Associations between baseline cognitive status and motor outcomes after treadmill training in people with Parkinson’s disease: a pilot study

Abstract Purpose To determine the effect of baseline cognition on gait outcomes after a treadmill training program for people with Parkinson’s disease (PD). Methods This pilot clinical trial involved people with PD who were classified as having no cognitive impairment (PD-NCI) or mild cognitive impairment (PD-MCI). Baseline executive function and memory were assessed. The intervention was a 10-week gait training program (twice-weekly treadmill sessions), with structured speed and distance progression and verbal cues for gait quality. Response to intervention was assessed by gait speed measured after week 2 (short-term) and week 10 (long-term). Results Participants (n = 19; 12 PD-NCI, 7 PD-MCI) had a mean (standard deviation) age of 66.5 (6.3) years, disease duration of 8.8 (6.3) years, and MDS-UPDRS III score of 21.3 (10.7). Gait speed increased at short-term and long-term assessments. The response did not differ between PD-NCI and PD-MCI groups; however, better baseline memory performance and milder PD motor severity were independently associated with greater improvements in gait speed in unadjusted and adjusted models. Conclusions These findings suggest that memory impairments and more severe motor involvement can influence the response to gait rehabilitation in PD and highlight the need for treatments optimized for people with greater cognitive and motor impairment. IMPLICATIONS FOR REHABILITATION Cognitive deficits in Parkinson’s disease (PD) could impact motor learning and gait rehabilitation, yet little is known about the effects of cognitive impairments on the response to rehabilitation in people with PD. This study demonstrates that the response to gait rehabilitation did not differ between people with PD who had no cognitive impairment and those with mild cognitive impairment. Across all participants, better baseline memory was associated with greater improvements in gait speed. Rehabilitation professionals should be mindful of PD severity, as those with more substantial memory and motor impairments may require additional dosing or support to maximize gait training benefits.


Introduction
Rehabilitation is integral to the clinical management of gait disturbances in people with Parkinson's disease (PD), but the response to gait rehabilitation may be complicated by cognitive impairments that are common in PD.Gait impairments are among the most consequential motor impairments in PD, presenting in 85% of people within three years of diagnosis [1] and contributing to significant disability, falls, and reduced quality of life [2][3][4].In early-stage disease, gait impairments include reduced speed, decreased step length, and increased gait variability.With disease progression, these impairments worsen and advanced-stage gait disturbances such as festination and freezing emerge, even as the benefits of pharmacological and surgical interventions for gait decline or become more variable [5].Given the limitations in current pharmacological and surgical management of gait impairments, there is substantial interest in the effects of rehabilitation and exercise to improve gait in PD.
While there are various approaches to gait rehabilitation in PD, treadmill training is one approach shown to be effective at improving gait speed and stride length [6].Treadmill training is often used in PD because it allows high-repetition practice, the moving belt provides a powerful external cue for stride length, and intensity can be easily adjusted by changing speed and/or incline.Treadmill training may be a particularly effective approach because aerobic exercise can be combined with verbal cueing, Gait training; rehabilitation; physiotherapy; physical therapy; cognition which can bring attention to gait pattern impairments [7][8][9], and with high-repetition, task-specific training, which is important to motor learning.Attentional strategies that incorporate verbal cues to correct posture, enhance arm swing, or increase stride length, are commonly incorporated into gait rehabilitation and are effective in improving gait in people with PD [10], consistent with the idea that people with PD increasingly rely on cognition to compensate for gait impairments [11].High-repetition, task-specific training may then contribute to increased automaticity of improved gait patterns, thereby reducing reliance on compensatory strategies.
Though gait interventions are a critical aspect of rehabilitation in PD, the impact of cognitive impairments on the response to gait interventions is not well understood, despite the fact that cognitive impairments are common in PD and increase in prevalence with age and disease duration [12].There is significant heterogeneity in the severity and nature of cognitive impairment in PD [12,13], and deficits in global cognition or in specific cognitive domains could limit the capacity to improve gait through rehabilitation.Many trials incorporating treadmill training in PD exclude people with cognitive impairments or target people with early PD, for whom cognitive impairments may be minimal [6].Prior research has shown that people with PD and cognitive impairment can improve gait speed in response to auditory cues combined with instructions to focus on temporal and spatial aspects of walking [14]; however, the lack of a PD group with normal cognition makes it difficult to know if cognitive impairment impacted the benefits of this approach to gait training.Motor learning processes are generally considered to be preserved in PD, though learning may require more practice for those with PD compared to adults without PD [8,15].Cognitive deficits in PD could further impact the rate or extent of motor learning, contributing to reduced benefits from gait training in those with cognitive deficits compared to those with intact cognition.In addition, cognitive impairments could limit improvements in gait automaticity such that people with PD continue to rely on cognitive compensations for gait control, even after training.
The purpose of this pilot study was to examine the effects of baseline cognition on the response to a 10-week treadmill training intervention.This study tested the hypothesis that those with cognitive impairment at baseline would demonstrate lesser improvements in gait, reflecting motor learning, and lesser transfer to turning and dual-task walking compared to those without cognitive impairment.Understanding the influence of cognition on the response to gait training in PD could aid clinicians in counseling early engagement in gait rehabilitation programs and could guide the development of interventions optimized for cognitive function.

