Physiotherapy in Parkinson’s Disease: A Meta-Analysis of Present Treatment Modalities

Search Strategy

A systematic literature search including articles published until June 2020 in the PubMed (Medline), CINAHL, Embase, and Web of Science databases was conducted. The search strategy was identical to the one used in the European Guideline. ¹⁴ The search criteria were kept broad and trials were included when they evaluated (a) any type of physiotherapy intervention, (b) people with PD, and (c) any outcomes related to motor function or quality of life. Search terms included “Parkinson’s disease,” “Parkinson disease,” “physiotherapy,” “randomized controlled trial,” “controlled clinical trial,” “systematic review,” and “guideline.” A flowchart describing the literature selection is presented in Figure 1. Full-text articles, published abstracts, and conference proceedings were included. In addition, reference lists were hand searched to identify further relevant articles.

OPEN IN VIEWER

Figure 1.

Flowchart describing the search strategy.

Inclusion Criteria

RCTs that studied a physiotherapy intervention compared to no intervention, sham therapy (eg, stretching exercise or usual care with no specific exercise component) or active therapy in people with PD were included, regardless of disease stage and severity. There was no start date for trials to be included; the oldest included trial is from 1994 and the most recent one from 2020. Trials that included subjects with atypical or secondary parkinsonism were not considered. Based on the categories used for the development of the European Guideline, we included conventional physiotherapy, treadmill training, strategy training, dance, martial arts, Nordic walking, whole body vibration, and massage. These categories were identified by representatives of 20 European professional associations and people with PD. ¹⁴ In addition, trials investigating one of the following categories were included: resistance training, aerobic exercises, hydrotherapy, balance and gait training, dual tasking, and exergaming. To be included, trials had to report a precise description of the applied exercise intervention in terms of training regime, frequency, intensity, duration, and progression. Trials also had to be written in English. Trials were excluded if exercise interventions (experimental or control) were not adequate for meta-analytic comparisons, for example, when mean (differences) and/or standard deviations were not described or if the intervention was not considered a physiotherapy intervention (eg, experimental trials with only 1 day of treatment). Also, multidisciplinary interventions were excluded because the effectiveness of physiotherapy alone cannot be derived from these trials.

Conventional physiotherapy was defined as all active (exercise) interventions traditionally used by physiotherapists to manage people with PD, such as traditional physiotherapy techniques or multifaceted interventions combining different physiotherapy techniques.^8,15 Interventions using transcranial direct current stimulation were also included in this category. Hydrotherapy and trials using dual task training were treated as separate categories. The strategy training category included strategies for complex motor sequences and cueing interventions that use temporal or spatial external stimuli associated with the initiation and ongoing facilitation of motor activity, particularly gait. In the dance category, various types of dance were included, such as tango, American ballroom, and waltz/foxtrot. Tai Chi and Qigong were included in the martial arts category. The treadmill training category included trials using treadmill training as a stand-alone intervention, just as the Nordic walking category included trials that used Nordic walking as a stand-alone intervention. Finally, we considered exercise or training interventions with a therapeutic goal as a physiotherapy intervention.

Outcome Measures

Based on the European Guideline, where representatives of 20 European professional associations and patients graded these as being critical, ¹⁴ the following outcomes were selected:

Data Extraction

For articles published until the end of 2012, SK and the Guideline Development Group performed the data extraction and subsequent analysis. For this present update, one reviewer (DR) screened all identified trials on title and abstract and any doubt was discussed with another independent reviewer (NdV). Final inclusion was based on consensus between the two reviewers. Data extraction was performed using a standardized data collection form. We extracted the main characteristics of each included study, including information about the sample (initial sample size, age, gender), intervention (experimental and control), and physiotherapy outcomes (assessment and follow-up). Adverse events were not taken into account. Only data assessed immediately after the intervention period were used; data from longer follow-up periods were not taken in to account due to the large variability in intervals. Moreover, only few trials reported data on longer-term follow-up (eg, 6 months or longer). For trials with more than 2 intervention arms, the 2 arms most likely to show the largest effect were included (eg, intervention group and no intervention control group, instead of sham therapy group). We contacted authors via email (sending 1 reminder) if descriptive information or data for statistical analysis were missing, such as mean (differences) and/or standard deviations regarding the included outcome measures.

