Effect of Anodal tDCS on Articulatory Accuracy, Word Production, and Syllable Repetition in Subjects with Aphasia: A Crossover, Double-Blinded, Sham-Controlled Trial

Twelve right-handed, native Portuguese speakers with chronic post-stroke aphasia (6 male and 6 female, aged 57.6 ± 12.7 years) were included in the study. These patients were recruited in treatment centers of a capital at the northeast region of Brazil. Single stroke occurred in 75% of the sample. All subjects presented right hemiparesis. Time after onset of stroke varied from 15 to 70 months (37 ± 16.8 months). All individuals, except one, presented arterial hypertension (91.6%). Before the stroke, 41.6% usually consumed alcohol (3 to 5 times per week) and 33.3% were smokers. None presented cardiac disease. Table 1 shows the characteristics of all participants.

Participants had never received tDCS. Subjects were first evaluated by a neurologist to define the stroke diagnosis using the parameters defined by the Trial of Org 10172 in Acute Stroke Treatment (TOAST) [23]. All had a lesion that occurred at least a year prior to the study and all had been receiving traditional speech-language therapy once a week for at least 6 months. The study excluded patients with moderate to severe cognitive deficit (according to DSM-IV criteria), receptive language deficit, auditory loss, cardiac disease, pacemaker use, seizures within the past 6 months, and orofacial malformation or malocclusion that could compromise speech praxia. Additional exclusion criteria applied after the beginning of the study included changes in pharmacological and non-pharmacological treatments, infection, head trauma, another stroke episode, and distress during tDCS sessions.

In order to classify the type of aphasia, a speech-language pathologist not involved in the experiment assessed all participants. The therapist used a Brazilian ecological test for classification of aphasia, the “Teste de Reabilitação das Afasias: Rio de Janeiro” [24]. This test includes picture naming, verbal and non-verbal apraxia, performance on repetition of words and phrases, reading, and spontaneous speech and writing ability. After receiving the aphasia classification, participants with speech articulatory deficits were included. As previous studies have presented results for heterogeneous samples, and due to the difficulty of recruitment for tDCS studies, we decide to include all types of expressive aphasia. Subjects received medical approval for participation and gave their written consent. All procedures performed in this study were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments. The study was approved by the institutional review board (IRB) of Hospital Santo Antonio, Salvador-Bahia, Brazil (#22/07).

This was a crossover, double-blinded, sham-controlled offline study. Figure 1 depicts the experimental design. Participants were randomly assigned to two groups with different orders of stimulation (sham followed by active tDCS or vice versa). Follow-up assessment was performed at the end of each intervention period. A qualitative follow-up occurred at 1 week and at 4 months after the end of the intervention.

tDCS was delivered by a battery-driven, constant-current stimulator through a pair of saline-soaked sponge electrodes (7 cm x 5 cm) which were kept in place by elastic bands. We used a tDCS device developed at Mackenzie University–Sao Paulo-SP, Brazil [25]. As the majority of previous studies have shown positive effects of anodal tDCS (A-tDCS), we chose this polarity [6, 12‐15, 18, 20, 26]. The anodal electrode was placed on the left side of patient’s head (Broca’s area) following the 10/20 EEG system [27]. Broca’s area was determined by the intersection of the lines formed by points Cz-F7 and Fz-T3. These points, previously used in other studies [22, 25] were identified using a measuring tape. The cathodic electrode was placed at the right supraorbital area.

The stimulation lasted 20 min using a current of 1 mA over the course of 10 days (5 days during a week, weekend interval, and another 5 days) [12, 28]. The duration and intensity level were chosen based on previous studies that assessed safety of tDCS in humans [18, 29]. At the onset of both active and sham stimulation, the current was increased in a ramp-like fashion eliciting a transient tingling sensation on the scalp that faded over 5 s [28, 30]. The researchers who conducted the assessments were blind to the study design. Since the goal was solely to investigate the effect of A-tDCS, an offline design (task and tDCS are not performed concurrently) was adopted [31].

After the washout period, patients treated with anodal stimulation received sham treatment and vice versa. Given the lack of a standard protocol, we defined the washout period based on the findings of previous studies. [12, 18]. This interval varied from 24 h to 3 weeks [13, 17]. To avoid carryover effects, we adopted a conservative washout period of 3 weeks.

