Parkinson’s disease (PD) is a complex neurodegenerative disorder (Rubenis 2007) with wide reaching implications for patients and their families. Although disability can occur at all stages of the disease (Deane 2001a), PD is progressive in nature, and so patients face increased difficulties with activities of daily living (ADL) (Kwakkel 2007) and various aspects of mobility such as gait, transfers, balance, and posture (Keus 2007b). Ultimately, this leads to decreased independence, inactivity, and social isolation (Keus 2007b), resulting in reduced quality of life (Schrag 2000).
The management of PD has traditionally centred on drug therapy, with levodopa viewed as the 'gold standard' treatment (Rascol 2002). However, even with optimal medical management, patients with PD experience deterioration in body function, daily activities, and participation (Nijkrake 2007). For this reason, support has been increasing for the inclusion of rehabilitation therapies as an adjuvant to pharmacological and neurosurgical treatment (Gage 2004; Nijkrake 2007), and a call for the move towards multidisciplinary management of this multidimensional condition (Robertson 2003; Rubenis 2007).
The physiotherapist is a member within this multidisciplinary team (Robertson 2008; Rubenis 2007), whose purpose is to maximise functional ability and minimise secondary complications through movement rehabilitation within a context of education and support for the whole person (Plant 2000; Deane 2001a). Physiotherapy for PD focuses on transfers, posture, upper limb function, balance (and falls), gait, and physical capacity and (in)activity by using cueing strategies, cognitive movement strategies, and exercise to optimise the patient’s independence, safety, and well-being, thereby enhancing quality of life (Keus 2004; Keus 2007a).
Referral rates to physiotherapy for people with PD have historically been low (Mutch 1986; Yarrow 1999). However, in recent years, the number of referrals has increased, with a survey by Parkinson’s UK in 2008 reporting that 54% of the 13,000 members surveyed had seen a physiotherapist compared with 27% in a survey undertaken in 1998 (PDS 2008; Yarrow 1999). This rise in referrals may be attributed to two factors. First, guidelines published by the National Collaborating Centre for Chronic Conditions (Nat Collab Centre for Chronic Conditions 2006) recommended that physiotherapy be made available throughout all stages of the disease, raising the profile of the profession. This has been further supported by the publication of Dutch physiotherapy guidelines (Keus 2004), which provide specific information for physiotherapists involved in the management of PD. Second, a substantial increase has been noted in the number of trials completed over the past decade (particularly in the last five years), offering supportive evidence for the inclusion of physiotherapy in the management of PD (Keus 2009).
This Cochrane review assessing the effectiveness of physiotherapy intervention versus no physiotherapy intervention in patients with PD was first published in 2001, and included only 11 randomised controlled trials with a total of 280 participants (Deane 2001a). Most of the trials in the review reported a positive effect in favour of physiotherapy, but few outcome measures were statistically significant. This, combined with the presence of methodological flaws, small sample sizes, and the possibility of publication bias, led Deane et al. to conclude that evidence was insufficient to support or refute the efficacy of physiotherapy for PD (Deane 2001a). This review updates the previous Cochrane review. We appraised and synthesised relevant randomised controlled trials, and we conducted a meta-analysis of outcomes where possible.
To compare the effectiveness of physiotherapy intervention versus no physiotherapy intervention in participants with PD.
To indirectly compare the different physiotherapy interventions used within the various trials.
Criteria for considering studies for this review
Types of studies
For inclusion in the review, we considered all randomised controlled trials (including the first phase of cross-over trials) comparing a physiotherapy intervention with no physiotherapy intervention (including placebo control). We included trials in which the no intervention arm used an active or credible placebo, as long as no physiotherapy was delivered to this group. We included only trials that implemented random methods of treatment allocation.
Types of participants
Participants with a diagnosis of PD (as defined by the authors of the studies):
- Any duration of PD.
- All ages.
- Any drug therapy.
- Any duration of physiotherapy treatment.
Types of interventions
Physiotherapy interventions aim to maximise functional ability and minimise secondary complications through movement rehabilitation within a context of education and support for the whole person. Physiotherapy encompasses a wide range of techniques, so we were inclusive in our definition of physiotherapy interventions (including those not delivered by a physiotherapist) with trials of general physiotherapy, exercise, treadmill training, cueing, dance, and martial arts included.
Types of outcome measures
Gait outcomes such as:
- Two- or six-minute walk test (m) － measures the number of metres a person can walk in two or six minutes, thereby providing a measurement of walking endurance (Kersten 2004).
- Walking speed
- Cadence (steps/min) － measures the number of steps taken in a given period, which is then converted into the number of steps taken per minute (Trew 2005).
- Stride length (m) － measures the average distance (in metres) between two successive placements of the same foot (Whittle 1996).
- Step length (m) － measures the average distance (in metres) between successive foot-to-floor contacts with opposite feet (Trew 2005).
- Freezing of Gait Questionnaire － validated questionnaire for the assessment of freezing of gait. The questionnaire consists of six items, and scores range from 0 to 24, with higher scores corresponding to more severe freezing of gait (Giladi 2000).
Functional mobility and balance outcomes such as:
- Timed Up & Go (s) － measures time taken in seconds for a person to get up from a chair, walk a certain distance (usually three metres), turn around, and walk back to the chair and sit down (Podsiadlo 1991).
- Functional Reach Test (cm) － “the maximal distance one can reach forward beyond arm’s length, while maintaining a fixed base of support in the standing position” (Duncan 1990).
Data on falls such as:
- Number of patients falling － e.g. falls diary.
- Falls Efficacy Scale － 10-item patient-reported questionnaire that measures how confident a person is at carrying out various activities of daily living (ADL). Items are rated from 1 to 10, with higher scores correlating with lower levels of confidence, and a total score of 70 or higher indicating that a person has a fear of falling (Tinetti 1990).
- Falls Efficacy Scale International － 16-item questionnaire that includes the 10 original items of the standard Falls Efficacy Scale, as well as six items regarding higher functioning and social activities. Each item is rated on a scale of 1 to 4, with 1 being ‘not concerned at all’ and 4 ‘very concerned’ (maximum score out of 64) (Yardley 2005).
Clinician-rated impairment and disability measures such as:
- Hoehn & Yahr － scale used to describe how symptoms of Parkinson's disease progress. Scale ranges from 0 to 5, with higher levels indicating greater disability (Hoehn 1967).
- Unified Parkinson's Disease Rating Scale (UPDRS) － designed to assess motor impairment and disability in Parkinson’s disease. Higher scores correspond to greater disability (Fahn 1987).
- Total － scores range from 0 to 176.
- Mental － scores range from 0 to 16.
- ADL － scores range from 0 to 52.
- Motor － scores range from 0 to 108.
- Webster Rating Scale － assessment of severity of disease and clinical impairment against 10 items using a scale of 0=normal to 3=maximum impairment: bradykinesia, rigidity, posture, upper extremity swing, gait, tremor at rest, facial expression, seborrhoea, speech, and self care. Scores range from 0 to 30, with higher scores indicating greater disease severity and disability (Webster 1968).
- Columbia University Rating Scale － assessment of motor impairment and activities of daily living against 13 items, using a five-point scale for each to yield a total score between 0=normal and 65=maximum disability (Yahr 1969).