Study design
This was a prospective, pilot clinical trial, with data collected between 2018 and 2020.All participants completed the same intervention, with the goal of determining whether baseline characteristics influenced the response to gait rehabilitation.Study assessments and training sessions occurred at the Veterans Affairs Puget Sound Health Care System and the University of Washington.Participants provided written informed consent prior to initiating study procedures in accordance with the institutional review board-approved procedures at these sites.The study protocol was registered with ClinicalTrials.gov(NCT03607695).

Participants
Participants were recruited through the Washington State Parkinson Disease Registry [16] and local PD groups.Eligibility criteria were: meeting the UK Parkinson's Disease Society Brain Bank (UKBB) clinical diagnostic criteria for PD [17], modified so that having more than one affected relative was not considered an exclusion criterion; modified Hoehn and yahr stage ≤3; the ability to walk 400 m without physical assistance; no dementia; and no health conditions that impacted the ability to safely participate in a moderate-intensity exercise program, such as cardiopulmonary, orthopedic, or neurologic conditions other than PD.Initial eligibility was assessed in a telephone screen prior to enrollment.Final eligibility was determined at a consensus diagnosis case conference and was based on clinical examination, comprehensive history taking, and cognitive assessment (Supplementary Table ).A diagnosis of dementia was exclusionary; published clinical diagnostic criteria were used to define dementia in the context of established PD (i.e., PD diagnosed for at least one year prior to dementia diagnosis) [18].Dementia was indicated by cognitive impairment (>1.5 standard deviations below normative means) in more than one cognitive domain, impairment that represented a decline from premorbid levels of cognitive function, and cognitive impairments severe enough to impact daily life.Participants completed all study sessions while taking their regular PD medications.

Intervention
All participants completed a 10-week intervention, consisting of twice weekly 1-h gait training sessions on a treadmill under the supervision of a licensed physical therapist or physical therapy assistant (Figure 1).In the first two weeks, each session included 30 min of walking at a moderate intensity (5 min of walking alternating with 5 min of rest, repeated six times), with verbal cues for improved gait quality (upright posture, heel strike, increased stride length and/or arm swing).Treadmill speed was limited to 120% of self-selected walking speed and increased only if gait quality was maintained.Intensity was monitored using age-adjusted heart rate maximum and perceived exertion at the end of each walking block.In weeks 3-10, walking duration, speed, and intensity were gradually increased, with a maximum of 60 min of walking in weeks 5-10.Treadmill speed was increased based on maintenance of gait quality and participant tolerance.Cues were gradually reduced in frequency in weeks 3-10.For each participant, the number of sessions completed and the total distance and time walked over all sessions were recorded.

