Jaana Paltamaa, PhD, PT1,2,3, Tuulikki Sjögren, PhD, PT3, Sinikka H. Peurala, PhD, PT3,4 ; and Ari Heinonen, PhD, PT3

From the 1JAMK University of Applied Sciences, School of Health and Social Studies, Jyväskylä, 2Peurunka Medical Rehabilitation Center, Laukaa, 3Department of Health Sciences, University of Jyväskylä, Jyväskylä and 4Lahti ; Rehabilitation Centre, Lahti, Finland

OBJECTIVE: To determine the effects of physiotherapy interventions on balance in people with multiple sclerosis.

DATA SOURCES: A systematic literature search was conducted in Medline, Cinahl, Embase, PEDro, both electronically and by manual search up to March 2011.

STUDY SELECTION: Randomized controlled trials of physiotherapy interventions in people with multiple sclerosis, with an outcome measure linked to the International Classification of Functioning, Disability and Health (ICF) category of “Changing and maintaining body position”, were included.

DATA EXTRACTION: The quality of studies was determined by the van Tulder criteria. Meta-analyses were performed in subgroups according to the intervention.

DATA SYNTHESIS: After screening 233 full-text papers, 11 studies were included in a qualitative analysis and 7 in a meta-analysis. The methodological quality of the studies ranged from poor to moderate. Low evidence was found for the efficacy of specific balance exercises, physical therapy based on an individualized problem-solving approach, and resistance and aerobic exercises on improving balance among ambulatory people with multiple sclerosis.

CONCLUSION: These findings indicate small, but significant, effects of physiotherapy on balance in people with multiple sclerosis who have a mild to moderate level of disability. However, evidence for severely disabled people is lacking, and further research is needed.

Key words: multiple sclerosis; systematic review; balance; physiotherapy; exercise training.

J Rehabil Med 2012; 44: 811–823

Correspondence address: Jaana Paltamaa, JAMK University of Applied Sciences, School of Health and Social Studies, PL 207, FI-40101 Jyväskylä, Finland. E-mail: jaana.paltamaa@jamk.fi

Submitted November 25, 2011; accepted June 7, 2012

Introduction

Multiple sclerosis (MS) is a chronic progressive disease of the central nervous system (CNS) that affects a wide range of neurological functions, including cognition, vision, muscle strength and tone, coordination and sensation (1). The many symptoms associated with MS cause mobility limitations (2), e.g. gait and balance disorders in later stages of the disease (1), and sometimes even early stages of the disease in recently diagnosed people with MS who present with no clinical disability (3, 4).

The maintenance of upright stance or balance requires the interaction of multiple sensorimotor processes (visual, vestibular, proprioception) to generate coordinated movements that maintain the centre of mass within the limits of stability (5, 6). Balance is an integrated component of physical function, and a product of the task undertaken and the environment in which it is performed (7). Therefore, the components of balance training include multisensory and motor strategy training, resistance and aerobic training and several neurotherapeutic approaches in individual tailored or group therapy (6, 7). According to the World Health Organization’s (WHO) International Classification of Functioning, Disability and Health (ICF), balance is operationalized as “Changing and maintaining body position” in the Mobility domain of the Activities and Participation component (2).

Abnormalities in balance and the underlying physical functions are common findings in people with MS (8–13). A recent review of postural control in MS demonstrated that people with MS have balance impairments characterized by increased sway in quiet stance, delayed responses to postural perturbations, and reduced ability to move towards their limits of stability (14). Many people with MS fall frequently (14–16), fear falling (17), and risk of fall-related injuries is increased (18, 19). Increased risk of fall has also been found in connection with various gait assessments (14) and with the use of a walking aid (20).

Many studies have examined the effects of exercise training on walking mobility in people with MS. The cumulative evidence of reviews (21, 22) and a meta-analysis (23) indicate that exercise training is associated with a small improvement in walking among individuals with MS. Although evidence-based rehabilitation techniques are of interest in the care of people with MS, there is a lack of information on the correlation of physiotherapy (PT) with balance disorders.