Patient-rated quality of life such as:
- PDQUALIF － PD-specific health-related quality of life questionnaire containing 32 items in seven dimensions and one item of global health-related quality of life. Total score ranges from 0 to 128, with higher scores indicating poorer quality of life (Welsh 2003).
- PDQL － PD-specific health-related quality of life questionnaire containing 37 items grouped into four subscales. Item scores range from 1 to 5. The PDQL-Summary Index ranges from 37 to 185, with higher scores reflecting better quality of life (Deboer 1996).
- Short Form-36 or -12 － generic short-form health survey consisting of 36 or 12 questions. The SF-36 consists of eight scaled scores assessing vitality, physical functioning, bodily pain, general health perceptions, physical role functioning, emotional role functioning, social role functioning, and mental health. Scores range from 0 to 100, with higher scores corresponding to better quality of life (Ware 1992).
Adverse events (e.g. fractures, pain).
Compliance (e.g. participant adherence, treatment fidelity).
Search methods for identification of studies
The review is based on the Movement Disorders Group search strategy and the following more general search strategy:
- Physiotherapy OR physical therapy OR exercise OR rehabilitation.
- Parkinson OR Parkinson's disease OR Parkinsonism.
- #a AND #b.
Further details on this search strategy are available in the Group's module within The Cochrane Library (www.cochrane.org). This includes explanations of the acronyms, sources, and Websites.
We undertook a systematic search of the literature up to the end of January 2012 for publications or abstracts describing relevant trials. This included searching:
- General biomedical and science electronic databases (without date limiters) including the Movement Disorders Review Group Specialized Register, The Cochrane Library, MEDLINE (1966-2012), EMBASE (1974-2012), CINAHL (1982-2012), and ISI-SCI (1981-2012); rehabilitation databases: AMED (1985-2012), REHABDATA (1995-2012), REHADAT, and GEROLIT (1979-2012); English language databases of foreign language research and third world publications: LILACS (1982-2012), MedCarib (17th Century-2012), and IMEMR (1984-2012).
- The Cochrane Controlled Trials Register, the CentreWatch Clinical Trials listing service, the metaRegister of Controlled Trials, ClinicalTrials.gov, RePORT, PEDro, NIDRR, and NRR.
- Handsearching of general (Lancet, BMJ, JAMA) and specific journals (Movement Disorders, Neurology, Archives of Physical Medicine and Rehabilitation, Clinical Rehabilitation, Physiotherapy, Physical Therapy) from 2001 to the end of January 2012.
- The reference lists of retrieved papers and review articles.
- Abstract books and conference proceedings. This included The XIII International Congress on Parkinson's Disease (1999), The International Congress of Parkinson's Disease and Movement Disorders (1990, 92, 94, 96, 98, 2000, 02, 04, 05, 06, 07, 08, 09, 10, 11), World Congress on Parkinson's Disease and Related Disorders (2009, 2012), and The American Academy of Neurology 51st Annual Meeting (1999).
- Grey literature databases (including theses): Conference Proceedings Citation Index (1982-2010), DISSABS (1999-2012), Conference Papers Index (1982-2012), Index to Theses (1970-2012), Electronic Theses Online Service (EThOS) (16th century-2012), and ProQuest dissertations and theses databases (1861-2012).
Data collection and analysis
Selection of studies
Abstracts of potentially relevant studies from search results were screened by two of the the four review authors involved in study selection (CT, SP, CH, LS). The full paper was obtained if the abstract did not provide sufficient information for investigators to determine eligibility for inclusion in the review. Disagreement was resolved by referral to an additional review author (RS). We contacted authors of potentially eligible studies for further information if details of the trial were unclear.
Data extraction and management
Four review authors (CT, SP, CM, and CH) independently assessed the identified papers and abstracts for trial details and outcome data, and each eligible study was considered by two of these four authors. This was validated by discussion, with any discrepancies resolved by consensus. We recorded trial details on a standard trial description form and included the following: trial name, trial group, authors, randomised comparison, treatment schedule (including duration, number of sessions, type of intervention), other therapy, eligibility criteria, method of randomisation, allocation concealment, blinding, accrual period, number of participants randomised, number of dropouts, duration of follow-up, outcomes reported, use of intention-to-treat analysis, and publication date(s). Outcome data extracted included data on gait, functional mobility and balance, falls, clinician-rated disability scale and patient-rated quality of life, adverse events, compliance/withdrawals, and health economics where available.
We contacted the authors of any eligible unpublished studies to ask whether further details and data for their trial could be provided.
Assessment of risk of bias in included studies
We assessed the full papers for methodological quality by recording eligibility criteria, methods of randomisation and blinding, concealment of allocation, similarity of participants in treatment groups at baseline, cointervention(s) constant, use of active or credible placebo, whether an intention-to-treat analysis was performed, and the numbers of participants lost to follow-up and missing values (see Risk of Bias tables under Characteristics of included studies).
We combined the results of all trials using standard meta-analytic methods to estimate an overall effect for physiotherapy intervention versus no physiotherapy intervention.
All outcomes with data available for meta-analysis were continuous variables, so we calculated the mean difference between treatment arms using mean difference methods (Fleiss 1993). In summary, this involved calculating for each trial the mean change (and standard deviation) from baseline to the postintervention time point for the intervention and no intervention groups. From these, the mean difference and its variance between arms for each trial could be calculated. In some studies, the standard deviation for the mean change was not reported; in these cases, we imputed this standard deviation using the standard deviations for baseline and final scores. To do this, we used the following formula to estimate the variance of the change in score:
These values were then combined using weighted mean difference methods to obtain the overall pooled estimate of the mean difference, with 95% confidence interval, for physiotherapy intervention versus no physiotherapy intervention (control).
If any trials with three or more intervention arms were identified, the following assumptions were made for the analysis:
- If the trial was comparing two or more physiotherapy interventions within the same classification (see subgroup analysis later) versus no intervention, then we combined the data for these physiotherapy interventions to give one comparison of physiotherapy intervention versus no intervention.
- If the trial was comparing two or more physiotherapy interventions in different classifications versus no intervention, then we included that trial in each relevant physiotherapy intervention classification. This meant that some trials were included multiple times in the analysis, and the control arms from these trials were counted more than once in the analysis.
The primary analysis was a comparison of physiotherapy intervention versus no physiotherapy intervention (control) based on change from baseline to the first assessment after the treatment period (which in most cases was immediately post intervention). This was chosen as the primary analysis for this review, as in most trials this was the main data analysis, and few trials reported data at longer-term assessment points (i.e. after six months). Also, some trials allowed participants in the 'no intervention' arm to receive physiotherapy intervention after this point. So this allowed a clean comparison of physiotherapy intervention versus no physiotherapy intervention.
Subgroup analysis and investigation of heterogeneity
The different trials implemented various types of physiotherapy intervention. Therefore trials were divided according to the type of intervention administered:
- General physiotherapy versus control.
- Exercise versus control.
- Treadmill versus control.
- Cueing versus control.
- Dance versus control.
- Martial arts versus control.