Assessments
At baseline, participants completed a targeted demographic and health interview.Motor severity was assessed using the Movement Disorder Society-sponsored Unified Parkinson Disease Rating Scale, Part III, Motor Subscale [19] and modified Hoehn and yahr stage [20].Levodopa-equivalent daily dose (LEDD) was calculated for each participant [21].Clinical measures of mobility included the Activities-specific Balance Confidence (ABC) Scale [22], the Mini-Balance Evaluation Systems Test (Mini-BEST) [23], and the 6-Minute Walk Test (6MWT) [24].Baseline physical activity was assessed using the International Physical Activity Questionnaire (IPAQ) [25].
Baseline cognition was assessed in two ways.First, a cognitive diagnosis of no cognitive impairment (PD-NCI) or mild cognitive impairment (PD-MCI) was determined using published consensus guidelines [26] at a consensus diagnosis panel of experts as previously described [27] using data from comprehensive cognitive testing performed at a separate visit, along with a neurological examination, detailed functional assessment, and careful clinical history-taking.Cognitive assessments included at least two validated measures from each of five domains (executive, attention/ working memory, memory, visuospatial, and language; Supplementary Table ).Second, a separate battery of global, executive, and memory tests was administered for analyses in the current study.Global cognition was assessed using the Montreal Cognitive Assessment (MoCA) [28].Executive function was assessed with semantic verbal fluency (fruits/vegetables or cities/towns) [29] and the Frontal Assessment Battery [30], which includes abstract reasoning, phonemic verbal fluency, motor programming, interference, and response inhibition.Additional executive function measures included the Pattern Comparison Processing Speed Test [31] and the Dimensional Change Card Sort Test [32] from the National Institutes of Health (NIH) Toolbox.Attention was assessed using the Flanker Inhibitory Control and Attention Test [32] from the NIH Toolbox, which examines the ability to inhibit visual attention to irrelevant stimuli.Verbal learning and memory were assessed using Craft Paragraph Recall, a measure of episodic verbal memory that includes both immediate and delayed recall [33].
Gait outcomes were measured at three time points (Figure 1).A pre-intervention assessment was completed at baseline, prior to any gait training sessions.An intermediate assessment was completed at the end of week 2 (after completion of the fourth training session) to evaluate short-term motor learning.A post-intervention assessment was completed at the end of week 10 (after completion of the final gait training session) to evaluate long-term motor learning and transfer to other mobility tasks.Instrumented gait analysis was performed using six opal inertial sensors (APDM Inc., Portland, oR, USA) placed on the feet, wrists, waist, and trunk.Each sensor included a tri-axial accelerometer, gyroscope, and magnetometer.Inertial sensor data were collected at 128 Hz, and gait metrics were analyzed using associated software, Mobility Lab, analysis version 4.0 (APDM Inc., Portland, oR, USA).
Gait outcomes reflecting motor learning were calculated during single-task walking, where participants walked at a self-selected speed for 2 min along a 7 m walkway, with 180° turns at each end of the walkway.For motor learning, the primary walking outcome was gait speed, which has high test-retest reliability among people with PD [34].Secondary walking outcomes included stride length, stride time, stride time variability (standard deviation [SD]), trunk rotation (transverse plane excursion), and arm swing excursion.Motor learning outcomes were calculated by averaging over all available strides for the linear walking portion of the task.
Transfer of treadmill training to other mobility tasks was assessed using the 180° turns (performed during single-task walking) and dual-task walking.Transfer to 180° turns was assessed with turn peak velocity, calculated as the average across all available turns.For dual-task walking, participants performed two 30-s trials of walking while performing serial-3 subtractions.Single-task performance of the cognitive task was assessed with two seated 30-s trials of serial-3 subtractions.A shorter trial length was used for dual-task walking trials compared to single-task walking trials in order to prevent fatigue or attentional lapses in performance of the cognitive task.A different starting number between 100 and 250 was used for all trials of serial-3 subtraction, with starting numbers standardized across participants and assessments.Cognitive task performance was measured as the correct response rate, defined as the number of correct responses per second.Transfer to dual-task walking was assessed using the gait outcomes noted above, calculated as the average over all available strides during the dual-task walking trials.Dual-task walking also serves as a measure of gait automaticity, as declines in gait during dual-task performance are thought to reflect continued reliance on cognitive compensations for gait control.

