Effects of HD-tDCS targeting prefrontal cortex, cerebellum, and motor cortex on dual-task gait in older adults

Abstract

Falls pose a significant health risk for older adults, with most incidents occurring during dual-task situations—when walking is combined with a secondary cognitive task. Reducing cognitive-motor interference and attentional demands during walking is therefore a key goal in effective fall-prevention and rehabilitation programs. High-definition transcranial direct current stimulation (HD-tDCS) is an emerging neuromodulation technique that may support this goal by enhancing neural efficiency. To date, no studies have directly compared the effects of HD-tDCS targeting different brain regions—such as the left dorsolateral prefrontal cortex (DLPFC), cerebellum, and primary motor cortex (M1)—to identify the most effective site for reducing dual-task cost (DTC) during walking in older adults. Thus, the present study aimed to compare the effects of these HD-tDCS montages on dual-task gait performance in older adults. This study utilized a randomized, double-blind, sham-controlled, counterbalanced crossover design. Each of the 19 older adult participants completed four sessions, with a minimum one-week washout period between sessions. In each session, HD-tDCS was applied for 20 min, targeting either the cerebellum, M1, left DLPFC, or a sham condition, with the order of stimulation counterbalanced across participants. Although the main effect of stimulation site was not significant (P ≥ 0.204), significant interaction effects were found for stride time variability under both single-task (P = 0.023) and dual-task (P < 0.001) conditions. Post hoc analyses showed significant reductions in stride time variability following stimulation of the left DLPFC, M1, and the cerebellum (P ≤ 0.001 for all), but not after sham stimulation (P ≥ 0.646). For the DTC of stride length variability, a significant interaction effect was also observed (P = 0.038), with reductions following M1 (P = 0.007) and cerebellar (P = 0.011) stimulation. Additionally, dual-task gait speed improved significantly after stimulation of the left DLPFC (P = 0.039) and M1 (P = 0.028). No significant effects were found for stride length, stride length variability, or the DTC of gait speed. Although M1, left DLPFC, and cerebellar stimulation showed favorable effects on gait parameters over time, no stimulation target demonstrated clear superiority. These findings suggest that HD-tDCS may modulate dual-task gait performance, but further research is required to confirm site-specific advantages.

RCT registration: On the Iranian Registry of Clinical Trials (IRCT20220718055490N1). Registration date: 1/10/2022.

Introduction

Walking is a complex functional activity that requires the coordination of multiple systems, including balance, motor control, and cognition¹. Given this complexity, any disturbances in gait can indicate impairments in one or more of these systems. Research has shown that such gait disturbances are strongly associated with an increased risk of falls, making them valuable markers for identifying individuals at higher fall risk (fallers) versus those who are not (non-fallers). Moreover, gait disturbances can serve as predictors of future recurrent falls, offering important insights for prevention strategies^2,3. It is well documented that reduced walking speed⁴, increased variability in stride time and length^2,3,5, and decreased stride length⁶ are associated with falls.

In everyday life, walking is rarely performed as a single task but often occurs alongside other tasks such as speaking, interpreting environmental cues, or planning activities⁷. Many falls happen under such dual-task conditions due to the limitations of shared attentional resources. Attention refers to the cognitive ability to selectively focus on relevant stimuli or tasks while suppressing irrelevant information⁸. When both cognitive and motor tasks are performed simultaneously, as in dual-task walking, competition for attentional resources can impair performance in one or both domains⁶.

Furthermore, evidence demonstrates a strong relationship between gait disturbances and cognitive deficits such as executive dysfunction and impaired attention in older adults. Gait parameters can even predict future cognitive decline^9,10. This interaction is known as cognitive-motor interference (CMI), which describes the competition between cognitive and motor processes during simultaneous task performance. CMI is commonly assessed using dual-task cost (DTC), which quantifies the decline in performance when a motor task (e.g., walking) is executed alongside a cognitive task, providing insights into an individual’s ability to allocate attention across tasks.

This issue is particularly relevant in daily life, where multitasking is common. In older adults, diminished cognitive resources and reduced attentional capacity heighten vulnerability to CMI and increase DTC during dual-task walking¹¹. As a result, older individuals are more susceptible to falls. Spatiotemporal gait parameters under dual-task conditions have been shown to be predictive of fall risk and frailty status¹². Therefore, reducing CMI and attentional demands during walking is a critical component of effective fall-prevention and rehabilitation programs for older adults.

Transcranial direct current stimulation (tDCS) is one of the neuromodulation techniques that is being developed as a promising adjunctive tool to facilitate rehabilitation, but the optimization and effectiveness of this treatment have not been completely clarified. There is evidence that anodal tDCS usually increases cortical excitability, while cathodal tDCS decreases it¹³. A meta-analysis has reported that in older adults, tDCS compared to sham significantly improves the Timed Up and Go test (TUG), the Berg Balance Scale (BBS), and standing postural sway (e.g., sway area) and gait (e.g., walking speed) with a large effect size¹⁴. However, in general, few studies have investigated the effectiveness of tDCS on balance in dual-task conditions^15–18. Among these studies, some have examined balance in static conditions^15–18, and only two studies have evaluated it in walking conditions^15,16. As previously mentioned, assessment of spatiotemporal parameters of gait in dual-task conditions should be an integral part of every comprehensive geriatric assessment because these parameters allow early detection of subtle gait deficits which under the single-task traditional gait assessments (TUG, 10-m walking) may remain undetected¹⁹.

The location of stimulation plays a pivotal role in determining the effectiveness of tDCS treatment¹⁶. Identifying the optimal brain region for stimulation is crucial, particularly when aiming to reduce DTC in older adults. A recent study explored the effects of stimulating the left dorsolateral prefrontal cortex (DLPFC) and primary sensorimotor cortex (SM1), both individually and simultaneously, on DTC during walking. The results demonstrated that tDCS targeting both the left DLPFC and SM1, or the DLPFC alone, significantly reduced DTC in walking compared to stimulating SM1 alone or using sham conditions¹⁶. This highlights the importance of selecting appropriate tDCS targets to maximize treatment efficacy. However, more research is needed to understand how different brain regions respond to tDCS, particularly in relation to DTC in older adults. Theoretically, regions such as the DLPFC, cerebellum, and primary motor cortex (M1) are critical for balance and motor control, but it remains uncertain which area is most effective for reducing DTC.