To assess for differences between the different types of interventions involved, we performed indirect comparisons using tests of heterogeneity and I
Description of studies
We identified 76 randomised trials of physiotherapy intervention in PD patients. We excluded 31 studies (see Characteristics of excluded studies). The reasons for excluding these trials were as follows: cross-over study with data not presented for the first treatment period or cross-over over a short period (e.g. 1 day) (n=6), not randomised or not properly randomised (n=7), no outcome measures relevant to our review (n=4), multidisciplinary therapy rehabilitation trial (n=4), study was confounded (n=2) and treatment given in trial was not usually used by physiotherapists (n=6), excessive number of withdrawals (n=1), and insufficient information (n=1). There were also six ongoing trials for which data were not yet available (see Characteristics of ongoing studies). Therefore, 39 trials were available for inclusion in the review compared with 11 in the 2001 review (Figure 1).
|Figure 1. Study PRISMA flow diagram.|
The number of participants randomly assigned into the 39 trials ranged from six to 153 participants, with 1827 participants randomly assigned in total (giving an average trial size of nearly 50 participants) (Characteristics of included studies). The assessment period ranged from three weeks to 12 months. The mean age of participants in the trials was 67 years, 64% were male, the mean Hoehn & Yahr stage was 2.4, and participants had had PD for approximately six years ( Table 1).
One trial compared walking on a treadmill listening to music versus walking on a treadmill without music versus listening to music alone (Shankar 2009). We excluded the treadmill without music arm of this trial from the analysis as this was a confounded comparison.
Two three-arm trials compared two exercise interventions with control. One compared exercise versus exercise and education versus control (Klassen 2007), and the other compared exercise versus PD SAFEx versus control (Sage 2009a). The exercise interventions being compared in these studies were considered suitably similar, so we combined the data from the two exercise arms within each trial to obtain one comparison of exercise versus control. Two three-arm trials compared two cueing interventions with control. The overground and treadmill walking groups, each with equally spaced transverse lines as cues, were combined to obtain a single cueing versus control comparison (Almeida 2012). Finger tapping and arm swing interventions were similarly combined (Haase 2011). One four-arm trial compared two types of dance (waltz/foxtrot and tango) and martial arts with control. We combined the two dance arms to obtain one comparison of dance versus control, as well as a martial arts versus control comparison (Hackney 2009).
Four other three-arm trials contributed data to two of the different physiotherapy intervention comparisons. Two of these were trials of cueing versus exercise versus control, which contributed to both the cueing versus control and exercise versus control comparisons (Mak 2008; Thaut 1996). Another trial was of treadmill versus general physiotherapy versus control, which contributed to both the treadmill versus control and general physiotherapy versus control comparisons (Fisher 2008). The last trial, which provided no analysable data, contributed information to two comparisons: general physiotherapy versus control and treadmill versus control (Talakad 2011). The 39 trials therefore contributed data to 44 comparisons within the six different types of physiotherapy interventions － general physiotherapy versus control (n=7), exercise versus control (n=14), treadmill versus control (n=8), cueing versus control (n=9), dance versus control (n=2), and martial arts versus control (n=4).
Below is a summary of the characteristics of included studies. Details of individual studies are given in Characteristics of included studies.
General Physiotherapy versus Control
The seven trials of general physiotherapy versus control involved 244 participants (Chandler 1999; Ellis 2005; Fisher 2008; Homann 1998; Keus 2007b; Stack 2012; Talakad 2011)). Sixty participants split between physiotherapy and treadmill categories are not included in this total as the group splits were not given (Talakad 2011). The mean participant age was 65 years, 69% were male, the mean Hoehn & Yahr stage was 2.4, and mean duration of PD was four years. All trials were of parallel group design, except one, which used a cross-over design (Ellis 2005). Treatment sessions took place over a period of four weeks to 12 months; duration of sessions was described by only two trials (Ellis 2005; Stack 2012). One trial used Bobath training for gait and posture (Homann 1998). The remaining trials provided multifaceted interventions encompassing movement strategies, exercise, hands-on techniques, education, and advice, targeting a wide range of areas including gait, balance, transfers, posture, and physical fitness. Thus, general physiotherapy is a holistic intervention and on the whole uses a combination of techniques that do not routinely include complementary and/or alternative medicine such as acupuncture or hypnotherapy.
Exercise versus Control
The 14 trials of exercise versus control involved 769 participants (Allen 2010; Ashburn 2007; Boehm 2011; Cerri 1994; Goodwin 2009; Klassen 2007; Mak 2008; Meek 2010; Sage 2009a; Schenkman 1998; Schilling 2008; Stozek 2003; Taheri 2011; Thaut 1996). The mean participant age was 69 years, 60% were male, the mean Hoehn & Yahr stage was 2.6, and mean duration of PD was six years. Thirteen trials were of parallel group design, and one used a cross-over design (Boehm 2011). Treatment sessions lasted from 30 minutes to two hours, and took place over a period of three to 24 weeks. Exercise involved a variety of different activities, including strengthening and balance training, walking, falls prevention, neuromuscular facilitation, resistance exercise and aerobic training, and education and relaxation techniques. Although sometimes multifaceted, the primary focus of these interventions was exercise delivery, and treatment was frequently categorised in this way by the trial authors.
Treadmill versus Control
The eight trials of treadmill versus control involved 179 participants (Cakit 2007; Canning 2008; Fisher 2008; Ganesan 2010; Kurtais 2008; Protas 2005; Shankar 2009; Talakad 2011). Sixty participants split between physiotherapy and treadmill categories are not included in this total, as the group splits were not given (Talakad 2011). The mean participant age was 68 years, 61% were male, the mean Hoehn & Yahr stage was 2.4, and mean duration of PD was five years. All trials used a parallel group design. Treatment sessions lasted from 30 to 60 minutes, and took place over a period of four to eight weeks. Treadmill training mainly involved participants walking on a treadmill with speed and/or incline adjustments. Three trials used body weight-supported treadmill training (Fisher 2008; Ganesan 2010; Talakad 2011), and two trials provided gait and step training (Kurtais 2008; Protas 2005).
Cueing versus Control
The nine trials of cueing versus control involved 371 participants (Almeida 2012; de Bruin 2010a; de Bruin 2010b; Haase 2011; Lehman 2005; Mak 2008; Nieuwboer 2007; Shankar 2008; Thaut 1996). The mean participant age was 67 years, 59% were male, the mean Hoehn & Yahr stage was 2.6, and mean duration of PD was seven years. Eight of the trials were of parallel group design, and one used a cross-over design (Nieuwboer 2007). Treatment sessions lasted from four to 30 minutes and took place over a period of a single session to 13 weeks. Three types of cueing were used in the trials: audio (music, spoken instructions), visual (computer images), and sensory (vibration). Six trials applied external cues during gait or gait-related activity, and Mak (Mak 2008) used cues for the rehabilitation of sit-to-stand transfers.
Dance versus Control
The two trials of dance versus control involved 120 participants (Duncan 2012; Hackney 2009). The mean participant age was 69 years, 63% were male, the mean Hoehn & Yahr stage was 2.3, and mean duration of PD was seven years. Both trials used a parallel group design. Dance classes lasted one hour over 12 to 13 weeks, with a trained instructor teaching participants the tango, waltz, or foxtrot.