Statistical analyses
Descriptive statistics were used to summarize participant demographic and clinical characteristics as well as intervention details.To determine the effects of baseline cognition on gait outcomes, two analyses were performed.First, linear regression models were used to examine potential differences between cognitive groups (PD-NCI, PD-MCI) in short-term motor learning effects (week 2-baseline), long-term motor learning effects (week 10-baseline), and transfer (week 10-baseline) for all outcomes.For transfer to dual-task walking, descriptive statistics characterized cognitive task performance for each group to provide context for interpretation of dual-task gait outcomes, but additional statistical analyses were not performed.As a sensitivity analysis, multiple linear regression models (with permutation tests) were used to determine if gait outcomes differed between cognitive groups when adjusting for age.Second, simple linear regression was used to examine a potential association between scores on baseline cognitive tests and gait outcomes.Multiple linear regression models (with permutation tests) were used to determine if the association between gait outcomes and baseline cognition differed while adjusting for age, sex, education, and disease duration, using separate models for each due to the small sample size.Linear regression models were also used to examine potential associations between disease characteristics (years since diagnosis and motor severity) and gait outcomes, with separate multiple linear regression models adjusted for age and LEDD.Due to the small sample size and potential violation of assumptions, the significance of all regression models was determined using permutation tests with Monte Carlo sampling and a two-sided significance level of 0.05.A permutation test uses random assignment of the sampled data, often thousands of iterations, to calculate the null distribution of a test statistic that does not rely on distribution assumptions of the sample [35].This approach is useful in cases such as this where small sample sizes may limit the ability to verify standard statistical assumptions about the underlying population and are too small to benefit from large sample properties.In accordance with this approach, confidence intervals are not provided, as the necessary parametric assumptions may be violated.Finally, the aggregated changes in gait outcomes were examined.As subgroups were not used in this analysis, the sample size was deemed sufficient to use paired t tests.only statistically significant differences are described below.

Participants
Figure 2 shows the numbers of individuals included at screening, enrollment, intervention, and analysis.Due to local and institutional restrictions implemented at the start of the CoVID-19 pandemic, the study was ended prematurely, with study procedures impacted for two individuals (described below).Telephone screens were initiated with 66 individuals, with 46 exclusions due to not meeting eligibility criteria, individuals declining participation due to travel and time requirements or lack of interest, lack of telephone response after initiating the screening process, or termination of the screening process due to CoVID-19 restrictions.Twenty participants were screened in person, with one person excluded due to a concurrent health condition that limited involvement in an exercise program.A total of 19 individuals with PD participated in the study, including 12 with a cognitive diagnosis of PD-NCI and 7 with a cognitive diagnosis of PD-MCI (Table 1, Figure 1).The mean age of all participants was 66.5 (6.3) years, with a disease duration of 8.8 (6.3) years, and MDS-UPDRS III score of 21.3 (10.7).Eight (42%) of 19 participants were females.Participants completed an average of 19.3 (1.9) intervention sessions.Research restrictions related to CoVID resulted in early termination of the intervention for two participants, both of whom were in the NCI group.one completed 18 intervention visits, and the second completed 12 intervention visits.Primary and secondary gait outcomes during single-task walking were collected at the final visit for both participants and included in all analyses related to motor learning and transfer to turning.outcomes examining transfer to dual-task walking were not available for these two participants, and they were excluded from all dual-task analyses.