Methodological Quality

For trials published until the end of 2012, SK and the Guideline Development Group assessed the methodological quality using the Grading of Recommendations Assessment Development and Evaluation (GRADE) quality assessment tool. ¹⁶ For this present update, one reviewer (DR) assessed the methodological quality using GRADE. A second reviewer (JD) assessed a sample of the included trials to check for consensus, independent from the first reviewer. This sample was selected randomly and included one-third of all included trials. The reviewers agreed on the final quality grading of all these trials. GRADE is endorsed by many major organizations, such as the Cochrane Collaboration, the World Health Organization, and the British Medical Journal. It is a valid measure of the methodological quality of clinical trials. Each study was rated based on the comprehensiveness of the reported information, which led to an overall methodological quality rating of high, moderate, low, or very low. RCTs started at the high level. Downgrading was based on incorrect or unclear description of the randomization procedures, (single) masked, concealment of allocation, eligibility criteria, differences between treatment groups at baseline and failure to use an intention to treat analysis. Finally, the GRADE quality rating was summarized per outcome measure in each category, also taking into account heterogeneity between trials, imprecision and publication bias (Appendix, Table 3).

Data Synthesis

We used the Cochrane Collaboration’s software Review Manager (RevMan version 5.3, Copenhagen) to perform the meta-analysis (available from the website for free: http://tech.cochrane.org/revman/download). All outcome variables used for analysis were continuous data and were entered in the database as mean values with standard deviations of postintervention outcome scores. The extracted continuous data from different scales were converted to standardized mean differences (SMD) with 95% confidence intervals (CIs) and adjusted for sample size. We calculated SMDs in order to combine data of outcomes with different units. For scales indicating improvement by decrease (eg, Unified Parkinson’s Disease Rating Scale, Timed Up and Go test), mean values were adjusted by multiplication with −1. ¹⁷ A significant effect size of less than 0.2 was considered a small effect, a size of 0.2 to 0.5 a moderate effect, and an effect size of more than 0.8 a large effect. ¹⁸ The meta-analysis results were expressed as pooled effects, with corresponding 95% CIs and P values.

Study Characteristics

We found 425 potentially eligible trials and these were screened for inclusion. A total of 191 full-text papers fulfilled the inclusion criteria and were included in the analysis (references in Supplementary Material 15). The remaining 234 studies were excluded for various reasons, mainly due to missing data for statistical analysis. The type of exercise interventions varied considerably. In general, the interventions included either a specific type of exercise (eg, Tai Chi, tango dance, or aerobic training) or a combination of various exercise components (eg, strength, flexibility, and balance training). Intervention characteristics also varied extensively; however, on average interventions lasted between 4 and 12 weeks, sporadically up to 6 months or even 2 years. Training sessions were organized between 1 and 7 times a week, but on average 3 to 4 times. Sessions lasted an average of 67 ± 52 minutes. The study characteristics of all included article are described in Supplementary through 12 in the Supplementary Table 1. There were insufficient data to perform a meta-analysis on whole body vibration, massage, boxing, yoga, and fall prevention programs. These interventions are briefly discussed in the section “Other Physiotherapy Techniques”. Among the eligible outcome measures, the (MDS)-UPDRS-III, TUG, and gait speed were most frequently reported. Table 1 shows the results of the meta-analysis regarding the selected types of interventions and main outcome measures. Supplementary File 3 shows the forest plots accompanying Table 1.

OPEN IN VIEWER

Table 1.

