Respiratory-Swallow Training and Variable Practice in Parkinson’s Disease: A Clinical Trial Pilot Study

Abstract

Introduction

Dysphagia in people with Parkinson’s disease (PwPD) is closely linked to impairments in respiratory-swallow coordination. Respiratory-swallow training (RST) is a promising, mechanism-based intervention for dysphagia, but, to date, only a single-subject study has examined RST in PwPD. As an early-stage next step in refining RST for PwPD, this clinical trial pilot study aimed to: (1) explore the preliminary associations between exhale-swallow-exhale (EX-EX) RST and swallowing safety, swallowing efficiency, and respiratory phase patterning; and (2) explore differences in practice structure (constant vs. variable practice) on RST-related motor learning outcomes.

Methods

PwPD completed pre-RST assessments, four once-weekly sessions of RST targeting only the EX-EX pattern, and post-RST assessments. Participants were randomized to constant practice or variable practice training conditions. Outcomes were measured using nasal airflow, respiratory inductive plethysmography, submental surface electromyography, and flexible endoscopic evaluation of swallowing. Multilevel statistical models were used to evaluate changes in EX-EX usage, pharyngeal residue, penetration, and aspiration.

Results

Twelve PwPD completed all study phases, yielding an analysis of 1,565 swallows. In this pilot sample, EX-EX usage was higher post-RST compared to pre-RST (p < 0.05), with EX-EX usage observed to be higher at 1-month follow-up in the variable practice group compared to the constant practice group (p < 0.05). Additionally, pharyngeal residue, laryngeal vestibule residue, vocal fold residue, and Penetration-Aspiration Scale scores were lower post-RST compared to pre-RST (p < 0.05).

Conclusion

This clinical trial pilot study provides preliminary evidence that RST may be associated with increased EX-EX usage, swallowing safety, and swallowing efficiency in PwPD. Preliminary evidence also suggests that variable practice may enhance retention and generalization of EX-EX usage, warranting further evaluation. Findings from this study will be used to inform refinement of RST for PwPD and the design of future confirmatory randomized clinical trials.

Introduction

Parkinson’s disease is the second most common neurodegenerative disorder worldwide and is projected to affect more than 25.2 million adults by 2050 [1]. It is classically characterized by degeneration of dopaminergic neurons within the basal ganglia, resulting in cardinal motor symptoms of tremor, rigidity, bradykinesia, and akinesia [2]. However, Parkinson’s disease involves widespread neurodegeneration beyond the basal ganglia, affecting cortical and brainstem regions that support motor learning, motor control, somatosensory processing, cognition, breathing, and swallowing [3–10].

Dysphagia is highly prevalent in people with Parkinson’s disease (PwPD), affecting up to 82% of individuals and often emerging in early disease progression [11–13]. Although dysphagia in PwPD is multifactorial, impairments in motor control are thought to play a central role, given the disease’s primary impact on basal ganglia circuitry and distributed motor networks [14–17].

One motor behavior critical for safe and efficient swallowing is respiratory-swallow coordination. This behavior comprises three related components: respiratory phase patterning (the unique grouping of inhalations and exhalations before and after swallowing), lung volume during swallowing (where within the lung volume range swallowing occurs), and respiratory pause duration (the amount of time breathing temporarily stops to accommodate the swallow). In healthy adults, respiratory-swallow coordination is typically characterized by an exhale-swallow-exhale (EX-EX) pattern, initiated at mid-lung volume, with a brief pause of approximately 1 s [18–25]. These patterns are hypothesized to facilitate safe and efficient swallowing by supporting timely laryngeal closure, effective pharyngeal constriction, and favorable pressure gradients between the pharynx and esophagus [26–29].

In PwPD, respiratory-swallow coordination is frequently disrupted. Compared to healthy adults, PwPD more often inhale before and after swallowing, initiate swallows at lower lung volumes, and exhibit longer respiratory pauses [30–34]. Importantly, these alterations – particularly atypical respiratory phase patterning – have been shown to be strongly associated with impaired swallowing safety (penetration and aspiration) and efficiency (pharyngeal residue) in PwPD, suggesting a mechanistic link between respiratory-swallow coordination and dysphagia in this population [35–37].

Respiratory-swallow training (RST) is a treatment paradigm designed to improve swallowing by targeting respiratory-swallow coordination behaviors, most commonly the EX-EX pattern. The effects of RST were first described in individuals with head and neck cancer, where 4–8 sessions of an RST approach that combined respiratory phase patterning and lung volume during swallowing resulted in improvements in swallowing physiology, pharyngeal residue, and penetration-aspiration outcomes [28]. This work provided the foundation for subsequent applications of RST in neurogenic populations [38, 39].

In PwPD, respiratory-swallow coordination has been shown to be immediately modifiable using behavioral cues, providing proof-of-concept that this behavior can be altered [30, 31]. Building on this evidence, a prior single-subject treatment study combining RST (targeting respiratory phase patterning and lung volume during swallowing) and cough skill training demonstrated large improvements in swallowing safety and efficiency in a person with mid-stage Parkinson’s disease and severe dysphagia [39]. While promising, the single-subject design limits generalizability, and key aspects of the RST paradigm remain untested in PwPD – including whether the structure of practice influences RST-related motor learning outcomes.

Because RST is fundamentally a skill-based intervention, the structure of practice is likely to influence learning, retention, and generalization. Motor learning frameworks describe several modifiable practice parameters (e.g., dose, schedule, variability, feedback) and consistently show that variable practice – which involves practicing multiple task variations – often promotes greater learning, retention, and generalization than constant practice, which involves repetition of a single task variant [40]. However, these principles have not been systematically examined in relation to swallowing rehabilitation or RST in PwPD.

Given these gaps, there is a clear need for early-phase, hypothesis-generating research to refine the RST paradigm for PwPD and to determine whether key components of the protocol – such as varying the structure of practice – are potentially impactful. Accordingly, the overall objective of this clinical trial pilot study was to preliminary evaluate an RST approach focused solely on EX-EX patterning in a small cohort of PwPD, generating initial effect estimates to inform the design of future confirmatory clinical trials. To meet this objective, we had two aims. Aim 1 was to explore the preliminary associations between EX-EX focused RST and swallowing safety, swallowing efficiency, and respiratory phase patterning in a small cohort of PwPD. Aim 2 was to explore whether practice structure (constant versus variable practice) differentially influenced acquisition, retention, or generalization of respiratory phase patterning to guide refinement of RST for future confirmatory clinical trials in PwPD.