Gait outcomes
Table 2 shows gait and transfer outcomes aggregated for all participants.Gait speed, stride length, trunk excursion, and arm swing excursion increased at short-term and long-term assessments compared to baseline.Stride time variability decreased at the long-term assessment compared to baseline.Transfer to dual-task walking was observed at the long-term assessment, with increased dual-task gait speed, stride length, and arm swing and decreased stride time.No transfer to turning was observed, and peak velocity during turning did not change from baseline to the long-term assessment.
Table 3 shows gait and transfer outcomes for the PD-MCI and PD-NCI groups.Changes in primary and secondary gait outcomes did not differ significantly between the PD-MCI and PD-NCI groups at either assessment compared to baseline.Transfer to turning and walking did not differ between the PD-MCI and PD-NCI groups.Correct response rate is also provided for single-task (seated) and dual-task (walking) cognitive task performance.At baseline, the correct response rate for the serial-3 subtractions was 0.50 (0.23) responses per second during single-task performance and 0.40 (0.19) responses per second during dual-task walking.At the long-term assessment, the correct response rate was 0.53 (0.27) responses per second during single-task performance and 0.46 (0.24) responses per second during dual-task walking.
Table 4 shows associations of baseline cognition and disease characteristics with long-term changes in gait speed.Better baseline memory performance (delayed paragraph recall) was associated with greater long-term increase in gait speed in unadjusted and adjusted models (Figure 3; separate models adjusted for age, sex, education, and disease duration).Lower MDS-UPDRS III scores were associated with greater long-term increase in gait speed in both unadjusted and adjusted models (Figure 3; separate models adjusted for age and LEDD).

