Physical exercise for people with Parkinson’s disease: a systematic review and network meta‐analysis
Physical exercise is effective in managing Parkinson's disease (PD), but the relative benefit of different exercise types remains unclear.
To compare the effects of different types of physical exercise in adults with PD on the severity of motor signs, quality of life (QoL), and the occurrence of adverse events, and to generate a clinically meaningful treatment ranking using network meta‐analyses (NMAs).
An experienced information specialist performed a systematic search for relevant articles in CENTRAL, MEDLINE, Embase, and five other databases to 17 May 2021. We also searched trial registries, conference proceedings, and reference lists of identified studies up to this date.
We included randomized controlled trials (RCTs) comparing one type of physical exercise for adults with PD to another type of exercise, a control group, or both.
Data collection and analysis
Two review authors independently extracted data. A third author was involved in case of disagreements.
We categorized the interventions and analyzed their effects on the severity of motor signs, QoL, freezing of gait, and functional mobility and balance up to six weeks after the intervention using NMAs. Two review authors independently assessed the risk of bias using the risk of bias 2 (RoB 2) tool and rated the confidence in the evidence using the CINeMA approach for results on the severity of motor signs and QoL. We consulted a third review author to resolve any disagreements.
Due to heterogeneous reporting of adverse events, we summarized safety data narratively and rated our confidence in the evidence using the GRADE approach.
We included 156 RCTs with a total of 7939 participants with mostly mild to moderate disease and no major cognitive impairment. The number of participants per study was small (mean 51, range from 10 to 474). The NMAs on the severity of motor signs and QoL included data from 71 (3196 participants), and 55 (3283 participants) trials, respectively. Eighty‐five studies (5192 participants) provided safety data. Here, we present the main results.
We observed evidence of beneficial effects for most types of physical exercise included in our review compared to a passive control group. The effects on the severity of motor signs and QoL are expressed as scores on the motor scale of the Unified Parkinson Disease Rating Scale (UPDRS‐M) and the Parkinson's Disease Questionnaire 39 (PDQ‐39), respectively. For both scales, higher scores denote higher symptom burden. Therefore, negative estimates reflect improvement (minimum clinically important difference: ‐2.5 for UPDRS‐M and ‐4.72 for PDQ‐39).
Severity of motor signs-The evidence from the NMA (71 studies; 3196 participants) suggests that dance has a moderate beneficial effect on the severity of motor signs (mean difference (MD) ‐10.32, 95% confidence interval (CI) ‐15.54 to ‐4.96; high confidence), and aqua‐based, gait/balance/functional, and multi‐domain training might have a moderate beneficial effect on the severity of motor signs (aqua‐based: MD ‐7.77, 95% CI ‐13.27 to ‐2.28; gait/balance/functional: MD ‐7.37, 95% CI ‐11.39 to ‐3.35; multi‐domain: MD ‐6.97, 95% CI ‐10.32 to ‐3.62; low confidence). The evidence also suggests that mind‐body training and endurance training might have a small beneficial effect on the severity of motor signs (mind‐body: MD ‐6.57, 95% CI ‐10.18 to ‐2.81; endurance: MD ‐6.43, 95% CI ‐10.72 to ‐2.28; low confidence). Flexibility training might have a trivial or no effect on the severity of motor signs (MD 2.01, 95% CI ‐4.82 to 8.98; low confidence). The evidence is very uncertain about the effects of strength/resistance training and "Lee Silverman Voice training BIG" (LSVT BIG) on the severity of motor signs (strength/resistance: MD ‐6.97, 95% CI ‐11.93 to ‐2.01; LSVT BIG: MD ‐5.49, 95% CI ‐14.74 to 3.62; very low confidence).