The trial was not registered because at the time the study was conducted, registration of clinical trials in Brazil was still not a consolidated practice. Reasonable effort has been made since to apply for retrospective trial registration, but has not been possible.

To assess the effects of A-tDCS, subjects in the two groups (anodal and sham) were evaluated before and after the 10-day intervention using the study naming and repetition tasks. Pre-assessment occurred at the same day participants received the first tDCS session, for both sham and anodal groups. Post-assessment occurred at the last day of intervention, 15–20 min after the session.

The picture-naming task of the Snodgrass test was used [32, 33]. The same 20 pictures were used for both sham and anodal groups, before and after the 10-day intervention period. These pictures were uploaded on Chronos software and were selected based on characteristics such as word length, visual complexity, and phonologic information (e.g. syllable articulatory complexity). Picture-naming tasks are widely used in tDCS studies to test individuals without [34] and with [10, 12] aphasia. A speech sample was recorded using the software Sony Sound Forge Pro 9.0. We collected these samples using a Stile VG Felitron voice tube headset, connected to a notebook with a Windows 2007 operational system.

Two measures related to the naming task were created: (1) number of words produced correctly, and (2) articulatory accuracy. Two blinded speech-language pathologists (SLPs) who received training in order to standardize the assessment and increase inter-rater reliability conducted the assessment of speech production and the overall speech praxis. Using the list of 20 words supplied in the task, they compared each word’s production before and after the 10-day stimulation period and rated the subsequent production as being “the same”, “better accuracy”, and “less accuracy”. The SLPs recorded the total number of words that presented “better accuracy” for each study participant (articulatory accuracy). The criteria used to judge accuracy was articulatory precision to the extent that the evaluator was able to understand the target word. Although articulatory accuracy is a subjective parameter, it allows judgment based on real-world experiences of subjects with aphasia. The measure translates how well a listener understands target words. This has a pragmatic and social value because we are observing language from a functional and interactional perspective [35, 36].

The non-parametric Mann–Whitney test for paired samples was used to compare pre- and post-stimulation scores on the naming test (number of words produced correctly and articulatory accuracy) and the syllable repetition task (number of syllables produced correctly). To calculate these scores, the difference between pre- and post-intervention for each subject in the intervention group was computed. This difference was then divided by the number of correct productions pre-intervention and averaged for each group. The Wilcoxon paired test was used to compare the average number of words produced with better accuracy at each treatment (sham and active A-tDCS). Cohen’s kappa was used to assess inter-rater agreement for the two referees. The same index was computed for inter-rater agreement at the qualitative assessment. For the qualitative assessment, we considered five categories, according to kappa values: no agreement (< 0), slight agreement (0–0.19), fair agreement (0.2–0.39), moderate agreement (0.4–0.59), substantial agreement (0.6–0.79), and almost perfect agreement (0.80–1) [37]. The software R version 2.11.1 [38] was used to perform the Mann–Whitney matched pairs test, and SPSS was used for all other analyses. Differences that achieved a p value ≤ 0.05 were considered statistically significant. We used the Bonferroni method to correct for multiple comparisons.

As shown in Table 2, non-parametric tests suggest that there was no statistical difference in the number of words (p = 0.1540) and syllables produced correctly (p = 0.7896) after the treatment.

The Wilcoxon paired test was used to compare the average number of words produced with better accuracy at each treatment (sham and active A-tDCS). The mean number of words produced with better accuracy after the intervention in the sham group was 2.17 (SD = 1.75), while it was 4.42 (SD = 3.40) for the A-tDCS group (Fig. 2). The difference was statistically significant, after correcting for multiple comparisons (V = 43, p = 0.01693). The value V = 43 corresponds to the sum of ranks assigned to the differences with positive sign. Cohen’s kappa test was used to compare the agreement rate between the two referees that judged the data. There was no statistical difference between them (anodal: p = 0.5941, placebo: p = 0.7918).

