Abstract
To examine the dose–response relationship between specific types of exercise for improving walking velocity in Parkinson’s disease (PD). This systematic review and network meta-analysis included searches of PubMed, Medline, Embase, PsycINFO, Cochrane Library, and Web of Science were searched from inception until February 18th, 2024. Data analysis was performed using R software with the MBNMA and RJAGS packages. Outcome indicators were measured as mean standard deviation (SMD) along with 95% confidence intervals (95% CrI). The network’s risk of bias was independently evaluated by two reviewers employing the ROB2 tool. Our review encompassed 54 studies with 2,828 PD patients, examining the dose–response effects of different exercises on walking velocity. Aerobic Exercise (AE) demonstrated the greatest improvement at an optimal dose of 1,400 METs-min/week (SMD:1.215, 95% Crl: 0.113 to 2.306). Both Multicomponent Exercise (Mul) (SMD: 1.202, 95% Crl: 0.193 to 2.231) and Sensory Exercise (SE) (SMD: 0.649, 95% Crl: 0.139 to 1.183) showed optimal outcomes at a dose of 1,000 METs-min/week. Resistance Training (RT) was most effective at 750 METs-min/week (SMD:0.778, 95% Crl: 0.062 to 1.549), while Mind–Body Exercise (MBE) yielded significant improvements at a lower optimal dose of 500 METs-min/week (SMD: 0.580, 95% Crl: 0.218 to 1.137), offering valuable insights for exercise prescription in PD management. Various types of exercise showed specific optimal benefits at corresponding doses, among which AE was the most effective in improving the walking speed of PD patients at 1,400 Mets*min/week.
Trial registration: PROSPERO (CRD42024506919).
Keywords: Parkinson’s disease, Exercise, Walking velocity, Dose–response, RCTs, Bayesian network meta-analysis
Subject terms: Neurology, Neurological disorders
Introduction
Walking is a hallmark of human evolution1, however, in people with Parkinson’s disease (PD), this natural rhythm is disrupted. Their gait is characterized by a slow, shuffling stride, stooped posture, and reduced arm swing. As the disease progresses, these gait abnormalities worsen, severely affecting independence and increasing the risk of falls. Walking velocity has garnered the most attention in gerontology literature. Numerous research findings indicated that older adults exhibited a walking velocity lower than 0.8 m/s, categorizing them as having “limited mobility in the community”2–4. As age increases, the walking velocity of 12 to 16% of older adults decreases by 0.02–0.03 m/s per year5. A rapid decrease in walking velocity is an important predictor of adverse prognoses, such as increased difficulty in activities of daily living and increased fall risk6, Not only that but decreased walking velocity is highly correlated with mortality risk7. Consequently, walking velocity is frequently regarded as a critical predictor of disability diagnosis. In PD, walking velocity declined notably faster, decreasing by approximately 0.02 m/s every 6 month8. Slow walking velocity is significantly correlated with clinical assessments of disease severity, compromised performance in daily activities, heightened overall disability, and diminished community ambulation6,9,10. At the same time, findings from a two-year prospective cohort study underscore the significance of walking velocity as a crucial indicator of disability8. Therefore, walking velocity in PD as a basis for treatment should receive more attention11.
Exercise has been shown to improve walking velocity in PD12. At the same time, two head-to-head RCT studies demonstrated that aerobic exercise and treadmill-based balance and gait training significantly improved walking velocity in PD13,14, and a meta-analysis of 20 studies also supported the finding that aerobic exercise improved walking velocity in PD (SMD: 0.49, 95% CI: 0.20 to 0.78, p = 0.0009)15. Although most of the focus was on aerobic exercise, the results of a network meta-analysis suggested that resistance training (SMD: 1.06, 95% CI: − 0.63 to -2.75) appears to be more effective than aerobic exercise (Not statistically significant) in improving walking velocity in PD16. Additionally, the network meta-analysis of Victor et al. (2020) provided new insights into improving walking velocity in PD. Results showed that aquatic therapy-based multicomponent exercise was the most effective way17, and mind–body exercises like Qigong, Tai Chi, and Yoga also had valuable reference18,19. Interestingly, sensory exercise (sensory exercises often refer to interventions designed to stimulate or improve sensory perception) in the treatment of patients with PD has also gained new development in the past decade. Recently, the network meta-analysis of Yang Y et al. (2020) showed that sensory exercise was the most effective way to improve walking velocity (included 199 original studies)20.
In previous dose–response meta-analyses, the effects associated with resistance exercise on healthy older adults had yielded good insights21,22. Additionally, the dose and response network meta of Daniel et al. opened a new direction in the study of dose–response relationships and confirmed for the first time that the relationship between exercise dose and cognition in older adults was nonlinear23. In conjunction with our research, we thought the variability in the effectiveness of different exercise interventions in enhancing walking velocity among patients with PD may primarily stem from differences in exercise dosage. However, in Parkinsonians, little was known about how exercise-specific dosage interventions can improve walking velocity. In our study, through Kevin Norton et al. standardized definition of exercise intensity24, we consider metabolic equivalents (METs) as a measure of exercise intensity (used to quantify physical activity and set limits, e.g., for patients undergoing rehabilitation25). At the same time, to accurately determine the optimal exercise dose to increase walking velocity in PD, the exercise dose was precisely calculated as the metabolic equivalents of the intervention task (number of interventions per week × minutes × METs of different forms of exercise = METs-min/week), to fill this gap.
In this systematic review and network meta-analysis, we utilize advanced techniques, including model-based dose–response network meta-analysis within a Bayesian framework26, to investigate the relationship between various exercise interventions and walking speed in PD patients. We acknowledge that statistically significant improvements in walking velocity may not always reflect clinically meaningful changes. Hence, our analysis emphasizes assessing clinical relevance alongside statistical significance. Additionally, we aim to explore the minimum clinically important difference (MCID) for walking velocity and identify optimal exercise dosages that might lead to clinically meaningful improvements. Our research seeks to provide evidence-based insights that can inform exercise guidelines for managing motor symptoms in PD, supporting healthcare professionals in making informed decisions.
