What telerehabilitation trials in Parkinson's disease actually show

The headline meta-analysis: telehealth versus control

If you only read one paper on this topic this year, make it Sun et al. (2026), a meta-analysis of 15 randomised controlled trials (765 participants) evaluating telehealth interventions for people with Parkinson's disease. Quality of life improved significantly on the SF-36 and Brunnsviken Brief Quality of Life Scale (SMD 0.39, 95% CI 0.06–0.72), with only a marginal, non-significant improvement on the PDQ-8.

Before you read that as "the phone beats the app", note the authors' own caveat: cognition was significantly enhanced, but with moderate heterogeneity and significant publication bias. The cognition result, in other words, should not be quoted without the bias warning attached.

Head-to-head: remote versus hospital-based care

The comparison that matters more to a clinic owner is not telehealth versus nothing — it is telehealth versus you. Prukviwat et al. (2026), in a nonrandomised controlled pilot study of 46 participants, had patients with idiopathic Parkinson's disease self-select into telerehabilitation or hospital-based rehabilitation over 16 sessions. Both groups improved significantly on the Berg Balance Scale (tele: post-pre = 3.50, p < 0.001; hospital: post-pre = 4.35, p < 0.001) with no significant difference between groups (mean difference: -0.85, p = 0.454), and the same pattern held for the Chula PMS (tele: 3.45; hospital: 5.70; between-group difference -2.25, p = 0.086). Attendance exceeded 90% in both groups. The authors' caveat matters: the BBS improvements did not reach the minimal clinically important difference, so both delivery models produced statistically significant but small balance gains.

What the home-based review adds — and what it cannot say

D'Souza et al. (2025) systematically reviewed 37 studies of home-based telerehabilitation in people with Parkinson's disease, 13 of which were eligible for meta-analysis. The meta-analysis revealed small but significant improvements in balance (SMD = 0.25; 95% CI: 0.04 to 0.45; p = 0.02) and functional mobility (SMD = -0.28; 95% CI: -0.52 to -0.05; p = 0.02), but no significant effect on quality of life (SMD = -0.08; 95% CI: -0.25 to 0.09; p = 0.35). The authors themselves note that the effect on quality of life is unclear and warrants further research. Treat this review as supporting context, not as proof of effect.

The delivery models actually being studied

How are these programmes run in practice? Naqvi et al. (2026), in a scoping review of home-based virtual reality telerehabilitation, reviewed 18 studies with 745 participants: 16 studies (89%) used fully home-based delivery, structured clinic-to-home transitions were rare, asynchronous supervision — mainly scheduled telephone calls — was used in most studies (50%), and only two (11%) employed real-time synchronous supervision. Adherence was generally high, but adverse event monitoring was inconsistently reported. The authors call for standardised dose reporting, structured adverse event monitoring, evidence-based supervision guidelines and geographically diverse trials before translation into routine clinical care.

On adherence more broadly, D'Alessandro et al. (2026) found adherence was generally higher in supervised and socially engaging interventions, while dropout was mainly associated with disease-related, logistical, technological, and psychosocial factors — though with considerable heterogeneity across intervention characteristics, exercise dose and outcome measures. And the field is clearly growing: publication volume on Parkinson's disease and exercise therapy has increased yearly with a notable surge after 2019, and telemedicine, telehealth and telerehabilitation are current research hotspots.

Where the wider evidence gets shaky

Two related literatures should temper any enthusiasm for over-claiming. In a related population of adults with various neurological disorders, Calabrò et al. (2026) found the point estimate favoured telerehabilitation for walking outcomes but was imprecise and compatible with no difference (SMD 1.06, 95% CI -0.23 to 2.35; three studies; 121 participants; I² = 57.6%), with similar uncertainty for global motor-functional and ADL outcomes (0.83, -0.35 to 2.00) — both rated very-low-certainty evidence. Separately, a review of wearable-based measurement in neurological rehabilitation found kinematic biomarkers dominated Parkinson's disease studies (9/10), but the authors note that methodological heterogeneity and limited longitudinal evidence mean clinical implementation remains at an early stage. In PD gait measurement specifically, pooled estimates came with prediction intervals far wider than their confidence intervals — gait speed pooled at 0.84 m/s with a 95% PI of 0.56–1.12 — precluding generalisation to future settings.

What this means for your practice

A reasonable reading of the central evidence: telehealth interventions show promise for people with Parkinson's disease — in one meta-analysis of randomised controlled trials they improved quality of life, motor symptoms and ADL, and in one nonrandomised pilot study, motion sensor telerehabilitation showed no statistically significant difference from hospital-based rehabilitation in balance and mobility gains, with attendance above 90% in both arms. In the head-to-head pilot study, telerehabilitation produced balance and mobility improvements similar to hospital-based rehabilitation, with no statistically significant difference between groups.

Three practical starting points emerge from these trials. First, supervision structure is a design decision worth making deliberately: most reviewed programmes used scheduled telephone calls rather than real-time synchronous sessions, and, in a separate scoping review of exercise interventions in general, adherence was generally higher in supervised and socially engaging interventions. Second, set outcome expectations carefully — in the head-to-head comparison, both groups improved significantly within themselves but the between-group differences were not statistically significant, and the balance gains did not reach the minimal clinically important difference. Third, if you measure gait remotely with wearables, treat your own numbers as within-patient trends rather than population benchmarks, because the pooled estimates are too dispersed to generalise.

A reasonable reading of the current literature is that telehealth interventions have been compared with controls in several trials, that head-to-head comparisons with in-person rehabilitation remain few and uncertain, and that measurement approaches across studies are still heterogeneous. If you want to see how remote exercise monitoring and phased plans fit into a clinic workflow, that is what we built PhysioSync for — but the trials above stand on their own.