The aim of this systematic review was to determine the effectiveness of PT interventions on balance in people with MS. Specifically, the evidence is based on sub-meta-analyses according to the intervention, content of the control group therapy and quality of the randomized controlled trials (RCTs) selected for review. The ICF classification was used as a framework for classifying the interventions and their outcomes in the RCTs.

Methods

Eligibility criteria

Studies that investigated the effect of a comprehensive combination of PT intervention were included in the review. Further eligibility criteria for inclusion in the review according to the PICOS (population, intervention, comparison, outcomes) were as follows:

• (P) subjects with MS;

• (I) a method of PT as a single discipline;

• (C) experimental vs control (placebo or no treatment) condition or 2 experimental; and

• (O) an outcome measure of balance linked to the ICF category of “Changing and maintaining body position” including both capacity and performance qualifiers.

Only RCTs published in English, Finnish, Swedish or German were included in the study.

Non-randomized and non-controlled pre-experimental studies, studies with a single session, abstracts and protocols were excluded. Studies including multiple diagnoses without separate analysis of MS and multidisciplinary rehabilitation studies without separate analysis of a single PT method were also excluded.

Search strategy

The following databases were searched from the beginning of each database to December 2008: OVID Medline, Cumulative Index to Nursing & Allied Health Literature (CINAHL) and Embase. An update search was conducted in OVID Medline and CINAHL databases for the period January 2009 to March 2011. Fig. 1 shows the combined flow chart for these searches.

Two information specialists performed the searches in the selected electronic databases in conjunction with the researchers. The search strategy was designed to include a broad range of research on PT interventions in people with MS. In addition, a supplementary manual search was conducted and, where appropriate, the authors of the relevant publications were contacted for further information.

The following key words were used: type of disease, i.e. multiple sclerosis, MS or demyelinating autoimmune diseases, CNS AND type of physiotherapy intervention AND type of study, i.e. randomized controlled trial or clinical trial. A comprehensive combination of keywords describing the PT intervention, e.g. exercise therapy, ambulation, balance, musculoskeletal equilibrium or postural stability, were used. Additional treatment methods, such as physical therapy modalities, were allowed. Search terms were entered into each database using either MeSH or keyword headings specific to the requirements of the database.

The full search strategies for each database are available on request from the corresponding author; the original Ovid Medline search strategy (Appendix S1 (available from http://www.medicaljournals.se/jrm/content/?doi=10.2340/16501977-1047)) is also available (24) on the following web link: https://helda.helsinki.fi/bitstream/handle/10138/24581/VAKE_liiteS32.pdf?sequence=35.

Review process

In accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for systematic reviews, two reviewers (JP and SHP/TS/EA) independently screened all the titles and abstracts of the articles. After these steps, potentially relevant articles were retrieved for full-text assessment. Two members of the research team (JP and SHP/TS/EA) independently evaluated all the potential full-text articles in order to identify potentially eligible studies. They also grouped the included studies according to the PT intervention. In the event of disagreement, a third reviewer (AH) evaluated the article to achieve a joint consensus.

Rating of study quality

The methodological quality of RCTs was rated using van Tulder scale (25). This scale rates RCTs based on random concealed allocation of participants, the similarities of participants at baseline, blinding the patient, care provider and outcome assessor, co-interventions, compliance, the dropout rate, similar timing of outcome assessment and the use of intention-to treat analysis (Table I). All 11 items were scored positive (“yes”) if the criterion was fulfilled, negative (“no”) if it was not fulfilled, or unclear (“don’t know”). If the article did not contain information on the methodological criteria, the authors were contacted for additional information. If the authors could not be contacted, or if the information was no longer available, the criteria were scored as “unclear”. The methodological quality and content analysis were evaluated by two blinded and independent reviewers (JP with undergraduate or doctoral students). Any disagreements were resolved by seeking a consensus between the reviewers, while a third reviewer (TS or SHP) was brought in to help resolve any remaining disagreements. A total score was computed by counting the number of positive scores. The maximum score was 11. The RCTs were considered to be of high, moderate or poor level depending on the number of yes-rated items and the number of subjects (26) (Table I).