Methods
This systematic review was reported according to the 2020 Preferred Reporting Items for Systematic Review and Meta-analysis Protocols statement extension for the PRISMA-NMA checklist27. This systematic review and network meta-analysis was registered with the International Prospective Register of Systematic Reviews (PROSPERO) (CRD42024506919).
Data sources
The systematic search for studies was carried out through databases PubMed, Medline, Embase, PsycINFO, Cochrane Central Register of Controlled Trials, and Web of Science from inception to January 18, 2024. The search strategy was structured according to the Peer Review of Electronic Search Strategies 2015 guidelines28. We provide a complete list of search times, search entries, and search results (Supplementary File 1). Two investigators (JW and TW) independently screened the title/abstract/full text in duplicate, with any disagreements resolved by discussion or adjudication by a third author (YY).
Study selection and eligibility
We included if they met the following inclusion criteria:(1) Participants: diagnosed with PD, mean age ≥ 50 years, Hoehn and Yahr stage < 4; (2) Intervention: any type of exercise was used as an intervention (9 exercise forms in total, Supplementary File 2 for more details); (3) the comparator did not receive the intervention, usual care, knowledge education, or active control (another type of exercise that was different from that of the experimental group or the same type of exercise as the experimental group that but with a different exercise dose). (3) Outcome: must report at least one walk velocity (3-10 m walking test (time or speed), gait velocity). (5) Study design: the study was designed as a randomized controlled trial (RCT).
The exclusion studies follow the criteria: (1) Patients with PD who also have other diseases; (2) the studies that reported on the acute effects of exercise; (3) focused solely on the acute effects of exercise; (4) incorporated mixed interventions from different disciplines (e.g., combining exercise with repetitive transcranial magnetic stimulation); (5) studies with exercise doses that did not explicitly describe the type of exercise or could not be computed; (6) Not available mean and standard deviation for the baseline to endpoint; Following these criteria, two independent reviewers (JW and TW) meticulously screened the titles, abstracts, and full texts of potentially relevant studies to determine their suitability for inclusion.
Data extraction
Two reviewers (JW and TW) extracted the relevant publication information from the study (first author and year of publication), sample (sample size), age (patient age), intervention characteristics (invention duration/time, type of exercise), and outcome indicators (Supplementary File 3 for more details). For the effect size calculation, the change score (endpoint minus baseline score), standard deviations, and sample size of each group were extracted. Missing change means and standard deviations were converted according to the Cochrane Handbook29. To meet the data analysis requirements of the Dose–Response Network meta-analysis package in the R program, we also converted the standard errors (SE). SE = SD /SQRT (Sample size)30. If the data needed for the study cannot be extracted from the above methods, we will contact the authors and invite them to provide additional data.
Data coding and management
The interventions were coded according to the specific type of exercise performed: "Aerobic exercise (AE)", "Balance and gait training (BGT)", "Resistance Training (RT)", “Dance”, "Mul_C” (Mixed Exercise Program)" "Sensory Exercise (SE)", "Mind–body Exercise (MBE: including “Qigong”, “Tai Chi”, and “Yoga”). Then, different exercises were set by specific types and doses. And expressed as METs-min/week metabolic equivalents of task (METs), which were produced of the time of session, frequency, and intensity of a certain type of exercise31–33. To better promote the connectivity of the network, we performed an approximate value of 250, 500, 750, 1,000, or 1,500 METs-min/week for the exercise dose, which has been adopted in previous studies23,34, which is a necessary step for the network meta-analysis35.
Data synthesis
We completed the data analysis for this study in R software (V.4.3.0, www.r-project.org), and employed the “MBNMAdose” package and “rjags” package for conducting Model-Based Network Meta-Analysis (MBNMA), and predicted the results of our data36,37. We specifically implemented a random-effects Bayesian MBNMA approach to synthesize data and assess the dose–response relationship between exercise dose and walking velocity. The effect sizes measure chooses the standardized mean difference (SMD) of the change score (end end-point minus baseline score) because the studies use different rating scales or units of the outcome. We analyzed the bases of research hypotheses in network meta-analysis. Firstly, we checked the connectivity of the motion pattern grid for different motion doses based on the dose approximation38. Secondly, we assessed between the two models by comparing them with an uncorrelated mean-effects model (UME)39 (Supplementary File 5). Finally, we assessed transitivity using a node-splitting approach40 (Supplementary File 5). There are no parametric results that indicate a violation of the above key assumptions, and we did a detailed explanation of the statistical principles in different assumptions Supplement material. In the process of selecting an appropriate dose–response model, we compared fit indices for different models in common and random effects(Emax, Exponential, Restricted cubic spline, Non-parametric, and Linear models), including the Deviance Information Criterion (DIC), between-study standard deviation, number of parameters in the model, and residual values, as suggested by the research result of Nathan J. Evans et al.41. Ultimately, we opted for the random effects model for the restricted cubic spline to evaluate the non-linear dose–response association, as more detailed in Supplementary File 6.
To estimate the different exercise doses that resulted in the predicted maximum significant effect referred to as the ‘optimal exercise dose’ on walk velocity in PD. We summarized the result of the dose–response relationship by the MCMC (Markov Chain Monte Carlo iterations) model [3 chains, 20,000 iterations each (first 10,000 discarded), n. thin = 10] of the beta coefficients on the restricted cubic spline curves42. And we positioned the three nodes at the 10th, 50th, and 90th percentile of the exercise dose to visualize our model-fitting results. We provided an inclusion of study-specific data in supplementary File 3.
