Optimal dose and type of exercise to improve motor symptoms in adults with Parkinson's disease: A network meta-analysis

Abstract

Objectives

This study aimed to evaluate the dose-response relationship between different exercise types and the alleviation of motor symptoms in Parkinson’s Disease patients.

Design

A systematic review and network meta-analysis were conducted to compare the effects of 12 exercise types on motor symptoms in Parkinson’s Disease patients using randomized controlled trials.

Methods

A systematic search was conducted across PubMed, Medline, Embase, PsycINFO, Cochrane Library, and Web of Science until September 10, 2024. A total of 81 trials involving 4596 patients were included. Mean differences with 95 % credible intervals were calculated, and evidence quality was assessed using Confidence in Network Meta-Analysis. The Minimum Clinically Important Difference was used to assess clinical efficacy.

Results

The optimal exercise dose for overall motor symptom improvement was 1300 MET-min/week (mean difference: − 6.07, 95 % credible intervals: − 8.10 to − 4.01). Dance at 850 MET-min/week provided the greatest improvement (mean difference: − 11.18, 95 % credible intervals: − 16.01 to − 6.22). Significant improvements were seen with doses as low as 60–100 MET-min/week for body weight support training, dance, resistance training, and sensory exercise. The Minimum Clinically Important Difference was achieved with doses exceeding 670 MET-min/week for overall exercise, and at lower doses for specific types: aerobic exercise (1100 MET-min/week), body weight support (420 MET-min/week), and dance (230 MET-min/week).

Conclusions

The optimal exercise dose for alleviating motor symptoms in Parkinson’s Disease patients is 1300 MET-min/week, with dance being most effective at 850 MET-min/week. These findings provide evidence-based recommendations for Parkinson’s Disease management.

1 Introduction

Parkinson's disease (PD) is a chronic neurodegenerative disorder characterized by dopamine (DA) deficiency, ^{1 affecting more than 1 % of the global population over the age of 65 years and 5 % of individuals up to 85 years of age. 2 , 3 The prevalence of DA deficiency is increasing, particularly in light of the global shift toward an aging population. It is predicted that the number of people with PD will reach 12.9 million worldwide by 2040. 4 PD is characterized by motor symptoms such as bradykinesia, rigidity, tremor, gait abnormalities and postural instability. These symptoms significantly reduce the quality of life and functional autonomy of PD patients. 5 Motor symptoms, including bradykinesia, rigidity, and postural instability, are among the most debilitating features of PD. The Movement Disorder Society-sponsored Unified Parkinson's Disease Rating Scale Part III (MDS-UPDRS III) is a validated and widely used clinical tool for assessing these motor symptoms.}

Treatment methods for PD typically include medications (e.g., levodopa) and exercise, ^{5 , 6 both of which aim to alleviate motor symptoms. However, medications often cause adverse side effects and may fail to correct balance and gait abnormalities, increasing the risk of falls and serious complications. 7 , 8 Exercise is increasingly recognized as an effective treatment for PD, particularly for alleviating motor symptoms. 9 , 10 Numerous studies have demonstrated the beneficial impact of various exercise interventions on motor function in PD patients, improving their quality of life and functional autonomy. 11 , 12 To determine the most effective types of exercise, numerous researchers have conducted meta-analyses to examine the differential effects of different types of exercise on symptom relief. Ernst et al. found that most forms of exercise had a positive effect on motor symptoms, 13 while Yang et al. identified strength training and treadmill exercise as the most effective for symptom relief. 14 Álvarez-Bueno et al. advocate a fusion of more complex and demanding exercises, including endurance, resistance, and dance, as the most effective intervention for mitigating PD's motor symptoms. 15 Collectively, these studies seem to offer a plethora of viable exercise options for both clinicians and patients. Yet, the question of how to choose the most appropriate dose of exercise remains unclear. Different doses of the same type of exercise are a primary factor affecting the alleviation of motor symptoms in PD patients, 14 and the relationship between exercise and its health benefits is not linear, meaning more is not necessarily better. 16 Cui et al. found that exercise doses with high adherence to the American College of Sports Medicine (ACSM) guidelines resulted in more significant improvements in PD motor symptoms. 17 However, identifying the specific dose–response relationship for PD symptom relief remains an area of ongoing investigation. As noted by the World Health Organization (WHO) 18 and the American Physical Activity Guidelines Advisory Committee, 19 exploring the dose–response relationship is paramount to determining the minimum effective dose, the optimal dose, and the maximum safety threshold for improving population health. 18}