Martial Arts versus Control
The four trials of martial arts versus control involved 143 participants (Hackney 2009; Marjama-Lyons 2002; Purchas 2007; Schmitz-Hubsch 2006). The mean participant age was 65 years, 74% were male, the mean Hoehn & Yahr stage was 2.1, and mean duration of PD was six years. All the trials were of parallel group design, except one, which used a cross-over design (Purchas 2007). Treatment lasted one hour and took place over a period of 12 to 24 weeks. Participants took classes on Tai Chi (three trials; Hackney 2009; Marjama-Lyons 2002; Purchas 2007) or Qigong (one trial; Schmitz-Hubsch 2006).
Risk of bias in included studies
|Figure 2. Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.|
|Figure 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study.|
Thirty-five trials had a parallel design and four had a cross-over design (Boehm 2011; Ellis 2005; Nieuwboer 2007; Purchas 2007). The cross-over trials had no washout period, with participants assessed at baseline, after the first treatment period, and then after the second treatment period. Most trials looked at the short-term effect of therapy by assessing participants at baseline and immediately or shortly after the physiotherapy intervention period (which ranged from two to 52 weeks). Ten of the parallel design trials (Almeida 2012; Ashburn 2007; Goodwin 2009; Klassen 2007; Lehman 2005; Mak 2008; Meek 2010; Schmitz-Hubsch 2006; Stack 2012; Stozek 2003) reported additional data at assessment points after the treatment period had finished; this may have been at only one week or up to 12 months after the end of the treatment period.
Only six studies (15%; Allen 2010; Ashburn 2007; Duncan 2012; Ellis 2005; Goodwin 2009; Nieuwboer 2007) reported a sample size calculation in the trial report, three of which failed to achieve their target (Ashburn 2007; Duncan 2012; Goodwin 2009).
Eligibility criteria for the trials were broad and varied considerably across trials. The level of detail provided in the eligibility criteria was also variable, with some studies providing a detailed description of the entry criteria, and others just stating “patients with Parkinson’s disease.” Only eight trials (Cakit 2007; de Bruin 2010a; Homann 1998; Keus 2007b; Nieuwboer 2007; Schmitz-Hubsch 2006; Shankar 2008; Stack 2012) stated that a diagnosis of PD by the United Kingdom Brain Bank Criteria (Gibb 1988) was required. It is vital that eligibility criteria are well defined, so that the trial participant population can be determined.
Randomisation Method and Concealment of Allocation
Only 18 trials (46%) described the randomisation method used, of which 11 trials used low-risk methods (e.g. block randomisation, computer random number generators). No details on the randomisation method used were provided for the remaining 21 trials. Further, only 14 trials (36%) stated or gave adequate information that allowed the assessment of whether an adequate concealment of treatment allocation procedure had been used. Five trials were considered to be low risk by virtue of the fact that they used a central randomisation service, and the other nine were considered high risk (i.e. concealment of treatment allocation was potentially compromised － sealed envelopes, picking card or picking from a hat).
Blinding of Assessors
It would be impossible to blind participants and therapists to randomised treatment allocation in trials of physiotherapy. Therefore, such trials are open label by nature, and are consequently liable to the possibility of both performance and attrition bias. However, assessors could be blinded to try to reduce the possibility of bias. Twenty-four (62%) of the thirty-nine studies used blinded assessors (although in one study, the assessors correctly guessed the treatment allocation in nearly 30% of patients; unclear risk; Ashburn 2007), three used unblinded assessors so were classed as high risk, and in the other 12 studies, this information was not provided (classed as unclear risk).
Description of the No Intervention (Control) Group
In most trials (n=34), the control group did not receive any physiotherapy treatment or intervention; however, in five trials (Allen 2010; Ashburn 2007; Fisher 2008; Haase 2011; Shankar 2009), an active placebo was used that attempted to control for the time and attention involved in receiving physiotherapy intervention compared with no treatment. This included contact with a PD nurse, education classes, advice on falls prevention, and listening to music. The control groups were followed-up and were assessed in the same manner as the intervention groups.
Information on cointerventions was provided in 23 trials (59%), with participants continuing with their standard PD medication. In 16 trials, this drug therapy was kept stable (low risk) throughout the duration of the trial, whereas seven trials allowed variation (unclear risk). The remaining 16 trials did not describe drug therapy (unclear risk).
Similarity of Treatment Groups at Baseline
A description of the baseline characteristics of the trial participants is important for determination of whether the trial results are generalisable and for comparison of the characteristics of the two arms to ensure that the randomisation methods were successful.
Six trials (de Bruin 2010b; Ganesan 2010; Homann 1998; Marjama-Lyons 2002; Taheri 2011; Talakad 2011) did not provide any information on the baseline characteristics of participants entered into the trial. Twenty-eight (of the 33) trials that reported baseline data gave this information split by treatment group and showed participants to be similar at baseline. In ten trials; the baseline characteristics of the withdrawn participants were not given (Cakit 2007; de Bruin 2010a; Haase 2011; Hackney 2009; Klassen 2007; Kurtais 2008; Mak 2008; Purchas 2007; Sage 2009a; Schenkman 1998). This, along with the six studies that did not supply baseline data, meant that 261 (14%) of the 1827 randomly assigned participants were not characterised.
Nine trials stated intention-to-treat as the primary method of analysis, although it was not always clear if patients who withdrew from the trial were included in the analysis. The number of patient withdrawals was classed as low risk (≤ 10% of trial participants withdrew) in seven of these nine trials. Three trials stated per protocol as the primary method of analysis. In the other 27 trials, the method of analysis was not described (unclear risk). Of these trials, 12 were considered high risk in terms of the proportion of patients that withdrew (i.e. > 10%), and in 14 trials, the number of participant withdrawals (if any) was not given (unclear risk).
Data Available for Analysis
Thirteen trials were reported in abstract form. We requested further information from authors; six (Boehm 2011; Haase 2011; Klassen 2007; Meek 2010; Purchas 2007; Shankar 2008;) provided additional information, and seven (Cerri 1994; de Bruin 2010b; Ganesan 2010; Homann 1998; Marjama-Lyons 2002; Shankar 2009; Talakad 2011) requests were unsuccessful. Sufficient data were available for meta-analysis for five of the 13 studies (Boehm 2011; Haase 2011; Klassen 2007; Meek 2010; Shankar 2009). Further, one trial had relevant data that could not be extracted as it was available only in graph form (Lehman 2005), and another trial published only median and interquartile range data, so their results could not be meta-analysed in this format (Stack 2012). Therefore data were not available for meta-analysis for ten trials, meaning that of the 39 trials, data available for analysis were provided by 29 trials.
Effects of interventions
Two- or Six-Minute Walk Test (m)
Data on the two- or six-minute walk test were available from six trials for seven comparisons within four physiotherapy interventions (exercise, treadmill, dance, and martial arts). (Note: Hackney 2009 contributed data to both the dance and martial arts comparisons.) Two hundred forty-two participants were included in this analysis. A benefit of borderline significance was identified, along with a greater increase in the distance walked in two or six minutes with physiotherapy intervention compared with no intervention (mean difference 13.37 m, 95% confidence interval (CI) 0.55 to 26.20; P = 0.04; see Figure 4). No evidence of heterogeneity was found between the individual trials (P = 0.44, I
|Figure 4. 2- or 6-Minute walk test (m).|
Meek 2010 contributed to 2-minute walk test. Hackney 2009, Schilling 2008, and Schenkman 1998 contributed to 6-minute walk test.