To further enhance the clinical relevance of our findings, we conducted an estimation of the exercise dosage or range of dosages required to achieve the Minimum Clinically Important Difference (MCID), as recommended by JA Bernstein and DT Mauger43. In our analysis, we applied a distribution-based method to establish a consolidated MCID value for the walking velocity, following the approach outlined by JA Watt, AA Veroniki, AC Tricco and SE Straus44. Our results indicated that the MCID for the walking velocity could be estimated as an increase of 0.19 m/s seconds when considering a 0.4 standard deviation (SD) threshold or an increase of 0.23 m/s at a 0.5 SD level. Then, we calculated the pooled effect size (SMD) of the studies that at least achieved the estimated pooled MCID (k = 8). Finally, we predicted at which dose(s) of treatment these effects were achieved for each type of intervention45,46.
Risk of bias
The risk of bias was assessed according to the Cochrane Risk of Bias Tool for Randomized Trials (RoB2) version 2 The results were visualized using the “robvis” package47. Two (YB and BQ) reviewers assessed the study-assessment items, including randomized sequence generation, bias due to deviation from the intended intervention, incomplete data, measurement bias, and selective bias in reporting results. In instances where discrepancies arose, they were resolved through thorough discussion or by seeking the input of a third reviewer, YY, to reach a consensus.
Additionally, At the same time, to assess the confidence level of each comparison compared with active control, two researchers (YB and BQ) used CINeMA (Confidence in Network Meta-Analysis)48. Two researchers judged each comparison for within-study bias, reporting bias, indirectness, imprecision, heterogeneity, and inconsistency, according to recommended guidelines.
Result
Basic characteristics of the included studies
A total of 8,258 articles were retrieved from six databases and hand-searched. After removing duplicates, we screened citations by title and abstract, we considered 589 potentially eligible study articles and searched for full-text articles. After removing studies that did not meet the inclusion criteria, 54 studies were included in our analysis (Fig. 1), and We provided an included studies list in Supplementary File 4. Involve 2,828 participants (male:1,352) were included in our studies. All patients included in the studies had PD and were aged between 53.4 and 74.3 years old, with a mean disease duration of 7.04 years (SD 1.34) and a mean of 3.11 (SD 0.47) for Hoehn and Yahr stages. The exercise period ranged from 2 to 48 (mean period 9.3 weeks, SD 6.1), and the frequency of exercise per week ranged from 1–7 (mean frequency mean frequency 2.06, SD 1.09). The total time of the single session ranged from 15–120 min (mean time 39.44 min, SD 11.85). We provided a basic characteristic of all included population studies in Supplementary File 3.
Fig. 1.
PRISMA Flow diagram of the search process for studies. RCT randomized controlled trials.
Network connectivity
Whether or not connectivity is met determines the basis of NMA. When direct comparison is not possible, a lack of connectivity can lead to low statistical power and misleading results40. The results showed that there was no connectivity deficit in the two networks, thus ensuring the accuracy of the analysis (Figs. 2 and 3).
Fig. 2.
Agent-level network plot. AE Aerobic Exercise, BGT Balance and Gait Training, MBE Mind–body Exercise, CON Control group, Mul Multicomponent Exercise Program, RT Resistance Training, SE Sensory Exercise.
Fig. 3.
Treatment-level and agent-level network plot. The first value indicates the specific intervention, and the second one is the corresponding dose of that intervention. AE Aerobic Exercise, BGT Balance and Gait Training, MBE Mind–body Exercise, CON Control group, Mul Multicomponent Exercise Program, RT Resistance Training, SE Sensory Exercise.
Dose–response relationship
Figure 4 shows the dose–response nonlinear relationship of different exercise types improving walking velocity. In 25 studies involving 382 PD patients who participated in AE, we found that AE did effectively improve the walk velocity until 1,400 METs-min/week, and the graphical results showed a nonlinear improvement in the mean value of walk velocity changes in PD with increasing exercise doses of AE. At the same time, the 1,400 METs-min/week (SMD:1.215, 95% Crl: 0.113 to 2.306) is also the optimal AE dose to improve walk velocity in PD.
Fig. 4.
Dose–response association between treatment-level dose and change in walking velocity in Parkinson’s disease patients.AE Aerobic Exercise, BGT Balance and Gait Training, MBE Mind–body Exercise, CON Control group, Mul Multicomponent Exercise Program, RT Resistance Training, SE Sensory Exercise. Note: The green bar graph represents the specific sample size at different dose levels, and the color depth reflects the number of people participating in a specific sport: the darker the green, the more people participate; the lighter the green, the fewer people participate. The black solid line represents the dose-response curve, showing the relationship between the dose level and the predicted response value. The black dotted line represents the 95% confidence interval (95% CI). When the dotted line crosses the "0" axis, it means that the result is not statistically significant. The horizontal axis (Exercise Dose) represents the dose level, and the vertical axis (Walking velocity Predicted response) represents the response value predicted by the model.
At the same time, in ten studies involving 191 patients with PD who participated in Mul_C and in 17 studies involving 557 patients with PD who participated in SE, we found that the maximum exercise dose for both Mul_C and SE was estimated at 1,000 METs-min/week. At the same time, these studies were worth attention we found that Mul_C and SE did effectively improve the walk velocity at any dose, and both the graphical results showed a nonlinear improvement in the mean value of walk velocity changes in PD with increasing exercise doses of Mul_C and SE. and the optimal dose of Mul_C was estimated at 1,000 METs-min/week (SMD: 1.202, 95% Crl: 0.193 to 2.231). And the optimal dose of SE was estimated at 1,000 METs-min/week (SMD: 0.649, 95% Crl: 0.139 to 1.183).
Finally, ten studies involving 206 patients with PD participated in BGT, five studies involving 96 patients with PD participated in Dance, eight studies involving 186 patients with PD participated in MBE, and 17 studies involving 378 patients with PD participated in RT. These studies were worth attention in that none of the exercise doses for BGT, MBE, and RT exceeded 750 METs-min/week. (Fig. 3 Shades of green represent the sample size for distributing exercise doses, with larger sample sizes being darker and vice versa). At the same time, we found that BGT and Dance did not effectively improve the walking velocity at any dose. However, it is interesting to note that MBE effective dose of MBE ranges from 390 to 500 METs-min/week, and we found that at 500 METs-min/week (SMD: 0.580, 95% Crl: 0.218 to 1.137) appears to be the optimal MBE dose for improving the walk velocity in PD. In addition, RT effective dose of RT ranges from 380 to 750 METs-min/week, the optimal dose of RT was estimated at 750 METs-min/week (SMD:0.778, 95% Crl: 0.062 to 1.549).