Existing studies have primarily focused on individual exercise types or combinations without systematically exploring the optimal dose across a wider range of exercise modalities, including less commonly studied interventions like sensory exercise and mind–body exercises such as Tai Chi and Qigong. ^{20 We chose to assess 12 different exercise types because current research either focuses on individual exercises or limited combinations, leaving gaps in understanding how a wider range of exercises affects motor symptoms. Furthermore, these studies have not fully employed advanced dose–response network meta-analysis techniques, which offer a more precise evaluation of the dose–response relationship. 14 This study addresses these gaps by incorporating a broader range of 12 exercise types and using Bayesian-based model network meta-analysis to quantify dose–response relationships. In addition, by applying the Minimum Clinically Important Difference (MCID), this study ensures that our results are not only statistically significant but also clinically relevant, offering more practical exercise prescriptions for PD patients.}

In this systematic review and network meta-analysis, using novel meta-analysis techniques (i.e., model-based dose–response network meta-analysis within a Bayesian framework) and evidence from existing randomized controlled trials (RCTs), this study aims to evaluate the dose–response relationship between different types and doses of exercise interventions and motor symptoms in PD patients. Furthermore, it's essential to recognize that statistically significant improvements in motor symptoms may not always correspond to clinically meaningful changes, underscoring the importance of evaluating clinical relevance in addition to statistical significance. With this in mind, this study also aims to estimate the MCID for motor symptoms and assess the optimal dose of exercise to achieve clinically significant improvements. This study's findings will contribute to the development of evidence-based exercise prescriptions for the management of motor symptoms in PD and provide a better evidence-based basis for decision makers.

2 Methods
2.1 Design

This systematic review was reported according to the 2020 Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) and PRISMA Extension Statement for Reporting of Systematic Reviews Incorporating Network Meta-analyses of Health Care Interventions, ^{21 , 22 and was registered at PROSPERO (CRD42023422762). This study was a meta-analysis and did not require ethics committee approval or a consent statement.}

2.2 Data sources and search strategy

Literature was retrieved by searching PubMed, Medline, Embase, PsycINFO, Cochrane Central Register of Controlled Trials, and Web of Science from the date of the respective database onset to September 10, 2024. Moreover, we screened the reference lists of all included studies and the bibliographies of systematic reviews published in the last 5 years to avoid omitting additional eligible studies. Title/abstract and full-text screening was conducted independently and in duplicate by investigators (XYF and HYZ), with disagreements resolved by discussion or adjudication by a third author (YY). The details of the search strategies for each database/platform are listed in Supplementary file 1.

2.3 Eligibility criteria and study selection

Studies were included if they met the following inclusion criteria: (1) participants were diagnosed with Parkinson's disease (PD), with a mean age ≥ 50 years and Hoehn and Yahr stages < 4; (2) the intervention involved any type of exercise (12 types in total, as detailed in Supplementary file 2); (3) the comparator group either received no intervention, usual care, or no knowledge education or an active control (e.g., a different type of exercise from the experimental group, or the same type of exercise with a different dose); (4) the study reported at least one global motor symptom, assessed using the Unified Parkinson's Disease Rating Scale (UPDRS) III ^{23 or the revised MDS-UPDRS III from the Movement Disorder Society (MDS) 24 to ensure clinical relevance and comparability across studies; (5) motor symptoms were assessed in the “ON” medication state; and (6) the study was designed as a RCT.}

Exclusion criteria included: (1) studies reporting only the acute effects of exercise (< 4 weeks) ^{25; (2) studies combining exercise interventions with non-exercise interventions (e.g., repetitive transcranial magnetic stimulation); (3) studies that did not clearly describe exercise types or lacked sufficient data to calculate exercise dose; (4) studies where mean and standard deviation data were unavailable or authors did not respond to data requests; and (5) studies that assessed global motor symptoms using scales other than UPDRS III or MDS-UPDRS III.}

Two independent reviewers (XYF and HYZ) screened titles, abstracts, and full texts of the potentially relevant articles to determine their eligibility based on the defined inclusion and exclusion criteria.