Research has shown that the prefrontal cortex, particularly the DLPFC, is crucial for managing cognitive and motor tasks under dual-task conditions. Notably, the cortical surface and gray matter in the DLPFC decline with age^20,21. Several studies have demonstrated that a single session of anodal tDCS over the left DLPFC can enhance cognitive functions such as working memory²², processing speed, and executive function²³, as well as improve gait performance, specifically DTC in gait speed^15,16, in older adults.

The cerebellum also plays a vital role, not only in motor planning, coordination, and execution but also in cognition^24,25. It integrates inputs from vestibular, somatosensory, visual, and auditory systems, controlling muscles involved in posture²⁶. Evidence shows that there are connections between the cerebellum and various brain regions such as the motor cortex, prefrontal cortex, and posterior parietal cortex which explain the extensive role of the cerebellum²⁴. Studies show age-related changes in cerebellar morphology, linking it to declines in both motor and cognitive function²⁷. Anodal tDCS targeting the cerebellum has been found to improve motor learning, processing speed, and postural control in older adults. In this regard, some studies have demonstrated that anodal cerebellar tDCS can improve motor learning²⁸, processing speed²⁹, and postural control¹⁸ in older adults.

Similarly, the M1 is crucial for voluntary movement execution and is involved in higher cognitive functions like attention and motor learning³⁰. A meta-analysis has shown that with advancing age, the excitability of the motor cortex is reduced. As a result, the strength of corticomotor projections to the muscles also declines³¹. In this regard, some studies have revealed that anodal M1 tDCS can improve dynamic balance³², lower-extremity somatosensory function, and mobility³³ as well as processing speed and executive function²³ in older adults.

While these studies suggest the potential advantages of tDCS in different contexts, research on its effectiveness in reducing DTC in older adults remains limited. Moreover, it is still uncertain which brain region plays a more significant role in decreasing DTC and enhancing spatiotemporal gait parameters in both single-task and dual-task conditions.

One of the methodological strengths of the current study is the use of high-definition tDCS (HD-tDCS). Because conventional tDCS uses larger electrodes compared to HD-tDCS, its electric fields are more widespread across the cortical surface, making it challenging to isolate and interpret the functions of specific brain regions. In contrast, HD-tDCS delivers more targeted stimulation, focusing precisely on the intended area for enhanced accuracy³⁴. Finite element method (FEM) models show that HD-tDCS creates a distinct electric field strength and distribution compared to conventional tDCS^35,36. Specifically, the electrode arrangement in HD-tDCS reduces the uncontrolled diffusion of the electric field, enhancing the spatial accuracy of stimulation to specific cortical regions³⁷. Furthermore, physiological studies have demonstrated that HD-tDCS offers greater precision in modulating neurophysiological components than conventional tDCS^34,38. Specifically, HD-tDCS applied to the motor cortex has been shown to result in longer-lasting after-effects compared to conventional tDCS³⁸.

To the best of our knowledge, there was no study investigating the effect of HD-tDCS optimized to facilitate the excitability of the left DLPFC, the cerebellum, and the M1 on DTC to walking in older adults. Thus, the current study aimed to compare the effects of different HD-tDCS montages on dual-task gait performance in older adults. Although all the selected brain regions are involved in both motor and cognitive functions, previous studies have highlighted the cerebellum’s crucial role in motor execution and cognition^24,25. Additionally, it integrates inputs from the vestibular, somatosensory, visual, and auditory systems to control posture-related muscles²⁶, and it has connections with various brain regions such as the motor cortex, prefrontal cortex, and posterior parietal cortex, which further underscores its extensive role²⁴. Based on this, we hypothesized that the cerebellum would play a more prominent role in reducing DTC and improving spatiotemporal gait parameters under both single-task and dual-task conditions.

Method

Study design

A randomized, double-blind, crossover-controlled design was employed in this study. Gait performance was assessed before and immediately after each stimulation session across four separate visits, with a minimum one-week washout period between sessions. HD-tDCS was applied for 20 min in a counterbalanced order to target one of four conditions: the cerebellum, M1, left DLPFC, or sham stimulation. This study was registered as a clinical trial in the Iranian Registry of Clinical Trials (IRCT20220718055490N1) on October 1, 2022.

The primary outcomes were the DTC values calculated from the spatiotemporal gait parameters (gait speed, stride length, stride time, stride length variability, and stride time variability). In contrast, the secondary outcomes were the raw values of these same gait parameters assessed separately under single-task and dual-task conditions.

Participants

The inclusion criteria were: (1) age 65 years and older; (2) being able to walk unaided for at least 10 m; (3) having an education level of more than 6 years (elementary school); (4) no cognitive impairment (Mini-Mental State Examination (MMSE) score ≥ 24 points); (5) not having a history of falling in the prior 6 months; (6) not having a self-reported history of the fractures, and surgery in the lower extremity; (7) not having psychological or neurological disorders; (8) no history of migraines; (9) no history of epilepsy or seizure; (10) no history of head injury resulting in a loss of consciousness; (11) no metallic implants, including surgical clips, intracranial electrodes, or a pacemaker; (12) not on prescription medication or self-medicating³⁹; participants were also naïve to tDCS.

Older adults were assessed for their balance ability using the BBS, functional mobility through the TUG, balance confidence with the Activities-specific Balance Confidence Scale (ABC), and cognitive function via the MMSE⁴⁰. These clinical characteristics are detailed in Table 1.

Table 1.

The demographical and clinical characteristics of subjects.

Female/male	15/4
Age (years)	71.22 ± 5.82
Height (cm)	161.21 ± 3.54
Weight (kg)	77.05 ± 17.03
Education Level (years)	9.95 ± 1.76
BBS (range 0–56)	50.3 ± 1.1
ABC (range 0–100)	83.55 ± 2.4
MMSE (range 0–30)	28.30 ± 0.78
TUG (s)	13.00 ± 4.15

BBS = Berg Balance Scale, ABC = Activities-specific Balance Confidence Scale, MMSE = Mini-Mental State Examination, TUG = Timed-Up & Go Test.

Of the 24 participants who met the inclusion criteria and provided written informed consent—approved by the Institutional Ethics Committee of Tehran University of Medical Sciences (IR.TUMS.FNM.REC.1400.193)—19 completed the treatment protocol (see Fig. 1). All procedures were conducted in accordance with relevant ethical guidelines and regulations. Three participants withdrew after the first session, and two others discontinued after the second session.

Fig. 1.

Flowchart of the study. A: cerebellum, B: M1, C: left DLPFC, D: sham stimulation.