Treatment dose and minimal clinically important difference
The pooled effect size equivalent to the estimated MCID for walking velocity was large (k = 8, SMD = 1.21; 95% CrI: 0.88 to 1.54). However, our findings indicated that no specific type of exercise, at the corresponding dosage, yields clinical benefits.
Risk of bias and quality of evidence
Overall, 22 studies (40%) were classified as having a low risk of bias, 20 studies (38%) were classified as having an unclear risk of bias, and 12 studies (23%) were classified as having a high risk of bias, Fig. 5 shows the result of the Cochrane Risk of Bias Tool. At the same time, when examining CINeMA, certain domains related to within-study bias, reporting bias, and imprecision raised “low risk” in most comparisons, with percentages of 84.6%, 94.32%, and 61.35% separately. On the other hand, the domains of indirectness, heterogeneity, and incoherence did not raise concerns in most comparisons, with percentages of 70.6%, 40.54%, and 84.2%, separately. We provided detailed reports’ results in Supplementary File 7.
Fig. 5.
Summary of the risk of bias assessment in the individual domains of the included studies.
Discussion
In this dose–response investigation incorporating 54 studies with 2,828 participants diagnosed with PD, we meticulously examined the nonlinear dose–response relationships between various exercise types and walking speed. Our analysis revealed that among the seven evaluated exercise modalities, five—namely AE, MBE, Mul_C, RT, and SE—significantly enhanced walking velocity in PD patients. Notably, BGT and Dance did not exhibit significant improvements. Aerobic exercise emerged as the superior modality, with AE, Mul_C, and SE demonstrating efficacy at any exercise dose. The effectiveness of these exercises on walking speed exhibited a nonlinear enhancement with dose escalation, pinpointing the optimal doses at 1,400 METs-min/week for AE and 1,000 METs-min/week for both Mul_C and SE. Moreover, MBE and RT were also effective within specific dose ranges, with MBE ranging from 390 to 500 METs-min/week and RT from 380 to 750 METs-min/week. Despite further analysis to determine the dose ranges necessary for achieving the MCID, we found that no particular exercise type at the corresponding dosage produced clinical benefits. These findings illuminate the intricate interplay between exercise type, dosage, and functional outcomes, guiding the development of optimized, evidence-based exercise recommendations for PD management.
This study compellingly establishes AE as the optimal intervention for enhancing walking speed in PD patients, with a pinpointed optimal dosage of 1,500 METs-min per week. Our findings resonate with previous research that underscored the efficacy of AE in mitigating PD symptoms and enhancing physical performance. Notably, studies such as those by Schenkman et al. have previously demonstrated the beneficial impacts of AE on motor control, cardiovascular health, and overall quality of life in PD patients49, thus reinforcing the pivotal role of AE in the therapeutic regime for PD. The effectiveness of AE is attributed to its multifaceted impacts on the human body. It has been shown that AE can induce neuroprotective effects, potentially through the upregulation of brain-derived neurotrophic factor (BDNF) and other growth factors that support neuronal health and function50. This neurobiological response is crucial for PD patients as it may mitigate the loss of dopaminergic neurons, thereby improving motor control and alleviating symptoms directly related to walking speed, such as bradykinesia and rigidity51. Furthermore, AE enhances cardiovascular health, thereby increasing blood flow to the brain and muscles52,53, which is essential for improving metabolic efficiency and endurance necessary for sustained physical activities. The systemic benefits of AE, including improved muscle strength and endurance as well as enhanced respiratory function, significantly contribute to its positive effects on walking velocity in PD patients54. The determination of 1,500 METs-min/week as the optimal dose further clarifies the intensity and duration necessary to achieve maximal benefits from AE, offering precise guidance for clinicians and patients in crafting exercise programs tailored to managing PD.
Our study reveals an intriguing pattern in the efficacy of both AE and SE for improving walking speed in PD patients. Initially, both AE and SE show a steep slope of improvement, indicating better efficacy, which then tapers off, suggesting a deceleration in the rate of improvement. Specifically, this transition occurs around 900 METs-min/week for AE and 500 METs-min/week for SE. Analyzing AE, prolonged and high-intensity aerobic activities may lead to fatigue in the central nervous system and respiratory system55,56, moderating the initial significant improvements in walking speed. This suggests that the beneficial impacts of AE on walking speed are most pronounced at moderate intensities and durations, beyond which the improvements begin to plateau due to physiological fatigue. When considering SE, the nature of these exercises involves prolonged sensory training, which demands continuous focus and engagement of visual and auditory senses57,58. This type of exercise might lead to fatigue sooner than AE, impacting its effectiveness in later stages. SE activities, such as balance exercises, proprioceptive training, or activities requiring focused attention to sensory inputs, can become particularly taxing over time, leading to a similar pattern of early gains followed by a plateau in improvements59. Furthermore, exercise modalities for AE and SE, such as jogging, cycling, or specific sensory-focused exercises, may undermine patients’ motivation and enthusiasm for participation during long-term interventions due to their inherent monotony and repetitiveness60. Particularly, as exercise duration or intensity increases, this singularity may further limit patients’ motivation to participate, thereby weakening the later effects of the intervention.