Results of the Outcome Measures Related to the Type of Intervention. ^a

	Motor symptoms	Balance outcomes		Gait outcomes					Quality-of-life outcomes
	(MDS-) UPDRS	TUG	BBS	6MWT	10MWT	Gait speed	Stride length	Cadence	PDQ-39
Conventional PT (n = 45)	0.48 [0.35, 0.60]** n = 26	0.11 [−0.07, 0.29] n = 14	0.03 [−0.25, 0.31] n = 3	0.13 [−0.01, 0.28] n = 6	0.30 [0.01, 0.59] n = 56**	0.24 [0.03, 0.45] n = 11	0.28 [−0.11, 0.67] n = 4	0.52 [0.11, 0.92] n = 4**	0.11 [0.01, 0.22] n = 17*
Resistance training (n = 17)	0.20 [−0.02, 0.42] n = 7	0.19 [−0.05, 0.43] n = 6	0.31 [−0.47, 1.09] n = 1	0.67 [0.09, 1.24] n = 2*		−0.07[−0.30, 0.16]n = 6	−0.63 [−1.20, −0.06]n = 2		0.23 [0.01, 0.43]n = 7
Treadmill training (n = 32)	0.10 [−0.09, 0.29]n = 16	0.07 [−0.19, 0.34]n = 8	0.21 [−0.13, 0.55]n = 5	0.29 [0.04, 0.55]* n = 9	0.47 [0.08, 0.85]n = 4*	0.52 [0.34, 0.69]n = 23**	0.20 [− 0.04, 0.44]n = 12	0.12 [−0.15, 0.39]n = 9	−0.06 [−0.49, 0.36]n = 4
Strategy training (n = 14)	0.43 [−0.32, 1.18]n = 1	0.53 [0.23, 0.82]n = 6**	0.12 [−0.43, 0.68]n=1	−0.02 [−0.90, 0.85]n = 1	−0.04 [−0.60, 0.51]n = 1	0.45 [0.13, 0.76]n = 6**	0.52 [0.11, 0.93]n = 4	0.47 [−0.01, 0.95]n = 3	0.17 [−0.34, 0.67]n = 2
Dance (n = 11)	0.72 [0.44, 1.01]n = 8**	0.49 [0.19, 0.80]n = 8**	0.59 [0.27, 0.91]n = 7**	0.51 [0.10, 0.91]n = 4*		0.32 [−0.02, 0.66]n = 6	0.11 [− 0.36, 0.58]n = 3	1.33 [0.31, 2.34]n = 1*	−0.13 [−0.63, 0.38]n = 3
Martial arts (n = 11)	0.26 [0.08, 0.43]n = 10*	0.56 [0.36, 0.77]n = 7**	0.24 [−0.02, 0.49]n = 4	0.20 [−0.15, 0.55]n = 3		0.29 [0.07, 0.52]n = 6*	0.32 [0.07, 0.56]n = 3*		−0.09 [−0.66, 0.47]n = 2
Nordic walking (n = 3)	0.74 [0.24, 1.24]n = 3**	0.55 [0.06, 1.04]n = 3*	0.99 [0.48, 1.50]n = 3**	0.94 [0.28, 1.60]n = 2*	0.37 [−0.15, 0.90]n = 2				0.46 [−0.18, 1.11]n = 1
Aerobic exercises (n = 5)	0.92 [0.61, 1.22]n = 4**	0.80 [0.44, 1.15]n = 1**	[−0.81, 0.83]n = 1	1.02 [0.69, 1.34]n = 3**					0.20 [−0.11, 0.52]n = 2
Balance and gait training (n = 28)	0.34 [0.11, 0.56]n = 12*	0.36 [0.15, 0.58]n = 11**	0.57 [0.35, 0.79]n = 12**	−0.12 [−0.48, 0.23]n = 4	0.15 [−0.56, 0.85]n = 2	0.28 [0.12, 0.44]n = 17**	0.36 [0.12, 0.59]n = 10*	0.24 [−0.01, 0.48]n = 8	0.28 [−0.04, 0.60]n = 6
Hydrotherapy (n = 8)	−0.11 [−0.41, 0.19]n = 5	0.50 [0.25, 0.75]n = 8**	0.31 [0.04, 0.59]n = 7						0.39 [−0.01, 0.79]n = 3
Dual task (n = 3)	−0.18 [−0.79, 0.42]n = 2	−0.36 [−1.39, 0.66]n = 1				−0.25 [−0.85, 0.34]n = 2	−0.08 [−0.46, 0.30]n = 1	−0.29 [−0.65, 0.07]n = 2	−0.04 [−0.78, 0.71]n = 1
Exergaming (n = 9)		0.58 [0.29, 0.87]n = 7**	0.47 [0.17, 0.77]n = 6*	0.23 [−0.20, 0.67]n = 3	−1.66 [−2.84, −0.48]n = 1*	0.30 [−0.02, 0.62]n = 4	0.77 [−0.07, 1.60]n = 1		0.45 [0.13, 0.77]n = 4*

Abbreviations: n, number of included studies; PT, physiotherapy; (MDS-)UPDRS-III, Movement Disorder Society–sponsored revision of the) Unified Parkinson’s Disease Rating Scale part III; TUG, Timed Up and Go; BBS, Berg Balance Scale; 6MWT, 6-Minute Walking Test; 10MWT, 10-Minute Walking Test; PDQ-39, Parkinson’s Disease Questionnaire–39; PDQL, Parkinson’s Disease Quality of Life Questionnaire.

a

Data are presented as standardized mean difference [95% CI]. Positive effect estimates mean positive results on each outcome measure. Red cells indicate small significant effect (≤0.2), orange cells indicate moderate significant effect (0.2-0.5), light green cells indicate moderately large significant effect (0.5-0.8), and dark green cells indicate large significant effect (≥0.8).

*

Statistically significant effect, P ≤ .05.

**

Statistically significant effect, P ≤ .001.