Data extraction

Seven out of 11 studies included in the qualitative analysis were accepted for the meta-analysis. All of these studies presented PT as a single discipline. The meta-analyses were performed in the following subgroups according to the intervention: specific balance exercises, resistance and aerobic training, whole-body vibration, group therapy and neurotherapeutic approaches.

Standardized outcome measures excluding quality of life questionnaires were linked to the ICF category of “Changing and maintaining body position”. Measures were linked to the ICF domains according to the international guidelines (27). Both capacity (which refers to the ability to execute a task or action if the environment were uniform or standard) and performance (which relates to what a person does in the area in question, in the environmental context in which they actually live) qualifiers were used.

In the meta-analyses, Cochrane Collaboration’s Review Manager 5.1.3 program was used to calculate pooled effect estimates for the combinations of single effects of RCTs. To calculate standardized mean differences (effect size; ES), follow-up values adjusted for baseline values were used. If adequate pre-post treatment values (mean (standard deviation, SD)) were not reported, a request was sent to the authors to supply this information. If two requests were not answered, the RCT was not taken into account in the meta-analyses. An ES of approximately 0.2 was considered a small effect, approximately 0.5 a medium effect, and ≥ 0.8 a large effect.

For all the analyses, the inverse variance weighted random effects method was used. This incorporates heterogeneity into the model. The centre of this distribution describes the average of the effects and its width the degree of heterogeneity. In multiple comparisons with two or more intervention groups, the number of controls was divided among the comparisons to ensure that we counted the control participants only once in the meta-analysis. The overall effect was tested with the z-test, where the null hypothesis is that there is no difference between the intervention group and the control group. Results were considered to be significant at an alpha level of < 0.05. In the results section of this systematic review and meta-analysis, the findings from the meta-analysis are presented using forest plots of the standardized mean differences for each subgroup.

The findings are summarized in the discussion section. Evidence was categorized into 4 levels depending on the quality and number of RCTs, as follows: high (at least two high-quality RCTs with parallel results), moderate (one high-quality RCT or several high-quality RCTs with some contradictions in the results or several acceptable-quality RCTs with parallel results), low (high-quality RCTs with notable contradictions in the results or at least one acceptable study) and no evidence (poor-quality RCT or no RCTs).

Results

Study selection and characteristics of participants with MS

The initial search yielded 8,376 publications and the update search 76 publications. Screening of 233 full-text papers revealed 81 potentially eligible publications. A total of 11 studies (28–38) published during the years 1998–2011 fulfilled the inclusion criteria and were deemed relevant for this review. Fig. 1 presents a flow chart of the structured review.

Collectively, the selected papers included a total of 340 persons with MS who completed the interventions (Table II). The mean age of the participants across the studies was 46 years. Approximately 68% of the participants were women. Five studies out of 11 included both relapsing-remitting and progressive forms of MS (28, 29, 33, 35, 37), one RCT only progressive forms (36), and one RCT only relapsing-remitting MS (31). Overall, 4 papers did not report the course of the MS (30, 32, 34, 38). In all the studies, the persons with MS were ambulatory. Six studies used the Expanded Disability Status Scale (EDSS) (39) to describe disease severity. The EDSS scores was ≤ 6.5, i.e. all these studies included participants with a minimal to moderate level of disability (29, 30, 33, 35, 36, 38). The other 5 studies (28, 31, 32, 34, 37) included participants who were able to walk with or without an assistive device or whose Hauser ambulation index ranged from 1 to 6 (40).

Methodological quality of the selected studies

Van Tulder score ranged from 2 to 6 out of 11 points (mode 4). Overall, the methodological quality was poor, with none of the studies scored as high quality (Table I). Five RCTs were classified as poor and 6 as moderate quality. The most common methodological flaw was selection bias. While half of the studies used an adequate randomization method (item A: 55%), only one of the studies used concealed treatment allocation (item B: 9%). Most of the studies used similar timing of outcome assessment (item J: 91%), but only 45% of the studies conducted blinding of outcome assessor (item F). In addition, none of the studies conducted intention-to-treat analysis (item K), and many studies also had a small sample size.