Methods

Participants

All PwPD were prospectively recruited for study enrollment and were blinded to the study aims. Participants were recruited through the Laboratory for the Study of Upper Airway Dysfunction and via referrals from local neurologists and Parkinson’s disease support groups in the New York City metropolitan area. Inclusion criteria were as follows: (1) a diagnosis of Parkinson’s disease from a fellowship-trained movement disorders neurologist using UK’s Parkinson’s Disease Society Brain Bank diagnostic criteria; (2) EX-EX on < 85% of swallows during the baseline flexible endoscopic evaluation of swallowing (FEES); and (3) penetration or aspiration on ≥1 trial during the baseline FEES. Exclusion criteria were as follows: (1) neurologic disease other than Parkinson’s disease; (2) history of head/neck/lung cancer; (3) history of respiratory disease not otherwise associated with Parkinson’s disease; and (4) current enrollment in swallowing therapy. The <85% threshold reflects the upper bound of EX-EX frequency in healthy adults [18] and was selected to ensure that all enrolled participants demonstrated at least some departure from the optimal EX-EX pattern.

Demographic, disease, and dysphagia-related information were collected at the first assessment. Parkinson’s disease severity was characterized using the Hoehn and Yahr Stage (HYS) and part III of the Unified Parkinson’s Disease Rating Scale (UPDRS-III). Swallow severity was characterized using the Functional Oral Intake Scale (FOIS) [41], the Eating Assessment Tool (EAT-10) [42], and the Dynamic Imaging Grade of Swallowing Toxicity (DIGEST) for Flexible Endoscopic Evaluation of Swallowing (DIGEST-FEES) [43, 44]. DIGEST-FEES grades (0–4) were derived using the Visual Analysis of Swallowing Efficiency and Safety (VASES) [45, 46].

Study Design

This prospective clinical trial pilot study included two pre-RST assessment visits, four once-weekly in-person RST sessions, and two post-RST assessment visits (Fig. 1). Participants were randomized 1:1 to constant practice or variable practice RST. Randomization used a computer-generated, permuted block sequence (block size = 6; three allocations per arm per block) applied sequentially to consecutively enrolled participants; the list was accessed only at enrollment to prevent foreknowledge of assignment. Participants were aware they were enrolling in a swallowing treatment study but were blinded to the aim comparing constant and variable practice. Participants taking PD-related medication confirmed being in the medication “on” state at each visit. An additional objective of this clinical trial pilot study was to evaluate the feasibility of training graduate speech-language pathology (SLP) clinicians to deliver RST with acceptable fidelity using structured performance bandwidth rules and clinician-perceptual monitoring, without instrumented biofeedback.

Fig. 1.

Study design.

Clinician Roles and Blinding

Five SLP graduate student interns were trained by the first author to conduct the assessments and deliver treatment. Each clinician served in either an assessing or treating role for any given participant; no clinician performed both roles for the same participant, ensuring separation between those blinded to time point and those delivering the intervention. Assessing clinicians were blinded to visit type (pre vs. post) and practice allocation; treating clinicians were blinded to pre-/post-assessment findings.

Treating clinicians used performance bandwidth rules (online suppl. material A; for all online suppl. material, see https://doi.org/10.1159/000552197) to guide the type, timing, and frequency of verbal feedback. All sessions were video recorded. The first author reviewed a 20% sample per clinician monthly to calculate treatment fidelity (protocol adherence), defined as: [# trials with protocol-concordant feedback ÷ total observed trials] × 100, based on audio/video review (feedback content, timing, and dose).

RST

The RST protocol followed core principles described in prior RST work [28] but was simplified to prioritize consistent EX-EX performance and minimize extraneous task demands in PwPD. Lung volume during swallowing was not trained because current evidence does not show a clear relationship between lung volume during swallowing and functional swallowing outcomes [35]. Sessions were delivered once weekly for four consecutive weeks.

Session 1 began with ∼10 min of education and awareness building. Education included three parts: (1) the fundamentals of RSC and its proposed relevance; (2) a general description of the participant’s baseline patterning (inhaling before and/or after swallowing more often than typical); and (3) the training goal (to breathe out before and after swallowing during eating/drinking). Awareness building included guided awareness of breathing movements (hand on chest/abdomen to notice ribcage/abdominal motion), coached performance of four respiratory phase patterns using a breath-hold in place of a swallow (exhale-hold-exhale; inhale-hold-exhale; exhale-hold-inhale; inhale-hold-inhale) with brief reflective prompts, and repetition of the same sequences while incorporating a saliva swallow.

The remainder of session 1 and all subsequent sessions involved completing 60 trials of EX-EX during natural sips of water (self-administered, self-selected volumes of room temperature tap water while seated upright). Practice was organized into 12 sets/session with 5 trials/set. The constant practice used a cup for all 12 sets. The variable practice group used cup (four sets), straw (four sets), and spoon (four sets), with set order randomized. Consistent with prior RST protocols, the thinnest liquid consistency without aspiration during the pre-RST FEES was used for training [28]. If aspiration occurred on thin, mildly thick, and moderately thick liquids, training was completed using saliva swallows only. Treating clinicians sat across participants and provided instructions, prompts, and verbal feedback based on real-time, visual-perceptual judgments of respiratory phase patterning. No home practice sets were prescribed; however, participants were encouraged to use the EX-EX during daily swallowing.

Pre- and Post-RST Assessments

The assessment schedule was designed to evaluate both the feasibility of the study protocol in terms of participant burden and retention and to generate preliminary outcome estimates across clinically relevant time points. Swallowing safety, swallowing efficiency, and RSC were assessed at four time points: 1-month pre-RST, immediate pre-RST (same day as session 1), immediate post-RST (same day as session 4), and 1-month post-RST (Fig. 1). The 1-month pre-RST served as a no-treatment within-subject control phase (test-retest stability). The 1-month post-RST served as a no-treatment retention phase.

