Study design

This study is a single-blind, prospective, single-center, randomized clinical trial (RCT), conducted in the Neurorehabilitation Unit of Policlinico San Marco in Osio Sotto, Bergamo, Italy, a specialized center for assisting to individuals with PD, beginning in May 2023. The study was carried out according to the principles of the Declaration of Helsinki and good clinical practice standards and in line with the Standard Protocol Items Recommendations for Interventional Trials (SPIRITS) guidelines. The Ethics Committee of the ASST Papa Giovanni XXIII Hospital, Bergamo, approved all the experimental procedures on 11th May 2023. All the individuals who agreed to participate provided written informed consent.

Participants

A group of 20 individuals with PD afferent to the outpatient clinic of Neurology, the VAMP Center, and the O.U. of Neurological Rehabilitation of Policlinico San Marco was enrolled.

Individuals were included if they had: idiopathic PD diagnosis according to the UK PD Society Brain Bank criteria, presence of FoG based on the neurologist’s clinical observation, disease staging ≥ 2 points according to the Hoehn & Yahr stage [20], age between 40 and 85 years, not exhibiting other associated neurological diseases and/or musculoskeletal and cardiorespiratory conditions, and no relevant cognitive deficits on the Montreal Cognitive Assessment (MoCA) [21] test. The exclusion criteria included a diagnosis of atypical PD, neuropsychiatric comorbidities, age younger than 18 years, not having undergone treatment rehabilitation in the previous three months, prior medical history of epilepsy and traumatic brain injury, neurosurgery, and the presence of pacemakers and DBS.

Experimental procedure

The enrolled individuals with PD exhibiting FoG underwent gait training on the C-Mill VR+ associated with or not associated with ctDCS stimulation. The treatment consisted of one 20-minute session per day for 10 days, from Monday to Friday each week. All assessments were conducted by physicians and neuropsychologists with expertise in the management of PD and cognitive and motor assessments.

The participants were randomly allocated to either a group receiving motor-cognitive training on the C-Mill paired with ctDCS or with sham tDCS and were assessed at 3 timepoints: pre-intervention (T0), post-intervention (T1), and 4 weeks after the last intervention session (T2). Randomization for assignment to the group was performed via an online generator (https://www.random.org/lists/), with odd and even numbers identifying subjects undergoing ctDCS and sham-tDCS treatment, respectively. The participants and caregivers were not aware of group allocation throughout the study. Physical therapists were the only ones aware of the type of treatments, as they must set the stimulation based on the randomization list. Notwithstanding, the treatment executor was instructed not to reveal the group assignment to anyone.

Anodal ctDCS and motor-cognitive training

The participants underwent a comprehensive treatment program consisting of one 20-minute session per day for 10 days. These sessions involved the application of tDCS combined with a gait rehabilitation program. During the sessions, a portable battery-powered neurostimulator (BrainStim, E.M.S srl) delivered a direct current of 2 mA through two electrodes positioned at the cerebellar level, on the posterior cranial fossa (anode), and at the right arm (cathode). The neurostimulator was placed in a backpack to allow movement during treadmill training. The electrodes, each measuring 35 cm2, were arranged in a unipolar montage and covered with saline-soaked sponges and electrogel to optimize conductivity [22]. This ctDCS protocol has been used in other studies previously [22, 23].

For those in the sham group, the stimulation program consisted of 2 s of initial stimulation followed by a decrease until its shutdown, which was intended to mimic the initial sensations of real stimulation but without delivering continuous stimulation. This approach was adopted to provide a similar sensory experience to the participants while not providing therapeutic current.

The safety of the ctDCS application was a priority and was therefore assessed at each session by collecting information on perceived sensations, possible discomfort, or side effects.

In association with tDCS, each participant underwent a structured gait rehabilitation program via the C-Mill, which involved one 20-minute sessions per day, structured as follows: 10 min of gait training and 10 min of motor-cognitive training. During gait training on the C-Mill, participants engaged in various exercises designed to improve their walking pattern and adaptability. This included the following C-Mill exercises: (1) gait assessment (2 min); (2) gait adaptability (3 min); (3) stepping stones random (3 min), in which the subject had to reach and step rectangular-shaped targets projected on the treadmill, which could transform into an obstacle, necessitating adjustment in their walking pattern to avoid it; and (4) speed adaptability (2 min), in which the subject has to walk inside a green rectangle that moves back and forth on the carpet, forcing the subject to speed up and slow down to stay inside the rectangular perimeter.

The motor-cognitive training component of the program focused on enhancing participants’ ability to perform dual tasks while walking. This included the following C-Mill exercises: (1) Trace (3 min): subjects had to interact with objects appearing on the treadmill while walking, by stepping onto them or avoiding them; (2) Soccer walking (3 min): while walking, subjects had to control a virtual ball using lateral movements, without dropping the ball to the ground; and (3) Italian Alps (4 min): while walking, subjects collected ingredients to make pizza by moving left and right.

Outcome measures

Assessments were performed on day 1 (T0), before the intervention began; on day 10 (T1), immediately after the last intervention session; and four weeks after the completion of the treatment (T2). All assessments were performed in the ON state of PD medication.