In contrast to the patterns observed with AE and SE, our findings for MulC reveal a different trajectory in the improvement of walking speed among PD patients. Initially, the slope of improvement at dosages below 650 METs-min/week was gradual. however, beyond this threshold, the slope became steeper, indicating enhanced efficacy in the latter phases. Remarkably, at 1,000 METs-min/week, the effectiveness of MulC in improving walking speed matched that of AE at 1,500 METs-min/week, suggesting a potent benefit at relatively lower dosages. Previous research has highlighted the substantial benefits of MulC for PD patients, emphasizing the significant improvements attributable to this mixed modality of exercise61,62. Such studies corroborate our findings, supporting the notion that a diversified exercise regimen, incorporating elements of strength, flexibility, balance, and aerobic training, yields considerable enhancements in motor function and mobility63. In contrast, the diversity of MulC procedures was able to significantly increase patient interest and engagement during the intervention. By combining multiple exercise elements, MulC not only effectively avoids the stagnation effect caused by monotony but also continues to motivate patients over time. More importantly, the diverse stimulation provided by MulC can target multiple levels of PD pathology and bring about more comprehensive physiological and functional improvements, thereby compensating for the limitations of a single form of exercise61,64. Furthermore, the sustained improvement observed up to 1,000 METs-min/week with MulC underscores its potential for long-term efficacy in enhancing walking speed among PD patients. However, it’s important to acknowledge that the studies included in our analysis are limited, and further empirical research is necessary to explore the advantages of MulC at higher dosages extensively.
Our study has elucidated that both MBE and RT significantly enhance walking speed in PD patients, particularly beyond a certain threshold of exercise dosage. Notably, significant improvements for MBE and RT were initiated at comparable dosage thresholds, around 390 METs-min/week for MBE and 380 METs-min/week for RT, indicates a parallel level of efficiency at these initial doses. This suggests that both modalities begin to exert a substantial impact on walking speed at relatively moderate exercise intensities. Analyzing the commonalities between these two forms of exercise, MBE, including practices such as Tai Chi and Qigong, focuses on balance, flexibility, and mental health62,65. Despite their gentle nature, these exercises involve isometric contractions to maintain posture and dynamic movements to sustain stability and balance, thereby indirectly contributing to muscle strength and endurance66. This aspect of muscle engagement and participation, albeit through different mechanisms, brings MBE and RT into a complementary relationship in addressing muscle weakness and motor control issues, thus improving the biomechanical aspects of walking67. Moreover, the requirement for moderate-intensity exercise to facilitate improvements highlights the critical importance of personalized exercise prescriptions. This finding emphasizes the potential of both MBE and RT when appropriately applied, to effect positive changes in walking speed, thereby contributing to the overall management and quality of life of PD patients.
Despite the significant enhancements in walking speed observed across a spectrum of exercise modalities (AE, SE, MulC, MBE, and RT) for PD patients, it is pivotal to note that none of these interventions met the MCID, indicates a possible discrepancy between the measurable improvements in walking speed and the perceptible changes deemed significant by patients and clinicians. This gap may be attributable to several factors, including the intrinsic limitations of exercise interventions in addressing the complex and multifaceted nature of PD, variability in individual responses to exercise, and potentially the need to recalibrate our expectations for what constitutes a clinically meaningful change within this population. This should be addressed by recognizing the need for a holistic management of PD that encompasses not only physical interventions but also the neurobiological, psychological, and sociological aspects of the disease68,69. In addition, more sensitive and specific outcome measures should be developed to capture the full range of benefits from exercise rather than limiting them to walking speeds and other routine metrics70. Furthermore, achieving the MCID may require more personalized exercise programs that are tailored to the disease stage, symptomatology, and individual patient preferences, potentially incorporating a combination of exercise types and intensities to optimize outcomes71. Such a customized approach, coupled with ongoing research into the optimal dosage, timing, and combination of exercises, is crucial for enhancing the effectiveness of exercise interventions in PD. While the absence of MCID achievement in this study may initially appear disheartening, it actually serves as a critical learning opportunity. It compels us to reassess our measurement standards and treatment strategies, fostering a deeper understanding of the role of exercise in PD management and encouraging the continuous search for innovative, patient-centered approaches to improve the lives of those affected by Parkinson’s Disease.
This systematic review and network meta-analysis present several strengths and limitations worth noting. A significant strength lies in our use of advanced Bayesian network meta-analysis techniques, which facilitated a detailed comparison across various exercise types, providing valuable insights into their relative efficacy and optimal dosages. Moreover, the adoption of stringent inclusion criteria and a rigorous methodological approach enhances the reliability and validity of our findings, contributing to a more robust body of evidence within the field. However, this study is not without its limitations. Firstly, despite our efforts to define METs for each exercise based on the activity and type, the METs values, calculated based on average data, might not accurately reflect the variability among individuals due to factors such as age, gender, weight, fitness level, and health status, with disease progression also potentially causing differences in METs. Secondly, the impact of environmental factors was not incorporated into the METs definition and analysis. Variables such as temperature, humidity, and whether the activity was conducted indoors or outdoors could influence energy expenditure and outcome measures for PD patients. Thirdly, the reliance on walking speed as the primary outcome measure, while significant as an indicator of mobility and overall functional capacity, may not fully capture the broad spectrum of benefits exercise provides to PD patients, including improvements in balance, cognitive function, and quality of life. This highlights the necessity of employing a multidimensional assessment approach to evaluate the comprehensive impact of exercise interventions. In essence, while this study offers significant contributions to understanding the role of exercise in managing Parkinson’s Disease, the noted limitations underscore the need for continued research. Future studies should aim to account for individual variability and environmental factors more accurately, explore a wider array of outcome measures, and investigate the long-term effects of different exercise modalities.
Conclusion
In this dose–response study involving 2,828 PD patients across 54 studies, we explored the nonlinear relationships between exercise types and walking speed. Our findings revealed significant walking velocity improvements with AE, MBE, MulC, RT, and SE, identifying optimal doses at 1,400 METs-min/week for AE and 1,000 METs-min/week for MulC and SE. MBE and RT also showed effectiveness within specific dose ranges. However, no exercise type at the analyzed dosages achieved the MCID, highlighting the need for tailored, evidence-based exercise prescriptions in PD management to optimize functional outcomes.