Assessment of Swallowing Safety and Efficiency

FEES was used to assess swallowing safety and efficiency. FEES were conducted by trained SLP graduate interns following hands-on transnasal endoscopy practice in normal and dysphagic adults until procedural independence was demonstrated. Equipment included a 3.0 mm diameter flexible distal chip laryngoscope (ENT-5000; Cogentix Medical, New York, NY, USA) and video system with integrated LED light source LCD display (Cogentix Medical, DPU-7000A). The system was white balanced at the start of each exam. The scope was passed transnasally without topical anesthetics or vasoconstrictors; water was used to lubricate the scope. Participants were seated upright and instructed to avoid talking unless needed; coughing/throat clearing was permitted as usual.

A standardized FEES protocol (23 trials across 11 swallowing conditions) was planned, including thin and mildly thick liquids, purees, and mixed consistencies across volumes (online suppl. material B). If aspiration occurred on mildly thick liquids, moderately thick liquids were also tested.

Videos were recorded and analyzed offline using VASES [45, 47]. Safety outcomes included ratings of laryngeal vestibule residue (0–100), vocal fold residue (0 to ≥100), subglottic residue (0 to ≥100), and Penetration-Aspiration Scale (PAS; 1–8) [48]. Efficiency outcomes included oropharyngeal residue (0 to ≥100) and hypopharyngeal residue (0 to ≥100). Residue ratings of 0 indicate “absent,” whereas >0 indicates “present,” with higher values reflecting a greater (worse) amount of residue.

Two independent raters (SLP graduate interns trained in VASES) rated clips while blinded to participant identity and time point. For each clip, raters assigned PAS and residue ratings (oropharynx, hypopharynx, laryngeal vestibule, vocal folds, subglottis). Agreement was defined as: (a) PAS – exact match; and (b) residue – agreement on presence vs. absence (0 vs. >0) and, if present, absolute difference by ≤10 percentage points. When agreement criteria were met, residue ratings were averaged. Disagreements (any PAS mismatch, 0 vs. >0 mismatch, or >10-point residue difference) were resolved by the first author, who was blinded to the paired ratings and all participant/treatment information.

Assessment of Respiratory Phase Patterning

Two swallow sets were used to support distinct aims: (1) non-FEES swallows (10 self-selected sips of clear water via cup, no endoscope in situ), to mirror training and assess task-specific changes; and (2) FEES swallows (23 standardized trials varying bolus volume, consistency, and delivery; online suppl. material B), to assess generalization to similar but untrained tasks while concurrently measuring swallowing safety and efficiency. The thinnest liquid consistency without evidence of aspiration on the immediate Pre-RST FEES was used for non-FEES swallows.

RSC was assessed using respiratory inductive plethysmography (Respitrace System; Ambulatory Monitoring, Inc.), nasal airflow (nasal cannula coupled to a spirometry pod; ADInstruments), and submental surface electromyography (Octal Bio Amps, ADInstruments). Signals were digitized and synchronized using PowerLab (PL3516 PowerLab 16/35; ADInstruments), with plethysmography signals at 2 kHz, airflow at 4 kHz, and EMG signals at 10 kHz. LabChart software (LabChart v8; ADInstruments) was used for acquisition and analysis. Signals were synchronized with FEES by pairing an external camera (Microsoft LifeCam Cinema HD) to LabChart. Synchronization was calibrated using LabChart’s built-in video time-alignment function by directing the FEES endoscope toward the LabChart display while black-white synchronization markers flashed at a fixed temporal rate, enabling automatic correction of FEES-to-physiological signal latency.

Semi-automated RSC analysis was performed using LabChart and R [49], yielding respiratory pause duration, respiratory phase patterning, and lung volume during swallowing. Only respiratory phase patterning was analyzed inferentially because it was the sole behavior explicitly trained in this RST study. Patterns were coded binary as either EX-EX or non-EX-EX.

Statistical Analyses

Analyses were conducted using R version 4.4.3 [50]. Multilevel models were fit with time (assessment visit) as a fixed effect and participant as a random intercept to account for repeated swallows within individuals. Model assumptions were evaluated for each model (distribution, link, residual diagnostics) with no violations observed. Consistent with early-phase pilot study objectives, statistical models were used to generate preliminary effect size estimates – specifically odds ratios (ORs) with 95% confidence intervals – to inform the design and power calculations of future confirmatory trials, rather than to establish efficacy or test prespecified hypotheses. The absence of correction for multiple comparisons, while appropriate given exploratory intent of this study [51, 52], further underscores that all findings should be interpreted as provisional and hypothesis-generating. Three planned contrasts were tested: (1) delayed baseline (1-month pre-RST vs. immediate pre-RST), (2) treatment (immediate pre-RST vs. immediate post-RST), and (3) retention (immediate post-RST vs. 1-month post-RST). Six models were fit per contrast (one per outcome). Statistical significance was set at p < 0.05. ORs and their 95% confidence intervals were used as standardized effect sizes for all statistically significant findings.

For airway residue outcomes (laryngeal vestibule, vocal folds, subglottis), multilevel zero-inflated beta regressions modeled (a) the probability (0 vs. >0) and (b) amount when present, reflecting the clinical distinction between whether material entered the airway and how much. For pharyngeal structures (oropharynx, hypopharynx), multilevel beta regressions modeled amount only. Percent outcomes were rescaled to 0, 1. Before rescaling, values of 0 were set to 0.001 and values ≥100 were set to 99.999, only when required by the beta family modeling. PAS was treated as an 8-point ordinal scale and was modeled using an ordinal cumulative-link mixed model.

Respiratory phase patterning was treated as a binary variable and was modeled using multilevel binomial logistic regression models to evaluate changes in EX-EX frequency across all FEES and non-FEES swallows. To explore practice effects, the multilevel binomial logistic regression models tested the practice condition (constant vs. variable) by time interaction. For practice comparison, EX-EX was analyzed separately for non-FEES swallows (task-specific learning) and FEES swallows (generalization).