Components of the interventions

Five groups (a–e) were identified according to the type of PT intervention: (a) specific balance exercises, i.e. multisensory and motor strategy training (n = 1) (28), (b) resistance and aerobic training (n = 4) (29–32), (c) whole-body vibration (n = 2) (33, 34), (d) group therapy (n = 1) (35), and (e) neurotherapeutic approaches (n = 3) (36–38). Sample sizes were small in most studies, varying from 12 (35) to 50 (28, 32) participants. The mean duration of the interventions was 9.0 weeks, ranging from 3 to 20 weeks. Table II presents a summary of the interventions in detail.

One RCT reported a single PT intervention group (E) compared with placebo or no treatment (C) (33), 5 RCTs had two intervention groups (E1, E2) compared with placebo or no treatment (C) (28–31, 38), and 5 RCTs compared two PT interventions (E1 vs. E2) (32, 34–37).

A total of 11 different outcome measures were identified from the selected 11 studies (column 6 of Table II presents the measures used in each study). There were a mean of two measures per study (range 1–4) in the aforementioned ICF category. The most common measures were the Timed Up and Go test (TUG) (41), Berg Balance Scale (BBS) (42), timed one-leg stance (43), Functional Reach test (FR) (44) and Dynamic Gait Index (DGI) (45). In addition, 3 self-report questionnaires were used. Two studies used the Activities-specific Balance Confidence (ABC) scale (46), one the Falls Efficacy Scale (FES) (47) and 1 the Modified Dizziness Handicap Inventory (mDHI) (48, 28). A higher score on the BBS, FR, ABC scale and mDHI and longer one-leg stance time denote better function. A lower TUG time and lower FES score denote better function.

Effects of the PT interventions

Of the 11 analysed RCTs, the data needed for the estimations of ES were reported in 5 articles, and in 2 others the authors provided these data on request. Thus the meta-analyses were conducted for 7 RCTs (28–30, 35–38) with 230 participants (Figs 2–5). The findings from the 4 remaining RCTs (31–34) could not be entered into the model. However, the results were quantitatively analysed (Table II), and the findings are discussed below.

Specific balance exercises

With regard to the effects of specific balance exercise, one RCT of moderate methodological quality was conducted (28). Our meta-analysis of this study evaluating inpatient training of specific motor and sensory strategies compared with placebo treatment indicates that there was a small but significant effect on balance (ES 0.34; 95% confidence interval (CI), 0.01–0.67) (Fig. 2A). The strongest effect was found on the BBS, but no significant effect was observed on the DGI and the two self-report tests, i.e. the mDHI and the ABC scale. No effect was found when training of motor strategies was compared with training of sensory strategies (Fig. 2B).

Resistance and aerobic training

A small, but non-significant, overall effect on balance was found when outpatient and home-based resistance and aerobic training were compared with no treatment (Fig. 3A). However, a significant effect on FR (ES 0.56; 95% CI, 0.02–1.11) was found (Fig. 3A). Furthermore, a significant overall effect on balance (ES 0.55; 95% CI, 0.14–0.97) was found when outpatient cycling progressive resistance training was compared with a home-based exercise programme (Fig. 3B). Our analysis revealed that simultaneous electro-stimulation during an American College of Sports Medicine (ACSM)-based resistance training programme did not enhance training efficiency, as no significant overall effect on functional balance was reported (Fig. 3C). The results of Harvey et al. (31) and Plow et al. (32) were excluded from our meta-analysis because adequate pre-post treatment values were not obtained. The authors’ (31, 32) analysis showed no significant differences between interventions (Table II).