Supplementary Information
Acknowledgements
We thank all study authors who responded to our data requests.
Abbreviations
- RCT
Randomized controlled trials
- SMD
Standardized mean difference
- SD
Standardized deviation
- SE
Standard errors
- AE
Aerobic exercise
- BGT
Balance and gait training
- CON
Control group
- MBE
Mind–body exercise
- MulC
Mixed training
- RT
Resistance training
- SE
Sensory exercise
- DIC
Deviance information criterion
- UME
Uncorrelated mean-effects model
- MCMC
Markov chain monte carlo
- MCID
Minimal clinically important difference
- METs
Metabolic equivalent of task
- PD
Parkinson’s disease
Author contributions
Concept and design (S.X., S.L.), acquisition of data (Y.Y., J.W., T.W.), data collection (J.W., T.W.), quality assessment (Y.B., B.Q.), statistical analysis (S.X.), interpretation of data (S.X., Y.Y., J.W., Y.B., T.W., B.Q., Y.Y., S.L.), drafting of manuscript (S.X., Y.Y.) and critical revision of manuscript (S.X., Y.Y., J.W., Y.B., T.W., B.Q., Y.Y., S.L.). All authors approved the final manuscript and agreed to be accountable for all aspects of the work.
Data availability
All data generated or analyzed during this study are included in this published article and its supplementary information file.
Declarations
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Contributor Information
Yong Yang, Email: yangyong@henu.edu.cn.
Shu-Cheng Lin, Email: s0975835@gmail.com.
Supplementary Information
The online version contains supplementary material available at 10.1038/s41598-025-85456-7.
References
- 1.Demes, B., Thompson, N. E., O’Neill, M. C. & Umberger, B. R. Center of mass mechanics of chimpanzee bipedal walking. Am. J. Phys. Anthropol.156(3), 422–433 (2015). [DOI] [PubMed] [Google Scholar]
- 2.Hass, C. J. et al. Quantitative normative gait data in a large cohort of ambulatory persons with Parkinson’s disease. (2012). [DOI] [PMC free article] [PubMed]
- 3.Plummer-D’Amato, P., Altmann, L. J. & Reilly, K. Dual-task effects of spontaneous speech and executive function on gait in aging: exaggerated effects in slow walkers. Gait Posture33(2), 233–237 (2011). [DOI] [PubMed] [Google Scholar]
- 4.Hollman, J. H., McDade, E. M. & Petersen, R. C. Normative spatiotemporal gait parameters in older adults. Gait Posture34(1), 111–118 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Himann, J. E., Cunningham, D. A., Rechnitzer, P. A. & Paterson, D. H. Age-related changes in speed of walking. Med. Sci. Sports Exerc.20(2), 161–166 (1988). [DOI] [PubMed] [Google Scholar]
- 6.Elbers, R. G., van Wegen, E. E., Verhoef, J. & Kwakkel, G. Is gait speed a valid measure to predict community ambulation in patients with Parkinson’s disease?. J. Rehabil. Med.45(4), 370–375 (2013). [DOI] [PubMed] [Google Scholar]
- 7.White, D. K. et al. Trajectories of gait speed predict mortality in well-functioning older adults: the health, aging and body composition study. J. Gerontol. Series A: Biomed. Sci. Med. Sci.68(4), 456–464 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Ellis, T. D. et al. Identifying clinical measures that most accurately reflect the progression of disability in Parkinson disease. Parkinsonism Relat. Disord.25, 65–71 (2016). [DOI] [PubMed] [Google Scholar]
- 9.Tan, D., Danoudis, M., McGinley, J. & Morris, M. E. Relationships between motor aspects of gait impairments and activity limitations in people with Parkinson’s disease: a systematic review. Parkinsonism Relat. Disord.18(2), 117–124 (2012). [DOI] [PubMed] [Google Scholar]
- 10.Brusse, K. J., Zimdars, S., Zalewski, K. R. & Steffen, T. M. Testing functional performance in people with Parkinson disease. Phys. Ther.85(2), 134–141 (2005). [PubMed] [Google Scholar]
- 11.Nisenzon, A. N. et al. Measurement of patient-centered outcomes in Parkinson’s disease: what do patients really want from their treatment?. Parkinsonism Relat. Disord.17(2), 89–94 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Ni, M., Hazzard, J. B., Signorile, J. F. & Luca, C. Exercise guidelines for gait function in Parkinson’s disease: a systematic review and meta-analysis. Neurorehabilit. Neural Repair32(10), 872–886 (2018). [DOI] [PubMed] [Google Scholar]
- 13.Rawson, K. S. et al. Exercise and Parkinson disease: comparing tango, treadmill and stretching. J. Neurol. Phys. Ther.: JNPT43(1), 26 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Rawson, K. S. et al. Exercise and Parkinson disease: comparing tango, treadmill, and stretching. J. Neurol. Phys. Ther.43(1), 26–32 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Zhen, K. et al. A systematic review and meta-analysis on effects of aerobic exercise in people with Parkinson’s disease. npj Parkinson’s Dis.8(1), 146 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Zhou, X., Zhao, P., Guo, X., Wang, J. & Wang, R. Effectiveness of aerobic and resistance training on the motor symptoms in Parkinson’s disease: systematic review and network meta-analysis. Front. Aging Neurosci.14, 935176 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Hvingelby, V. S. et al. Interventions to improve gait in Parkinson’s disease: A systematic review of randomized controlled trials and network meta-analysis. J. Neurol.269(8), 4068–4079 (2022). [DOI] [PubMed] [Google Scholar]
- 18.Li, X., Lv, C., Liu, X. & Qin, X. Effects of Health Qigong Exercise on Lower Limb Motor Function in Parkinson’s Disease. Front. Med. (Lausanne)8, 809134 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Li, F. et al. A randomized controlled trial of patient-reported outcomes with tai chi exercise in Parkinson’s disease. Mov. Disord.29(4), 539–545 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Yang, Y. et al. Efficacy and evaluation of therapeutic exercises on adults with Parkinson’s disease: a systematic review and network meta-analysis. BMC Geriatr.22(1), 813 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Nicola, F. & Catherine, S. Dose–response relationship of resistance training in older adults: a meta-analysis. Br. J. of Sports Med.45(3), 233–234 (2011). [DOI] [PubMed] [Google Scholar]