Interrater Reliability

Reliability analyses were conducted to evaluate the consistency of outcome measurement procedures, an important feasibility consideration for confirmatory trial planning. Twenty percent of the final swallow-level dataset – individually coded swallowing events selected randomly and stratified across assessment time points and participants – were re-rated by the first author, who was independent of the original raters for those events, for interrater reliability analyses. Full reliability coefficients and 95% confidence intervals are reported in online supplementary material C. Intraclass correlation coefficients were used to assess reliability for the five VASES residue outcomes using a mean rating (k = 1), absolute agreement, two-way mixed-effects model. Reliability of EX-EX patterning was assessed using Cohen’s Kappa (κ). PAS reliability was assessed with quadratic weighted kappa (κ_w) to reflect the ordered scale and to weight larger disagreements more heavily than adjacent ones.

Results

Study enrollment began November 1, 2019 and was discontinued early on March 6, 2020, due to the onset of COVID-19 and subsequent institutional closure. During this period, 24 PwPD were screened, 21 met eligibility criteria and enrolled, 8 participants were unable to start RST due to institutional shutdown, and 1 participant reported a change in diagnosis (progressive supranuclear palsy) at their final study visit. Therefore, 12 participants had sufficient data for inclusion in analyses.

All 12 participants completed the immediate post-RST assessment prior to institutional closure; seven also completed the 1-month post-RST retention assessment. No other dropouts or voluntary withdrawals were noted. Six participants were randomized to variable practice and six participants were randomized to constant practice. All participants completed all four training sessions and all planned repetitions. Four participants aspirated on thin liquids during pre-RST FEES and therefore completed RST using mildly thick liquids.

Participant characteristics are summarized in Table 1. Most participants presented with early-to-mid stage Parkinson’s disease (HYS II–IV) and moderate dysphagia severity (DIGEST Grade 2). Across the two pre-RST assessments, participants exhibited EX-EX for 31.4–38.4% of swallows, consistent with prior PwPD findings [34] and well below typical rates in healthy adults [18].

Table 1.

Demographic, disease, and dysphagia baseline data

​	Constant (N = 6)	Variable (N = 6)	Overall (N = 12)
Age, years
Mean (SD)	68.0 (14.2)	65.7 (5.65)	66.8 (10.4)
Median [Q1, Q3]	69.0 [58.0, 79.3]	67.0 [63.5, 69.8]	67.0 [60.0, 72.5]
Min, max	49.0, 84.0	56.0, 71.0	49.0, 84.0
Sex, n (%)
Male	5 (83.3)	5 (83.3)	10 (83.3)
Female	1 (16.7)	1 (16.7)	2 (16.7)
Disease duration, years
Mean (SD)	8.25 (5.17)	10.7 (5.04)	9.46 (5.03)
Median [Q1, Q3]	7.00 [5.25, 8.75]	10.0 [7.63, 14.6]	8.00 [5.75, 12.9]
Min, max	3.50, 18.0	4.00, 17.0	3.50, 18.0
UPDRSPart 3 (0–132)
Mean (SD)	30.0 (14.6)	37.5 (11.8)	33.8 (13.3)
Median [Q1, Q3]	26.5 [21.5, 32.3]	37.5 [34.3, 43.8]	33.5 [22.5, 41.3]
Min, max	16.0, 57.0	18.0, 53.0	16.0, 57.0
Hoehn and Yahr Stage (1–5), n (%)
1 – early stage	1 (16.7)	0 (0)	1 (8.3)
2	4 (66.7)	2 (33.3)	6 (50.0)
3 – mid stage	0 (0)	2 (33.3)	2 (16.7)
4	1 (16.7)	2 (33.3)	3 (25.0)
5 – late stage	0 (0)	0 (0)	0 (0)
DIGEST Safety Grade (0–4), n (%)
0 – no impairment	0 (0)	0 (0)	0 (0)
1 – mild impairment	1 (16.7)	2 (33.3)	3 (25.0)
2 – moderate impairment	3 (50.0)	2 (33.3)	5 (41.7)
3 – severe impairment	1 (16.7)	1 (16.7)	2 (16.7)
4 – profound impairment	1 (16.7)	1 (16.7)	2 (16.7)
DIGEST Efficiency Grade (0–4), n (%)
0 – no impairment	0 (0)	0 (0)	0 (0)
1 – mild impairment	5 (83.3)	3 (50.0)	8 (66.7)
2 – moderate impairment	1 (16.7)	1 (16.7)	2 (16.7)
3 – severe impairment	0 (0)	2 (33.3)	2 (16.7)
4 – profound impairment	0 (0)	0 (0)	0 (0)
DIGEST Total Grade, n (%)
0 – no impairment	0 (0)	0 (0)	0 (0)
1 – mild impairment	1 (16.7)	0 (0)	1 (8.3)
2 – moderate impairment	3 (50.0)	4 (66.7)	7 (58.3)
3 – severe impairment	2 (33.3)	1 (16.7)	3 (25.0)
4 – profound impairment	0 (0)	1 (16.7)	1 (8.3)

SD, standard deviation; Q1, 25th percentile; Q3, 75th percentile; UPDRS_{Part 3}, part 3 of the Unified Parkinson’s Disease Rating Scale; DIGEST, Dynamic Imaging Grade of Swallowing Toxicity.

Treatment fidelity ranged from 94.3% to 100% across the five treating clinicians, suggesting that RST can be delivered with high protocol adherence by trained SLP clinicians. Reliability coefficients for FEES outcomes were 0.73 (hypopharyngeal residue), 0.87 (oropharyngeal and vocal fold residue), 0.91 (PAS), and ≥0.97 (laryngeal vestibule and subglottic residue). Reliability coefficients for RSC outcomes were ≥0.98 for all outcome measures (online suppl. material C).

RST and Swallowing Safety

No changes in laryngeal vestibule residue were observed during the delayed baseline (pre-RST) no-treatment control phase (Fig. 2; Tables 2, 3). However, during this phase, amount (but not presence) of vocal fold residue was observed to decrease over time (OR = 0.49; 95% CI: 0.35–0.71; p = 0.0001).