Whole-body vibration

Both the studies of physical therapy modalities (33, 34) were acceptable quality RCTs and used whole-body vibration (WBV) training combined with exercises in outpatient setting. They were excluded from our meta-analysis because adequate pre-post treatment values were not obtained. They both reported statistically unchanged TUG performance following additional WBV training performance compared with no PT (33) or exercise alone (34) (Table II).

Group therapy

One RCT, of poor methodological quality, was conducted on the effects of group therapy (35). Our meta-analysis indicated a small, but non-significant, overall effect on balance when Awareness Through Movement (Feldenkrais) classes were compared with educational sessions (Fig. 4).

Neurotherapeutic approaches

All 3 studies of the overall PT intervention group focused on outpatient rehabilitation for 4–8 weeks periods, using different neurotherapeutic approaches. Our analysis revealed a significant effect on the timed one-leg stance test (ES 0.63; 95% CI, 0.36–0.91) when outpatient PT and home exercises based on an individualized problem-solving approach were compared with no treatment (Fig. 5A), but no significant effect when hospital outpatient PT was compared with home exercises (Fig. 5B). Furthermore, both Lord et al. (37) and Wiles et al. (38) compared a facilitation approach with functional exercises using different outcome measures. In line with our meta-analysis (Fig. 5B), they both reported statistically unchanged performance on the BBS and timed one-leg stance test following the facilitation approach when compared with functional exercises (Table II). Following neuromuscular rehabilitation (proprioceptive neuromuscular facilitation, Frenkel Coordination Exercise, postural stability and balance training, walking exercise) with Johnstone Pressure Splints, a significant effect was found on the timed one-leg stance test compared with neuromuscular rehabilitation alone (ES 2.23; 95% CI, 1.52–2.95) (Fig. 5B).

Discussion

The primary goal of this meta-analysis was to evaluate of the effects of PT interventions on balance in people with MS. Studies evaluating the effects of PT intervention on balance in people with MS showed heterogeneous results. The studies included in this systematic review and meta-analysis had various aims and a range of different outcome measures. This makes direct comparison between the studies, and hence meta-analyses, difficult, and consequently the overall evidence from a single study is weak. However, some general conclusions can be drawn.

Our meta-analysis of 7 RCTs found low evidence for positive effects on various balance outcomes of specific balance exercises (28), PT based on an individualized problem-solving approach (38), and of resistance and aerobic exercises (29, 30) in ambulatory people with MS. There was also low evidence that resistance and aerobic training appears to be more effective than home-based exercise (30). One RCT including progressive forms of MS found that neuromuscular rehabilitation accompanied with Johnstone Pressure Splints (36) is more effective than neuromuscular rehabilitation alone. Furthermore, inspection of the results based on the authors’ analyses, as reported in the original articles, indicated a significant effect of Awareness Through Movement (Feldenkrais) groups on the modified Clinical Test of Sensory Interaction in Balance (mCTSIB) and ABC scale compared with educational sessions (35), but no effect of whole-body vibration training on balance (33, 34). Also, there was insufficient evidence that whole-body vibration (33, 34), or electro-stimulation (29) improved functional balance outcomes for people with MS. The results of the analyses conducted with studies rated as yielding low evidence should be viewed critically, as new high-quality studies may change the magnitude and/or direction of this evidence.

Overall, the methodological quality of the studies was poor and sample sizes were small. The sample sizes of the RCTs in this review ranged from 4 to 40 per group, with most studies lacking sufficient statistical power. It has been stated that studies need to be adequately powered, or designed to fit into a meta-analysis, in which case the inclusion and exclusion criteria, outcomes, and time-points for assessment need to be established a priori (49). Obviously, it is very difficult, and perhaps impossible, to blind patients and care providers in studies of exercise therapy. Thus, a comprehensive criterion was adopted in order to determine high-quality methodological study (see Table I). Information on several methodological quality items was missing from most of the articles, thus decreasing the level of quality of the original paper. This should be taken into account when planning new studies and reporting results. In addition, we found that randomization procedures and concealed treatment allocation were poorly reported, as was the exact content of the interventions. Consequently, for the reasons given above, along with the methodological differences between the studies, such as the use of different outcome measures and types of intervention, meaningful comparison between the RCTs is severely limited.