Methodological Quality

The GRADE assessment showed that the methodological quality of the included trials varied considerably, with several quality indicators not fully discussed in many publications. Only 132 (66.7%) trials reported the randomization method used. Information on concealment of treatment allocation was poorly reported (57, 28.9%). Masked assessors were reported in 146 (73.8%) trials, and masking of patients was not possible in most trials due to the intervention characteristics. Finally, 45 (22.8%) trials described whether they used an intention to treat analysis, 185 (93.4%) trials reported on the eligibility criteria and in 169 (85.3%) trials the intervention and control group were similar at baseline. Table 3 in the Appendix shows the summary of findings table according to GRADE with detailed evidence per outcome measure. In summary, the evidence for conventional physiotherapy, resistance training, and balance and gait training, was considered to be moderate to high; for treadmill training and martial arts was moderate; for strategy training, dance, dual-task training, Nordic walking, aerobic exercise, hydrotherapy, and exergaming was low to moderate.

Limitations

A first major strength of the present analysis is the fact that different commonly used types of physiotherapy were included, while other trials mostly focused on specific areas of physiotherapy, such as strength training ²⁶ or treadmill training. ²⁷ We provide a comprehensive overview of the effectiveness of physiotherapy for people with PD, thus enabling an easy comparison of all available interventions. A second strength is that the outcome measures used were previously graded as being critical by 20 European professional associations and patients in the European Guideline. ¹⁴ Finally, the rigorous GRADE standards were used to systematically appraise the methodological quality of all included trials.

It can be difficult to perform a meta-analysis on a heterogeneous subject such as physiotherapy. First, included trials varied enormously in the number of exercise sessions per week, exercise intensity, and duration of the intervention. Also, many different interventions were used for the control groups, ranging from usual care to stretching exercises or doing low-intensity exercises at home. All these trials were included since the exercise interventions of these active control groups were expected to have no significant impact on outcome measures. Second, there was great heterogeneity across trials regarding the outcomes used, even to assess the same variable. For example, “number of falls” was measured over different time frames, while near-falls were variably included. Furthermore, “muscle strength” was measured using highly variable methods, which made it difficult to compare trial results. Therefore, trials using these outcomes were excluded from the meta-analysis. Third, the methodological quality and reporting was highly variable, and often inadequate. Lack of information does not necessarily indicate poor quality of the trial, but the level of possible bias was simply difficult to assess. Finally, most trials had relatively small sample sizes, and many evaluated short-term interventions (<12 weeks). Follow-up durations varied as well, at best spanning 6 months to several years, but mostly considerably shorter. Therefore, it was not possible to perform a meta-analysis on long-term effects of physiotherapy interventions. However, a recent review did evaluate the long-term effects of exercise and physiotherapy interventions for people with PD. ²⁸ The authors suggest a potential long-term effect on motor symptoms and physical functioning. Balance training had the longest carry-over effect (up to 12 months), followed by gait and Tai Chi training (up to 6 months). It appears that a training period longer than 12 weeks is needed to achieve clinically meaningful improvement in motor function (MDS-UPDRS-III).

The present meta-analysis found that physiotherapy is effective in improving specific outcomes. Therapists and patients can therefore choose from a variety of treatment strategies, based on the specific symptoms they seek to improve and based on the patients’ personal preferences regarding what exercise modality they relate to. This will facilitate a more patient-centered approach, where patients and therapists have an evidence-based choice between interventions. Being able to adjust the treatment according to patient preferences will be important to enhance motivation and increase long-term adherence to therapy. However, different strategies were only rarely compared back to back, hence any specific choice should be made cautiously until direct comparisons of the various treatment modalities become available.

Future studies should focus on improved methodological quality, including larger sample sizes, longer follow-up and use of relevant, reliable, and sensitive outcomes. Recently, the International Consortium for Health Outcomes Measurement published a consensus set of outcomes for PD and this may be a valuable tool for future trials. ²⁹ Moreover, this meta-analysis highlights the variety of physiotherapy interventions used to treat PD. Specific trials are needed to directly compare different interventions with standardized characteristics regarding training regime, frequency, intensity, duration, and progression. Such work is needed to underpin the most appropriate intervention, including for specific PD subgroups. Elaborating on this, young-onset PD patients are typically underrepresented in research and future work should also focus on the effects of different physiotherapy interventions in younger cohorts. Future studies should also assess the economic benefit and address the risk of adverse events (eg, possible increase in number of falls when patients are stimulated to move more) and determine adequate dosing. Finally, more work is needed to study the appropriate implementation strategies in the community to improve long-term adherence.