Fig. 2.

Swallowing safety. For the laryngeal vestibule (a) and vocal fold (b) residue figures, ORs < 1.0 represent improved swallowing safety. This includes lower odds of residue being present (yellow) and/or lower odds of a high residue rating when residue was present (blue). For the Penetration-Aspiration Scale figure (c), odds ratios < 1.0 also represent improved swallowing safety (i.e., lower odds of higher [worse] PAS). Figures display odds ratio point estimates (dots) and their 95% confidence intervals. Confidence intervals that cross the 1.0 marker are considered effects that are not statistically significant. With each figure, the Y-axis displays findings changes in outcomes at the immediate pre-RST assessment relative to the 1-month pre-RST assessment (a), immediate post-RST assessment relative to the immediate pre-RST assessment (b), and the 1-month post-RST assessment relative to the immediate post-RST assessment.

Table 2.

Swallowing safety and efficiency results – full cohort (n = 12)

​	1-month	Immediate	Immediate
	pre-RST (N = 255)	pre-RST (N = 255)	post-RST (N = 255)
Oropharyngeal residue
Median [Q1, Q3]	9.00 [6.00, 18.0]	8.00 [4.00, 14.0]	5.00 [3.00, 9.00]
Hypopharyngeal residue
Median [Q1, Q3]	6.00 [4.00, 11.0]	5.00 [3.00, 9.00]	4.00 [2.00, 5.00]
Laryngeal residue
Absence vs. presence, n (%)
Absent	128 (50.2)	123 (48.2)	161 (63.1)
Present	117 (45.9)	117 (45.9)	82 (32.2)
Amount when present
Median [Q1, Q3]	11.0 [5.00, 28.0]	7.00 [4.00, 22.0]	6.00 [4.00, 10.8]
Vocal fold residue
Absence vs. presence, n (%)
Absent	175 (68.6)	171 (67.1)	200 (78.4)
Present	70 (27.5)	69 (27.1)	43 (16.9)
Amount when present
Median [Q1, Q3]	18.5 [7.00, 36.3]	7.00 [4.00, 18.0]	7.00 [4.00, 11.5]
Subglottic residue
Absence vs. presence, n (%)
Absent	206 (80.8)	210 (82.4)	218 (85.5)
Present	39 (15.3)	30 (11.8)	25 (9.8)
Amount when present
Median [Q1, Q3]	19.0 [8.50, 36.0]	9.00 [5.25, 36.5]	15.0 [7.00, 19.0]
Penetration-Aspiration Scale (PAS), n (%)
1 – no airway invasion	121 (47.5)	118 (46.3)	156 (61.2)
2 – shallow penetration, ejected	6 (2.4)	3 (1.2)	4 (1.6)
3 – shallow penetration, not ejected	48 (18.8)	55 (21.6)	41 (16.1)
4 – deep penetration, ejected	1 (0.4)	1 (0.4)	0 (0)
5 – deep penetration, not ejected	28 (11.0)	35 (13.7)	19 (7.5)
6 – aspiration, ejected	2 (0.8)	2 (0.8)	0 (0)
7 – aspiration, not ejected	24 (9.4)	9 (3.5)	9 (3.5)
8 – silent aspiration	15 (5.9)	17 (6.7)	14 (5.5)

This table presents data from the 12 people with Parkinson’s disease who completed the 1-month pre-RST, immediate pre-RST, and immediate post-RST assessments, used for statistical analyses when examining changes in the pre-RST no-treatment control phase and the treatment phase.

Table 3.

Swallowing safety and efficiency results – partial cohort (n = 7)

​	Immediate	1-month
	post-RST (N = 141)	post-RST (N = 142)
Oropharyngeal residue
Median [Q1, Q3]	6.00 [3.00, 11.0]	5.00 [3.00, 10.0]
Hypopharyngeal residue
Median [Q1, Q3]	4.00 [2.00, 5.00]	3.00 [1.00, 6.00]
Laryngeal residue
Absence vs. presence, n (%)
Absent	84 (59.6)	86 (60.6)
Present	57 (40.4)	51 (35.9)
Amount when present
Median [Q1, Q3]	6.00 [4.00, 12.0]	6.00 [3.00, 7.50]
Vocal fold residue
Absence vs. presence, n (%)
Absent	109 (77.3)	106 (74.6)
Present	32 (22.7)	31 (21.8)
Amount when present
Median [Q1, Q3]	7.00 [5.00, 9.50]	6.00 [4.50, 8.50]
Subglottic residue
Absence vs. presence, n (%)
Absent	121 (85.8)	118 (83.1)
Present	20 (14.2)	19 (13.4)
Amount when present
Median [Q1, Q3]	15.0 [7.75, 17.5]	6.00 [4.00, 13.0]
Penetration-Aspiration Scale (PAS), n (%)
1 – no airway invasion	82 (58.2)	85 (59.9)
2 – shallow penetration, ejected	1 (0.7)	0 (0)
3 – shallow penetration, not ejected	27 (19.1)	21 (14.8)
4 – deep penetration, ejected	0 (0)	0 (0)
5 – deep penetration, not ejected	11 (7.8)	14 (9.9)
6 – aspiration, ejected	0 (0)	1 (0.7)
7 – aspiration, not ejected	7 (5.0)	4 (2.8)
8 – silent aspiration	13 (9.2)	12 (8.5)

This table presents data from the 7 people who completed all assessments. Data from their immediate and 1-month post-RST assessments were used to analyze changes in the retention phase.

Laryngeal vestibule residue, vocal fold residue, and PAS ratings were observed to be smaller (better) for immediate post-RST compared to immediate pre-RST. Specifically, the odds of laryngeal vestibule residue being present vs. absent were lower (better) for immediate post-RST compared to immediate pre-RST (OR = 0.71; 95% CI: 0.56–0.90; p = 0.0006), and when residue was present, the amount was also lower (OR = 0.51; 95% CI: 0.40–0.64; p = 0.005). Similarly, the odds of vocal fold residue being present were lower (better) for immediate post-RST (OR = 0.49; 95% CI: 0.35–0.68; p = 0.002), and when present, the amount also tended to be less (OR = 0.71; 95% CI: 0.52–0.99; p = 0.045). PAS scores were also lower (better) at immediate post-RST (OR = 0.55; 95% CI: 0.38–0.79; p = 0.001).

