All participants were recruited from the Dementia Center, Shin-Kong Wu Ho-Su.

Memorial Hospital, Taiwan. Sociodemographic variables such as age, sex, body weight, height, and body mass index, years of education and handedness were obtained from interviews and medical records. The inclusion criteria were as follows: (1) subjects aged 65 years and over; (2) a diagnosis of MCI according to Petersen’s criteria (Petersen, 2004); (3) a global rating of 0.5 on the Clinical Dementia Rating (CDR) scale; and (4) able to walk more than 10 m without walking aids. The exclusion criteria included (1) dementia; (2) brain tumors; (3) previous cerebral infarction or hemorrhage; (4) other known neurodegenerative or neuropsychiatric conditions; (5) the presence of an unstable orthopedic disease interfering with participation in the study; and (6) an education level of less than 6 years (elementary school).

This study was a double-blinded (assessor) randomized controlled trial. The participants were randomly assigned to either the anodal tDCS or sham tDCS group via a sealed envelope. The subjects participated in 40 min of TC training in each session. During the first 20 min of training, they simultaneously received either real or sham tDCS. Both groups attended sessions three times a week for 12 weeks for a total of 36 sessions. An experienced physical therapist conducted individual TC training, and all participants received the intervention individually. The assessor, who was always blinded to the group assignments, measured the outcomes at baseline and between 24 and 48 h (anodal tDCS group: 29.8 ± 8.3 h; sham group: 30.9 ± 8.1 h) after completing the 36 sessions. The gait assessment and cognitive assessment were independently performed with gait performance first followed by the cognitive assessments. The study was approved by the ethics committee of Shin-Kong Wu Ho-Su Memorial Hospital, and all subjects gave their informed consent prior to the beginning of the study.

A battery-operated constant direct-current Stimulator Plus (NeuroConn, Ilmenau, Germany) was used in the present study. The current of the stimulation was administered at 2 mA via a saline-soaked pair of sponge electrodes with dimensions of 5 × 7 cm (35 cm²) to optimize the stimulation at the left DLPFC. The anode electrode was centered over F3 based on the 10–20 international system, whereas the cathode electrode was centered at the contralateral (right) supraorbital area, i.e., Fp4. For the anodal tDCS group, 2 mA of direct current was delivered for 20 min with a 10-s ramp-up period. For the sham condition, while the stimulation parameters and electrode montage remained the same as those for the anodal condition, the current was delivered only for the initial 30 s and then ramped down to 0 mA. This simulated the tingling sensation of active stimulation that was indistinguishable from anodal tDCS but with presumably negligible effects on the brain. All participants were tDCS-naïve subjects who, in both the tDCS and sham groups, were led to believe that they were receiving real tDCS. For the purpose of simultaneously delivering tDCS during TC, all subjects wore a semitransparent backpack that carried the tDCS device during practice. In all sessions, the impedance of all electrodes was monitored throughout the entire period of stimulation to maintain values under 5 kΩ.

The participants were taught the 24 forms of Yang Style TC for 12 weeks. Yang Style TC is a popular and widely practiced form of TC characterized by gentle, broad, slow, and open movements, with an emphasis on breath and motor awareness. Forty minutes of TC included warming up, learning new forms, practicing learned forms, and cooling down. TC was taught and led by a certified coach, and the participants learned and practiced 2–3 new forms of the 24-form TC weekly. The BORG Perceived Exertion Scale after the TC session to control the exercise intensity was set at a score of 13–14 (somewhat hard) (Stuckenschneider et al., 2020). Both the anodal tDCS group (a-tDCS+TC) and the sham group (sham+TC) received online stimulation during the first 20 min of TC training by the methods described above.

Gait parameters were measured by the Gait Up system (Gait Up, Lausanne, Switzerland), which is a wearable device with good validity and reliability (Mariani et al., 2010; Dadashi et al., 2013). Two wireless inertial sensors with the triaxial Gait Up accelerometers were placed on the dorsal side of the shoes. Three conditions were designed for assessing gait performance: (1) the single-task condition: walking at the subject’s preferred speed; (2) the cognitive dual-task condition: walking while executing a serial 3′s subtraction task that starts from a randomized 3-digit number (e.g., 100, 97, 94…) and (3) the motor dual-task condition: walking while carrying a tray with glasses of water. The participants were asked to walk three trials in each condition, while spatiotemporal parameters recorded during each trial, namely, speed (cm/s), stride length (cm), and cadence (step/min), were averaged from the three trials for the statistical analysis. Dual-task interference was quantified by the dual task cost, which was calculated with the following equation: [dual task cost speed (%) = dual-task walking speed – single-task walking speed]/single-task walking speed * 100% (Plummer and Eskes, 2015).