2.4 Data extraction

Two reviewers (XYF and HYZ) extracted the relevant publication information (e.g., author, title, year, journal), number of patients, patient characteristics (e.g., age and sex), interventions considered and outcome measures (Supplementary file 3). For each exercise type, descriptive statistics (mean ± SD) for participant age, years of diagnosis, session duration, training frequency, and intervention duration were extracted and summarized to provide context for the characteristics of included studies ( ^{Table 1} ). 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 handbook. ^{26 If the data needed for the study cannot be extracted from the above methods, we will ask the authors about the data at least 4 times within 6 weeks.}

2.5 Risk of bias and quality of evidence

Two reviewers (HYZ and XYF) assessed and rated the studies according to the Cochrane Risk of Bias 2.0 criteria. ^{27 Similarly, any disagreements were resolved by discussion or by inviting a third reviewer (YY) to achieve a consensus. The quality of the evidence for each outcome was assessed using the CINeMA (Confidence of Network Meta-Analysis) web application. The CINeMA approach included six domains: Within-study bias, reporting bias, indirectness, imprecision, heterogeneity, and incoherence, which allows the confidence of the results to be graded as high, moderate, low, and very low. 28}

2.6 Data coding and management

First, the interventions included in the studies were divided into exercise versus control groups, and based on this, we assessed the relationship between actual exercise dose and motor symptoms in Parkinson's disease patients. We included a broad range of exercise types to ensure a comprehensive analysis. Based on the types and characteristics of exercise modalities reported in the literature, we categorized the interventions into 12 common exercise types, facilitating comparative analysis across studies. The interventions were then categorized by the specific type of exercise performed: “Aerobic exercise (AE)”, “aquatic exercise (AQE)”, “balance and gait training (BGT)”, “body weight support exercise (BWS)”, “dance”, “mixed exercise (a combination of two or more exercise types)”, “Qigong”, “resistance training (RT)”, “sensory exercise (SE)”, “stretch”, “Tai Chi (TC)”, and “yoga”.

Next, the exercise intensity (e.g., heart rate) was used to calculate METs (Metabolic Equivalent of Tasks) based on the prescribed exercise dose (frequency, duration, and intensity) as reported in the included studies. ^{29 Due to limited adherence data, the calculation did not account for the actual exercise performed by participants. To quantify the exercise dose, all interventions were converted to MET-min/week by multiplying the duration, frequency, and intensity of each exercise session. 30 , 31 For example, an exercise dose of 500 MET-min/week could be derived from multiple sessions of resistance training. To ensure network connectivity, a necessary step for conducting network meta-analysis, exercise doses were approximated to the nearest pre-specified categories: 0 (control group), 500, 750, 1000, 1500, or 2000 MET-min/week. This categorization, used in previous studies, facilitates consistency and comparison. 32 , 33}

2.7 Data synthesis

Based on the R statistical environment (V.4.3.0, www.r-project.org), we used the ‘MBNMAdose’ package ^{34 to perform Model-Based Network Meta-Analysis (MBNMA) and random-effects Bayesian MBNMA to summarize the dose–response association between exercise dose and motor symptoms. To assess heterogeneity across the studies, we calculated the I 2 statistic, which quantifies the proportion of variation in effect estimates attributable to between-study heterogeneity. An I 2 value greater than 50 % was considered indicative of substantial heterogeneity. 35 We assessed network connectivity by examining the structural linkages between each exercise type and its associated dose across the network, ensuring that all nodes were connected. This step is critical, as disconnected nodes could lead to biased or misleading results. 36–38 Connectivity was validated through visual inspection of the network diagrams, and the network plots demonstrated full connectivity ( Fig. 2, Supplementary files 4 and 5). Additionally, model adequacy and data consistency were assessed using various fit indices, such as the Deviance Information Criterion (DIC), 39 between-study standard deviation, number of parameters in the model, and residual values. We converted the UPDRS III to MDS-UPDRS III revised score, 40 and all effect sizes were reported as mean differences (MDs) with 95 % credible intervals (CrIs) to evaluate the reliability of our estimates. Finally, we selected restricted cubic splines for evaluating the non-linear dose–response association (Supplementary file 6).}