No statistically significant differences were observed between immediate post-RST and 1-month post-RST for laryngeal vestibule residue, vocal fold residue, or PAS, suggesting maintenance of these post-training improvements over the 1-month retention interval. Subglottic residue did not differ across any comparison (p > 0.05; online suppl. material D).

RST and Swallowing Efficiency

No exploratory changes in oropharyngeal or hypopharyngeal residue were observed during the delayed baseline no-treatment phase, suggesting stability in their swallowing efficiency. In contrast, pharyngeal residue ratings were observed to be lower (better) at immediate post-RST compared to immediate pre-RST (Fig. 3; Tables 2, 3). These exploratory observations included smaller amounts of both oropharyngeal residue (OR = 0.66; 95% CI: 0.58–0.75; p < 0.0001) and hypopharyngeal residue (OR = 0.60; 95% CI: 0.52–0.70; p < 0.0001). Furthermore, no statistically significant differences in pharyngeal residue were observed between immediate post-RST and 1-month post-RST.

Fig. 3.

Swallowing efficiency. For the oropharyngeal (a) and hypopharyngeal (b) residue figures, odds ratios < 1.0 represent improved swallowing efficiency (i.e., lower odds of having a high residue rating). Figures display odds ratio point estimates (dots) and their 95% confidence intervals. Confidence intervals that cross the 1.0 marker are considered effects that are not statistically significant. With each figure, the Y-axis displays findings changes in outcomes at the immediate pre-RST assessment relative to the 1-month pre-RST assessment (a), immediate post-RST assessment relative to the immediate pre-RST assessment (b), and the 1-month post-RST assessment relative to the immediate post-RST assessment.

RST and Respiratory Phase Patterning

EX-EX usage was observed to be higher at immediate post-RST compared to immediate pre-RST (Fig. 4; Tables 4, 5). Specifically, the odds of inhaling before and/or after (i.e., not exhibiting the optimal EX-EX pattern) were lower at immediate post-RST compared to immediate pre-RST (OR = 0.22; 95% CI: 0.16–0.31; p < 0.0001). No differences were observed in the frequency of EX-EX usage during the no-treatment retention phase (p > 0.05), suggesting the higher rate of EX-EX usage observed after RST was maintained 1 month later.

Fig. 4.

Respiratory phase patterning. In this figure, odds ratios < 1.0 improved represent improved respiratory phase patterning (i.e., lower odds of not performing the EX-EX pattern). Figure display odds ratio point estimates (dots) and their 95% confidence intervals. Confidence intervals that cross the 1.0 marker are considered effects that are not statistically significant. With each figure, the Y-axis displays findings changes in outcomes at the immediate pre-RST assessment relative to the 1-month pre-RST assessment (top), immediate post-RST assessment relative to the immediate pre-RST assessment (middle), and the 1-month post-RST assessment relative to the immediate post-RST assessment.

Table 4.

Respiratory-swallow coordination – full cohort (n = 12)

​	1-month	Immediate	Immediate
	pre-RST (N = 385)	pre-RST (N = 370)	post-RST (N = 375)
Respiratory phase patterning, n (%)
EX-EX	121 (31.4)	142 (38.4)	269 (71.7)
Non-EX-EX, n (%)
Inhale-swallow-exhale	124 (32.2)	118 (31.9)	54 (14.4)
Exhale-swallow-inhale	75 (19.5)	62 (16.8)	38 (10.1)
Inhale-swallow-inhale	59 (15.3)	41 (11.1)	14 (3.7)
Exhale-swallow-cough	5 (1.3)	3 (0.8)	0 (0)
Inhale-swallow-cough	1 (0.3)	4 (1.1)	0 (0)
Pre-pause phase, n (%)
Exhalation	201 (52.2)	207 (55.9)	307 (81.9)
Inhalation	184 (47.8)	163 (44.1)	68 (18.1)
Post-pause phase, n (%)
Exhalation	245 (63.6)	260 (70.3)	323 (86.1)
Inhalation	134 (34.8)	103 (27.8)	52 (13.9)
Cough	6 (1.6)	7 (1.9)	0 (0)
Cough	6 (1.6)	7 (1.9)	0 (0)

This table presents data from the 12 people with Parkinson’s disease who completed the 1-month pre-RST, immediate pre-RST, and immediate post-RST assessments, used for statistical analyses when examining changes in the pre-RST no-treatment control phase and the treatment phase.

Table 5.

Respiratory-swallow coordination – partial cohort (n = 7)

​	Immediate	1-month
	post-RST (N = 211)	post-RST (N = 212)
Respiratory phase patterning, n (%)
EX-EX	140 (66.4)	125 (59.0)
Non-EX-EX, n (%)
Inhale-swallow-exhale	33 (15.6)	39 (18.4)
Exhale-swallow-inhale	26 (12.3)	38 (17.9)
Inhale-swallow-inhale	12 (5.7)	10 (4.7)
Exhale-swallow-cough	0 (0)	0 (0)
Inhale-swallow-cough	0 (0)	0 (0)
Pre-pause phase, n (%)
Exhalation	166 (78.7)	163 (76.9)
Inhalation	45 (21.3)	49 (23.1)
Post-pause phase, n (%)
Exhalation	173 (82.0)	164 (77.4)
Inhalation	38 (18.0)	48 (22.6)
Cough	0 (0)	0 (0)

This table presents data from the 7 people who completed all assessments. Data from their immediate and 1-month post-RST assessments were used to analyze changes in the retention phase.

A smaller difference in respiratory phase patterning was also observed during the delayed baseline no-treatment phase assessments (OR = 0.70; 95% CI: 0.50–0.98; p = 0.039). The magnitude of the post-RST effect exceeded that observed during the delayed baseline phase, as reflected by nonoverlapping 95% confidence intervals (0.16–0.31 vs. 0.50–0.98).