Randomization and blinding

The randomization was performed using a web-based randomization service (https://statpages.info/latinsq.html) by a third party not involved in the study. Participants were assigned to the cerebellum, M1, left DLPFC, or sham stimulation in an order of ABDC, BCAD, CDBA, or DACB according to the Williams Square crossover design⁴¹. In this study, in addition to participants, the assessor collecting the outcome measures was blinded to the group allocation. At the end of the study (i.e., after the final session), participants were asked to guess which session involved sham stimulation. A binomial test was conducted to determine whether the proportion of correct guesses differed significantly from chance. The expected chance level for correct guesses was 25% (or 0.25). Only 3 out of 19 participants (16%) guessed the stimulation order correctly, and the binomial test produced a p-value of < 0.001, indicating that participants’ ability to correctly guess the session order was significantly below random chance. This result strongly suggests that the blinding was effective, as participants were largely unable to identify the correct session order.

Sample size

Since there were no studies with an identical design to ours, we based our sample size calculation on the most comparable study by Manor et al.¹⁵. In their study, the interaction between two conditions (prefrontal stimulation and sham) and time (pre- and post-intervention) for the dual-task cost of gait speed yielded an effect size of Cohen’s d = 0.61. We converted Cohen’s d to Cohen’s F using the formula f = d/2. An a priori power analysis using G*Power 3.1.9.2 software⁴², with a power of 0.95 and α = 0.05, determined that at least 19 participants per group were needed. To accommodate potential dropouts, 24 participants were recruited.

Procedure

Gait was evaluated both before and immediately after brain stimulation in four separate sessions, under both single-task and dual-task conditions. The HD-tDCS configuration was alternated in a counter-balanced way across sessions to stimulate different brain regions, including the cerebellum, M1, left DLPFC, or a sham condition, for 20 min. Participants were instructed to refrain from any intense physical or mental exertion 24 h prior to the intervention sessions and to maintain their usual daily activities throughout the study period, as they had before participating in the study. Prior to baseline assessments, all participants underwent a familiarization session with the testing procedures to minimize potential learning effects and ensure consistent performance across visits. During this session, participants were guided through each assessment task until they demonstrated comfort and proficiency with the procedures. No data were collected during the familiarization session. This approach was incorporated to reduce variability due to task novelty, enhance the reliability and validity of baseline measurements, and ensure that subsequent performance changes could be attributed to the intervention rather than practice or adaptation effects.

Gait

Gait parameters were recorded using an inertial measurement unit (IMU)-based Motion Capture System (Apex, BSN, Iran), which included a tri-axial accelerometer, gyroscope, and magnetometer (weighing 35 g, with dimensions of 60 mm × 35 mm × 15 mm) at a sampling rate of 500 Hz. Seven wireless sensors were attached with Velcro straps to the dorsal side of the feet, lateral sides of the shanks and thighs, and the dorsal side of the sacrum. Each participant completed two walking trials under two conditions: (1) a single-task condition, in which they walked at their preferred speed on a 20-meter walkway, and (2) a dual-task condition, in which they walked the same 20-meter distance while repeatedly subtracting 7 from a randomly selected 3-digit number provided before each trial. Spatiotemporal gait parameters—including speed (m/s), stride length (m), stride time (ms), stride length variability (%), stride time variability (%), and the DTC of these parameters—were measured. Stride length was defined as the distance between consecutive heel contacts of the same foot, and stride time was the time between those contacts. Variability was quantified using the coefficient of variation (CV), which indicates the degree of variability relative to the mean. In gait analysis, a higher CV suggests greater inconsistency in the parameter (e.g., more variable stride times), potentially signaling instability or inefficiency in gait. CV was calculated using the following equation⁴³:

Where:

• Standard deviation represents the variation in a gait parameter (e.g., stride length or stride time) and indicates how much an individual’s gait differs from their average performance over several strides.

• Mean refers to the average value of a gait parameter across multiple strides, determined by summing all the values and dividing by the total number of measurements⁴³.

Additionally, CMI was assessed using DTC, which quantifies the relative change in performance when a cognitive task is performed simultaneously with walking. DTC was calculated using the following formula¹⁵:

Whether a higher or lower DTC indicates improvement depends on the nature of the parameter:

• For variables where higher raw values reflect poorer performance (e.g., stride time, stride length variability, and stride time variability), more negative DTC values indicate better performance.

• For variables where higher raw values represent better performance (e.g., gait speed and stride length), more positive DTC values indicate improvement.

Therefore, in this study, a smaller absolute DTC value—irrespective of its direction—was interpreted as indicative of reduced CMI and improved dual-task performance. This interpretation applies specifically to the percentage-based DTC values, not to the raw gait measurements.

For example, a reduction in DTC from − 15% to − 5% or from + 12% to + 6% both indicate improvement, as they reflect smaller deviations from single-task performance and better task integration.

The first and last two steps were excluded from the analysis, and the average of the left and right gait parameters was calculated for statistical analysis. This study examined both spatial and temporal gait parameters. Spatial parameters, like stride length, evaluate physical aspects of walking, while temporal parameters, such as stride time, focus on timing, rhythm, and coordination. Together, they provide a comprehensive view of gait health^44,45. Furthermore, increased variability in stride length and stride time is an indicator of compromised gait stability, which is closely associated with a higher risk of falls. Researchers have suggested that gait variability may be more sensitive than other parameters, and can be used as a sensitive and clinically relevant parameter in the investigation of fall risk and the response to treatment⁴⁶. Additionally, some researchers have found that various gait parameters correlate with different cognitive activities and specific brain regions, a subject that is further examined in the discussion Sects.^47,48. In summary, the choice of different walking parameters was made because each one evaluates a unique aspect of gait, linked to different cognitive activities and specific areas of the brain. Additionally, these parameters differ in their sensitivity to detecting changes.

High-definition transcranial direct current stimulation (HD-tDCS)

HD-tDCS was delivered by a battery-driven Starstim system (Neuroelectrics, Barcelona, Spain). A 4 × 1 Ag/AgCl ring electrode montage (one anode electrode and four cathode electrodes) was used. The size of the electrodes was 3.14 cm², which were filled with conductive electrolyte gel. A realistic head model, including the scalp, skull, cerebrospinal fluid, and brain, was used for the finite element analysis.