10- or 20-Metre Walk Test(s)
Data on the 10- or 20-metre walk test were available from four trials for two physiotherapy interventions (exercise and treadmill). One hundred sixty-nine participants were included in the analysis. Borderline significance was reported in favour of no intervention for the time taken to walk 10 or 20 metres (0.40 s, CI 0.00 to 0.80; P = 0.05; see Figure 5). No evidence of heterogeneity between individual trials was obtained (P = 0.19, I
|Figure 5. 10- or 20-m walk test (s).|
Kurtais 2008 contributed to 20-m walk test. Meek 2010, Schenkman 1998, and Stozek 2003 contributed to 10-m walk test.
Data on speed were available from 15 trials for 19 comparisons within all six physiotherapy interventions. (Note: Fisher 2008; Hackney 2009; Mak 2008; and Thaut 1996 all contributed data to two physiotherapy comparisons.) Eight hundred fourteen participants were included in this analysis. A significant benefit was reported for physiotherapy, with speed increased by 4 cm/s with a physiotherapy intervention compared with no intervention (0.04 m/s, CI 0.02 to 0.06; P = 0.0002; see Figure 6). No evidence of heterogeneity was obtained between the individual trials (P = 0.55, I
|Figure 6. Speed (m/s).|
Data on cadence were available from seven trials for nine comparisons within four physiotherapy interventions (general physiotherapy, exercise, treadmill, and cueing). (Note: Fisher 2008 and Thaut 1996 contributed data to two physiotherapy comparisons). Three hundred fifty participants were included in this analysis. No significant difference in cadence was observed between the two treatment arms (-1.57 steps/min, CI -3.81 to 0.67; P = 0.17).
Stride Length (m)
Data on stride length were available from six trials for nine comparisons within all six physiotherapy interventions. (Note: Fisher 2008, Hackney 2009, and Thaut 1996 contributed data to two physiotherapy comparisons.) Two hundred twenty-five participants were included in this analysis. No difference in stride length was reported between the two treatment arms (0.03 m, 95% CI -0.02 to 0.08; P = 0.24).
Step Length (m)
Data on step length were available from five trials for six comparisons within four physiotherapy interventions (general physiotherapy, exercise, treadmill, and cueing). (Note: Fisher 2008 contributed data to both the general physiotherapy and treadmill comparisons.) Three hundred eighty-three participants were included in this analysis. No difference in step length was noted between the two treatment arms (0.02 m, 95% CI 0.00 to 0.04; P = 0.06).
Freezing of Gait Questionnaire
Data from the Freezing of Gait Questionnaire were available from four trials for three physiotherapy interventions (exercise, cueing, and dance). Two hundred ninety-eight participants were included in this analysis. A borderline significant benefit was noted, with freezing of gait questionnaire score improved by 1.4 points with a physiotherapy intervention compared with no intervention (-1.41, 95% CI -2.63 to -0.19; P = 0.02, see Figure 7). No evidence of heterogeneity between the individual trials was found (P = 0.74, I
|Figure 7. Forest plot of comparison: 1 Gait Outcomes, outcome: 1.7 Freezing of Gait Questionnaire.|
Functional Mobility and Balance Outcomes
Timed Up & Go (s)
Data on the Timed Up & Go test were available from nine trials for ten comparisons within four physiotherapy interventions (exercise, cueing, dance, and martial arts). (Note: Hackney 2009 contributed data to both the dance and martial arts comparisons.) Six hundred thirty-nine participants were included in this analysis. Overall, the time taken to complete the Timed Up & Go test was significantly improved (i.e. reduced) with physiotherapy intervention compared with no intervention (-0.63 s, 95% CI -1.05 to -0.21; P = 0.003; see Figure 8). No heterogeneity was observed between the individual trials (P = 0.12, I
|Figure 8. Timed Up & Go (s).|
The results for the Hackney et al. martial arts comparison were heavily weighted in the analysis (48.8%) by very small standard deviations (Hackney 2009) compared with the other studies. It was also noted that in the trial publication, a nonsignificant (P = 0.093) effect of martial arts intervention was reported － a finding that contrasted with our data analysis, which reported a significant improvement (P = 0.003). The author was contacted to check whether the data reported in the paper were in fact standard errors, but they were confirmed as standard deviations. We therefore performed a sensitivity analysis to remove this study and found that the overall result became not significant (-0.38 s, 95% CI -0.96 to 0.21; P = 0.21), so this result should be interpreted with caution.
Functional Reach Test (cm)
Data on the Functional Reach Test were available from four trials for two physiotherapy interventions (exercise and cueing). Three hundred ninety-three participants were included in this analysis. Functional reach was significantly improved with physiotherapy intervention compared with no intervention (2.16 cm, 95% CI 0.89 to 3.43; P = 0.0008, see Figure 9). No evidence suggested heterogeneity between the individual trials (P = 0.15, I
|Figure 9. Functional Reach (cm).|
Berg Balance Scale
Data on the Berg Balance Scale were available from five trials for six comparisons within four physiotherapy interventions (exercise, treadmill, dance, and martial arts). (Note: Hackney 2009 contributed data to both the dance and martial arts comparisons.) Three hundred eighty-five participants were included in this analysis. The Berg Balance Scale was significantly better after physiotherapy intervention (3.71 points, 95% CI 2.30 to 5.11; P < 0.00001; see Figure 10). No evidence of heterogeneity between the individual trials was noted (P = 0.06, I
|Figure 10. Berg Balance Scale.|
Activity-Specific Balance Confidence
Data on activity-specific balance confidence were available from three trials for two physiotherapy interventions (exercise and cueing). Sixty-six participants were included in this analysis. No difference between the two treatment arms was noted (2.40 points, 95% CI -2.78 to 7.57; P = 0.36).
Number of Falls
Seven trials (Ashburn 2007; Goodwin 2009; Marjama-Lyons 2002; Meek 2010; Nieuwboer 2007; Protas 2005; Purchas 2007) attempted to record the number of falls during the trial period.This was usually done by means of a falls diary, which can be difficult to analyse and is subject to bias. Nevertheless, most of the individual trials reported a general trend for a reduction in the number of falls with intervention. However, when compared with the no intervention arm, this finding was not significant, except in one trial. Marjama-Lyons 2002 reported a significant decrease in the chance of fall frequency with Tai Chi intervention when compared with no intervention.
Falls Efficacy Scale
Data on the Falls Efficacy Scale were available from four trials for four comparisons within two physiotherapy interventions (exercise and cueing). Three hundred fifty-three participants were included in this analysis. No difference in the Falls Efficacy Scale was found between the two treatment arms (-1.91 points, 95% CI -4.76 to 0.94; P = 0.19).
Only data on the Unified Parkinson’s Disease Rating Scale were available for meta-analysis.