Of the 12 participants, five showed qualitative improvement at follow-up assessments. One week after the end of all intervention, at the first follow-up assessment (evaluation 4 in Fig. 1), four subjects had improved repetition ability. The assessments yield to a change in their aphasia’s classification. They initially had Broca’s aphasia, but recovered into motor transcortical aphasia at follow-up 1 due to significant improvement of word repetition. One subject entirely suppressed the anomia deficit at follow-up 1 and was no longer classified as having aphasia. At a second follow-up assessment 4 months later (evaluation 5 in Fig. 1), these five participants presented the same results (Table 3). The same five subjects described above showed improvement on social communication as perceived by themselves, their families, and therapists (Fig. 3). Patient 6, for instance, reported great improvement in spontaneous speech and expressed desire to have more A-tDCS sessions: “quero fazer mais vezes” (“I want to have more”). The same baseline assessments were used in both follow-up sessions.

According to self-perception, eight of 12 participants experienced improvement after A-tDCS or sham. Six family members perceived improvement in the speech of their related participant after A-tDCS or sham. For SLPs, improvement was reported for five participants after A-tDCS or sham and for three participants after A-tDCS only. Agreement among these three groups of raters was low (kappa = 0.105) (Fig. 3). None of the participants, family members, and SLPs reported speech deterioration after either sham or A-tDCS sessions.

It was observed that articulatory accuracy improved after A-tDCS. The data suggest that ten sessions of 1 mA A-tDCS over Broca’s area can improve motor patterns related to language, as measured by the accuracy assessment.

The results of the potential improvement in articulatory accuracy were evaluated since many studies already suggest the effectiveness of A-tDCS in the improvement of naming ability [6, 12‐16, 18, 20]. To date, we only found three studies assessing articulatory accuracy after tDCS [22, 39, 40]. These three experiments corroborate our findings on articulatory accuracy improvement influenced by tDCS over Broca’s area.

First, Marangolo et al. demonstrated that A-tDCS over Broca’s area improved the articulatory ability of subjects with both aphasia and apraxia [39]. They hypothesized that A-tDCS elicits a prolonged increment in cortical plasticity, probably because of changes in synaptic connections of N-methyl-D-aspartate (NMDA) receptors. Also, stimulation over the left inferior frontal gyrus (IFG) determines long-term effects in the recovery of speech apraxia in patients with chronic aphasia. Second, Marangolo et al. evaluated the number of correct syllables and words produced by subjects with brain damage after bilateral tDCS–A-tDCS over Broca’s area and C-tDCS on a homologue area in the right hemisphere [40]. They concluded that bilateral stimulation restored interhemispheric balance, promoting the best outcomes for language recovery. The authors suggested that the improvement in articulatory accuracy was based on the notion that up-regulating the excitability of integral portions of the lesioned hemisphere and down-regulating the excitability of the contralesional hemisphere would lead to the greatest recovery of language.

Finally, Fiori et al. concluded that stimulus-tDCS over Broca’s area improves articulatory accuracy in healthy subjects [22]. They suggested that the frontal area could co-activate other premotor areas, leading to improvement in repetition ability and other language tasks. The connectivity of Broca’s area with other cortical regions offers the neuroanatomical substrate to explain this effect. The authors also observed a long-term effect of A-tDCS due to a hypothesized lasting depolarization of neurons in the Broca’s area.

Performance did not change in the word production and syllable repetition tasks after A-tDCS (p = 0.15 and p = 0.79, respectively) (see Table 2). Although contrary to many studies [12‐16, 20], this study is not the first to find negative results for the effect of tDCS on naming function. Polanowska et al. applied A-tDCS in 37 subjects with aphasia and did not observe improvement of naming function after tDCS [41]. The same research group in another study with 24 subjects with aphasia also used A-tDCS over Broca’s area and found weak evidence for A-tDCS-related language improvement [42]. The systematic review and meta-analysis conducted by Elsner et al. evaluated tDCS against sham tDCS in 12 studies and found no evidence to support the efficacy of intervention as a therapeutic tool for aphasia [6].