RST Practice Structure: Constant versus Variable Practice

Relative to the constant practice, variable practice was associated with more frequent EX-EX usage during FEES swallowing trials at the 1-month post-RST assessment, suggesting greater retention during similar but untrained tasks (OR = 0.23; 95% CI: 0.07–0.69; p = 0.009; Fig. 5). No other significant differences were observed between the two practice conditions.

Fig. 5.

Constant vs. variable practice (exploratory aim). For the task-specific (a) and generalized (b) RST motor learning figures, odds ratios > 1.0 represent greater odds of performing an EX-EX for the constant practice group, whereas odds ratios < 1.0 represent greater odds of performing an EX-EX for the variable practice group. Figures display odds ratio point estimates (dots) and their 95% confidence intervals. Confidence intervals that cross the 1.0 marker are considered effects that are not statistically significant. With each figure, the Y-axis displays findings changes in outcomes at the immediate pre-RST assessment relative to the 1-month pre-RST assessment (a), immediate post-RST assessment relative to the immediate pre-RST assessment (b), and the 1-month post-RST assessment relative to the immediate post-RST assessment. Task-specific RST Learning (a) refers to using the EX-EX pattern during the non-FEES swallowing trials (i.e., the task-specific RST training target). Generalization of RST to similar but untrained targets (b) refers to using the EX-EX pattern during the FEES swallowing trials (i.e., swallowing conditions not targeted during RST).

Discussion

This small clinical trial pilot study extends prior single-subject work by exploring the preliminary associations between RST, respiratory phase patterning, swallowing safety, and swallowing efficiency in a small cohort of PwPD. Consistent with previous observational research [34], participants demonstrated high rates of suboptimal respiratory phase patterning at baseline, with fewer than half of swallows exhibiting the EX-EX pattern. Following RST, use of the EX-EX pattern increased immediately post-training and remained elevated at 1-month follow-up, suggesting that respiratory phase patterning is modifiable in PwPD at the group level.

Preliminary analyses also indicated concurrent improvements in swallowing safety. Specifically, RST was associated with reduced odds of airway invasion, lower amounts of laryngeal vestibule and vocal fold residue when present, and lower PAS scores, with changes persisting at 1-month follow-up. A reduction in vocal fold residue was also observed during the delayed baseline no-treatment phase, which we interpret as test-retest variability rather than a treatment effect. Although physiologic observations cannot be largely observed with FEES, improved laryngeal vestibule closure during EX-EX may be one plausible physiologic pathway linking respiratory phase patterning to improved swallowing safety [28, 36].

These findings of improved swallowing safety are consistent with prior RST studies in head and neck cancer and with earlier single-subject investigations in Parkinson’s disease and anoxic brain injury, which reported favorable changes in swallowing safety and efficiency following RST [28, 38, 39]. The present clinical trial pilot study adds to this literature by providing preliminary cohort-level data suggesting that RST focused exclusively on respiratory phase patterning – without explicit lung volume training targets – may still be associated with favorable swallowing outcomes in PwPD. Importantly, these results should be interpreted as exploratory rather than confirmatory evidence of treatment efficacy.

Subglottic residue did not change following RST, suggesting RST may have limited influence on clearance of material that has already entered the lower airway. This finding contrasts with recent observational research linking respiratory phase patterning to aspirated residue in PwPD and highlights the need for larger studies to clarify whether and how RST may influence aspiration severity [35].

Swallowing efficiency outcomes showed a pattern similar to safety outcomes, with exploratory reductions in oropharyngeal and hypopharyngeal residue following RST that were maintained at 1-month follow-up. Together with emerging research linking respiratory phase patterning to pharyngeal residue [35], these findings support further investigation of RST as a potential mechanism-based intervention for dysphagia in PwPD, while acknowledging that confirmatory trials are required.

Exploratory comparisons of practice structure suggested that both constant and variable practice training conditions were associated with higher EX-EX usage immediately after RST, whereas variable practice appeared to promote greater retention and generalization of the EX-EX training target. These observations align with established motor learning theory and prior RST-related work but should be interpreted cautiously given the small sample size [39, 40, 53–56]. The purpose of this comparison was to inform protocol refinement rather than establish superiority of one practice structure over another. Our practice manipulation was intentionally limited to variability (constant vs. variable) to support protocol refinement and feasibility testing in PwPD. Other motor learning parameters (e.g., progressive task complexity; immediate, delayed, or self-controlled feedback; part vs. whole practice; instrumental-based biofeedback) warrant systematic evaluation in future trials but were beyond the scope of this early-phase clinical trial pilot study.

From a feasibility standpoint, all enrolled participants completed all four RST sessions and planned repetitions with no dropouts or voluntary withdrawals, and 12 of 12 completed the immediate post-RST assessment – suggesting the protocol was well tolerated and that the assessment schedule was feasible in terms of participant burden and retention. The high treatment fidelity observed across clinicians (94.3%–100%) further suggests that RST can be delivered reliably by trained graduate SLP clinicians using structured performance bandwidth rules and clinician-perceptual monitoring of respiratory phase patterning without instrumentation – an important consideration for clinical translation given that instrumented biofeedback is not routinely available in most clinical settings. Whether the absence of real-time instrumented feedback limits the precision of RST delivery and, consequently, motor learning outcomes remains an open question that future trials should directly address. Future trials should also evaluate whether augmenting clinician-delivered RST with instrumentation-based biofeedback enhances learning and treatment outcomes beyond what is achievable with perceptual monitoring alone, and whether explicitly targeting lung volume during swallowing – in addition to respiratory phase patterning – confers additional benefit over a phase-only approach. More broadly, the findings reported here are hypothesis-generating and should not be interpreted as evidence of clinical efficacy; replication in adequately powered, confirmatory randomized trials is required before conclusions about treatment effectiveness can be drawn.