Quality of life-The evidence from the NMA (55 studies; 3283 participants) suggests that aqua‐based training probably has a large beneficial effect on QoL (MD ‐14.98, 95% CI ‐23.26 to ‐6.52; moderate confidence). The evidence also suggests that endurance training might have a moderate beneficial effect, and that gait/balance/functional and multi‐domain training might have a small beneficial effect on QoL (endurance: MD ‐9.16, 95% CI ‐15.68 to ‐2.82; gait/balance/functional: MD ‐5.64, 95% CI ‐10.04 to ‐1.23; multi‐domain: MD ‐5.29, 95% CI ‐9.34 to ‐1.06; low confidence). The evidence is very uncertain about the effects of mind‐body training, gaming, strength/resistance training, dance, LSVT BIG, and flexibility training on QoL (mind‐body: MD ‐8.81, 95% CI ‐14.62 to ‐3.00; gaming: MD ‐7.05, 95% CI ‐18.50 to 4.41; strength/resistance: MD ‐6.34, 95% CI ‐12.33 to ‐0.35; dance: MD ‐4.05, 95% CI ‐11.28 to 3.00; LSVT BIG: MD 2.29, 95% CI ‐16.03 to 20.44; flexibility: MD 1.23, 95% CI ‐11.45 to 13.92; very low confidence).
Adverse events-Only 85 studies (5192 participants) provided some kind of safety data, mostly only for the intervention groups. No adverse events (AEs) occurred in 40 studies and no serious AEs occurred in four studies. AEs occurred in 28 studies. The most frequently reported events were falls (18 studies) and pain (10 studies). The evidence is very uncertain about the effect of physical exercise on the risk of adverse events (very low confidence).
Across outcomes, we observed little evidence of differences between exercise types.
We found evidence of beneficial effects on the severity of motor signs and QoL for most types of physical exercise for people with PD included in this review, but little evidence of differences between these interventions. Thus, our review highlights the importance of physical exercise regarding our primary outcomes severity of motor signs and QoL, while the exact exercise type might be secondary. Notably, this conclusion is consistent with the possibility that specific motor symptoms may be treated most effectively by PD‐specific programs. Although the evidence is very uncertain about the effect of exercise on the risk of adverse events, the interventions included in our review were described as relatively safe. Larger, well‐conducted studies are needed to increase confidence in the evidence. Additional studies recruiting people with advanced disease severity and cognitive impairment might help extend the generalizability of our findings to a broader range of people with PD.
Moritz Ernst, Ann-Kristin Folkerts, Romina Gollan, Emma Lieker, Julia Caro-Valenzuela, Anne Adams, Nora Cryns, Ina Monsef, Antje Dresen, Mandy Roheger, Carsten Eggers, Nicole Skoetz, Elke Kalbe
Plain language summary
Physical exercise for people with Parkinson's disease: what type of exercise works best?
Parkinson's disease (PD) is a progressive disorder of the nervous system that mostly affects people over 60. Symptoms begin gradually and include movement issues, such as trembling, stiffness, slowness of movement and balance, and coordination issues. People with PD can also have emotional and mood problems, fatigue, sleep problems, and thinking difficulties. The disorder cannot be cured, but the symptoms can be relieved, for example, with medicine or surgery. Moreover, people with PD may benefit from physiotherapy or other forms of physical exercise, such as dancing. But it remains unclear if some of these exercise types work better than others.
What was our aim?
We wanted to find out what type of physical exercise works best to improve movement and quality of life for people with PD. We also wanted to find out what type of exercise causes the least unwanted effects.
What did we do?
We searched for studies that compared physical exercise with no physical exercise or with another physical exercise type. We compared and summarized their short‐term results, and rated our confidence in the evidence, based on factors such as study methods and number of people included. We only studied short‐term results.
What did we find?
We found 156 studies on different physical exercise types for people with PD. The studies included a total of 7939 people. The smallest study was conducted with 10 people and the biggest with 474 people. The average participant age was between 60 and 74 years. The studies were conducted in countries around the world, with the highest number (34) in the USA. Of the included studies, 71 (3196 people) provided information on movement; 55 (3283 people) provided information on quality of life, and 85 (5192 people) provided information on unwanted effects.