Participants’ responses differ due to variations in the mechanism of brain plasticity involved in language recovery, which helps in interpreting negative results. Broca’s area, located at the frontal lobe and involved in naming processing, was stimulated. However, the temporal area is also linked to naming capability [43]. Location and size of the brain injury are other factors that could explain the lack of effect in our sample [44]. Electrode placement was the same for all individuals (Broca’s area), which did not necessarily match the location and extension of each patient’s lesion. Moreover, Fridriksson et al. recently found an interaction between a specific genotype of the BDNF gene and the tDCS polarity for treatment-related naming improvement. According to the authors, subjects that carry the genotype val/val BDNF are more prone to benefit from A-tDCS [45].

Placement of the active electrode is an additional source of mixed results. For instance, Shah-Basak et al., studying seven adults with chronic aphasia, observed improvement on naming tests for three subjects that received anodal stimulation on the left hemisphere, while three others responded positively to cathodal stimulation on the left hemisphere and one subject improved after cathodal stimulation on the right hemisphere [44]. The authors defend individualized use of tDCS electrode montages for stimulation because of individual differences.

Recent evidence suggests that A-tDCS when performed simultaneously with behavioral therapy can improve aphasia-related outcomes [46, 47]. The offline methodology (stimulation without any other intervention simultaneously) adopted in this study may justify the absence of significant results in word production and syllable repetition.

Several studies suggest that the mechanisms of brain reorganization in patients with lesions caused by a stroke remain largely unknown [48‐50], limiting the generalization of tDCS protocols, as a recent meta-analysis showed [6]. More studies are necessary to explore different tDCS electrode montages (size, position, current type) and current intensity (1–2 mA), as well as the use of neuroimaging techniques and transcranial magnetic stimulation (TMS) to help choose the best parameters to treat patients with aphasia. Some authors have begun using functional magnetic resonance imaging (fMRI) concomitantly with tDCS to define the best area to be stimulated [43].

Studies looking at the effect of tDCS in the language of subjects with aphasia use several secondary outcomes, complicating a holistic data analysis. Choosing parameters to represent functional improvement is a great challenge. The most common outcomes (e.g., naming accuracy and naming speed) measure isolated word production. Motor planning and organizing during the dialogic context are complex tasks that cannot be fully represented by the usual surrogate outcomes measured in single word production. Interpretation of these outcomes in combination with subjective perception of patients and people surrounding them may give a better understanding of clinical language improvement.

Qualitative improvement was observed even in patients with longer disease onset (patients 3, 4, and 9), when not much improvement would normally occur. A recent study including patients with post-stroke chronic aphasia also found that a more severe baseline language profile was associated with larger improvements in one aspect of speech production (fluency) after tDCS [51].

Language is a complex process involving different neural circuits and includes abilities beyond verbal production and comprehension. Little has been investigated about tDCS effects on suprasegmental aspects of language, such as intonation and pitch during speech production [6]. Future tDCS trials should consider investigating these subtle aspects of language both quantitatively and qualitatively.

The current study has limitations that need attention. We had a small sample size, which may have caused a type II error, preventing our ability to find differences between active and sham stimulation for the word production and syllable repetition tasks. A post hoc power analysis, considering the mean and the standard deviation of word production and syllable repetition, indicated we achieved small power for both outcomes (0.1027 for word production and 0.1025 for syllable repetition, respectively). A sample size of 95 subjects per group would be necessary to provide power of 0.80. Nonetheless, other tDCS clinical trials have included a small number of subjects as well, given the difficulties in recruiting and retaining participants with a similar design [12, 13, 17, 22]. The crossover design was used to reduce this limitation. Placement of electrodes was theoretically determined for all subjects, as already mentioned above. Ideally, this choice should be determined by neuroimaging assessment to better suit each patient’s needs. The lack of effect in this study may have occurred due to placement of the anodal electrode over mostly lesioned brain tissue or necrotic cavities.

We planned to analyze whether there was a carryover effect by comparing the baseline values before active and sham interventions for the group that received active intervention first in the crossover design. However, because of the randomization process, the group that received tDCS treatment first had only two subjects, while the group that received sham treatment first had 10 subjects. This large difference between group sizes hindered our ability to compare the groups regarding the carryover effect of tDCS. Finally, the absence of neuroimaging assessment was an important limitation in this study. This information would have allowed for exploratory analysis regarding location and size of the lesion. We had planned to request these results, but participants came from different healthcare clinics, and most did not have copies of these exams.