Half of the studies included in this review did not report the course of the MS of the participants and half of the studies had a mixed group of MS types. Thus, most of the results could be generalized to all forms of MS, i.e. both relapsing-remitting and progressive forms. When the determinants of balance (6, 7) and the previous results on balance problems in people with MS (8–13) are considered, the form of MS does not seem to be a crucial factor in balance. However, we would encourage future researchers to study the effect of balance interventions separately for relapsing-remitting and progressive MS. More importantly, the level of disability of the participants must be considered when evaluating the results. The analysed studies all comprised only ambulatory people with MS, with the result that the EDSS scores varied widely (EDSS 1–6.5) within and between the studies. In other words, people with no disability (EDSS score 1) and people who used constant bilateral support (EDSS score 6.5) both received the same intervention. As studies have not taken mobility levels into account, the optimal type of intervention and its frequency and duration is unknown.

In general, PT interventions for the improvement of balance have adopted various theoretic approaches, e.g. motor and sensory strategies (28), Feldenkrais (35) and neuromuscular facilitation (37, 38). Some significant effects on balance compared with no/placebo treatment (28, 38) were found. When two treatments were compared (E1 vs E2), no significant effects were found (28, 35, 37, 38). Thus, the optimal type of intervention for people with MS remains unclear. In this situation, the PT should choose the most appropriate method, or combination of methods, on the basis of a careful assessment.

It is known that there is a need for specificity of training, e.g. specific balance exercises to improve balance among older adults (50). Adequate balance relies on inputs from the visual, somatosensory and vestibular systems (6), which are frequently impaired in people with MS (21). Surprisingly, only one study (28) was based on this theoretical framework. Muscle weakness and spasticity have been found to further compromise the ability to balance, as they affect the sequencing and force of muscle contraction (21). Four RCTs included in this review studied the effects of resistance and aerobic training on balance (29–32).

The strongest evidence for positive outcomes in the present analyses are associated with interventions based on the theoretical background of balance (i.e. specific balance exercises) and interventions using a well-defined progressive exercise training programme. Future studies could provide a better estimate of the effects of PT interventions on balance if the interventions were planned and described in more detail (e.g. type of exercise, duration of intervention, weekly frequency of exercise, amount of exercise per session).

Therapeutic exercise is a method of general, non-specific, active, functional therapy. In another part of this larger study, 3 occupational therapy (OT) studies were identified after screening of 35 full-text papers (24). All of these studies dealt with an energy conservation course in persons with MS, and none of them used an outcome measure linked to the ICF category of “Changing and maintaining body position”. In addition, multidisciplinary rehabilitation is an important component of symptomatic and supportive treatment for MS. Overall, 13 multidisciplinary treatment interventions were taken into consideration in our review, but none of them fulfilled inclusion criteria. That is, no separate analysis of a single method was conducted and no outcome measures linked to the ICF category of “Changing and maintaining body position” were found. The Cochrane Review on multidisciplinary rehabilitation for adults with MS (51) expressed no outcome measures linked to the ICF category of “Changing and maintaining body position”. Overall, further studies focusing on balance and different therapies in people with MS are needed.

Balance control is an integral component of all daily activities, but its complex and flexible nature makes it difficult to assess adequately (7). Understanding the biomechanical and information processing demands imposed by the task and by the environmental context allows us to evaluate their probable impact on motor performance and balance. Drawing on this model (7), most of the functional measures (e.g. BBS, FR, DGI, timed standing) used in the studies included here have a closed task condition and simple and stable environmental conditions; however, they differ in terms of the base of support (stationary or moving) and balance mechanism (predictive or proactive) used. Only one study (35) used the mCTSIB, which features more constraints from the environmental context. The Cochrane Review recommended that the WHO’s ICF classification (2) should be used as a basis for outcome measurement (52). Most of these measures represent the ICF “Changing and maintaining body position” category and capacity qualifiers. In addition, 3 self-report questionnaires that were used to assess the participants’ performance in their current environment were included in the meta-analysis. The studies analysed here also, to a varying extent, included outcome measures of the ICF components of body functions, such as muscle strength or spasticity, but they were not included in the analysis. The ICF is closely related to the concept of well-being and it contains the content of items of instruments to address Health-Related Quality of life (53). However, there are non-health aspects which are part of the universe of well-being not covered in the ICF. Therefore quality of Life (QoL) measures were not included in our review. A further review of the effects of balance exercises on QoL is needed. In agreement with a previous review (54), our review also highlights the need for the use of more consistent measures across studies, allowing comparison of results.