Discussion
This pilot study examined how cognitive function at baseline impacts the response to a 10-week treadmill training intervention.The response to gait rehabilitation was quantified as changes in gait outcomes, reflecting motor learning, and as transfer of training to turning and dual-task walking.When categorized as PD-NCI or PD-MCI, both groups demonstrated similar motor learning and transfer, contrary to the study hypothesis.However, when examining baseline cognition within specific cognitive domains, better memory was associated with larger gait speed improvements.Aggregated across all participants, statistically significant improvements in gait, including increased speed, stride length, stride time variability, trunk excursion, and arm swing, were observed.Benefits of treadmill training did not transfer to turning, potentially because training was inherently limited to linear walking without any directional changes.However, benefits transferred to dual-task walking, suggesting task-specific improvements in the automatization of walking in response to this gait intervention.Cognitive task performance changes did not suggest that gait improvements came at the expense of the cognitive task, providing further support that this intervention improved gait automaticity.These improvements in motor learning and transfer outcomes reinforce the recognized benefits of gait training and exercise in PD for improving both cognitive and automatic aspects of gait control.
In addition, lower PD motor severity, as measured by MDS-UPDRS III scores, was associated with greater increases in gait speed.This reinforces the importance of early rehabilitation interventions for people with PD, with greater potential for gait improvements early in the disease process [7].
These findings highlight the need for interventions better optimized for people in later stages of PD, including those with cognitive impairments.An interest in the effects of cognitive impairment on the response to gait rehabilitation in people with PD stemmed from the knowledge that cognitive impairment is a common feature in PD, and its prevalence increases with disease duration.The prevalence of MCI in PD is approximately 25% at time of diagnosis, and having PD-MCI is a risk factor for developing dementia [12].In this study, PD-NCI and PD-MCI groups demonstrated similar increases in gait speed, contrary to the expected findings.However, profiles of cognition are variable among those with PD-MCI.Impairment has been documented both in single domains and in multiple domains of cognitive function [12], with learning, memory, and attention deficits among the most common in those with PD-MCI [13].In the current study, poorer baseline memory performance was associated with less benefit from gait training.Given the prevalence of memory deficits in PD and their potential impact on the response to gait rehabilitation, these findings suggest the need for research to determine optimal rehabilitation interventions for those with memory impairments.Research among older adults without PD suggests that those with cognitive impairment may require added structure and support to successfully participate in physical activity and exercise programs [36], and these same strategies may benefit those with PD.
Research consistently highlights relationships between cognition and gait in PD [37,38], potentially because of shared neural substrates.Early in the disease process, cognitive pathways may also be helpful in compensating for gait impairments [11].With disease progression, involvement of these shared substrates may result in increasing cognitive and gait impairment, and the compensatory mechanisms used earlier in the disease process may fail.Thus, the recommended use of cues or attentional strategies to improve movement quality and gait speed [9] may be less effective for those with cognitive impairment.Many trials examining the effects of treadmill training or exercise in PD exclude people with cognitive impairment based on MMSE or MoCA scores [39,40].However, it is worth noting that the MMSE and MoCA are designed as screening instruments and may lack the sensitivity and specificity to detect cognitive impairment compared to existing guidelines [26].Furthermore, screening tools may not fully capture cognitive status in PD, as even people with high scores on the MMSE can have a wide range of cognitive impairment [41].While the MoCA appears superior to the MMSE at detecting mild cognitive impairment in PD, both tests require additional assessments due to their low specificity at the recommended screening cutoff points [42].
Established relationships between cognition and gait also suggest the potential for synergistic benefits of interventions targeting cognition and gait, either independently or in combination.Exercise interventions, including treadmill training [43], have been shown to improve aspects of cognition as well as gait in people with PD [44,45].The current findings also indicate a possible role for cognitive rehabilitation approaches targeting either multiple cognitive domains or specific domains.While the evidence supporting cognitive rehabilitation is mixed [46,47], recent research demonstrates improved memory performance in response to cognitive training in people with PD-MCI [48].The effects of cognitive rehabilitation on gait are not clear, but improved memory could enhance the benefit of gait training.Future research is needed to determine whether interventions aimed at cognitive function impact the responsiveness to gait training.Dual-task training, which involves training the simultaneous performance of cognitive and motor tasks, has been demonstrated to improve both cognition and gait in people with PD [49,50].There may be additive benefits with rehabilitation approaches that combine cognitive and gait training, either in sequence or simultaneously.Continued research is needed to understand the optimal sequence and composition of coordinated cognitive and gait rehabilitation for people with PD, particularly for individuals with PD who have more advanced cognitive and motor impairments.
This study found that memory impairments were associated with reduced gait speed benefit in response to a gait rehabilitation intervention.Though there are few studies examining the effects of baseline cognition on exercise training, at least one previous study is consistent with these findings, showing that better cognitive scores at baseline predicted greater improvement in dual-task gait velocity after the training interventions [51].A geriatric rehabilitation study in a non-PD population also found that better cognitive function at admission predicted greater improvement in gait speed [52].In contrast, a recent clinical trial found that people with PD who had more severe motor and cognitive symptoms demonstrated greater improvements in balance (as measured by the Mini-BEST) in response to a group exercise intervention targeting both mobility and cognition [53].Similarly, a study of rehabilitation in people with PD found that those with lower scores on an attention task at baseline had greater improvements in balance and gait speed [54].This latter finding suggests that attention deficits, which are often assessed with declines in dual-task walking, may not interfere with motor learning for balance and gait.The findings of this study suggest that memory may be more critical to effective gait rehabilitation than other cognitive domains.Future rehabilitation trials in PD should carefully quantify cognitive function to understand the impact of specific cognitive impairments on the response to rehabilitation.