Unified Parkinson’s Disease Rating Scale (UPDRS)
Data on the total UPDRS score were available from three trials for four comparisons within three physiotherapy interventions (general physiotherapy, exercise, and treadmill). (Note: Fisher 2008 contributed data to both the general physiotherapy and treadmill comparisons.) Two hundred seven participants were included in this analysis. Overall, the UPDRS total score was significantly improved with physiotherapy intervention compared with no intervention (-6.15 points, 95% CI -8.57 to -3.73; P =< 0.00001; see Figure 11). Evidence of borderline heterogeneity was observed between the individual trials (P = 0.03, I
|Figure 11. UPDRS － total.|
Data on the mental sub-scale of the UPDRS were available from two trials for three comparisons within two physiotherapy interventions (general physiotherapy and treadmill). (Note: Fisher 2008 contributed data to both the general physiotherapy and treadmill comparisons.) One hundred five participants were included in this analysis. No difference in UPDRS mental score was reported between the two treatment arms (-0.44, 95% CI -0.98 to 0.09; P = 0.10).
Activities of Daily Living (ADL)
Data on the ADL sub-scale of the UPDRS were available from three trials for four comparisons within three physiotherapy interventions (general physiotherapy, treadmill, and dance). (Note: Fisher 2008 contributed data to both the general physiotherapy and treadmill comparisons.) One hundred fifty-seven participants were included in this analysis. Overall, the UPDRS ADL score was significantly improved with physiotherapy intervention compared with no intervention (-1.36 points, 95% CI -2.41 to -0.30; P = 0.01; see Figure 12). No evidence of heterogeneity was observed between the individual trials (P = 0.28, I
|Figure 12. UPDRS － ADL.|
Earhart 2010, MDS-UPDRS.
Data on the motor sub-scale of the UPDRS were available from 12 trials for 14 comparisons within all six physiotherapy interventions. (Note: Fisher 2008 and Hackney 2009 contributed data to two physiotherapy interventions.) Five hundred ninety-three participants were included in this analysis. Overall, the UPDRS motor score was significantly improved with physiotherapy intervention compared with no intervention (-5.01 points, CI -6.30 to -3.72; P < 0.00001; see Figure 13). Evidence indicated significant heterogeneity between the individual trials (P = 0.0009, I
|Figure 13. UPDRS － Motor.|
Earhart 2010, MDS-UPDRS.
Patient-rated Quality of Life
Only data on the Parkinson’s Disease Questionnaire–39 (PDQ-39) for the mobility domain and the summary index were available for meta-analysis.
Parkinson’s Disease Questionnaire–39 (PDQ-39)
Data on the Summary Index of the PDQ-39 were available from seven trials for eight comparisons within all six physiotherapy interventions. (Note: Hackney 2009 contributed data to both the dance and martial arts comparisons.) Four hundred five participants were included in this analysis. No difference between treatment arms was observed in patient-rated quality of life after physiotherapy intervention (-0.38 points, 95% CI -2.58 to 1.81; P = 0.73).
Data on the mobility domain of the PDQ-39 were available from two trials for three comparisons within three physiotherapy interventions (general physiotherapy, dance, and martial arts). (Note: Hackney 2009 contributed data to both the dance and martial arts comparisons.) One hundred five participants were included in this analysis. No difference in the PDQ-39 mobility score was observed between the two treatment arms (-1.43, 95% CI -8.03 to 5.18; P = 0.67).
No trials reported data on adverse events, and only one commented on adverse events, stating that none had occurred during treatment sessions (Goodwin 2009).
Only fourteen of the thirty-nine trials discussed patient compliance, with twelve (Allen 2010; Canning 2008, Duncan 2012, Ellis 2005; Goodwin 2009, Keus 2007b; Klassen 2007; Kurtais 2008; Meek 2010; Sage 2009a; Schenkman 1998; Schmitz-Hubsch 2006) quantifying it in some form; however, this was difficult to analyse.
No trials reported data on health economic outcomes.
Only one outcome, the UPDRS motor sub-scale, showed significant heterogeneity between the treatment effects of the different classes of interventions. In all other cases, no evidence of any differences was found. However, one outlying trial was the cause of this heterogeneity in the motor score (Boehm 2011); when this trial was excluded from the analysis, the result remained significant (−3.77 points, 95% CI -5.15 to -2.39; P < 0.001), but the test for between-trial and between-subgroup heterogeneity was no longer significant (P = 0.44 and P = 0.08, respectively).
Summary of main results
This review updates the previous Cochrane review published in 2001 (Deane 2001a) comparing physiotherapy intervention versus no physiotherapy intervention for the treatment of PD. The review now includes 39 randomised trials and 1827 participants (compared with 11 trials and 280 participants in the 2001 review). It also compares the different types of physiotherapy interventions used in the treatment of PD, thus providing a comprehensive assessment of physiotherapy treatment. Many recent systematic reviews have focused on specific areas of physiotherapy such as exercise and cueing (Crizzle 2006; Goodwin 2008; Lim 2005; Nieuwboer 2008). Nowadays, physiotherapy for PD encompasses a wide range of methods and techniques ranging from standard NHS physiotherapy to exercise regimens and martial arts. Therefore, it is important that all forms of physiotherapy intervention are included, so that the true benefit (if any) of physiotherapy can be assessed. The review also includes a more comprehensive range of outcome measures compared with previous reviews (18 outcomes assessing gait, functional mobility and balance, falls, clinician-rated Unified Parkinson’s Disease Rating Scale (UPDRS), and patient-rated quality of life), thus providing the most reliable summary available of the current published evidence.
Physiotherapy Intervention versus No Physiotherapy Intervention
This review provides evidence of the short-term (< three months) benefit of physiotherapy in the treatment of PD. All outcomes showed improvement with physiotherapy intervention compared with no intervention (except the 10- or 20-metre walk test). However, significant benefits after physiotherapy intervention were observed only for the gait outcomes of speed, the two- or six-minute walk test, and the Freezing of Gait questionnaire; the functional and mobility outcomes of the Timed Up & Go test, Functional Reach Test, and Berg Balance Scale; and the clinician-rated UPDRS. It is of interest that the direction of the treatment effect favoured physiotherapy intervention in all outcome measures, except one. The absence of evidence in these outcomes is not necessarily evidence of the absence of benefit for physiotherapy. One possible reason for this may be the lack of data. More than 1800 participants were randomly assigned into the 39 trials included in this review, and 29 trials and 1577 participants (86% of total) provided data for analysis. However, the greatest quantity of data were provided for analysis of the outcome speed, and this included just 15 trials and 814 participants (52% of the total number of participants providing data). This general lack of extractable data means that results of this meta-analysis should be interpreted with caution.
People with PD frequently have problems with gait, and treatment is usually targeted toward maximising exercise tolerance, improving the gait pattern, maintaining or increasing independence regarding mobility, and reducing the risk of falls. The most significant improvement among the outcomes assessing gait involved speed. In light of previous experimental evidence, it may be hypothesised that the improvement in speed is linked to an increase in step or stride length, or both, and that this in turn leads to a compensatory decrease in cadence (Morris 1994; Morris 1996). In this review, although a significant improvement in speed was observed, we found no difference in step length, stride length, or cadence. This could again be due to lack of data, as a smaller number of studies reported step and stride length and cadence (up to seven studies) compared with speed (15 studies). Thus, further data on the possible link between speed, cadence, step, and stride length are required.
Freezing of gait is a prevalent motor disturbance within PD, and it is known to have a detrimental impact on quality of life, as well as on gait and mobility (Moore 2007). We found a borderline significant difference in scores derived from the Freezing of Gait Questionnaire, but this was measured in only four trials (298 participants), again highlighting the need for further data in this important area.