Sociodemographic, neuropsychological and gait data analyses were performed using SPSS 20.0 software (SPSS Inc., Chicago, IL, United States). Descriptive statistics were generated for all variables, and variable distributions are expressed as the means ± standard deviations or as numbers. Intergroup differences in baseline characteristics were analyzed using independent t-tests or chi-square tests. Two-way ANOVAs with repeated measures were used to determine the effects of TC on cognitive function assessed by neuropsychological tasks and gait parameters assessed by Gait Up. We assessed the interaction between Intervention Group (a-tDCS + TC, sham + TC) and Assessment Time (baseline, post-intervention). Significance was set at p < 0.05. For the similar tasks involving executive function (TOL, TMT, SCWT), we performed Bonferroni corrections due to multiple comparisons. The significance was set at p < 0.017.

The results of the single- and dual-task performance are shown in Table 2 and Figure 2. In regard to cognitive dual-task performance, we observed significant interaction effects across both cadence [p = 0.006*, F_{(1, 18)} = 9.743, η² = 0.351] and speed outcomes [dual task cost of speed, p = 0.041*, F_{(1, 18)} = 4.837, η² = 0.212]. In the a-tDCS+TC group, post hoc tests showed significant improvements after training in speed (p = 0.048), cadence (0.003) and dual task cost of speed (0.002) of the cognitive dual-task gait.

The cognitive task performance before and after the intervention is illustrated in Table 3. There was some evidence of an interaction effect in the TMT-B [p = 0.018, F_{(1, 18)} = 6.706, η² = 0.271], however this was not significant when adjusting for multiple comparisons (α = 0.017).

Only a few studies have examined the enhancing effects of tDCS on dual-task performance in healthy adults (Zhou et al., 2014; Wrightson et al., 2015; Ljubisavljevic et al., 2019; Schneider et al., 2021) and patients with Parkinson’s disease (Schabrun et al., 2016). Most of these studies have explored the immediate effects after stimulation (offline) (Zhou et al., 2014; Wrightson et al., 2015; Manor et al., 2016, 2018; Schneider et al., 2021), with one study evaluating the effects during the application of tDCS (online) (Ljubisavljevic et al., 2019). Unlike these studies, which focused on the immediate effects of a single tDCS session, our study applied multisession tDCS and demonstrated the effects of coupling tDCS and TC 24–48 h beyond the entire period of intervention. Previous studies have shown that a combination of tDCS and aerobic exercise might provide acute synergic effects on cognition (Steinberg et al., 2018), our study expanded this type of investigation to gait and their long-lasting effects. Nevertheless, our study design cannot answer the question of whether multisession a-tDCS coupled with TC would exert additional effects compared to a-tDCS alone. According to another study, tDCS targeting the left DLPFC immediately mitigated dual-task costs to walking (Zhou et al., 2021). Although our experimental design did not include a “tDCS only” group, we may speculate that multisession a-tDCS alone could exert a similar beneficial effect on dual-task gait performance.

The strength of our study is that our subjects were able to receive tDCS and TC concurrently. Recent studies have shown that tDCS tends to modulate the same brain networks that are already activated by a task or brain event, suggesting that the impact of tDCS on cognitive or motor tasks is state-dependent (Stagg et al., 2011; Martin et al., 2014). Thus, coupling tDCS and TC together could potentially enhance the same TC-related brain networks that are targeted by tDCS and create a synergistic effect in cortical excitability and neuroplasticity (Martin et al., 2014). Another strength of the current study is that the augmenting effects of combining TC and tDCS were transferred to other unlearned tasks involving similar cognitive domain, i.e., cognitive dual-task gait. This suggests that the beneficial effects of combining TC and tDCS could potentially be transferred and applied to other daily activities in patients with MCI and thus improve their quality of life.