To further enhance the clinical utility of our results, we estimated the dose (or range of doses) at which the exercise type was able to achieve the Minimum Clinically Important Difference (MCID). ^{41 We used a distribution-based approach to derive a pooled MCID for MDS-UPDRS III. 42 In this study, the MDS-UPDRS III was estimated to be − 3.4 points at 0.4 SD (which is similar to Horváth, Aschermann, Ács, Deli, Janszky, Komoly, Balázs, Takács, Karádi, and Kovács 43 study result: − 3.25 points for detecting minimal) and − 4.3 points at 0.5 SD. However, the scale part of data merging was transformed from UPDRS III. In order to give more rigorous advice to clinicians, we finally chose − 4.3 points at 0.5 SD as the MCID of this study.}

3 Results
3.1 Risk of bias and quality of evidence

Out of the 81 included studies, 71 (87.6 %) were assessed as having a low risk of bias, 5 (6.2 %) had some concerns, and 5 (6.2 %) were rated as high risk of bias (Supplementary file 7). Regarding the confidence in the evidence for comparisons with control groups, 71.1 % of the evidence was judged to be of high confidence, 18.4 % of moderate confidence, 7.9 % of low confidence, and 2.6 % of very low confidence, as determined using the CINeMA framework (Supplementary file 8).

3.2 Characteristics of included studies

The flow diagram of the search process for the systematic review is presented in ^{Fig. 1} . We identified 8474 potentially eligible studies through an initial electronic search. After screening citations by title and abstract, 891 potentially eligible studies were retrieved for full-text review. Following the application of inclusion criteria, 81 studies with a total of 4596 participants (2355 males) were included in the final meta-analysis. The sample size of the included studies ranged from 4 to 71 participants, with a mean disease duration of 6.51 years (SD 2.66), and a mean Hoehn and Yahr stage of 2.35 (SD 0.49).

The studies encompassed various types of exercise interventions, providing a diverse dataset for analysis. ^{Table 1} outlines the distribution of intervention types, age, disease duration, Hoehn and Yahr stages, and study characteristics. Among the 81 studies, 20.3 % involved AE, 20.3 % involved BGT, 12.8 % involved RT, and 8.3 % involved BWS. Other interventions included dance (5.3 %), TC (6.8 %), Qigong (2.3 %), and yoga (3.0 %). The session duration ranged from less than 11 min to over 60 min, with 75.3 % of studies reporting a session duration of 31–60 min. The frequency of training varied, with 40.7 % of studies involving twice-weekly sessions and 33.3 % involving thrice-weekly sessions.

The exercise intervention period ranged from 4 to 96 weeks (mean duration 10.52 weeks, SD 8.40). The frequency of training per week ranged from 1 to 7 times, with a mean frequency of 2.85 sessions per week (SD 1.25). The total time of individual sessions varied from 20 to 120 min, with a mean session length of 54.62 min (SD 18.25). The full breakdown of study characteristics, including the number of studies and participants per exercise type, can be found in ^{Table 1} and Supplementary file 3.

3.3 Heterogeneity assessment

This study evaluated heterogeneity across the included studies. The I ² statistic, which quantifies the proportion of variability in effect estimates due to heterogeneity rather than random chance, revealed substantial heterogeneity (I ² = 96.1 %). Additionally, the between-study variance was quantified using tau, ^{2 which was calculated at 10.4490. Given the high degree of heterogeneity observed, indicating significant variability in study populations, intervention types, and exercise doses, we conducted further dose–response analyses for different exercise types. These analyses allow us to better understand the specific impact of each exercise type on motor symptom relief, while accounting for the variability across studies. The high heterogeneity highlights the importance of these additional analyses to ensure the robustness and clinical relevance of our findings.}

3.4 Network connectivity

Connectivity is a key assumption in network meta-analysis, where evidence of unconnectedness can lead to low statistical power and biased results. In our study, we assessed network connectivity by visualizing the connections between exercise types and doses in ^{Fig. 2} . The lines between different exercise types represent direct comparisons made across studies, and the thickness of the lines indicates the number of studies comparing each pair of interventions, with thicker lines corresponding to more comparisons. Strong network connectivity is achieved when all nodes (exercise types and doses) are linked, either directly or indirectly, allowing for robust statistical comparisons. Our results show no instances of poor network connectivity, as all nodes were connected, ensuring comprehensive and reliable comparisons between exercise interventions.