Limitations
As a pilot study, the main limitation was the relatively small sample size.While the use of a more robust statistical analysis with permutation tests allowed for appropriate inference, the power to detect significant differences may be low.The number of comparisons performed, in combination with the small sample size, may increase the risk of type I error.As these results are intended to provide a foundation for larger trials, where replication of these findings with larger sample sizes could inform our understanding of how cognitive function impacts the response to gait training, no adjustments for multiple comparisons were done.Due to potential violation of model assumptions, confidence intervals are not reported with the exception of the analysis for the overall effect of intervention.However, permutation tests still allow for valid inference, therefore p values are reported throughout to aid in interpreting results.Another limitation is that gait training was conducted exclusively on a treadmill, which may have reduced transfer to turning as observed in this study.Complex walking tasks, such as changing directions, obstacle avoidance, or walking on uneven surfaces, are critical to mobility in day-to-day life, and interventions incorporating over-ground walking may improve transfer to such walking tasks.Finally, while gait improvements transferred to dual-task walking, the shorter trial duration in the dual-task walking condition could impact the reliability and accuracy of some gait measures, such as variability.
A major strength of this study was the use of a detailed neuropsychological assessment to characterize cognitive status beyond the scope of screening tests like the MoCA or MMSE.While formal neuropsychological testing is necessary to fully characterize the deficits experienced by people with PD, it is also time-intensive and therefore not common in clinical practice.Therefore, in a clinical setting, the MoCA remains a useful screening test to identify cognitive dysfunction in PD.However, future exercise and gait training trials would benefit from a more detailed characterization of cognitive impairment and its impacts on the response to interventions.

Conclusions
This pilot clinical trial found that memory impairments and motor severity may impact the response to gait training interventions in PD.An improved understanding of how cognitive status and other clinical characteristics influence the response to gait training is critical to optimizing gait rehabilitation programs for persons with PD.While it is critical to continue ongoing rehabilitation trials targeting those with early PD, when the potential for learning and recovery may be highest, there is also a pressing need to understand how gait training and other rehabilitative interventions can be optimized for those with more advanced cognitive and motor impairments.

Figure 1 .
Figure 1.schematic showing the timing of gait assessments and the overall structure of the gait rehabilitation intervention.

Figure 2 .
Figure 2. Diagram illustrating participant flow, including number of participants screened for eligibility, number assigned to intervention, number receiving the intended intervention, and number included for each analysis.

Figure 3 .
Figure 3. Relationships of memory (a) and motor severity (b) with the change in gait speed in response to gait training.

Table 1 .
Participant characteristics and intervention details.
Notes: Values shown are mean (standard deviation), except those indicated as number (n), percent (%).6MWt: 6-Minute Walk test; abC scale: activities-specific balance Confidence scale; % hR max: percent of age-adjusted heart rate maximum, averaged across all sessions in weeks 3-10; iPaQ: international Physical activity Questionnaire; leDD: levodopa-equivalent daily dose; MDs-UPDRs, Part iii: Movement Disorder society-Unified Parkinson Disease Rating scale, Part iii: Motor examination; Mini-best: Mini-balance evaluation systems test; MoCa: Montreal Cognitive assessment.

Table 2 .
short-term and long-term effects and transfer outcomes across all participants.Values shown are mean difference and 95% confidence interval (Ci).p Values for paired t tests examining differences in short-term (st) effects and long-term (lt) effects relative to baseline (bl) for all participants, with values < 0.05 shown in bold.Dt: dual-task condition.

Table 3 .
short-term and long-term effects and transfer outcomes, by cognitive diagnosis.
Notes: Values shown are mean (standard deviation).p Values for unadjusted and age-adjusted regression models examining differences between cognitive diagnosis groups for short-term (st) effects and long-term (lt) effects relative to baseline (bl).Dt: dual-task condition; st: single-task condition.Cognitive task correct response rates are provided for st (seated) and Dt (walking) conditions.

Table 4 .
associations between baseline cognition and disease characteristics and long-term changes in gait speed.Values shown are regression coefficients and p values for simple linear regression models and multiple linear regression models adjusting for baseline characteristics, with values < 0.05 shown in bold.MoCa: Montreal Cognitive assessment; PCPst: Pattern Comparison Processing speed test; DCCst: Dimensional Change Card sort test; Flanker: Flanker inhibitory Control and attention test; leDD: levodopa-equivalent daily dose; MDs-UPDRs, Part iii: Movement Disorder society-Unified Parkinson Disease Rating scale, Part iii: Motor examination.