Observed differences in the three significant gait outcomes (speed, the two- or six-minute walk test, and freezing of gait) were relatively small. Therefore, their relevance and benefit to patients with PD must be put into context in terms of what is considered a minimally clinically important change (MCIC). Speed was significantly improved with physiotherapy intervention by 0.04 metres/s. Data on what is considered an MCIC are lacking for PD patients, but some data have been reported in stroke patients. In one study, it was reported that an increase in speed of just 0.03 and 0.13 metres/s could translate into a change from a limited household to an unlimited household walker, and from an unlimited household walker to a most-limited community walker, respectively (Perry 1995). Our data are consistent with the findings reported by Perry (Perry 1995). For the two- or six-minute walk test and freezing of gait, participants who received physiotherapy intervention were able to walk further over two or six minutes (by 13 m) and their Freezing of Gait score was improved by 1.4 points. Data on the MCIC are lacking for these outcomes, but although a 13-m increase in distance walked would probably be considered clinically important, the importance of a 1.4-point improvement in freezing of gait is less clear.
Functional Mobility and Balance
Changes in functional mobility and balance within PD have been well documented (Bloem 2001). Of the functional mobility and balance outcomes assessed within this review, significant improvements were observed in the Timed Up & Go test, Functional Reach Test, and Berg Balance Scale. The time taken to complete the Timed Up & Go test was significantly improved by 0.63 seconds with physiotherapy. Despite this significant change, the MCIC in PD patients is thought to be 11 seconds (Steffen 2008). Therefore, the small change observed within this review may not translate into a noticeable improvement in a person’s functional mobility.
A five-point change is the MCIC on the Berg Balance Scale (Steffen 2008). In this review, a significant four-point improvement in the Berg Balance Scale was noted after physiotherapy intervention. A greater evidence base is required to support or refute the clinical significance of this result. A significant improvement of 2 cm was also noted in the Functional Reach Test, but this is somewhat lower than the MCIC of 9 cm and 7 cm for the forward and backward Functional Reach Test (Steffen 2008).
Falls are a common and disabling problem within PD (Bloem 2001), with high clinical impact and serious cost implications to society. They are also a recurrent problem, with up to 51% of those falling reporting two or more falls per year (Wood 2002). Fear of falling has been recognised as a contributing factor to recurrent falls (Mak 2009). Within this review, fear of falling has been captured through the Falls Efficacy Scale (standard and international). No difference between treatment arms was observed for this outcome. This might be attributed to the small number of trials (and therefore participants) included within these analyses, but could also indicate that an improvement in balance does not automatically result in increased confidence in an individual’s ability not to fall. In turn, it could be hypothesised that improvement in balance does not directly equate to improved levels of mobility and independence. Although fear of falling was not reduced with physiotherapy within this review, it would be of interest to assess whether the number of falls was reduced, as this may be more relevant to patients. Unfortunately, data on this were poorly reported and were measured too variably within the trials; therefore, they could not be meta-analysed. However, in the seven trials in which data on the number of falls were reported, a general trend toward a reduction in the number of falls with physiotherapy intervention was seen, but with no difference between the two treatment arms.
Significant improvements after physiotherapy intervention were also observed for the clinician-rated UPDRS (total, ADL, and motor scores). The UPDRS total score was improved by 6.2 points, the ADL score by 1.4 points, and motor score by 5.0 points. The MCIC for the UPDRS have been reported in two studies. One analysed data from two independent randomised controlled trials and concluded the MCIC to be eight points for the UPDRS total score, between two and three points for the ADL score, and five points for the motor score (Schrag 2006). The second study performed a cross-sectional analysis on 653 PD participants, and reported MCIC of 2.3 to 2.7 points for motor and 4.1 to 4.5 points for total UPDRS (Shulman 2010). If the recommendations of both Schrag (Schrag 2006) and Shulman et al (Shulman 2010) are taken into account, it can be concluded that the significant improvements observed within this review are approaching or are MCICs (the MCICs for the UPDRS total, ADL, and motor scores lie within the confidence interval). This suggests that physiotherapy intervention is beneficial in improving motor symptoms and may positively impact ADL.
Patient-Rated Quality of Life
No significant benefit of physiotherapy intervention for overall patient-rated quality of life (measured using the Parkinson’s Disease Questionnaire (PDQ)-39 Summary Index) or the mobility domain of the PDQ-39 was noted, which is surprising in light of the significant improvements seen in UPDRS scores. Another study (Chandler 1999) assessed patient quality of life using the generic Short Form-36 and also showed no effect of physiotherapy intervention.
Comparison of Different Physiotherapy Interventions
Although we found short-term benefit for physiotherapy intervention in the treatment of PD, what is less clear is whether a certain type of physiotherapy intervention may provide greater benefit. This information would be of interest to both clinicians and patients, so that appropriate physiotherapy interventions that provide greater benefit can be delivered to patients with PD. To assess this, we categorised the various physiotherapy interventions used in the trials included in this review according to the type of treatment administered, and then compared them using tests for heterogeneity. We found no real evidence of any differences in the treatment effect between the different physiotherapy interventions used for any of the outcomes assessed. However, these were based on indirect comparisons (with limited data within each physiotherapy intervention) so should be interpreted with caution. They would be better assessed in trials directly comparing different types of physiotherapy interventions.
This lack of difference between the different types of physiotherapy intervention is perhaps not surprising. The content and delivery of the interventions used in the trials included within this review are diverse in nature and, although attempts were made to compare trials 'like for like' through the creation of different categories, the interventions delivered varied substantially within these categories. The variety in the therapies delivered is perhaps not surprising. By nature physiotherapists are autonomous professionals with differing sets of skills who work within their own scope of practice (Chartered Society of Physiotherapy), and so this variation in the interventions delivered within clinical trials may actually reflect clinical practice. Second, and perhaps more important, PD is recognised as a complex condition with an individualised presentation (van der Marck 2009). For this reason, Morris et al (Morris 2010) recognises the importance of the physiotherapist's understanding the specific experience of PD in each patient, and advocates that treatment is tailored to fit the individual’s complaints, lifestyle, and personal interests, as opposed to a 'one size fits all' approach. Over the past decade, steps have been taken to try to provide best practice consensus in the form of the Dutch KNGF guidelines for physical therapy in patients with Parkinson’s disease (Keus 2004). However, this publication provides a guidance framework rather than a 'recipe' for treatment. It is therefore important that physiotherapy interventions are compared against each other within rigorous trial designs to determine which are most effective. This will provide therapists with a menu of treatment strategies that are known to be effective, from which they can devise individualised interventions.
Quality of the evidence
Improvement in trial methodological quality and reporting has been noted since the last Cochrane review (Deane 2001a). The use of more robust randomisation methods, blinding, and intention-to-treat analyses had increased since the previous review but was still inadequate. Only 18 of the 39 trials provided information on the randomisation method (of which eleven were considered low risk), and only five used a central randomisation procedure to ensure concealment of treatment allocation. Twenty-four used blinded assessors and nine reported using intention-to-treat analysis. The lack of information on this in many trial reports may not necessarily indicate lack of implementation within the trial, but without this information, the level of bias within the individual trials is difficult to assess. This does, therefore, reduce the amount of confidence that can be placed in the results of this meta-analysis. The need for further improvement in the methodological quality of trials in physiotherapy for PD was noted in another recent systematic review (Kwakkel 2007). Future trials need to ensure that their designs fulfil the requirements of a methodologically sound, large, randomised controlled trial, and that the reporting follows the CONSORT guidelines (Schulz 2010).