Direct current was delivered for 20 min with an intensity of 2 mA (current density of 0.637 mA/cm²) at the anodal electrode, and − 0.5 mA (current density of − 0.159 mA/cm²) at each cathodal electrode. To minimize discomfort, the current was gradually ramped up and down over the first and last 30 s (fade-in/fade-out).

Montage Configurations:

• For the left DLPFC, the anode was placed at F3 of the 10/20 EEG template, and the cathodes were located at Fz, C3, FP1, and F7⁴⁹.

• For M1, the anode was positioned at Cz, with cathodes at Fz, C3, C4, and Pz^49,50.

• For the cerebellum, the anode was placed at the Iz location, defined as the point on the midline positioned 10% of the inion-to-nasion distance inferior to the Oz electrode. Return electrodes were placed at Oz, O2, P8, and PO8^49,50.

Research has shown that working memory and executive function tasks are associated with activation of the posterolateral cerebellar hemispheres, predominantly localized in Lobule VII, and extending to Lobule VI⁵¹. Complementing this, computational modeling studies using the Cerebellar Lobules Optimal Stimulation (CLOS) framework by Rezaee and Dutta (2019) demonstrated that a stimulation montage with the anode at Iz and return electrodes at Oz, O2, P8, and PO8 generates maximal electric field intensities over the right posterolateral cerebellum, particularly targeting Crus I, Crus II, and Lobule VI⁵².

This stimulation strategy is further supported by physiological evidence. For instance, Zhang et al. (2020) demonstrated that this cerebellar tDCS montage, even at low intensities, can modulate the spiking precision of Purkinje cells, reinforcing its neurophysiological plausibility⁵³.

Consistent with these findings, our FEM results revealed the highest electric field intensities in the right posterolateral cerebellar hemisphere, particularly within Lobules VI and VIII, followed by Lobule VII. While the vermal regions, especially Lobules VI–VII, also received some stimulation, the field intensity there was notably lower compared to the lateral hemispheric targets (Fig. 2).

Fig. 2.

The electric field distributions for the three stimulation montages—targeting the left dorsolateral prefrontal cortex (DLPFC), primary motor cortex (M1), and cerebellum—are illustrated. These simulations were generated by the authors using SimNIBS (Simulation of Non-invasive Brain Stimulation), version 4.0.0 (https://simnibs.github.io/). For the left DLPFC montage, the anode was placed at F3 according to the 10–20 EEG system, with return electrodes positioned at Fz, C3, FP1, and F7. For the M1 montage, the anode was placed at Cz, and return electrodes were located at Fz, C3, C4, and Pz. For the cerebellar montage, the anode was placed at Iz, while Oz, O2, P8, and PO8 served as return sites. Warmer colors indicate regions of higher electric field intensity, while cooler colors represent areas with lower field strength. A sagittal view is provided for the DLPFC and M1 montages, and an axial view is shown for the cerebellar montage.

For sham stimulation, one of the electrode locations (left DLPFC, M1, or cerebellum) was randomly selected. In these cases, the current was ramped up to 2 mA over 30 s (fade-in) and then immediately discontinued, with no further stimulation administered. This brief exposure is intended to mimic the initial tingling sensation associated with active stimulation. Notably, previous studies have shown that such short-duration stimulation (≤ 30 s) does not induce meaningful or lasting changes in cortical excitability and is therefore considered a valid placebo control⁵⁴.

The FEM analysis was conducted using SimNIBS (Simulation of Non-invasive Brain Stimulation), version 4.0.0, available at https://simnibs.github.io/simnibs/, to optimize electrode placement and current parameters. SimNIBS computes electric field distribution based on the stimulation parameters using a standard head model from the Montreal Neurological Institute (MNI), which includes scalp, skull, CSF, and brain tissues. In the generated simulations, warmer colors represent areas with higher electric field components, while cooler colors indicate lower intensity. Figure 2 presents the electric field distributions for each montage.

Statistical analysis

For statistical analysis, IBM SPSS Statistics version 23.0 was used. Normality of the data distribution was assessed using the Shapiro–Wilk test, which indicated that some variables were not normally distributed. To ensure comparability across baseline measurements and to evaluate the potential presence of carry-over effects, we conducted the Friedman test for non-normally distributed variables and repeated-measures ANOVA for normally distributed ones (α = 0.05).

To examine the main effects of stimulation site (left DLPFC, M1, cerebellum, or sham), time (pre vs. post), and their interaction, a generalized linear mixed model (GLMM) was employed. Stimulation site, time, and their interaction were treated as fixed effects, with participants modeled as random effects to account for within-subject correlations. A gamma distribution with a log link function was applied to non-normally distributed outcomes, and a normal distribution with an identity link was used for normally distributed variables. Where significant interactions emerged, post hoc analyses were conducted within each stimulation condition to explore specific effects. Effect sizes were reported using partial eta squared (η²), interpreted as small (0.01–0.06), moderate (0.06–0.14), and large (≥ 0.14).

Results

Table 2 summarizes the descriptive data, including the means and standard deviations of the spatiotemporal gait parameters for the different brain stimulation montages. Table 3 presents the results of baseline comparisons across stimulation sites, confirming that there were no significant baseline differences and that the one-week washout period was sufficient. Finally, Table 4 displays the main effects of time, stimulation site, and their interactions.

Table 2.

Mean and standard deviation of gait Spatiotemporal parameters at different montages of brain stimulation.