- 22.Borde, R., Hortobágyi, T. & Granacher, U. Dose–response relationships of resistance training in healthy old adults: a systematic review and meta-analysis. Sports Med.45, 1693–1720 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Gallardo-Gomez, D. et al. Optimal dose and type of exercise to improve cognitive function in older adults: A systematic review and bayesian model-based network meta-analysis of RCTs. Ageing Res. Rev.76, 101591 (2022). [DOI] [PubMed] [Google Scholar]
- 24.Norton, K., Norton, L. & Sadgrove, D. Position statement on physical activity and exercise intensity terminology. J. Sci. Med. Sport13(5), 496–502 (2010). [DOI] [PubMed] [Google Scholar]
- 25.Ainsworth, B.E., Haskell, WL., Whitt, MC., Irwin, ML., Swartz, A.M., Strath, S.J., O. Brien, W.L., Bassett, D.R., Schmitz, K.H., Emplaincourt PO: Compendium of physical activities: an update of activity codes and MET intensities. Medicine and science in sports and exercise 2000 32 (9; SUPP/1):S498-S504 [DOI] [PubMed]
- 26.Pedder, H: MBNMAdose: An R package for incorporating dose-response information into Network Meta-Analysis. In: Evidence Synthesis and Meta-Analysis in R Conference 2021: 2021
- 27.Page, M. J. et al. PRISMA 2020 explanation and elaboration: updated guidance and exemplars for reporting systematic reviews. bmj10.1136/bmj.n160 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.McGowan, J. et al. PRESS peer review of electronic search strategies: 2015 guideline statement. J. clini. Epidemiol.75, 40–46 (2016). [DOI] [PubMed] [Google Scholar]
- 29.Higgins, J.P., Thomas, J., Chandler, J., Cumpston, M., Li, T., Page, M.J., Welch, V.A (2019) Cochrane handbook for systematic reviews of interventions (John Wiley & Sons) [DOI] [PMC free article] [PubMed]
- 30.Watt, J. A. et al. Incorporating dose effects in network meta-analysis. Bmj376, e067003 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Ainsworth, B. E. et al. 2011 Compendium of Physical Activities: a second update of codes and MET values. Med. Sci. Sports Exerc.43(8), 1575–1581 (2011). [DOI] [PubMed] [Google Scholar]
- 32.Wasfy, M. M. & Baggish, A. L. Exercise dose in clinical practice. Circulation133(23), 2297–2313 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Ferguson, B., (2014) acsm’s guidelines for exercise testing and prescription 9th ed. 2014 The Journal of the Canadian Chiropractic Association 58 3 328
- 34.Yuan, Y. et al. Effective dosage and mode of exercise for enhancing cognitive function in Alzheimer’s disease and dementia: a systematic review and Bayesian Model-Based Network Meta-analysis of RCTs. BMC Geriatrics24(1), 1–13 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Higgins, J. P. T. et al. Consistency and inconsistency in network meta-analysis: concepts and models for multi-arm studies. Res Synth Methods (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Pedder, H., Dias, S., Bennetts, M., Boucher, M. & Welton, N. J. Joining the dots: linking disconnected networks of evidence using dose-response model-based network meta-analysis. Med. Decis. Mak.41(2), 194–208 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Shim, S. R., Kim, S.-J., Lee, J. & Rücker, G. Network meta-analysis: application and practice using R software. Epidemiol. Health41, e2019013 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Higgins, J. P. T. et al. Consistency and inconsistency in network meta-analysis: concepts and models for multi-arm studies. Res. Synth. Method.3(2), 98–110 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Wei, J., Zeng, C. & Lei, G. H. Heterogeneity, consistency and model fit should be assessed in Bayesian network meta-analysis. Ann. Rheum. Dis.75(1), e5 (2016). [DOI] [PubMed] [Google Scholar]
- 40.Rouse, B., Chaimani, A. & Li, T. Network meta-analysis: an introduction for clinicians. Int. Emerg. Med.12, 103–111 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Evans, N. J. Assessing the practical differences between model selection methods in inferences about choice response time tasks. Psychon. Bull. Rev.26, 1070–1098 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Desquilbet, L. & Mariotti, F. Dose-response analyses using restricted cubic spline functions in public health research. Stat. Med.29(9), 1037–1057 (2010). [DOI] [PubMed] [Google Scholar]
- 43.Bernstein, J. A. & Mauger, D. T. The Minimally Clinically Important Difference (MCID): What difference does it make?. J. Allergy Clin. Immunol. Pract.4(4), 689–690 (2016). [DOI] [PubMed] [Google Scholar]
- 44.Watt, J. A., Veroniki, A. A., Tricco, A. C. & Straus, S. E. Using a distribution-based approach and systematic review methods to derive minimum clinically important differences. BMC Med. Res. Methodol.21(1), 41 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Jette, M., Sidney, K. & Blumchen, G. Metabolic equivalents (mets) in exercise testing, exercise prescription and evaluation Of Functional-Capacity. . Clin. Cardiol.13(8), 555–565 (1990). [DOI] [PubMed] [Google Scholar]
- 46.Partlett, C. & Riley, R. D. Random effects meta-analysis: coverage performance of 95% confidence and prediction intervals following REML estimation. Stat. Med.36(2), 301–317 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Minozzi, S., Cinquini, M., Gianola, S., Gonzalez-Lorenzo, M. & Banzi, R. The revised Cochrane risk of bias tool for randomized trials (RoB 2) showed low interrater reliability and challenges in its application. J. Clin. Epidemiol.126, 37–44 (2020). [DOI] [PubMed] [Google Scholar]