3.5 Dose–response relationships

^{Fig. 3} illustrates the non-linear dose–response relationship between exercise dose and motor symptoms in PD patients. The results revealed a U-shaped relationship, with exercise beginning to alleviate motor symptoms at a dose of 60 MET-min/week. The optimal dose for symptom relief was identified at 1300 MET-min/week (MD: − 6.07, 95 % CrI: − 8.10 to − 4.01). Beyond this point, the relief effect gradually diminished. In ^{Figs. 3 and 4} , the solid lines represent the MD effect sizes, while the dashed lines show the 95 % confidence intervals. The green shading indicates the distribution of sample sizes across different exercise doses, with darker shading representing larger sample sizes.

^{Fig. 4} presents the dose–response relationship for different types of exercise in alleviating motor symptoms in PD patients. The results showed that, with the exception of AQE and stretch, all other exercise types effectively reduced motor symptoms. Dance demonstrated the greatest symptom relief, with a U-shaped dose–response relationship similar to that observed in the overall exercise analysis. The effective dose range for dance was between 30 MET-min/week and 1400 MET-min/week, with the optimal dose identified at 850 MET-min/week (MD = − 11.18, 95 % CrI: − 16.01 to − 6.22). Additionally, BWS at 60 MET-min/week, dance at 30 MET-min/week, RT at 70 MET-min/week, and SE at 90 MET-min/week all exhibited significant symptom relief at relatively low doses (< 100 MET-min/week).

Regarding the MCID, our study demonstrated that clinically meaningful improvements, as defined by achieving the MCID, were observed when the overall exercise dose exceeded 670 MET-min/week, reaching the MCID value of − 4.3 points calculated in this study. Among the different exercise types, clinically meaningful improvements were achieved with AE at 1100 MET-min/week, BWS at 420 MET-min/week, dance at 230 MET-min/week, SE at 760 MET-min/week, TC at 390 MET-min/week, and yoga at 290 MET-min/week.

4 Discussion

This study yields several important findings that carry significant implications for clinical practice and future exercise guidelines for PD patients. Firstly, we identified a non-linear dose–response relationship between exercise dose and motor symptom relief in PD patients, with a U-shaped curve demonstrating that symptom alleviation begins at 60 MET-min/week and reaches an optimal dose at 1300 MET-min/week. Beyond this dose, the effects of symptom relief gradually diminished, highlighting the importance of balancing exercise intensity and duration. Secondly, the analysis revealed that different exercise types have varying degrees of effectiveness in alleviating motor symptoms. Dance emerged as the most effective, with an optimal dose range between 30 and 1400 MET-min/week, and a peak effect at 850 MET-min/week. In addition, exercise types such as BWS, RT, and SE also demonstrated significant motor symptom relief at relatively low doses (< 100 MET-min/week). Thirdly, regarding clinically meaningful improvements, the study found that overall exercise doses exceeding 670 MET-min/week led to improvements defined by achieving the MCID. Various exercise types, including AE, BWS, dance, SE, TC, and yoga, all reached their respective MCID thresholds at specific dose ranges. These findings underscore the clinical relevance of tailored exercise prescriptions, which not only achieve statistical significance but also provide meaningful improvements in patients' quality of life.