What are the key results?
Many types of physical exercise worked well for people with PD compared to no physical exercise.
Dance has a moderate beneficial effect on movement. Aqua‐based training, gait/balance/functional training, and training that consists of several exercise types (i.e. multi‐domain training) might have a moderate beneficial effect on movement. Mind‐body (e.g. tai chi or yoga) and endurance training might have a small beneficial effect on movement. Flexibility training might have little to no effect on movement. We are very uncertain about the effects of strength/resistance training and the PD‐specific physical therapy "Lee Silverman Voice training BIG" (LSVT BIG) on movement.
Aqua‐based training probably has a large beneficial effect on quality of life. Endurance training might have a moderate, and gait/balance/functional and multi‐domain training might have a small beneficial effect on quality of life. We are very uncertain about the effects of mind‐body training, gaming, strength/resistance training, dance, LSVT BIG, and flexibility training on quality of life.
Our confidence in the effects ranged from high to very low. When our confidence was reduced, it was often because of two reasons. First, not all of the studies provided information on movement or quality of life from all the people who participated. Second, studies were very small.
Only 85 studies provided some information about unwanted effects, and mostly only for the physical exercise groups, not the groups who did not do exercise. No unwanted effects were reported in 40 studies. No serious unwanted effects were reported in four studies. Unwanted effects were reported in 28 studies. The unwanted effects reported most frequently were falls (18 studies) and pain (10 studies). We could not say what type of exercise causes the least unwanted effects because studies did not provide information about everything we needed. That is why we are very uncertain about the results on unwanted effects.
What does this mean?
We found that many types of physical exercise can help improve movement and quality of life for people with PD. We found scant evidence that certain exercise types work better than others. Therefore, for movement and quality of life, we think physical exercise is important, but the exact exercise type might be less important. Still, it is possible that some symptoms may be relieved best with specific types of training made for people with PD. The types of training we included seemed to be quite safe.
Larger, well‐designed studies are needed to increase our confidence in the evidence. Also, more research is required to understand the features that influence the effects of exercise. More studies involving people who have worse symptoms could help extend the results to more people with PD.
How up to date is this review?
The evidence is up to date to May 2021.
Moritz Ernst, Ann-Kristin Folkerts, Romina Gollan, Emma Lieker, Julia Caro-Valenzuela, Anne Adams, Nora Cryns, Ina Monsef, Antje Dresen, Mandy Roheger, Carsten Eggers, Nicole Skoetz, Elke Kalbe
Implications for practice
We provide evidence of beneficial effects on the severity of motor signs, quality of life (QoL), and functional mobility and balance for most types of physical exercise for people with Parkinson's disease (PD) included in this systematic review. We also observed evidence of superiority in some effects of aqua‐based training compared to effects of other interventions (i.e. an effect on QoL superior to the effects of gait/balance/functional training and multi‐domain training; and an effect on functional mobility and balance superior to the effects of gait/balance/functional training, strength/resistance training, and multi‐domain training). We did not identify any further evidence of differences between the exercise types. Also, while some interventions were among the three highest‐ranked exercise types multiple times (i.e. aqua‐based training, dance, mind‐body training, and strength/resistance training), these results should be interpreted carefully due to the lack of full data on all interventions in some analyses, and large confidence intervals in the effect estimates.
