We conducted a single-center, parallel, phase 2, randomized, active-controlled clinical trial that was retrospectively registered at ClinicalTrials.gov (NCT05728099) in 2023, after recruitment began in 2022 but before trial completion in 2025, and reported in accordance with the CONSORT-EHEALTH (Consolidated Standards of Reporting Trials of Electronic and Mobile Health Applications and Online Telehealth) guidelines (Checklist 1) [29]. The institutional review board (IRB; the ethics committee of the General University Hospital in Prague) confirmed that the protocol and study design reported in this paper are consistent with the study protocol and design that were reviewed and approved by the ethics committee before the trial started. A side-by-side comparison of key protocol elements across the IRB submission, the ClinicalTrials.gov record, and this paper is provided in Multimedia Appendix 1. This trial was conducted at the Department of Neurology, First Faculty of Medicine, Charles University and General University Hospital in Prague, Czech Republic, between March 2022 and January 2025. Participants were randomly assigned, with allocation concealment, to either the control group following the conventional EMST protocol or to the experimental group, following the same protocol enhanced with the SpiroGym app.

Participants were recruited through convenience sampling from the Movement Disorders Center, Department of Neurology, First Faculty of Medicine, Charles University and General University Hospital in Prague, Czech Republic, between March 2022 and August 2024. The inclusion criteria were: (1) a diagnosis of PD confirmed by movement disorder specialist using the UK Brain Bank criteria [30], (2) a modified Hoehn and Yahr stage of I-IV in ON medication state, (3) an age of 40 to 80 years, and (4) stable dopaminergic medication (stable dose for at least 1 mo). The exclusion criteria were: (1) suspected parkinsonism due to causes other than idiopathic PD, (2) significant cognitive impairment (Montreal Cognitive Assessment score <19), (3) respiratory disorders or diseases, (4) current or previous history of head and neck cancer, (5) smoking within the past 5 years, (6) uncontrolled hypertension, and (7) previous experience with EMST.

A simple computer-generated random allocation sequence (1:1 ratio) was generated by the trial statistician prior to study initiation using a web-based randomization tool. All outcome assessors and data analysts were blinded to group allocation. However, the physiotherapists overseeing the EMST training could not be blinded, as the use of the SpiroGym app in the experimental group was apparent. Participants were informed that the study compared 2 EMST programs. They were not given specific details regarding the nature of each program or the hypothesis that one approach might be superior, thereby maintaining participant blinding as effectively as possible under these circumstances.

The primary aim of the study was to compare EMST adherence during the unsupervised home-training phase (weeks 8‐24) between a control group and an experimental group among patients at risk for nonadherence, defined as a SEHEPS score below 59 points at the 8-week visit. The decision to identify at-risk participants at week 8 was based on previous research [26], demonstrating no correlation between baseline SEHEPS scores and later adherence, indicating that week 8 is a more reliable point to detect adherence risk. Because the SEHEPS classification was obtained at week 8, randomization at baseline could not be stratified by SEHEPS. Allocation was not altered thereafter, and participants remained in their originally randomized arms. Secondary outcomes included changes in MEP after the intensive EMST phase (weeks 0‐8) and the maintenance phase (weeks 8‐24), as well as changes in SEHEPS scores following the intensive phase (weeks 0‐8). Since the risk of nonadherence was defined at week 8, and all participants proceeded from that visit into the unsupervised phase, we did not exclude those classified as not at risk. Their data from week 8 onward are reported alongside the at-risk group to provide a comprehensive view.

The same assessment protocol was completed at pre- and posttreatment visits. Outcome examinations were carried out at the same time of day in the ON medication state (1 h after regular dopaminergic medication) at the Department of Neurology, First Faculty of Medicine, Charles University and General University Hospital in Prague, Czech Republic.

Adherence during the unsupervised home-training phase (weeks 8‐24) was defined as the cumulative number of completed expiratory maneuvers. The prescribed total was 800 maneuvers. In the experimental group, adherence was monitored using the SpiroGym app, which automatically recorded the number of completed EMST maneuvers. In the control group, adherence was tracked via self-reported training diaries (Multimedia Appendix 2), where participants recorded the number of EMST maneuvers completed.

MEP assessments were performed using a flanged rubber mouthpiece connected to a pressure manometer (Micro RPM, Vyaire Medical). The procedure followed standards for respiratory muscle testing [31]. Seated upright, participants inhaled to total lung capacity and then exhaled forcefully into the manometer. Participants received verbal encouragement throughout the MEP assessment. Nose clips were applied to prevent air leakage through the nose. Participants were explicitly instructed to support their cheeks and lips with their hands during the MEP maneuver, as any cheek bulging and lip leakage can dissipate pressure and result in artificially reduced values [32]. One week before baseline testing, a training MEP session was conducted to minimize learning effects [33]. Before each formal assessment, participants completed a warm-up of 6 MEP maneuvers at 50% maximal effort [34] to account for neural facilitation from repeated efforts [33]. At least 3 measurements with a less than 10% variation between them were taken, and the highest value (cmH₂O) was recorded as the MEP.