The trials included in the review were relatively small, with most assessing the effects of physiotherapy intervention versus no physiotherapy intervention over a short period with limited follow-up. The overall size of trials has increased (with an average of 46 participants per trial in this review compared with 25 in the previous review), but the number of small and underpowered trials remains a problem. Small trials may be subject to ‘random error’ (Doll 1980), and consequently may give rise to false-negative or -positive results. To highlight this point, this review illustrates that any differences observed in the various outcome measures showing benefit for physiotherapy were quite small. So trials need to be large enough to detect these small but possibly clinically important differences.
Further, it must be noted that only 14 of the 39 trials discussed participant compliance. This is surprising in that compliance can be an important determinant of the outcomes measured in trials. Therefore, it would be beneficial if the level of compliance is measured in future trials.
Another limitation is that the follow-up period in the trials included in this review was relatively short. Outcome measures were assessed by all trials at baseline and immediately or shortly after intervention had ceased (one or two weeks with one trial (Goodwin 2009) assessing at 10 weeks post intervention). Thus, this review is able to provide conclusions only on the short-term benefits of physiotherapy. It is also important to consider results alongside the possibility of a so-called honeymoon effect (Goetz 2008) in the period during or just after physiotherapy, which may inflate the treatment effect in favour of physiotherapy. Parkinson’s disease is a long-term neurodegenerative disease, so it is important that the long-term effect of treatment be assessed. Only 12 of the 39 trials followed-up participants and reported further data during the post-treatment period (but this could have been only one week or up to six months post the treatment period). The recommendations of the previous review were that participants should be followed-up for at least six months, but only one trial (Schmitz-Hubsch 2006) reported follow-up data at six months post treatment completion. Long-term data will provide valuable information about the duration of any improvement following therapy.
The outcome measures included in this review are standard physiotherapy and PD outcomes. However, PD is a multidimensional disease, and many important outcomes were poorly reported or were not reported, this includes data on the number of falls, depression and anxiety, adverse events, and the health of the caregiver supporting the person with PD. Further, no health economics analysis of physiotherapy intervention was reported; therefore little is known about the cost-effectiveness and economic value of this therapy. Future trials should include these outcomes.
In summary, this review provides evidence of the short-term (< three months) benefit of physiotherapy intervention for the treatment of PD. It is important to note that although most of the observed differences between the two treatments were small, the improvements observed for speed, Berg Balance Scale, and UPDRS scores were at levels considered to be of clinical importance. To clarify the long-term (if any) benefit of physiotherapy, additional large, well-designed randomised trials with a follow-up of at least 12 months, alongside a health economics assessment, are needed to assess the impact of this treatment on all aspects of a patient's PD.
Implications for practice
Physiotherapy provides short-term benefit in the treatment of PD. Significant benefits with physiotherapy intervention were observed for the following outcomes: two- or six-minute walk test, speed, Freezing of Gait questionnaire, Timed Up & Go test, Functional Reach Test, Berg Balance Scale, and UPDRS total, ADL, and motor scores. Although most of the observed differences between the two treatment arms were small, the improvements seen for speed, Berg Balance Scale, and UPDRS scores occurred at levels that may be considered to be of clinical importance. These benefits should be interpreted with caution, however, because the quality of most of the included trials was not high.
The long-term, if any, benefit of physiotherapy remains unidentified, as does which type of physiotherapy intervention should be delivered. Therefore, although this review has provided evidence that physiotherapy intervention may be of benefit to PD patients, it has also highlighted that further evidence is needed before firm conclusions can be made on the long-term benefit and on which physiotherapy intervention should be used.
Implications for research
Most of the studies in this review were small and had a short follow-up period. It is clear that larger randomised controlled trials are required, particularly focusing on improving trial methodology and reporting. Rigorous methods of randomisation should be used and the allocation adequately concealed. Data should be analysed according to intention-to-treat principles, and trials should be reported according to the guidelines set out in the CONSORT statement (Schulz 2010).
A large variety of outcome measures were assessed in these trials, but data were sufficient only for meta-analysis to be performed for eighteen outcomes. This variation in outcome selection and lack of extractable data resulted in a small proportion of included trials contributing to each outcome. This review illustrates the need for the universal employment of relevant, reliable, and sensitive outcome measures. Additionally, only one trial looked at the longer-term benefit of physiotherapy intervention. To assess whether, or how long, any improvements due to physiotherapy intervention may last, it is important that long-term follow-up is performed.
No evidence indicates the best form of physiotherapy intervention. Comparisons of the different physiotherapy interventions described in this review were based on indirect comparisons between individual trials. A more reliable comparison would be obtained in large, randomised trials that directly compare different physiotherapy interventions.
This review highlights the variety of physiotherapy interventions being used in the treatment of PD. More specific trials with improved treatment strategies are needed to underpin the most appropriate choice of physiotherapy intervention and the outcomes measured.
We recognise the contributions of all the original trialists and of the individuals who performed the trials that contributed to this meta-analysis, and we thank the patients who agreed to help improve the assessment of PD treatment by taking part in these trials.
We acknowledge Parkinson’s UK for funding, as well as the Department of Health, whose core support for BCTU made this review possible.
We would also like to thank Alex Furmston, Kinga Malottki, Mohammad Tokhi, and Manijeh Ghods, who provided translations for foreign papers.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Index terms
Last assessed as up-to-date: 31 January 2012.
Protocol first published: Issue 4, 2000
Review first published: Issue 3, 2001
Contributions of authors
Claire Tomlinson was involved in searching and selection of studies, data extraction, and analysis and interpretation of the review.
Smitaa Patel was involved in selection of studies, data extraction, and analysis and interpretation of the review.
Charmaine Meek was involved in data extraction and provided expert physiotherapy input into interpretation of the review.
Clare Herd was involved in searching and selection of studies, data extraction, and analysis and interpretation of the review for trials published from 2011 onwards.
Carl Clarke contributed to the design of the protocol and was involved in the interpretation of the review, providing clinical input.
Rebecca Stowe contributed to the design of the protocol and was involved in searching and selection of studies and analysis and interpretation of the review.
Laila Shah was involved in searching and selection of studies for the review.
Catherine Sackley contributed to the design of the protocol and provided expert physiotherapy input into the interpretation of the review.
Katherine Deane undertook the 2001 Cochrane Review and was involved in the interpretation of this review.
Keith Wheatley contributed to the design of the protocol and was involved in the interpretation of this review.
Natalie Ives contributed to the design of the protocol and was involved in the analysis and interpretation of this review.
Declarations of interest
Carl Clarke, Natalie Ives, Charmaine Meek, Smitaa Patel, Catherine Sackley, and Keith Wheatley are recruiting or are involved in the running of the UK PD REHAB trial.
Sources of support
- No sources of support supplied
- Parkinson's UK, UK.
- Department of Health, UK.
Medical Subject Headings (MeSH)
MeSH check words
* Indicates the major publication for the study