In summary, the overall pattern of results across outcomes and interventions provides only little evidence of differences between the exercise types included in this review. Thus, our systematic review highlights the importance of physical exercise for people with PD in general, while the exact exercise type might be secondary with respect to the rather global outcome measures severity of motor signs and QoL. Therefore, the personal preferences of people with PD should be given special consideration. Nevertheless, fundamental principles of exercise should be taken into account when establishing an individual training routine. For example, the World Health Organization (WHO) guidelines for physical activity for adults living with disability recommend that a variety of exercise types are undertaken, including aerobic physical activity, muscle‐strengthening activities, and multi‐component physical activity that emphasizes functional balance and strength training (WHO 2020). Moreover, people improve at what they practice. Thus, people with PD might be encouraged to select among the diverse landscape of available exercise programs according to their personal preferences, and establish a training routine that includes a variety of modes and addresses their individual goals, impairments, and activity limitations. Our results are consistent with the possibility that specific motor symptoms in PD (e.g. freezing of gait) may be treated most effectively with PD‐specific programs rather than with 'any kind' of physical exercise. Overall, people with PD should be advised to seek professional advice, including assessment of motor and non‐motor symptoms, in order to develop a training agenda based on their individual needs.
In this review, we observed up to large beneficial effects of physical exercise on the severity of motor signs. When expressed as mean differences on the motor scale of the Unified Parkinson Disease Rating Scale (UPDRS‐M), the point estimates of the beneficial effects compared to a passive control group ranged between ‐10.32 (dance) and ‐5.49 (Lee Silverman Voice training BIG [LSVT BIG]). Comparing these effects to effects of pharmacotherapy in people with PD is difficult, because all participants in our review received pharmacological treatment, and physical exercise is always an "add on". Moreover, the interpretability of the effect sizes is limited due to the imprecision of the estimates. However, descriptively, the beneficial effects of physical exercise found in our review were within the range or exceeded the point estimates of the beneficial effects of pharmacological agents compared to placebo that were reported in a network meta‐analysis on the efficacy and tolerability of the most frequently used drugs in the treatment of people with PD (range between ‐6.05 for levodopa and ‐1.60 for cabergoline; Zhuo 2017). Notably, comparable to our results, the drug effects were also highly heterogeneous.
In addition to the evidence of the efficacy, no major safety concerns were raised for the interventions included in our review. Therefore, several exercise programs may be selected from when establishing a training routine, provided that there are no individual safety concerns. Importantly, as the disease progresses, safety concerns may increase and the availability of safe exercise options may decrease.
We observed evidence of beneficial effects in favor of physical exercise compared with a passive control group, more frequently when analyzing studies with an intervention lasting for 12 weeks or longer compared to studies with an intervention lasting for a shorter period. Although these results should be interpreted carefully, longer training periods might have a positive impact on the effects of exercise for people with PD. As we only analyzed outcomes assessed shortly after the intervention, our review does not allow us to draw conclusions on the sustainability of the effects of physical exercise. Nevertheless, one might assume that people with PD would benefit from exercising continuously over the course of disease in order to maintain beneficial effects. Also, while eligible studies in this review had to consist of at least five directly supervised sessions, and the impact of remote supervision could not be investigated, it might be helpful in maintaining the individual training routine.
When interpreting the results of this systematic review, it should be recognized that network meta‐analyses cannot replace direct head‐to‐head comparisons. Furthermore, although we observed only very little evidence of differences in the effects of different types of exercise, we cannot rule out the possibility that differences exist between or within the categories we used that might be clinically relevant for individuals.
Clinicians and other health professionals informing people with PD about the beneficial effects observed in this review may increase their motivation to perform physical exercise. Also, given the fact that various types of exercise show positive effects for people with PD, it would be helpful to provide regional information on the availability of specific exercise offers; for example, by self‐help groups.
Implications for research
Larger, well‐conducted studies are needed to increase the confidence in the evidence. In particular, 49% of study results had a high risk of bias. Most frequently, we had concerns regarding bias due to deviations from the intended interventions, as the results reported by trialists frequently lacked data from a substantial proportion of participants who had been randomized. Therefore, in order to reduce bias and increase the confidence in the effects, trialists should report results from intention‐to‐treat analyses and include data from all participants randomized. Furthermore, our confidence in the effects was frequently limited due to large confidence intervals, large prediction intervals, or both; these primarily affected our assessment of imprecision and heterogeneity and, may be, in part, a result of small sample sizes. In fact, the studies were usually small: on average, only 51 participants were randomized per study (range from 10 to 474), and data for the analyses were provided for only 21 participants per trial arm (range from 4 to 115). Among the 156 studies included in our review, only 77 studies (49%) described considering test power a priori. Thirty‐nine studies considered test power after conducting the trial, and 40 studies did not address this issue. Therefore, in order to increase the confidence in the evidence, more trialists should consider test power when designing a trial, and intend to recruit larger samples.
