm conclusions regarding the importance of effects demonstrated to date on real-life communication ability, the primary aim of the small early phase studies conducted so far has been to establish safety and prove that particular stimulation protocols can modulate language abilities in aphasia. For these early stages of methods development, it is reasonable to focus on sensitive measures like picture naming. Larger phase II and phase III trials in the future should focus on more clinically relevant outcome measures after factors like location, duration, and type of stimulation are optimized.

Overall, the most consistently successful brain stimulation treatments for aphasia to date have utilized low frequency rTMS or cathodal tDCS aiming to inhibit the RH, most typically the right IFG [50]. This consistency may be somewhat misleading, however, as RH inhibition has been the most commonly used treatment strategy to date, based on the pioneering work of Naeser and colleagues. Indeed, some crossover studies have shown greater benefit for other modes of stimulation, including stimulation intended to inhibit the LH [51] or excite the right [46]. Despite these conflicting findings, many have taken the beneficial effects of low frequency rTMS and cathodal tDCS over the RH as evidence that involvement of the RH in aphasia recovery is maladaptive. However, these conclusions must be considered in the larger context of neuroscience research on aphasia recovery, in which there is still a great deal of debate about the role of the RH. It is thus worth considering the practical and scientific motivations for inhibiting the RH to facilitate aphasia recovery, along with supporting and contradictory neuroscientific evidence for this treatment strategy.

Why Inhibit the RH?

Targeting the intact RH provides distinct practical advantages for brain stimulation treatments of LH stroke survivors with aphasia. Since stroke locations differ within the LH between patients, stimulating the LH requires individualized targeting to ensure that TMS or tDCS is administered to intact brain tissue rather than an area of encephalomalacia. In addition, because reorganization of LH language circuits likely differs depending on stroke location, extra techniques like fMRI may be needed to ensure that spared areas of the LH are involved in language processing [52]. There are also theoretical safety concerns for stimulating tissue around the lesion, including current shunting through cerebrospinal fluid cavities and seizure induction from excitation of epileptogenic tissue, although the risk of significant adverse events with either TMS or tDCS is extremely low regardless of the brain area stimulated [53, 54]. Targeting the intact RH allows for the possibility of identifying a single target that can be used across groups of people with aphasia, without these complications [13]. The simplicity of this approach could be key to making brain stimulation treatments for aphasia accessible for widespread clinical use in the future, just as the simplicity of the TMS protocol used for depression has led to FDA clearance and more widespread use than would be possible with a more complicated approach [55].

From a neurobiological perspective, the hypothesis that inhibiting the intact RH might improve aphasia outcomes is derived primarily from the motor literature, based on the so called “theory of interhemispheric inhibition.” In the motor system, transcallosal inhibitory connections between the primary motor cortices of the two hemispheres may help to coordinate bimanual movement [56]. After a stroke involving the motor cortex, the interhemispheric inhibitory balance is disrupted and the intact motor cortex in the hemisphere opposite the stroke inhibits the injured side [57, 58], contributing to deficits [59, 60]. Inhibition of the intact motor cortex using rTMS or tDCS can increase cortical excitability on the lesioned side and improve clinical motor function [61–64], suggesting that this transcallosal inhibitory imbalance is clinically important and modifiable. The theory of interhemispheric inhibition thus is well supported in the human motor system.

The design of Naeser and colleagues’ original rTMS study was based on the hypothesis that the principles of interhemispheric inhibition apply to language systems as well. The specific hypothesis was that LH damage releases transcallosal inhibition on the RH homolog to Broca’s area, allowing it to suppress surviving tissue around Broca’s area in the LH and hence impede aphasia recovery. A logical treatment intervention to remedy this right-to-left suppression would be to inhibit the RH homolog to Broca’s area, restoring proper interhemispheric balance and allowing Broca’s area and surrounding tissue to play a larger role in language. Since the success of the original rTMS protocol for aphasia, the theory of interhemispheric inhibition has served as the framework guiding many similar protocols using rTMS and tDCS to inhibit the RH in aphasia [65–67]. The success of these protocols has been taken as evidence that interhemispheric inhibition plays a key role in language reorganization in aphasia [66, 42].