Parameters	Conditions	Left DLPFC (N = 19)		M1 (N = 19)		Cerebellum (N = 19)		Sham (N = 19)
		Pre	Post	Pre	Post	Pre	Post	Pre	Post
Stride time [ms]	Single task	914.05 ± 261.53	873.19 ± 303.53	966.97 ± 265.62	945.98 ± 268.00	933.63 ± 194.41	904.92 ± 227.43	936.84 ± 205.01	940.81 ± 202.87
	Dual task	1049.72 ± 340.13	979.38 ± 351.76	1141.34 ± 347.74	1045.65 ± 332.95	1036.34 ± 220.29	994.18 ± 245.02	1041.27 ± 236.40	1052.44 ± 245.06
DTC to stride time [%]		14.58 ± 11.89	12.58 ± 14.50	18.01 ± 14.92	9.93 ± 13.11	13.03 ± 9.43	10.60 ± 19.00	11.73 ± 10.97	12.22 ± 12.42
Stride time variability [%]	Single task	6.00 ± 2.50	4.70 ± 2.15	6.78 ± 2.36	6.01 ± 2.28	6.32 ± 2.41	5.05 ± 2.09	6.49 ± 3.35	6.60 ± 3.49
	Dual task	7.28 ± 2.69	5.20 ± 2.32	8.24 ± 2.28	6.75 ± 2.19	7.62 ± 3.09	5.45 ± 2.19	8.52 ± 5.57	8.36 ± 5.26
DTC to stride time variability [%]		24.16 ± 18.01	13.70 ± 20.23	25.55 ± 19.24	15.31 ± 13.83	21.23 ± 11.59	9.82 ± 9.60	28.33 ± 29.44	31.06 ± 54.32
Stride length [m]	Single task	1.22 ± 0.31	1.30 ± 0.41	1.16 ± 0.31	1.19 ± 0.35	1.17 ± 0.26	1.22 ± 0.32	1.17 ± 0.28	1.17 ± 0.27
	Dual task	1.10 ± 0.30	1.20 ± 0.40	1.02 ± 0.31	1.12 ± 0.38	1.07 ± 0.24	1.14 ± 0.35	1.07 ± 0.24	1.07 ± 0.26
DTC to stride length [%]		− 9.92 ± 7.80	− 8.15 ± 9.31	− 11.93 ± 9.19	− 6.73 ± 9.29	− 8.09 ± 5.27	− 6.73 ± 9.92	− 8.11 ± 6.96	− 8.36 ± 7.71
Stride length variability [%]	Single task	7.96 ± 6.98	6.77 ± 6.29	8.13 ± 6.61	7.85 ± 7.16	7.23 ± 4.75	6.37 ± 4.92	7.45 ± 4.09	7.36 ± 4.50
	Dual task	8.60 ± 7.18	7.14 ± 6.64	8.88 ± 6.67	8.01 ± 7.28	8.06 ± 4.98	6.70 ± 5.04	8.15 ± 4.08	8.09 ± 4.78
DTC to stride length variability [%]		12.11 ± 12.33	5.46 ± 16.53	13.88 ± 12.96	2.91 ± 9.46	12.96 ± 8.94	6.34 ± 10.68	13.33 ± 12.86	17.96 ± 12.79
Gait speed [m/s]	Single task	1.23 ± 0.64	1.45 ± 0.91	1.12 ± 0.66	1.20 ± 0.78	1.12 ± 0.57	1.24 ± 0.71	1.43 ± 0.90	1.41 ± 0.88
	Dual task	1.01 ± 0.53	1.24 ± 0.82	0.89 ± 0.60	1.08 ± 0.82	0.95 ± 0.49	1.10 ± 0.81	1.16 ± 0.69	1.08 ± 0.79
DTC to gait speed [%]		− 18.28 ± 14.20	− 14.81 ± 17.63	− 21.64 ± 16.02	− 12.18 ± 17.84	− 15.25 ± 9.31	− 12.08 ± 17.38	− 16.52 ± 13.40	− 16.84 ± 14.49

DTC = Dual-Task Cost; DLPFC = Dorsolateral Prefrontal Cortex; M1 = Primary Motor Cortex.

Table 3.

Results of baseline comparison across stimulation sites.

Parameters	Conditions	F/χ²	P
Stride time [ms]	Single task	1.042	0.791
	Dual task	0.579	0.631
DTC to stride time [%]		4.200	0.241
Stride time variability [%]	Single task	1.635	0.651
	Dual task	1.042	0.791
DTC to stride time variability [%]		1.295	0.730
Stride length [m]	Single task	0.910	0.823
	Dual task	0.532	0.912
DTC to stride length [%]		2.874	0.412
Stride length variability [%]	Single task	0.537	0.911
	Dual task	0.916	0.822
DTC to stride length variability [%]		0.537	0.911
Gait speed [m/s]	Single task	7.232	0.065
	Dual task	5.190	0.158
DTC to gait speed [%]		1.105	0.776

DTC = Dual-Task Cost; DLPFC = Dorsolateral Prefrontal Cortex; M1 = Primary Motor Cortex.

Table 4.

The results of the generalized linear mixed model analysis.

Parameters	Conditions	Time effect			Stimulation site effect			Interaction effect
		F	P	η2	F	P	η2	F	P	η2
Stride time [ms]	Single task	0.299	0.585	0.002	0.465	0.707	0.010	0.057	0.982	0.001
	Dual task	1.060	0.305	0.007	0.603	0.614	0.012	0.229	0.876	0.005
DTC to stride time [%]		2.926	0.104	0.140	0.270	0.847	0.015	0.891	0.452	0.047
Stride time variability [%]	Single task	31.421	< 0.001*	0.636	1.084	0.364	0.057	3.456	0.023*	0.161
	Dual task	143.120	< 0.001*	0.888	1.701	0.204	0.086	8.249	< 0.001*	0.314
DTC to stride time variability [%]		3.364	0.083	0.157	1.947	0.133	0.098	1.209	0.315	0.063
Stride length [m]	Single task	0.597	0.441	0.004	0.639	0.591	0.013	0.129	0.942	0.003
	Dual task	1.493	0.224	0.010	0.510	0.676	0.011	0.220	0.883	0.005
DTC to stride length [%]		3.723	0.070	0.171	0.491	0.690	0.027	0.977	0.410	0.051
Stride length variability [%]	Single task	0.415	0.521	0.003	0.270	0.847	0.006	0.073	0.974	0.002
	Dual task	0.939	0.334	0.006	0.218	0.884	0.005	0.109	0.955	0.002
DTC to stride length variability [%]		8.349	0.010*	0.317	2.503	0.069	0.122	3.508	0.038*	0.163
Gait speed [m/s]	Single task	0.639	0.425	0.004	1.026	0.383	0.021	0.152	0.928	0.003
	Dual task	9.079	0.007*	0.335	0.221	0.882	0.012	2.680	0.050*	0.130
DTC to gait speed [%]		4.309	0.053	0.193	0.463	0.709	0.025	0.880	0.457	0.047

DTC = Dual-Task Cost.

Asterisks (*) indicate statistically significant effects (α = 0.05).

Effect sizes are reported using partial eta squared (η²), interpreted as small (0.01–0.06), moderate (0.06–0.14), and large (≥ 0.14).

Stride time

The GLMM analysis revealed that time, stimulation site, and their interaction had no statistically significant effects on stride time under either single- or dual-task conditions, nor on the DTC of stride time (Table 4; Fig. 3).

Fig. 3.