Despite the fact that many people with MS fall frequently, fear falling and are at increased risk of fall-related injuries (14–20), falls were not a primary focus of any of the studies. The results of our review reveal that, while motor and sensory strategies training (28) had no significant effect on the self-reported tests, they reduced the fall rate. In addition, both cycling progressive resistance training and a home-exercise programme (30) reduced FES and Feldenkrais-based group therapy (35) yielded significant improvements in balance confidence (ABC-scale). Consequently, we recommend further interventions focusing on fall prevention in people with MS.

Previously, several meta-analyses have been conducted to evaluate the overall effects of exercise on walking mobility (23) and quality of life (55) in MS. Some reviews have focused on mobility (21, 22) and one on hippotherapy (56). Bronson et al. (56) found that hippotherapy had a positive effect on balance in people with MS, but this finding was based on only 3 non-RCT studies. To our knowledge, our study was the first meta-analysis to explore the effect of PT interventions on balance.

The search strategy was designed to include a broad range of research regarding PT interventions in people with MS. Inclusion criteria were set to ensure rigorous searching and to enable this review to be replicated by others. Although our results are mostly positive, the limitations of this meta-analysis should be taken into account. Table I shows that the studies included in this review have a moderate or high risk of bias, e.g. selection, performance, attrition and detection bias. However, it should be noted that all the PT treatments studied were well tolerated by the participants and had no negative effects. Therefore, further high-quality studies are needed in order to develop treatment recommendations for clinicians treating people with MS.

Implications for future studies

Recommendations for future research, based on this review, are similar to those of the Cochrane Review on exercise therapy (52) and from mobility reviews (21–23). Methodological weaknesses, such as small sample sizes, lack of adequate randomization and blinding, and inadequate reporting of intervention protocols, continue to be as evident now as they were in 2004 (52). This review reiterates the need for better quality studies that address these weaknesses.

The present results indicate clearly that it is of the utmost importance to conduct studies that stratify people with MS according to their mobility level. It is also important that future studies carefully consider the sample size required to detect any potential between-group differences. Only then will it be possible to determine the most effective intervention for treating people with MS who have different mobility problems. To allow for data pooling, future studies need to standardize the relevant methodological aspects, e.g. inclusion and exclusion criteria for subjects, outcome measures and follow-up time.

Conclusion

The evidence assembled here suggests that there is a need for specificity of training, e.g. specific balance exercises to improve balance. In addition, there is some evidence that progressive resistance and aerobic training have positive effects on balance in people with MS whose level of disability is mild or moderate. Evidence for severely disabled people is lacking. The review emphasizes the need for high-quality RCTs with larger numbers of participants and a longer follow-up period. There is a need for more rigorous, scientifically sound research that is based on the theoretical background of balance. The use of standardized assessment instruments would ultimately improve the quality of MS research and would enable comparisons across studies.

Acknowledgements

We thank information specialists Anitta Pälvimäki, Jyväskylä University library, and Irma Helldan, Social Insurance Institute of Finland, and Eeva Aartolahti (EA), MSc, Janne Halonen, MSc, and Sari Hakkarainen, PT, Department of Health Sciences, University of Jyväskylä, for their help during the data collection. We also thank Atte Hänninen, MSc, Department of Mathematical Information Technology, University of Jyväskylä, for his statistical assistance.

This work was supported by the Social Insurance Institution of Finland [grants # 31863].

RefereNces