Transcallosal fiber pathways between language areas do exist, providing the anatomical basis for interhemispheric inhibition [68] although it remains unclear whether or not these connections are predominantly inhibitory. Two rTMS studies in the subacute phase of stroke recovery have included PET activity during a verb generation task as an outcome measure in order to assess the biological effects of treatment [67, 69]. In these cases, laterality indices of activity demonstrated a leftward shift after rTMS that did not occur with sham rTMS. These findings could support the theory of interhemispheric inhibition, but a number of caveats limit their interpretation. For example, PET scans restrict the ability to control for task difficulty (see below for more discussion on this issue). Further, the use of laterality indices of activity precludes localization of activation changes, making it difficult to know whether changes occurred primarily in the RH, LH, or both, as would be predicted by the theory of interhemispheric inhibition. As such, it remains unclear whether the findings of these studies truly support the theory of interhemispheric inhibition. Moreover, studies have not assessed neurobiological outcome measures in the chronic phase of aphasia recovery, from which most clinical evidence for efficacy of rTMS and tDCS derives. Thus, despite the success of rTMS and tDCS studies aiming to inhibit the RH to improve aphasia, the existence and importance of interhemispheric inhibitory interactions in language networks still remains largely theoretical, based mainly on extrapolation from the motor system.

Apart from rTMS and tDCS studies, several recent functional imaging studies do provide some support for the notion that RH activity in aphasia is maladaptive, although not for interhemispheric inhibition specifically. These studies have noted increased RH activity during incorrect naming responses or inverse relationships between activity and performance across groups, suggesting that some RH areas recruited in aphasia might be ineffective, inefficient, or maladaptive [70–74]. In some longitudinal functional imaging studies, RH recruitment has peaked early in recovery or immediately after training and has diminished over time in association with clinical improvements, suggesting that “turning off” the RH might improve long-term recovery [15, 75–77].

Contradictory Evidence Regarding the Role of the RH in Aphasia Recovery

In contrast to the results of rTMS and tDCS studies and the recent functional imaging studies above, multiple older lines of evidence suggest instead that the RH compensates for LH damage and supports recovery from aphasia. This evidence begins with Barlow’s 1877 case of a boy who became aphasic after a small stroke to the left posterior IFG, then recovered, but worsened again after a small symmetrical stroke in the RH [78]. More recently, similar adult cases have been reported in which a first LH stroke caused aphasia and after partial recovery a second RH stroke worsened language performance [79, 80]. Several other lines of evidence have also suggested RH compensation in aphasia: a relationship between poor aphasia outcomes from LH stroke and “clinically silent” RH strokes [81], worsening of language performance in aphasic patients after right carotid anesthesia [82], and left visual field and left ear advantages in people with aphasia [83–85]. These sources lack the spatial resolution to implicate particular parts of the RH in aphasia recovery but suggest that overall the RH contributes to language ability after damage to the native LH language network.

More recently, some functional imaging and electrophysiological studies of aphasia have reported activity in RH areas that mirror typical LH language areas, corresponding to a clinical response to treatment [86, 87]. Based on evidence for RH compensation in aphasia, some successful forms of speech-language therapy have been designed to engage the RH in language processes, for example Melodic Intonation Therapy. These methods have been shown to induce remodeling of the RH as predicted [88, 89], and in a pilot study, applying anodal tDCS to enhance right IFG activity during treatment improved fluency of speech output compared to sham tDCS [90].