Limitations and Future Directions

Several limitations should be considered when interpreting these findings. First, this small clinical trial pilot study included a cohort of 12 PwPD with predominantly early-to-mid stage Parkinson’s disease and moderate dysphagia severity. Recruitment was curtailed by COVID-19-related institutional closure, resulting in only 12 participants rather than the originally planned sample of 30. These factors limit the extent to which findings can be extrapolated beyond this cohort and further underscore the exploratory nature of the results. Future confirmatory studies should prioritize larger samples spanning a broader range of disease severity and swallowing impairment to better characterize the therapeutic potential of RST in PwPD. The effect size estimates generated here – including ORs for pre-to-post-RST changes in EX-EX usage, pharyngeal residue, and penetration – will inform a priori power calculations for such trials. Additionally, because the <85% EX-EX threshold permitted enrollment across a range of baseline RSC impairment levels, future research should explore whether participants with lower baseline EX-EX rates respond differently to RST than those closer to the eligibility threshold. As a provisional hypothesis, a stricter criterion – such as <70%, approximating the lower bound of what is typically observed in healthy adults – may improve sensitivity to treatment effects and warrants evaluation in future confirmatory trials.

Second, although a delayed baseline no-treatment within-subject control phase was included to characterize test-retest stability and within-subject variability, the study did not include a randomized no-treatment or alternative-treatment control group. As such, regression to the mean, placebo effects, or nonspecific treatment effects cannot be fully excluded. Future studies should incorporate concurrent control conditions. Future studies should also incorporate longer follow-up periods beyond the 1-month retention window used here and evaluate whether booster sessions or adaptive training schedules are necessary to maintain any treatment-related changes over time.

Third, clinicians relied on visual-perceptual judgments to guide feedback during training sessions. This was planned to maximize generalization of findings to the methods of RST that are likely to be used in clinical practice. However, this approach may have introduced variability between perceived and objectively measured performance. Future work should examine whether incorporating biofeedback or instrumentation-based monitoring alters learning and treatment outcomes.

Finally, the study focused on functional swallowing outcomes and did not include direct physiologic measures such as laryngeal vestibule closure, pharyngeal constriction, or pressure generation. Incorporation of physiologic instrumental assessments (e.g., high-resolution manometry or videofluoroscopic swallow studies) will be essential for identifying mechanisms of change and refining patient selection in future trials.

Conclusion

This small clinical trial pilot study explored the preliminary effects of RST on EX-EX usage, swallowing safety, and swallowing efficiency in 12 PwPD. Exploratory findings revealed higher EX-EX patterning usage and greater swallowing safety and efficiency immediately after and 1 month after four sessions of RST. Exploratory comparisons of practice structure also appeared to suggest that variable practice may support greater retention and generalization of respiratory phase patterning compared to constant practice, although these findings are preliminary.

Taken together, the exploratory findings from this clinical trial pilot study are hypothesis-generating and support continued investigation of RST as a mechanism-informed intervention for dysphagia in Parkinson’s disease. Specifically, this study generates three provisional hypotheses for future confirmatory investigation: (1) four sessions of EX-EX focused RST are associated with meaningful increases in EX-EX patterning in PwPD, with rates above approximately 70% potentially achievable; (2) post-RST improvements in swallowing safety outcomes co-occur with increases in EX-EX usage, suggesting a mechanistic rather than nonspecific treatment effect; and (3) variable practice conditions may confer greater long-term retention of RST gains compared to constant practice. These hypotheses are provisional and require replication in adequately powered confirmatory randomized trials incorporating longer follow-up, concurrent control conditions, and physiologic-based outcome measures.

Acknowledgments

The authors would like to thank the participants and caregivers for the time and effort they contributed to this study. The authors would also like to thank the speech-language pathology clinicians and research assistants who were involved with data collection and data analyses (in alphabetical order): Samantha Ayala, Sydney Brooks, Julianna Chapman, Crystal Diaz, Alyssa Dicker, Shannon Dinizo, Victoria Gonzalez, Theresa Lee, Belle Maniago, Sayume Romero, Gabriela Pereira, and Kaylee Viets. Lastly, the first author would like to thank the members of his PhD Dissertation Committee who provided invaluable support, mentorship, and educational training that helped make this study possible: Drs. Jessica Huber, Sonja Molfenter, Richard Magill, Lori Quinn, Lisa Edmonds, and Michelle Troche.

Statement of Ethics

This study was performed in accordance with the Declaration of Helsinki. This human study was approved by Teachers College, Columbia University Institutional Review Board – approval: 19-469. The study was not registered as a clinical trial because this was a clinical trial pilot study completed as part of doctoral dissertation research. All adult participants provided written informed consent to participate in this study.

Conflict of Interest Statement

The authors have no relevant financial or nonfinancial conflicts of interest associated with this study.

Funding Sources

This study was supported in part by a Clinical Research Training Scholarship in Parkinson’s Disease awarded to Dr. James Curtis from the American Brain Foundation and the Parkinson’s Foundation in collaboration with the American Academy of Neurology (Grant # 2360). This publication was also made possible with partial support from the National Institute on Deafness and Other Communication Disorders of the National Institutes of Health (Grant # F31DC0119281-01) awarded to Brianna Kiefer.

Author Contributions

All authors met the minimum criteria for authorship status, as proposed by the International Committee of Medical Journal Editors. Authorship contributions are outlined using the following CRediT roles (https://credit.niso.org/) – James A. Curtis: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, visualization, writing – original draft, and writing – review and editing; James C. Borders: formal analysis, methodology, software, validation, and writing – review and editing; Brianna Kiefer: formal analysis, funding acquisition, software, and writing – review and editing; Roy N. Alcalay: formal analysis, methodology, and writing – review and editing; Richard Magill, Lori Quin, Lisa Edmonds, and Sonja Molfenter: methodology and writing – review and editing; and Michelle S. Troche: funding acquisition, methodology, resources, software, supervision, and writing – review and editing.

Funding Statement

This study was supported in part by a Clinical Research Training Scholarship in Parkinson’s Disease awarded to Dr. James Curtis from the American Brain Foundation and the Parkinson’s Foundation in collaboration with the American Academy of Neurology (Grant # 2360). This publication was also made possible with partial support from the National Institute on Deafness and Other Communication Disorders of the National Institutes of Health (Grant # F31DC0119281-01) awarded to Brianna Kiefer.

Data Availability Statement

All data and R code associated with this study are openly available in the Open Science Framework repository at: https://osf.io/vnskj/. All further inquiries can be made to the corresponding author.