This study highlights the non-linear dose–response relationship between exercise dosage and motor symptom relief in PD patients, emphasizing that the benefits of exercise are not strictly proportional to the dose. Rather than a linear relationship, our findings suggest a U-shaped curve, indicating that there is an optimal range of exercise that maximizes therapeutic benefits without pushing patients into overtraining territory. This underscores the need for a balanced approach to exercise prescription, ensuring that patients receive adequate stimulus to improve motor symptoms while avoiding the risks associated with excessive physical activity. Our results demonstrate that motor symptoms begin to improve at an exercise dose of 60 MET-min/week, with the most pronounced symptom relief occurring at 1300 MET-min/week. However, beyond this threshold, the effectiveness of exercise in alleviating motor symptoms begins to taper off, further emphasizing the clinical importance of achieving the MCID to gauge meaningful improvements. The observed uncertainty at higher exercise doses, represented by wider credible intervals in ^{Figs. 3 and 4}, reflects the smaller sample sizes available for these doses, underscoring the need for caution when interpreting results at the right end of the dose–response curves. This non-linear pattern may be attributed to the potential for overtraining, where excessive exercise places PD patients at risk of insufficient recovery, leading to damage to skeletal muscles, connective tissues, and bones. The resulting trauma could trigger the release of pro-inflammatory cytokines, which negatively affect the central and sympathetic nervous systems, exacerbating rather than alleviating symptoms. ^{44 , 45}

A systematic review further supports this non-linear relationship, showing a positive correlation between exercise dosage and neural plasticity, but only in healthy young adults. In contrast, no such improvements were observed in older adults or patients with neurological disorders, challenging the assumption that higher doses of exercise necessarily promote better neuroplasticity in these populations. ^{16 This suggests that the benefits of exercise may plateau or even decline at higher doses for PD patients, particularly when exceeding the recommended upper limit of 1300 MET-min/week. While 1300 MET-min/week produced the greatest symptom relief in this study, it exceeds the WHO recommended range of 600–1200 MET-min/week, 18 potentially increasing the risk of overexertion, fatigue, and difficulty maintaining regular activity. Considering these factors, we advocate for a more conservative target of 1200 MET-min/week, which aligns with WHO guidelines, achieves clinically meaningful improvements (as defined by MCID), and promotes long-term adherence to exercise regimens in PD patients. 46 Adopting this approach may result in sustained improvements in the management of PD symptoms over time, ensuring both safety and efficacy in clinical practice.}

The non-linear dose–response relationship identified in this study was also observed across different exercise types. Among these, dance was found to be the most effective in alleviating motor symptoms in PD patients, with an optimal dose of 850 MET-min/week (MD = − 11.18, 95 % CrI: − 16.01 to − 6.22). Dance demonstrated the greatest improvement in motor symptoms compared to other exercise types, making it a particularly valuable intervention for PD patients. These findings reaffirm that dance not only provides statistically significant benefits but also achieves clinically meaningful improvements, as demonstrated by surpassing the MCID threshold at a dose of 230 MET-min/week. ^{47 This reinforces the growing body of evidence that sensory exercises, including dance, play a critical role in managing PD motor impairments. 48 From a physiological and biomechanical perspective, dance may offer superior benefits for several reasons. Firstly, dance involves complex movements that challenge coordination, balance, and rhythm, which are particularly beneficial for PD patients who often experience deficits in these areas. 49 , 50 These movements require the integration of multiple sensory inputs and motor outputs, enhancing sensorimotor integration, which is often impaired in PD patients. 51 Secondly, dance incorporates dynamic postural adjustments and continuous weight shifting, which can improve balance and reduce the risk of falls—one of the major concerns in the PD population. 52 , 53 Finally, the rhythmic nature of dance, particularly with music accompaniment, has been shown to activate brain regions involved in movement timing and coordination, such as the basal ganglia and cerebellum, areas that are typically affected in PD. 54 This rhythmic entrainment may facilitate motor control and movement fluidity in PD patients, helping to alleviate bradykinesia and other motor symptoms. Therefore, our findings provide a concrete dose recommendation for dance that ensures clinically meaningful improvements while minimizing the risk of overexertion.}