The enhancing effects of TC on dual-task gait performance (gait speed and stride time variability) has been reported in previous studies (Wayne et al., 2015; Li et al., 2019), yet their training effects required as long as 6 months to achieve. In comparison, our relatively shorter period of TC practice, i.e., 3 months, might not have been sufficient to yield a significant impact on gait performance. As such, we were unable to see substantial dual-task gait improvements in the sham group. Interestingly, significant intergroup improvements in the a-tDCS+TC group may be attributed to the facilitating effect of tDCS. A recent meta-analysis showed that tDCS was able to induce considerable improvement in balance control and dual-task gait conditions in older adults in a short time period (from 1 to 16 sessions) (Guo et al., 2020). Our more extended period of intervention, i.e., 36 sessions in 3 months, may enhance the facilitating effects of tDCS and led to more robust improvements than TC alone.

We chose the left prefrontal cortex as our tDCS target because it has been implicated in the allocation of cognitive resources between two concurrently performed tasks (Collette et al., 2005), notably during dual-task gait (Doi et al., 2013; Al-Yahya et al., 2016). As such, we hypothesized that the tDCS online effect over the augmented neural circuit of the prefrontal cortex under TC practice might improve cognitive capacity and reduce the attention needed to perform the cognitive task, thereby permitting greater attention to be shifted toward performing a concurrent task (e.g., walking). The bottleneck theory of dual-task interference suggests that when two tasks are performed simultaneously, one task involving the same neural network will be postponed until the other prioritized task is completed (Ruthruff et al., 2001). Based on this theory, tDCS-related improvements in our study may have resulted from improved processing speed and shortened time delay between the cognitive and walking tasks (Redfern et al., 2002).

Walking while performing serial subtraction in older adults with MCI is a highly demanding task that requires alternating momentary processing capacity and filtering out all signals that are irrelevant to the task (Montero-Odasso et al., 2009). We speculate that without the augmenting effect of tDCS, the improvements in cognitive function in the sham group may not have been substantial enough to transfer to improvements in dual task gait. A higher intensity, duration, and challenge might be required in future TC studies.

There was some evidence of an interaction effect in the TMT-B, however, this was not significant when adjusting for multiple comparisons. The TMT-B and SCWT can be seen as similar tasks associated with cognitive flexibility and working memory (Scarpina and Tagini, 2017). Arguably, the augmenting effects of tDCS and TC may not have consistently transferred to all tasks we tested for executive function. While the mechanisms leading to selective enhancement of individual tasks engaging the same cognitive domain remain unclear, the positive aftereffects of tDCS on performance tended to be more prominent if the task closely resembled the task performed during tDCS application (Stagg et al., 2011; Martin et al., 2014). Since TC is a dual task resembling cognitive dual task gait, the latter was subject to the greatest aftereffect of tDCS. Interestingly, the TMT-B shares some similarities with dual task gait in that both tasks assess divided attention—the ability to simultaneously attend to two different stimuli and respond to the multiple demands of the surroundings. Indeed, a recent study showed that poor TMT performance was directly associated with altered dual task prioritization in elderly individuals (Hobert et al., 2011). The acute effects of tDCS on enhancing divided attention assessed via the TMT has been demonstrated in patients with Parkinson’s disease and MCI (Cruz Gonzalez et al., 2018a; Bueno et al., 2019). Our finding showing a favorable trend of tDCS relative to sham in TMT performance warrants future studies on a larger cohort of MCI subjects.

Limitations of this study include the lack of a “no intervention” group and a “tDCS only” group, which may further clarify the mechanisms underlying our results. Second, trials with larger samples and longer periods of follow-up are needed to confirm the prolonged effects of tDCS+TC on cognition and dual task gait performance. Third, the non-significant training effects of the motor dual task gait in both groups might be due to the relatively simple motor task (carrying a glass of water). Our finding supports a previous study that gait performance in patients with MCI was more susceptible to cognitive than motor secondary task (Hunter et al., 2018). Future studies may involve more complicated motor tasks to assess motor complexity influences on gait performance. Fourth, although we adopted a traditional tDCS montage with the anodal electrode over the left DLPFC and the reference electrode over the contralateral supraorbital region, we were unable to demonstrate the precise distribution and dosage of electric current in specific brain regions. Future studies should incorporate brain imaging and computational models to determine the accurate location and dosage of current, as well as changes in brain activities following this novel combination of non-pharmacological interventions for MCI. Lastly, the large number of statistical tests and the small sample size may have increased Type I errors based on the multiple comparisons in gait and cognitive outcome measurements.