The exercise self-efficacy was assessed using the SEHEPS [35]. The test-retest reliability of SEHEPS has been demonstrated as excellent with an intraclass correlation coefficient (ICC) of 0.96 (95% CI 0.91‐0.98) in patients with PD [26]. The minimal detectable change at the 95% confidence limit (MDC₉₅) for SEHEPS is 12 points [35]. A total self-efficacy score is derived by summing responses to 12 items, resulting in a range from 0 to 72, where higher scores indicate greater exercise self-efficacy. A SEHEPS score of less than 59 points identifies individuals at risk of nonadherence (defined as <70% of exercise adherence) to home exercise programs [35]. At baseline, patients completed the SEHEPS based on their confidence in undertaking any long-term home exercise program. At week 8, they completed the SEHEPS again, focusing specifically on their confidence in continuing long-term EMST.

Sample size calculations were based on the primary outcome (adherence to unsupervised EMST). Because no prior controlled studies had compared EMST alone with SpiroGym-assisted EMST, we estimated the effect size from an internal pilot study. We analyzed week 8 to 24 adherence in the first 10 PD participants in each arm with SEHEPS less than 59 at week 8. Their adherence data yielded a Cohen d of 0.92 (unpublished internal data). A power analysis using G*Power (version 3.1.9.3; Heinrich Heine University Düsseldorf) [37] indicated that 16 participants at risk for nonadherence per group would achieve 80% power at a 1-sided 5% type 1 error rate to detect this effect size. Accounting for a 20% attrition rate, each group was thus targeted to include 19 participants.

Safety was assessed by monitoring adverse events throughout the study. During the intensive phase, safety was monitored via bi-weekly in-person visits. During the maintenance phase, participants were instructed to contact the study team by phone if any adverse effects occurred.

Baseline variables were tested for normality (Kolmogorov-Smirnov test). Normally distributed data were compared with independent 2-tailed t tests; nonnormal data with the Mann-Whitney U test. Sex distribution was analyzed by the Fisher exact test. A linear mixed-effects model was used to analyze adherence across time intervals, groups, and SEHEPS subgroups, accounting for repeated measures within individuals. The model formula was:

where group (control/experimental), interval (wk 0-8/8-24), and SEHEPS group (<59/≥59) were included as fixed effects and their interactions, and participant ID was included as a random intercept to account for within-individual correlation. Model assumptions were checked using residual diagnostics (Shapiro-Wilk and Breusch-Pagan tests). Post hoc analyses were performed to assess group differences at each interval. Adherence was compared between groups for weeks 0 to 8 and 8 to 24, and within SEHEPS subgroups (<59 vs ≥59) using the Mann-Whitney U test. Four-week interval (8-12, 12-16, 16-20, and 20-24) adherences were also analyzed using the same approach. MEP changes (0‐8, 8‐24, and 0‐24 wk) were assessed with paired t tests within groups and independent t tests between groups. Effect size was expressed as Cohen d (95 % CI), interpreted as moderate (≥0.5) and large (≥0.8). Associations between adherence and SEHEPS were examined with Spearman ρ. Our sample-size planning used a 1-sided hypothesis based on prior evidence [26] supporting a directional expectation (SpiroGym >control). However, for the inferential analyses, we adopted 2-sided tests to provide a more conservative interpretation. Missing data were examined for randomness using the Little Missing Completely at Random test to determine whether the missingness was random or systematic. The missing data originated from participants who did not complete the therapy. P<.05 was considered statistically significant. All analyses were performed in Python (version 3.12; Python Software Foundation) using the scipy and statsmodels packages.

This study was conducted in accordance with the principles of the Declaration of Helsinki. The study was approved prospectively by the IRB of the General University Hospital in Prague (protocol 223/20 S-IV). All participants provided written informed consent prior to any study procedures, including consent for collection, analysis, and publication of deidentified data. Participation was entirely voluntary, and participants could withdraw at any time without consequence. Participants did not receive any financial compensation. All data were deidentified before analysis and stored in password-protected files accessible only to the research team in accordance with institutional data protection policies. The paper and supplementary materials contained no identifiable images of participants. However, Multimedia Appendix 3 shows an identifiable study researcher (not a participant). Written consent for publication from this individual has been obtained.