Bar plots showing changes in dual-task cost (DTC) for spatiotemporal gait parameters following stimulation of the left dorsolateral prefrontal cortex (DLPFC), primary motor cortex (M1), cerebellum, and sham. Data are presented as mean values, with error bars representing standard deviation (SD). Asterisks and connecting lines indicate statistically significant post hoc comparisons (p < 0.05). Absolute DTC values are shown, where lower values indicate reduced cognitive-motor interference and improved dual-task performance. A significant reduction in stride length variability DTC was observed following stimulation of M1 (p = 0.007) and the cerebellum (p = 0.011), while no significant changes were found following stimulation of the DLPFC (p = 0.173) or under the sham condition (p = 0.147).

Stride time variability

The GLMM analysis revealed a significant interaction effect for stride time variability under both single- and dual-task conditions (P = 0.023 and P < 0.001, respectively), indicating that the change in this variable over time differed across stimulation sites. However, the main effect of stimulation site was not statistically significant (P ≥ 0.204) (Table 4). Post hoc analyses within each stimulation site revealed that, under the single-task condition, stride time variability significantly decreased following stimulation of the left DLPFC (P = 0.001, η² = 0.454), M1 (P < 0.001, η² = 0.574), and the cerebellum (P < 0.001, η² = 0.594), while no significant change was observed in the sham condition (P = 0.835, η² = 0.002). A similar pattern was observed under the dual-task condition, with significant reductions following stimulation of the left DLPFC (P = 0.001, η² = 0.667), M1 (P < 0.001, η² = 0.843), and the cerebellum (P < 0.001, η² = 0.680), whereas the sham condition again showed no significant change (P = 0.646, η² = 0.012). In contrast, for the DTC of stride time variability, no significant main effects or interaction effects were found for time, stimulation site, or their interaction (Table 4; Fig. 3).

Stride length

The GLMM analysis revealed that time, stimulation site, and their interaction had no statistically significant effects on stride length under either single- or dual-task conditions, nor on the DTC of stride length (Table 4; Fig. 3).

Stride length variability

The GLMM analysis revealed a significant interaction effect for the DTC of stride length variability (P = 0.038), indicating that changes over time in this variable differed across stimulation sites. However, the main effect of stimulation site was not statistically significant (P = 0.069) (Table 4). Post hoc analyses within each stimulation site showed that the DTC of stride length variability significantly decreased following stimulation of M1 (P = 0.007, η² = 0.341) and the cerebellum (P = 0.011, η² = 0.308). In contrast, no significant changes were observed following stimulation of the left DLPFC (P = 0.173, η² = 0.101) or in the sham condition (P = 0.147, η² = 0.113) (Table 4; Fig. 3).

However, for stride length variability under both single- and dual-task conditions, there were no statistically significant main effects of time or stimulation site, nor was there a significant interaction between them.

Gait speed

Regarding gait speed under the dual-task condition, the interaction effect between time and stimulation site approached statistical significance (P = 0.050), while the main effect of stimulation site was not (P = 0.882) (Table 4). Post hoc analyses within each stimulation site revealed significant improvements in dual-task gait speed following stimulation of the left DLPFC (P = 0.039, η² = 0.216) and M1 (P = 0.028, η² = 0.242). No significant changes were observed following cerebellar stimulation (P = 0.145, η² = 0.114) or in the sham condition (P = 0.053, η² = 0.202). In contrast, neither the single-task gait speed nor the DTC of gait speed showed significant main or interaction effects (Table 4; Fig. 3).

Discussion

The findings of this crossover study showed that although the main effect of stimulation site was not statistically significant, significant interaction effects suggest that certain regions may influence gait performance differently over time. Specifically, M1 stimulation resulted in reduced stride time variability under both single- and dual-task conditions, improved dual-task gait speed, and decreased the DTC of stride length variability. Stimulation of the left DLPFC led to improvements in stride time variability and dual-task gait speed, while cerebellar stimulation was associated with reductions in stride time variability and DTC of stride length variability. Sham stimulation did not produce notable changes in any parameter. These site-specific trends suggest that HD-tDCS targeting M1, DLPFC, or the cerebellum may more effectively modulate gait performance under cognitively demanding conditions. However, as overall group differences were not significant, these findings should be interpreted as relative, not conclusive evidence of one site’s superiority.

Gait variability—defined as stride-to-stride fluctuations—is a clinically meaningful indicator of locomotor control and stability⁴⁶. Compared to traditional spatiotemporal parameters such as gait speed or stride length, gait variability is more strongly associated with fall risk and higher-order motor control processes^3,55,56. Moreover, this parameter likely has higher statistical sensitivity, making it better able to detect changes induced by brain stimulation. This could explain why the stimulation-induced effects were more pronounced for gait variability, highlighting its greater sensitivity as a clinical marker relative to other gait parameters.

Furthermore, gait variability is closely linked to executive function—a cognitive domain essential for maintaining performance in attentionally demanding contexts such as dual-task walking^46,57. Although higher gait variability has been consistently associated with impaired cognitive function, neurodegeneration, and increased fall risk^3,58, overly low variability has also been considered maladaptive, reflecting reduced flexibility in adapting to environmental challenges. According to the optimal movement variability framework proposed by Stergiou et al., healthy motor behavior lies within a functional range—neither excessively variable nor overly rigid⁵⁹. From this perspective, the reductions in variability observed after stimulation of the M1, DLPFC, and cerebellum may reflect a shift toward more optimal, stable, and adaptable gait patterns under cognitive load. These findings underscore the potential of targeting these regions to enhance dynamic balance and cognitive-motor efficiency in older adults.

The observed improvement in CMI and increased gait automaticity following left DLPFC stimulation aligns with previous neuroimaging research demonstrating this region’s central role in dual-task processing. For instance, Stelzel et al. reported increased functional connectivity between the DLPFC and sensory areas during tasks with high temporal overlap, highlighting enhanced task shielding— a process whereby the brain selectively allocates cognitive resources to prioritize a primary task while suppressing interference from secondary tasks, thus maintaining accuracy and efficiency in multitasking contexts⁶⁰. Similarly, improved dual-task performance following training has been associated with greater gray matter volume⁶¹, increased neural processing efficiency⁶², and more distinct task separation within frontoparietal-subcortical networks, where the left DLPFC functions as a critical hub⁶³.