The samples of most studies included in our review were limited to people with mild to moderate PD and without severe cognitive impairment or dementia. Additional studies on physical exercise recruiting people with advanced disease severity and cognitive impairment might help extend the generalizability of our findings to a broader range of people with PD.
The efficacy outcomes analyzed in this review consisted of two primary outcomes (i.e. the severity of motor signs and QoL), for which we performed additional evaluations (i.e. sensitivity analyses and in‐depth risk‐of‐bias assessments), and two secondary outcomes (i.e. freezing of gait and functional mobility and balance). The severity of motor signs, usually measured using the UPDRS‐M, was the most frequently reported outcome in the included studies. However, as the UPDRS‐M was designed to measure changes in motor tone and amplitude throughout the entire body, it may not highlight other important aspects, such as gait and balance. Furthermore, in this review, we did not analyze other well‐established tools to measure aspects related to gait and balance, such as the Berg Balance Scale (Berg 1989) or the Falls Efficacy Scale (Tinetti 1990). Depending on the intervention, effects on specific outcomes may be expected to different degrees: for example, gait training may primarily affect freezing of gait while other exercise programs may have a more global effect on disease severity. Thus, investigating the differential effects of exercise more precisely would require extending our focus beyond global effects. Therefore, our results may be complemented by future evidence syntheses that address other specific effects of exercise by analyzing other outcomes and tools included in core outcome sets.
Judgments about the general and comparative safety of different types of physical exercise for people with PD based on our review remain very uncertain, as we could not conduct any quantitative analyses due to limited, incomplete, and heterogeneous reporting of the occurrence of adverse events. In order to facilitate the conduct of evidence syntheses beyond a narrative report of the data, trialists should consider reporting the safety of interventions more consistently and completely for all study arms, including control groups. This could increase the confidence in the evidence of the safety of different physical interventions for people with PD.
As we focused on the evaluation of outcomes assessed up to six weeks after the intervention, our conclusions are limited to the short‐term impact of physical exercise for people with PD. Therefore, more researchers conducting evidence syntheses in the future should consider analyzing the medium‐ and long‐term effects of physical exercise as well.
Finally, comparing groups of interventions that represented different exercise types, we investigated the effects of physical exercise using a relatively broad unit of specification. Furthermore, we conducted subgroup analyses only by the length of intervention, while we could not study the impact of further effect modifiers specified in the protocol (i.e. age, sex, cognitive stage). As a result, our review cannot address questions on the impact of several features that might moderate the effect of physical exercise, such as characteristics of the population (e.g. age, sex, cognitive stage, duration and/or severity of disease, phenotype, skills, or personal preferences and joy), parameters of the intervention (e.g. mode, intensity, frequency, complexity, supervision and feedback, specificity, personalization, or use of technology), or aspects of the study design (e.g. timing of assessment relative to medication status). More nuanced approaches accounting for these features are needed to better understand their role in the effects of exercise for people with PD. First, more investigators could study the effects of physical exercise in specific populations (e.g. people with severe cognitive impairment or dementia). Second, more investigators could conduct studies that directly compare interventions varying in one or more potentially relevant features. Third, evidence syntheses could study the impact of one or more of these features by defining more specific research questions, selecting other effect modifiers, and/or defining other subgroups (i.e. conduct subgroup analyses in studies that include people with mild cognitive impairment, and in studies that include people with severe cognitive impairment or dementia). Investigating these features may help to better understand the effects of exercise for people with PD and to improve the design of individually‐tailored exercise programs.