No study-related adverse events occurred during the study period. To achieve the target of 19 participants with PD in each arm with low self-efficacy, recruitment continued until this quota was achieved, which did not occur until enrolling 75 individuals. Patients in the control group (n=38) and SpiroGym-assisted EMST group (n=37) were well-matched at baseline. Baseline demographics also did not differ significantly between the at-risk (SEHEPS <59) and not-at-risk (SEHEPS ≥59) subgroups. A summary of demographic and clinical characteristics is provided in Table 1. A total of nine participants from the control group and seven participants from the experimental group did not complete the study (detailed reasons are provided in Figure 1), resulting in missing data. The Little Missing Completely at Random test was nonsignificant (P>.05), indicating that the data were missing completely at random. Over the 24-week period, the mean MDS-UPDRS-III motor score increased by 1.8 (SD 5.2) points in the experimental group and by 1.7 (SD 3.9) points in the control group, with no significant difference between the groups (P=.95).

The primary analysis using a linear mixed-effects model did not show a significant 3-way interaction between group, time interval, and baseline risk for nonadherence (group×interval×SEHEPS group; P=.14; Multimedia Appendix 4). However, a planned secondary analysis of participants at high risk for nonadherence (SEHEPS <59 at week 8) revealed a significant group×interval interaction. In this subgroup, the intervention group showed a significantly smaller decline in adherence during weeks 8 to 24 compared to the control group, as confirmed by a post hoc test (β=496.9, 95% CI 130.7-863.3; P=.008), indicating that the intervention was effective in sustaining adherence specifically in this at-risk population.

Among participants at risk for nonadherence (n=34), adherence during the maintenance phase (wk 8‐24) was significantly higher in the experimental group than in the control group (P=.04; Cohen d=0.80; Figure 2 and Table 2). Within this group, 47% (8/17) of SpiroGym participants versus 18% (3/17) of controls achieved the minimum prescribed 800 or more repetitions during weeks 8 to 24. A 4-week interval analysis revealed that between-group differences in adherence reached significance during weeks 12 to 16 (P=.04; Cohen d=0.84) and weeks 20 to 24 (P=.01; Cohen d=1.03), with no significant differences during weeks 8 to 12 or 16 to 20 (Figure 3). Among patients (n=25) without risk for nonadherence, no difference in adherence was observed during the maintenance period between the control and experimental groups (P=.34; Cohen d=0.22; Table 2; Multimedia Appendix 5). For the overall sample (n=59), adherence during the intensive phase (wk 0‐8) did not differ significantly between groups (P=.05; Cohen d=0.49), while during the maintenance phase (wk 8‐24), the experimental group achieved significantly greater adherence (P=.02; Cohen d=0.58; Table 2).

This phase 2 RCT compared the efficacy of a conventional EMST protocol with a SpiroGym-assisted protocol. For the primary objective, participants at risk for EMST nonadherence who trained with SpiroGym sustained greater adherence during the unsupervised phase than those following conventional EMST. Regarding secondary objectives, the SpiroGym group achieved greater improvements in MEP and a clinically meaningful increase in self-efficacy, whereas the control group resulted in smaller but still significant gains.

Building on earlier findings [26] demonstrating sustained user engagement with the SpiroGym app during long-term EMST, our study adds further evidence that SpiroGym can effectively support adherence over time. Notably, participants with sufficient self-efficacy (SEHEPS ≥59) performed 30% more expiratory maneuvers than the minimum prescribed target during the maintenance period in both groups. In contrast, among those with low self-efficacy (SEHEPS <59), adherence in the control group dropped from 83% (wk 8‐12) to 37% (wk 20‐24), while the SpiroGym group consistently (every 4-wk interval) maintained adherence 20% above the prescribed maneuver target. These findings align with Chung et al [38], who reported higher adherence and enhanced self-efficacy in an mHealth-supported exercise program compared to a paper-based approach. Similarly, Sørensen et al [39] reported that feedback-based inspiratory muscle training yielded better adherence than self-reported protocols over a 12-week unsupervised period.

In line with our prior study [26], we observed a positive correlation between SEHEPS scores at week 8 and adherence during weeks 8 to 24 in both groups, suggesting that self-efficacy assessed after the intensive phase may predict long-term EMST compliance. Although SEHEPS scores improved significantly in both protocols, only the SpiroGym group exceeded the 12-point MDC₉₅ threshold [35], indicating a true increase in exercise self-efficacy beyond measurement error—a finding supported by previous research [40].