Participants were assessed at T0, T1 and T2 via the following scales/instruments: part 3 of the Unified Parkinson’s Disease Rating Scale (UPDRS-III) [20] to evaluate general motor functions; the Freezing of Gait Questionnaire (FOG-Q) [24]; the 6 min Walk Test (6MWT) [25] to assess endurance; the Borg Category Ratio Scale 0–10 (BORG) [26] to assess the perception of exertion; the Mini Balance Evaluation Systems Test (Mini-BESTest) [27] Italian version, to detect balance impairments; the Timed Up and Go (TUG) test [28] to assess functional mobility, balance, walking, and fall risk; the MoCA test [21] and Mini Mental State Examination (MMSE) [29] to evaluate overall cognitive performance; the Frontal Assessment Battery (FAB) [30]; the Parkinson’s Disease Questionnaire-8 (PDQ-8) [31] to assess quality of life; the Beck Depression Inventory-II (BDI-II) [32] Italian version to detect the severity of depressive symptoms; and the Barthel Index (BI) [33], the Activities of Daily Living (ADL) [34] and Instrumental Activities of Daily Living (IADL) [35] to monitor functional changes and autonomy in daily life.

Motor function was also assessed by C-Mill assessments at T0 and T1, which allow the evaluation of postural control and gait through the recording of the following parameters:

Stand assessment for postural control:

Limits of Stability, which measures how far a subject can lean safely in different directions without losing balance as an indicator of dynamic stability. The recorded indicators include the surface area within which the Center of Pressure moves as a person tries to maintain balance (Sup CoP), as well as its oscillations along the medio-lateral (ML) and antero-posterior (AP) axis.

Postural Stability, which evaluates body sway (CoP velocity) in various static positions as an indicator of balance and postural control while standing: body sway on the right while standing with eyes open (OAD) and on the left (OAS), body sway on the right lower limb while standing with eyes closed (OCD) and on the left (OCS), body sway on the right in tandem stance (TD) and on the left (TS), and single-leg stance to the right leg (Dx) and to the left leg (Sin).

Gait assessment:

Gait Assessment, which measures the subject’s walking pattern on a treadmill during a 5-minute walking, during which the C-Mill records parameters such as right step length (LPDx), left step length (LPSx), step width (AmP), right weight distribution (DPDx), and left weight distribution (DPSx).

Gait Adaptability, which assesses the ability to navigate obstacles on the treadmill. The recorded parameters include right step length with obstacles (LPDxO), left step length with obstacles (LPSxO), step width with obstacles (AmPO), right weight distribution with obstacles (DPODx), and left weight distribution with obstacles (DPOSx).

StatisticsCognitive and motor-functional outcomes

The majority of the cognitive and motor-functional outcomes across the three time-points, proved to distribute Normally (i.e., skewness and kurtosis values >|1| and >|3|, respectively [36]), except for the BI, TUG, ADL and MMSE scores, which were effectively normalized via a reverse transformation. Hence, the effects of Time, Group and their interaction (Time*Group) on each cognitive and motor-functional outcome were explored via either linear or generalized linear mixed models by assuming different underlying data-generating processes on the basis of empirical data distributions. More specifically, the UPDRS-III, FOG-Q, MoCA, FAB, PDQ-8, BDI, BORG, Mini-BEST, 6MWT and IADL scores were analyzed via linear mixed models (i.e., assuming an underlying Normal distribution), whereas the BI, TUG, ADL scores were analyzed via generalized linear mixed models underlying Negative Binomial distributions, and the MMSE score was analyzed via a generalized linear mixed model underlying a Gamma distribution – which is a data-generating process suitable for empirical distributions characterized by floor-like effects and high interindividual variability [37]. Zero values are not allowed to fit a Gamma distribution, so a constant K = 0.01 was added to the MMSE score. Within all of these models, Subject was addressed as the cluster, whereas Time and Group were addressed as between- and within-subject factors, respectively. A random intercept was fitted within the Subject cluster. Bonferroni-corrected post-hoc comparisons were run for significant terms. We focused only on interaction effects.

C-Mill outcomes

The vast majority of C-Mill measures across the two time-points were distributed Normally, as yielded by the abovementioned descriptive analysis, with the exception of OAS, OCS, TD, TS, Dx, DPDx, and MLr, which were effectively normalized via a reverse transformation. Hence, the effects of Time, Group and their interaction (Time*Group) on each C-Mill outcome were explored via either linear or generalized linear mixed models. Different underlying data-generating processes were assumed based on empirical data distributions. More specifically, SupCoP, AP, OAD, OCD, Sin, LPDx, LPSx, AmP, DPSx, LPDxO, LPSxO, AmPO, DPODx, and DPOSx were analyzed via linear mixed models (i.e., assuming an underlying Normal distribution), whereas OAS, OCS, TD, TS, Dx, DPDx, and MLr, all being moderately to heavily right-skewed and overdispersed, were analyzed via generalized linear mixed models underlying a Gamma distribution, which is a data-generating process suitable for empirical distributions characterized by floor-like effects and high interindividual variability [37]. Zero values are not allowed to fit a Gamma distribution, so a constant K = 0.01 was added to the above variables. Within all of these models, Subject was addressed as the cluster, whereas Time and Group as between- and within-subject factors, respectively. A random intercept was fitted within the Subject cluster. Bonferroni-corrected post-hoc comparisons were run for significant terms. We focused only on interaction effects.

Analyses were run via IBM® SPSS® Statistics 29 (IBM Corp., 2023) and jamovi 2.3 (the jamovi project, 2022). Missing data were excluded pairwise in accordance with a per-protocol analysis.