- 48.Papakonstantinou, T., Nikolakopoulou, A., Higgins, J. P., Egger, M. & Salanti, G. Cinema: software for semiautomated assessment of the confidence in the results of network meta-analysis. Campbell Syst. Rev.16(1), e1080 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Schenkman, M. et al. Exercise for people in early- or mid-stage Parkinson disease: a 16-month randomized controlled trial. Phys. Ther.92(11), 1395–1410 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Lima Giacobbo, B. et al. Brain-Derived neurotrophic factor in brain disorders: focus on neuroinflammation. Mol. Neurobiol.56(5), 3295–3312 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Zhou, Z. D., Yi, L. X., Wang, D. Q., Lim, T. M. & Tan, E. K. Role of dopamine in the pathophysiology of Parkinson’s disease. Transl. Neurodegener.12(1), 44 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Querido, J. S. & Sheel, A. W. Regulation of cerebral blood flow during exercise. Sports Med. (Auckland, NZ)37(9), 765–782 (2007). [DOI] [PubMed] [Google Scholar]
- 53.Kleinloog, J. P. D. et al. Aerobic exercise training improves cerebral blood flow and executive function: A randomized, controlled cross-over trial in sedentary older men. Front. Aging Neurosci.11, 333 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Padilha, C. et al. Physical exercise and its effects on people with Parkinson’s disease: Umbrella review. PloS one18(11), e0293826 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Lira, F. S. et al. Low and moderate, rather than high intensity strength exercise induces benefit regarding plasma lipid profile. Diabetol. Metab. Syndr.2, 31 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56.Racil, G. et al. Plyometric exercise combined with high-intensity interval training improves metabolic abnormalities in young obese females more so than interval training alone. Appl. Physiol. Nutr. Metab. Physiol. Appl. Nutr. et Metabol.41(1), 103–109. 10.1139/apnm-2015-0384 (2016). [DOI] [PubMed] [Google Scholar]
- 57.Allison, L. K., Kiemel, T. & Jeka, J. J. Sensory-Challenge balance exercises improve multisensory reweighting in Fall-Prone older adults. J. Neurol. Phys. Ther.: JNPT42(2), 84–93 (2018). [DOI] [PubMed] [Google Scholar]
- 58.Beck, E. N., Wang, M. T. Y., Intzandt, B. N., Almeida, Q. J. & Ehgoetz Martens, K. A. Sensory focused exercise improves anxiety in Parkinson’s disease: A randomized controlled trial. PloS one15(4), e0230803 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59.Proske, U. Exercise, fatigue and proprioception: a retrospective. Exp. Brain Res.237(10), 2447–2459 (2019). [DOI] [PubMed] [Google Scholar]
- 60.Lambourne, K., Audiffren, M. & Tomporowski, P. D. Effects of acute exercise on sensory and executive processing tasks. Med. Sci. Sports Exerc.42(7), 1396–1402 (2010). [DOI] [PubMed] [Google Scholar]
- 61.Ernst, M. et al. Physical exercise for people with Parkinson’s disease: a systematic review and network meta-analysis. Cochrane Database Syst. Rev.1(1), Cd013856 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 62.Siega, J., Iucksch, D. D., Da Silva, A. Z., Zotz, T. G. G. & Israel, V. L. Parkinson’s disease and multicomponent aquatic exercise: Effects on motor aspects, functional mobility, muscle function and aquatic motor skills. J. Bodyw. Mov. Ther.27, 314–321 (2021). [DOI] [PubMed] [Google Scholar]
- 63.Linhares, D. G. et al. Effects of Multicomponent Exercise Training on the Health of Older Women with Osteoporosis: A Systematic Review and Meta-Analysis. Intern. J Enviro. Res. Public Health19, 21 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 64.Earhart, G. M. & Falvo, M. J. Parkinson disease and exercise. Compr. Physiol.3(2), 833–848 (2013). [DOI] [PubMed] [Google Scholar]
- 65.Song, R. et al. The impact of Tai Chi and Qigong mind-body exercises on motor and non-motor function and quality of life in Parkinson’s disease: A systematic review and meta-analysis. Parkinsonism Related Disord.41, 3–13 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 66.Penn, I. W. et al. Effects of individualized Tai-Chi on balance and lower-limb strength in older adults. BMC Geriatr.19(1), 235 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 67.Kwok, J. Y. Y., Kwan, J. C. Y., Auyeung, M., Mok, V. C. T. & Chan, H. Y. L. The effects of yoga versus stretching and resistance training exercises on psychological distress for people with mild-to-moderate Parkinson’s disease: study prxotocol for a randomized controlled trial. Trials18(1), 509 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 68.Ahmad, M. H., Rizvi, M. A., Ali, M. & Mondal, A. C. Neurobiology of depression in Parkinson’s disease: Insights into epidemiology, molecular mechanisms and treatment strategies. Ageing Res. Rev.85, 101840 (2023). [DOI] [PubMed] [Google Scholar]
- 69.Navarta-Sánchez, M. V. et al. Perspectives of people with Parkinson’s disease and family carers about disease management in community settings: A cross-country qualitative study. J Clin. Nursing32(15–16), 5201–5218 (2023). [DOI] [PubMed] [Google Scholar]
- 70.Awai, L. & Curt, A. Locomotor Recovery in Spinal Cord Injury: Insights Beyond Walking Speed and Distance. J. Neurotrauma33(15), 1428–1435 (2016). [DOI] [PubMed] [Google Scholar]
- 71.Onishi, A. et al. Efficacy of personalized exercise program on physical function in elderly patients with rheumatoid arthritis at high risk for sarcopenia: study protocol for a randomized controlled trial. BMC Musculoskelet. Disord.24(1), 280 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
All data generated or analyzed during this study are included in this published article and its supplementary information file.