Understanding the apparent contradictions in the literature on the role of the RH in aphasia recovery will be critical to optimizing brain stimulation treatments for aphasia. Multiple factors may contribute to the conflict in the field: methodological limitations in brain stimulation and neuroimaging studies, differences in the RH role in recovery of specific language functions and between the roles of specific RH brain areas, and individual differences in the brain basis of aphasia recovery. In terms of methodological limitations, the first concerns the poor understanding of the long-term neurobiological effects of RH inhibitory brain stimulation protocols. As noted above, it remains unclear that brain stimulation protocols known to induce short-term inhibition at the site of stimulation induce long-term inhibition lasting weeks to months in association with the behavioral benefits on language. The theory of interhemispheric inhibition specifically suggests that such long-term RH inhibition at the site of stimulation should be accompanied by enhanced LH engagement in brain areas directly opposite the stimulation site; this too remains unproven. In a case study on a chronic nonfluent aphasic patient who received a 2-week course of low frequency rTMS to the right pars triangularis and then unfortunately suffered a second stroke, this time affecting the RH, we found no evidence to support the theory of interhemispheric inhibition in either the fMRI activity or the behavioral effects of the second stroke [79]. It thus remains possible that a different unpredicted neurobiological effect accounts for the long-term effects of brain stimulation, at least in chronic aphasia. For instance, altering the inhibitory-excitatory balance within a reorganized bihemispheric language network might induce a renewed period of plasticity in the chronic phase, allowing for further optimization of network efficiency in both hemispheres [91]. Alternatively, multiple sessions of RH inhibition might induce a longer-term excitatory overshoot after treatment ends, such that enhanced RH compensation accounts for behavioral improvements seen after RH inhibitory brain stimulation.

Limitations of task-related brain activity, the primary metric used to quantify RH involvement in language, may also account for some inconsistent findings in the literature on aphasia recovery. In particular, the impact of performance and effort on task-related activity complicates the interpretation of many functional imaging results. Increasing effort on language tasks produces more activity in both hemispheres in both controls and people with aphasia [92, 93]. Thus, while inverse correlations between activity and performance may suggest maladaptive activity, an alternate interpretation is that individuals with large strokes and more severe aphasia must exert more effort to perform the task and thus engage the RH to a greater degree. In this context, when in-scanner performance is not controlled in longitudinal imaging studies, a decrease in RH activity may be a consequence of improved aphasia, not a cause.

Aside from methodological limitations, some of the inconsistencies between previous studies on the RH’s role in aphasia recovery may derive from a complicated relationship between the RH and recovery. For instance, different specific parts of the RH may play different roles in aphasia recovery, such that some areas of the RH compensate for damage to the LH language networks, while others are too inefficient to be effective, and others may indeed interfere with optimal recovery [13•, 79]. Likewise, the RH may be able to compensate more effectively for some specific language functions compared to others [94]. Finally, the specific pattern of language system reorganization likely differs somewhat between individuals, based on personal characteristics and features of the stroke [95, 96], and also may change substantially over time after stroke [97•].

Conclusions

Overall, we are still early in the development of noninvasive brain stimulation treatments for aphasia, but it appears that protocols aimed at suppressing the RH, particularly the pars triangularis of the right IFG, do improve specific language abilities. However, many questions remain unanswered. Too few studies have directly compared different brain stimulation protocols to determine whether alternate approaches might produce larger effects or whether personalizing stimulation protocols based on individual differences will maximize benefits. There is also inadequate evidence to determine the long-term biological changes caused by rTMS and tDCS treatments for aphasia, and so the mechanism of effect is poorly understood. For this reason, it seems inappropriate to use behavioral effects occurring days, weeks, or months after brain stimulation treatments to support particular neurobiological theories of aphasia recovery, most notably the theory of interhemispheric inhibition. Going forward, including neurobiological outcome measures in brain stimulation studies will produce a great deal more progress toward understanding the biological basis of aphasia recovery and optimizing brain stimulation treatments. In particular, because of the thorny dependence of task-related functional activity on effort and task performance, it will be particularly useful to examine more stable brain measures that do not depend on effort, such as gray and white matter morphology and resting functional connectivity. Finally, although some clinicians are now providing brain stimulation treatments for aphasia on an out-of-pocket fee-for-service basis, this is not yet clearly justified by the available data. Patients or families interested in brain stimulation treatments should instead be referred to clinicaltrials.gov or other resources to identify ongoing studies, and encouraged to participate, given the safety of these techniques and the promise of future benefits for people with aphasia.