These findings have direct clinical implications for tailoring EMST. Patients should be assessed for self-efficacy following the intensive phase. Those with low self-efficacy may benefit from enhanced supervision or motivational support, such as the SpiroGym app, to sustain adherence during the maintenance period. Conversely, conventional EMST may be sufficient for patients with higher self-efficacy. Beyond self-efficacy, other factors such as apathy or caregiver support may plausibly influence adherence and treatment outcomes and may guide treatment selection in practice [41,42]. Screening for these factors alongside SEHEPS may help triage patients to conventional versus supported EMST.

Another key finding relates to real-time visual feedback and its impact on MEP. After the 8-week intensive phase, both groups achieved meaningful MEP gains, but the increase was significantly larger in the SpiroGym group (P=.03; Cohen d=0.54). On average, MEP increased by 22% in the control group and 31% in the experimental group, consistent with prior reports of approximately 24% MEP gains in PD [8]. Based on the observation that, during weeks 0‐8, the experimental group showed greater MEP gains despite comparable adherence, we hypothesize that real-time visual feedback may enhance motivational aspects of learning and thereby facilitate better skill acquisition. This interpretation is consistent with prior studies showing amplified respiratory muscle activation under visual feedback conditions [43]. Given the role of EMST in airway protection [4,6], these superior MEP outcomes suggest that SpiroGym-assisted EMST may translate into improved clinical outcomes for individuals with dysphagia or impaired cough, although this remains to be confirmed in future research.

A further notable result pertains to long-term MEP maintenance in patients at risk for nonadherence (SEHEPS <59). While the control group showed an average 9% decline in MEP between weeks 8 and 24, the SpiroGym group maintained stable values. Previous studies report detraining-related MEP declines of 2% to 20% over 1 to 18 months in PD [9,44,45]. Long-term maintenance training at least twice per week is generally recommended [46]. In our study, control group participants missed an average of 11 out of 32 scheduled sessions, falling below the minimum recommended dose and likely contributing to the observed MEP reduction. In contrast, adherence in the experimental group was sufficient to prevent detraining effects. On the other hand, resistance-training adaptations typically extend beyond 8 weeks in both healthy adults and people with PD. One likely explanation for the plateau in MEP gains is that no participant requested recalibration during the maintenance period. As a result, training thresholds remained fixed, reducing progressive resistance. This contrasts with evidence from a 2-year randomized trial, in which Corcos et al [47] showed that people with PD who trained twice weekly with progressively increased resistance continued to improve upper-limb strength up to the 24-month visit, whereas those following a nonprogressive program did not. A second contributor may be the device load ceiling. In a subset of experimental participants (n=11), week-8 MEP exceeded 200 cmH_₂O, placing the 75% training target above the EMST150’s maximum load. To sustain MEP progression, future protocols should implement routine recalibration during the maintenance period and, for participants approaching the ceiling, employ devices with a higher resistance range.

This study has several limitations. First, adherence in the control group was based on self-reported training diaries, which might be prone to inaccuracy [48]. Second, we did not include objective measures of swallowing function or cough efficacy, limiting conclusions about the broader clinical impact of the observed increase in the MEP. Third, the sample primarily consisted of individuals with mild-to-moderate PD, which limits the generalizability of findings to those with more advanced PD or cognitive impairment. Fourth, we did not collect person-centered outcomes (eg, treatment burden and quality of life). Although a separate study reports acceptability of SpiroGym, it does not provide a head-to-head comparison with conventional EMST [25,26]. Consequently, it remains unclear whether SpiroGym-supported EMST reduces perceived burden or improves quality of life relative to standard care. Fifth, during the maintenance phase, participants were instructed to perform five sets of five expirations twice weekly, with the option to train more frequently to reflect real-world practice. This flexibility allowed some individuals to accumulate higher totals of expiratory maneuvers, which may have influenced the effect size.

Future studies should involve larger cohorts, longer follow-up periods, and objective assessments of airway protection. In addition, qualitative and mixed methods studies comparing SpiroGym with traditional EMST are warranted to understand patient experience (eg, satisfaction, motivation, enjoyment, and usability).

To support further research and clinical adoption, the SpiroGym app is freely available at [36], along with detailed hardware specifications. A user manual is included in Multimedia Appendix 6.

This is the first RCT to integrate mHealth with EMST. Unlike prior studies in the EMST field, we focused on long-term exercise adherence, which is critical for sustaining therapeutic effects. Compared with conventional EMST, SpiroGym-assisted training led to higher adherence, most notably among patients prone to nonadherence, and produced greater expiratory muscle strength gains. In routine PD care, evaluating self-efficacy after the supervised EMST phase may help identify patients who are likely to benefit from digital support, positioning mHealth-assisted EMST as a practical strategy to sustain exercise adherence.