Support for cerebellar involvement also comes from neuroimaging findings. Wu et al. used functional magnetic resonance imaging (fMRI) to investigate brain activity in healthy individuals during dual-task performance, which involved a motor task (self-paced alternating finger tapping) combined with a cognitive task (visual letter counting)⁶⁴. Their findings revealed activation of the cerebellum, with functional connectivity to key motor and cognitive areas such as the M1, supplementary motor area (SMA), pre-SMA, and superior parietal lobule. These results underscore the cerebellum’s integrative role in coordinating motor and cognitive networks to facilitate dual-task performance⁶⁴. Moreover, they observed cerebellar plasticity following dual-task training, suggesting that the cerebellum contributes to transitioning motor behavior from attentionally demanding to more automatic modes of control⁶⁴.

The present study demonstrated improvements in dual-task gait speed following stimulation of the left DLPFC, consistent with prior research highlighting the role of prefrontal regions in modulating cognitive-motor function. Manor et al.¹⁵ reported a substantial improvement in dual-task gait speed after tDCS over the left DLPFC compared to sham, with a large effect size (d = 3.19). Zhou et al.¹⁶, in contrast, examined the effect of stimulation on the DTC of gait speed and found greater reductions in DTC following stimulation of the left DLPFC—either alone or combined with SM1—than SM1 alone or sham, with a moderate effect size (d = 0.53). In the current study, the magnitude of improvement in dual-task gait speed between active and sham stimulation corresponded to a comparable moderate effect (d = 0.50), calculated based on the standardized difference in pre-post change scores across conditions. This approach captures both the direction and magnitude of the differential response, providing a meaningful estimate of clinical relevance. Collectively, these findings reinforce the potential of left DLPFC stimulation to modulate dual-task gait performance in healthy adults.

Moreover, previous studies have linked gait speed to executive functions, which are primarily mediated by the frontal cortex, particularly the DLPFC^48,57,65. Structural neuroimaging research has also shown a positive association between gait speed and gray matter volume in frontal regions among older adults⁴⁷. Emerging evidence further suggests that specific gait parameters—such as pace, rhythm, phase, and variability—are associated with distinct cognitive functions and neural substrates⁴⁸. However, the present study did not analyze these components separately and thus cannot identify which aspects of gait were most affected by stimulation. Future research should aim to disentangle these associations and clarify how neuromodulatory interventions differentially impact specific gait domains.

Furthermore, a study by Orcioli-Silva et al. that used anodal tDCS on both the prefrontal cortex (PFC) and M1 in combination with treadmill training found a negative correlation between changes in left PFC activity and stride length variability. Specifically, greater increases in PFC activity were associated with larger decreases in stride length variability. They suggested that enhanced PFC activity from the combined intervention is linked to improved gait automaticity by increasing the availability of executive-attentional resources for walking control¹⁰. Although our study did not find a significant effect of left DLPFC stimulation on stride length variability, reductions were observed in stride time variability following stimulation of the left DLPFC, M1, and cerebellum. These findings may similarly suggest that HD-tDCS targeting these brain regions could enhance gait automaticity and reduce the attentional demands of walking. The absence of significant effects on stride length variability may be due to differences in stimulation protocols, outcome measures, or population characteristics. Nonetheless, the observed improvements in stride time variability highlight the potential of these stimulation sites in modulating cognitive-motor integration during gait.

One possible explanation for the observed improvement in CMI is the enhancement of its core components—cognitive and motor functions—via brain stimulation. Previous studies have demonstrated that a single session of anodal tDCS over the left DLPFC can enhance a range of cognitive abilities, including working memory²², executive function²³, and processing speed⁴⁹. Similarly, anodal stimulation of the cerebellum and M1 has also been linked to improvements in processing speed and working memory^29,49, suggesting that modulating different nodes of the motor-cognitive network can enhance processing efficiency and indirectly benefit motor performance, particularly during cognitively demanding walking tasks.

The theoretical framework of dual-task interference further supports this interpretation. According to the capacity-sharing theory, the brain possesses limited cognitive resources that must be distributed between tasks. When the combined task demand exceeds this capacity, performance declines in one or both domains¹⁷. HD-tDCS, by modulating cortical excitability and enhancing functional connectivity, may expand this resource pool or increase access to executive-attentional reserves. As a result, the cognitive gains induced by HD-tDCS could reduce the competition for shared resources during dual-task walking, thereby improving CMI.

In line with the bottleneck theory, which posits that tasks requiring similar cognitive processes must be processed serially, HD-tDCS may reduce delays between decision-making processes by improving processing speed. This acceleration could allow for more efficient task switching or parallel execution, ultimately mitigating CMI¹⁷.

In addition to cognitive enhancements, tDCS also appears to confer motor benefits, especially in older adults. Previous research has shown that a single session of anodal tDCS over the left DLPFC can improve gait speed^15,16. Additionally, stimulation of the cerebellum and M1 has been associated with enhanced motor learning and adaptation²⁸. In the present study, stimulation of all three regions—the DLPFC, cerebellum, and M1—resulted in reduced stride time variability during single-task walking. This finding may be attributed to the strong functional connectivity of the DLPFC and cerebellum with core motor regions, including the M1, as well as higher-order motor areas such as the premotor and SMA, which themselves maintain extensive interactions with M1. These networks play a key role in the coordination and execution of complex motor behaviors.

The motor benefits observed following DLPFC stimulation are particularly noteworthy, given its traditional association with cognitive control rather than direct motor output. However, emerging evidence suggests a broader role for the DLPFC in motor control. Neuroimaging studies have shown that the DLPFC is activated not only during imagined movements but also during actual motor execution⁶⁶. Moreover, its anatomical and functional connections with subcortical structures such as the basal ganglia—critical for planning and sequencing movements—further support its involvement in motor planning and execution⁶⁷.

Therefore, the observed reduction in stride time variability during single-task walking following DLPFC stimulation may reflect improved motor planning, even in the absence of cognitive load. With aging, declines in the connectivity between the DLPFC and M1 have been linked to poorer motor coordination, prompting compensatory recruitment of non-motor areas like the DLPFC to support motor output—particularly under dual-task conditions⁶⁸. While direct evidence linking DLPFC stimulation to increased corticospinal excitability is limited, neurophysiological studies support the existence of functional connectivity between the DLPFC and motor-related regions, including M1^69–71. Based on this, it is hypothesized that DLPFC stimulation may influence motor network dynamics, possibly affecting corticospinal excitability and spinal circuitry through top-down modulation. These network-level changes may help explain the observed improvements in gait stability—such as reduced stride time variability—even during single-task walking.