In addition, we found that certain exercise types, such as BWS, dance, RT, and SE, were able to alleviate motor symptoms at relatively low doses, starting from < 100 MET-min/week. These exercises share several key characteristics, including their ability to engage multiple sensory and motor pathways, their emphasis on improving balance and coordination, and their incorporation of controlled, repetitive movements that may be easier for PD patients to execute consistently. The low dose required to initiate symptom improvement in these exercises may be attributed to their targeted nature, focusing on specific motor deficits that are prominent in PD. Previous studies have also supported these findings. For instance, research on BWS has demonstrated its effectiveness in improving gait and postural stability at low exercise doses by reducing the physical strain on patients while allowing for repetitive practice of motor tasks. ^{55 , 56 Similarly, studies on dance, RT, and SE have shown that even low levels of these activities can lead to significant improvements in motor function, highlighting their therapeutic potential in early-stage interventions for PD patients. 47 This suggests that the unique attributes of these exercises, which focus on reducing motor deficits and enhancing sensorimotor integration, can yield substantial benefits even at lower intensities.}

The mechanisms behind these effects may be linked to how these exercises stimulate neuroplasticity, improve neuromuscular control, and enhance proprioceptive feedback—areas that are often impaired in PD patients. BWS, for example, reduces the gravitational load, enabling patients to focus on correct gait patterns with less risk of falling, allowing for safer but effective motor practice. ^{57 Dance and SE emphasize rhythm and sensory feedback, which engage brain regions involved in motor control, such as the basal ganglia and cerebellum, helping to improve movement coordination. 58 Meanwhile, RT helps build muscle strength, improving rigidity and bradykinesia, which are common motor symptoms in PD. 59 These findings suggest that BWS, dance, RT, and SE offer significant therapeutic benefits even at lower doses, making them particularly suitable for PD patients who may face challenges with more intensive exercise regimens. This highlights the importance of personalized exercise prescriptions that consider the physical limitations of PD patients, ensuring symptom relief can be achieved without excessive physical strain.}

This study established the MCID as a critical benchmark for assessing improvements in motor symptoms, enhancing the clinical relevance of our findings. By incorporating MCID into the analysis, we ensured that the observed improvements in motor function were not only statistically significant but also meaningful from a clinical perspective. Our results demonstrated that significant clinical effects, defined by achieving the MCID, began to manifest when the weekly exercise dose exceeded 670 MET-min/week. Across different exercise types, we observed that several forms of exercise reached or exceeded the MCID at varying doses. Specifically, AE achieved MCID at 1100 MET-min/week, BWS at 420 MET-min/week, dance at 230 MET-min/week, SE at 760 MET-min/week, TC at 390 MET-min/week, and yoga at 290 MET-min/week, all showing clinically significant effects. These findings provide a valuable framework for both patients and clinicians, offering clear guidance on how to tailor exercise prescriptions to achieve clinically meaningful improvements in PD patients.

While this NMA primarily included moderate to high-quality randomized controlled trials, several limitations warrant consideration. First, this study only included trials that assessed motor symptoms in the “on-medication” state due to the limited availability of studies conducted in the “off-medication” state. This restricts the generalizability of our findings to patients in the “on-medication” state, potentially overlooking variability in motor symptoms under different medication conditions. Additionally, the study is subject to limitations inherent to the included trials, such as variability within PD patient cohorts, differences in exercise types and dosages, and diverse outcome measures. Furthermore, only published randomized controlled trials were included, which may introduce publication bias, as studies with significant findings are more likely to be published. Regarding the methodology used to estimate the optimal exercise dosage, METs were calculated based on the prescribed exercise doses (frequency, duration, and intensity) reported in the included studies. However, adherence data were not consistently available across studies, meaning that the actual amount of exercise performed by participants may not align with the prescribed dose. This introduces a potential discrepancy between the prescribed and actual exercise doses, which could impact the dose–response relationship observed in this study. Future research should prioritize the collection of adherence data to more accurately reflect the actual exercise performed, which would enhance the clinical relevance and precision of exercise dose calculations. Finally, although most studies were of moderate to high quality, a small proportion of studies were rated as low or very low quality according to the CINeMA assessment. This variability in study quality should be considered when interpreting the findings, particularly where evidence quality was lower. Additionally, the study did not account for individual patient characteristics (e.g., age, gender, disease progression) and their influence on exercise intervention efficacy, and no subgroup analysis was performed. Future research should focus on exploring personalized exercise regimens for PD patients, taking into account individual characteristics to optimize motor symptom relief and improve quality of life.