The cerebellum, a key node in the cerebello–thalamo–cortical circuit, influences motor output by modulating synaptic inputs to upper motor neurons in the M1, including those of the corticospinal tract, and indirectly affects the excitability of spinal alpha motor neurons⁷². Evidence from neurophysiological studies indicates that cerebellar inputs interact with distinct cortical interneuron circuits via different pathways, contributing to motor adaptation and learning⁷². This may underlie the observed improvements in gait stability following cerebellar HD-tDCS.

Furthermore, although the main effect of stimulation site was not statistically significant, significant time × stimulation site interaction effects indicate that M1 stimulation may modulate gait performance over time. Pre–post comparisons revealed that M1 stimulation led to reduced stride time variability in both single- and dual-task conditions, improved dual-task gait speed, and a reduced DTC of stride length variability. These findings suggest that M1 stimulation may have specific time-dependent effects on motor-cognitive performance, although no overall differences were observed across stimulation conditions within participants. Mechanistically, these effects are hypothesized to arise from increased corticospinal excitability through depolarization of neuronal resting membrane potentials, potentially enhancing motor control and gait stability⁷³. This increased efficiency in the motor system may reduce the cognitive demands of walking, thereby minimizing dual-task interference. Furthermore, M1 stimulation promotes motor learning by reinforcing synaptic plasticity within motor pathways^74,75, facilitating gait automaticity and reducing reliance on higher-order executive control. Given M1’s connectivity with regions such as the premotor cortex, SMA, and prefrontal cortex^76,77, stimulation of this area may enhance network-level coordination, further contributing to improved motor-cognitive integration and reductions in DTC and CMI.

While this study found trends suggesting differential effects of M1, DLPFC, and cerebellar stimulation on gait parameters, no statistically significant differences were observed across stimulation sites. Several factors may explain this outcome. First, the relatively small sample size (N = 19) may limit statistical power to detect subtle site-specific effects, even in a crossover design where each participant serves as their own control, which generally reduces within-subject variability and increases sensitivity. Second, the single-session HD-tDCS protocol may have produced transient or modest neuromodulatory effects insufficient to generate robust differences. Third, inter-individual variability in anatomy, baseline cortical excitability, and responsiveness to tDCS likely contributed to heterogeneous responses, diluting potential site-specific effects.

Moreover, although prior studies demonstrate functional specificity of M1, DLPFC, and cerebellar regions, gait is controlled by highly distributed and interconnected networks. Stimulation of one region may induce compensatory changes in connected areas, resulting in overlapping functional effects that mask clear site-specific differences. Additionally, it is possible that the intensity and cognitive demands of the dual-task—walking 20 m while repeatedly subtracting 7 from a three-digit number—may have reduced the sensitivity of our measures to detect subtle differences between M1, DLPFC, or cerebellar stimulation, as the high cognitive load could engage multiple neural networks simultaneously.

Taken together, these factors suggest that the lack of statistically significant differences across stimulation sites does not indicate absence of site-specific effects. Rather, it likely reflects a combination of sample size limitations, inter-individual variability, network-level compensatory mechanisms, and moderate sensitivity of behavioral measures. These findings still provide valuable insights into the potential of HD-tDCS to modulate gait performance and may guide the design of future studies to optimize stimulation parameters, session frequency, and outcome measures for detecting site-specific effects.

Limitations and future directions

This study has several limitations that should be considered when interpreting the findings. First, we did not assess the long-term carryover effects of stimulation; therefore, it remains unclear how long the observed improvements in dual-task gait performance persist beyond the intervention. Future studies should incorporate follow-up assessments to evaluate the durability of these effects over time. Second, certain outcome measures—such as neuroimaging—were not included, which could have provided deeper insight into the underlying neurophysiological mechanisms of HD-tDCS. Incorporating techniques like functional MRI or EEG is recommended to better understand how changes in cortical activity relate to improvements in cognitive-motor performance. Additionally, while this study examined a single cognitive task (serial subtraction during walking), future research should investigate a broader range of dual-task paradigms to enhance generalizability. Moreover, the intervention was delivered at rest and involved cognitively healthy older adults; further studies are needed to determine whether similar effects occur when stimulation is applied during active dual-task walking or in populations with cognitive impairment. Finally, although the Williams Square crossover design and a one-week washout period were implemented to minimize potential order and carry-over effects, and no baseline differences were observed across sessions, order effects were not explicitly modeled in the statistical analyses. While these design features and preliminary checks reduce the likelihood of meaningful order-related bias, the possibility of subtle sequence effects cannot be entirely excluded, particularly given the relatively small number of participants in each sequence group (approximately 4–5). Future studies with larger samples and explicit inclusion of sequence effects in the analyses may provide a more definitive assessment of potential order influences.

Conclusion

This crossover study demonstrates that HD-tDCS targeting M1, left DLPFC, or the cerebellum can enhance gait performance, particularly stride time variability and dual-task walking, even though no single stimulation site showed clear superiority. The absence of statistically significant differences across sites suggests that multiple nodes within the motor-cognitive network may contribute to gait modulation through overlapping or compensatory mechanisms, rather than any single region being uniquely critical. Clinically, this indicates that interventions targeting any of these regions could potentially improve walking automaticity, strengthen cognitive-motor integration, and reduce fall risk in older adults or individuals with neurological impairments.

Future research should employ larger samples and repeated stimulation sessions to increase statistical power and detect subtle site-specific effects. Investigating diverse dual-task paradigms, explicitly modeling site- and sequence-dependent effects, and integrating behavioral outcomes with neurophysiological or neuroimaging measures will help elucidate the underlying mechanisms. Furthermore, exploring multi-site stimulation protocols and accounting for individual variability may optimize HD-tDCS interventions and guide personalized rehabilitation strategies. Collectively, these considerations situate our findings in a broader context and provide more definitive insight into the contributions of different brain regions to gait modulation.

Author contributions

R.K. conceptualized and designed the study, obtained funding, supervised data collection, interpreted the results, and edited the manuscript. M.I. and V.R.M. contributed to data collection, conducted all analyses, interpreted the results, and drafted the main manuscript.

Data availability

All data generated or analyzed during this study are included in this article. Further enquiries can be directed to the corresponding author.

Declarations

Statement of ethics

All participants provided written informed consent. This study was approved by the Institutional Ethics Committee of the Tehran University of Medical Sciences (IR.TUMS.FNM.REC.1400.193). All procedures were conducted in accordance with relevant ethical guidelines and regulations.

Competing interests

The authors declare no competing interests.