5 Conclusion

This study identified a non-linear dose–response relationship between exercise dose and motor symptom relief in PD patients, with 1300 MET-min/week being optimal for symptom improvement. Dance emerged as the most effective exercise at a dose of 850 MET-min/week. Additionally, exercises like body weight support training, resistance training, and sensory exercise provided symptom relief at lower doses (< 100 MET-min/week). These findings offer valuable insights for personalized exercise prescriptions in the management of PD, aiming to improve motor symptoms and enhance patient quality of life.

Abbreviations

PD

Parkinson's disease

MDs

Mean differences

CrIs

Credible intervals

CINeMA

Confidence in Network Meta-Analysis

MCID

Minimum Clinically Important Difference

SE

Sensory exercise

DA

Dopamine

ACSM

American College of Sports Medicine

WHO

World Health Organization

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analysis

UPDRS

Unified Parkinson's Disease Rating Scale

MDS

Movement Disability Society

RCTs

Randomized controlled trials

AE

Aerobic exercise

AQE

Aquatic exercise

BGT

Balance and gait training

BWS

body weight support exercise

Mix

Mixed exercise

RT

Resistance training

SE

Sensory exercise

TC

Tai Chi

METs

Metabolic Equivalent of Tasks

NMA

Network meta-analysis

MBNMA

Model-Based Network Meta-Analysis

DIC

Deviance Information Criterion

Confirmation of ethical compliance

No ethical approval and consent was required for the current study.

Consent for publication

Not applicable.

Funding information

No funding.

CRediT authorship contribution statement

Junyu Wang: Methodology, Writing – original draft. Yuan Yuan: Software, Formal analysis. Ting Xie: Software, Formal analysis. Ligong Zhang: Resources, Formal analysis. Hong Xu: Data curation, Validation. Shu-Cheng Lin: Investigation, Validation. Yong Yang: Writing – review & editing. Dong Zhu: Resources, Funding acquisition. Jie Zhuang: Conceptualization.

Declaration of interest statement

The authors declare no competing interests.

Acknowledgment

Not applicable.

Appendix A Supplementary data

Supplementary material Image 1

Appendix A Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.jsams.2025.01.003.

Table 1

Descriptive characteristics of the included studies by exercise type.

Variable	AE	AQE	BGT	BWS	Dance	Mix	Qigong	RT	SE	Stretch	TC	Yoga
Age (years)	65.2 ± 7.5	66.3 ± 8.1	64.7 ± 7.9	65.5 ± 7.4	63.1 ± 6.8	65.0 ± 8.0	67.2 ± 6.5	64.8 ± 7.3	64.9 ± 7.6	66.5 ± 7.8	64.3 ± 6.9	63.8 ± 7.1
Years of diagnosis	5.8 ± 2.1	6.2 ± 2.3	6.0 ± 2.2	6.1 ± 2.0	5.5 ± 2.4	6.3 ± 2.1	6.5 ± 2.1	5.7 ± 2.0	5.9 ± 2.3	6.1 ± 2.2	5.6 ± 2.1	6.2 ± 2.0
Session duration (min)	55.3 ± 12.7	50.1 ± 10.5	58.9 ± 13.2	52.0 ± 11.8	60.2 ± 14.1	57.5 ± 12.3	45.0 ± 9.8	54.0 ± 11.5	52.7 ± 10.9	48.5 ± 9.7	59.2 ± 13.5	50.8 ± 11.3
Training frequency (sessions/week)	2.9 ± 0.8	2.7 ± 0.6	3.1 ± 0.7	2.8 ± 0.7	3.0 ± 0.9	2.9 ± 0.7	2.5 ± 0.5	3.0 ± 0.8	2.8 ± 0.6	2.6 ± 0.5	3.0 ± 0.7	2.7 ± 0.6
Intervention duration (weeks)	12.5 ± 4.9	11.0 ± 3.7	13.2 ± 5.1	10.8 ± 3.9	14.0 ± 5.2	12.0 ± 4.3	10.5 ± 3.5	12.8 ± 4.7	11.5 ± 3.6	10.0 ± 3.4	13.5 ± 4.9	11.2 ± 3.8