Exercise therapy for fatigue in multiple sclerosis

Martin Heine; Ingrid van de Port; Marc B Rietberg; Erwin EH van Wegen; Gert Kwakkel

doi:10.1002/14651858.CD009956.pub2

. 2015 Sep 11;2015(9):CD009956. doi: 10.1002/14651858.CD009956.pub2

Exercise therapy for fatigue in multiple sclerosis

Martin Heine ¹, Ingrid van de Port ¹, Marc B Rietberg ^2,^✉, Erwin EH van Wegen ², Gert Kwakkel ²

Editor: Cochrane Multiple Sclerosis and Rare Diseases of the CNS Group

PMCID: PMC9554249 PMID: 26358158

Abstract

Background

Multiple sclerosis (MS) is an immune‐mediated disease of the central nervous system affecting an estimated 1.3 million people worldwide. It is characterised by a variety of disabling symptoms of which excessive fatigue is the most frequent. Fatigue is often reported as the most invalidating symptom in people with MS. Various mechanisms directly and indirectly related to the disease and physical inactivity have been proposed to contribute to the degree of fatigue. Exercise therapy can induce physiological and psychological changes that may counter these mechanisms and reduce fatigue in MS.

Objectives

To determine the effectiveness and safety of exercise therapy compared to a no‐exercise control condition or another intervention on fatigue, measured with self‐reported questionnaires, of people with MS.

Search methods

We searched the Cochrane Multiple Sclerosis and Rare Diseases of the Central Nervous System Group Trials Specialised Register, which, among other sources, contains trials from: the Cochrane Central Register of Controlled Trials (CENTRAL) (2014, Issue 10), MEDLINE (from 1966 to October 2014), EMBASE (from 1974 to October 2014), CINAHL (from 1981 to October 2014), LILACS (from 1982 to October 2014), PEDro (from 1999 to October 2014), and Clinical trials registries (October 2014). Two review authors independently screened the reference lists of identified trials and related reviews.

Selection criteria

We included randomized controlled trials (RCTs) evaluating the efficacy of exercise therapy compared to no exercise therapy or other interventions for adults with MS that included subjective fatigue as an outcome. In these trials, fatigue should have been measured using questionnaires that primarily assessed fatigue or sub‐scales of questionnaires that measured fatigue or sub‐scales of questionnaires not primarily designed for the assessment of fatigue but explicitly used as such.

Data collection and analysis

Two review authors independently selected the articles, extracted data, and determined methodological quality of the included trials. Methodological quality was determined by means of the Cochrane 'risk of bias' tool and the PEDro scale. The combined body of evidence was summarised using the GRADE approach. The results were aggregated using meta‐analysis for those trials that provided sufficient data to do so.

Main results

Forty‐five trials, studying 69 exercise interventions, were eligible for this review, including 2250 people with MS. The prescribed exercise interventions were categorised as endurance training (23 interventions), muscle power training (nine interventions), task‐oriented training (five interventions), mixed training (15 interventions), or 'other' (e.g. yoga; 17 interventions). Thirty‐six included trials (1603 participants) provided sufficient data on the outcome of fatigue for meta‐analysis. In general, exercise interventions were studied in mostly participants with the relapsing‐remitting MS phenotype, and with an Expanded Disability Status Scale less than 6.0. Based on 26 trials that used a non‐exercise control, we found a significant effect on fatigue in favour of exercise therapy (standardized mean difference (SMD) ‐0.53, 95% confidence interval (CI) ‐0.73 to ‐0.33; P value < 0.01). However, there was significant heterogeneity between trials (I² > 58%). The mean methodological quality, as well as the combined body of evidence, was moderate. When considering the different types of exercise therapy, we found a significant effect on fatigue in favour of exercise therapy compared to no exercise for endurance training (SMD_{fixed effect} ‐0.43, 95% CI ‐0.69 to ‐0.17; P value < 0.01), mixed training (SMD_{random effect} ‐0.73, 95% CI ‐1.23 to ‐0.23; P value < 0.01), and 'other' training (SMD_{fixed effect} ‐0.54, 95% CI ‐0.79 to ‐0.29; P value < 0.01). Across all studies, one fall was reported. Given the number of MS relapses reported for the exercise condition (N = 25) and non‐exercise control condition (N = 26), exercise does not seem to be associated with a significant risk of a MS relapse. However, in general, MS relapses were defined and reported poorly.

Authors' conclusions

Exercise therapy can be prescribed in people with MS without harm. Exercise therapy, and particularly endurance, mixed, or 'other' training, may reduce self reported fatigue. However, there are still some important methodological issues to overcome. Unfortunately, most trials did not explicitly include people who experienced fatigue, did not target the therapy on fatigue specifically, and did not use a validated measure of fatigue as the primary measurement of outcome.

Plain language summary

Exercise therapy for fatigue in multiple sclerosis

Background

Multiple sclerosis (MS) is a chronic disease affecting over 1.3 million people globally. MS is characterized by diffuse damage to the central nervous system, leading to a wide range of different physical and cognitive (mental processes) symptoms. One of the most prominent and disabling symptoms of MS is fatigue. Currently, there is no effective medicine to reduce fatigue in people with MS. Treatment with exercise may be a way to reduce fatigue either directly by changing how the body works, for example hormonal function, or indirectly through improved physical activity and general health.

Study characteristics

We searched scientific databases for clinical trials comparing exercise to no exercise or other treatments in adults with MS. The evidence is current to October 2014.

Key results

We found 45 trials, involving 2250 people with MS, assessing the effect of exercise therapy using self reported fatigue. We used 36 studies, involving 1603 people with MS, in an analysis. Combined, these 36 trials supported the idea that exercise therapy may be a promising treatment to reduce fatigue without side events. This finding seems especially true for endurance training, mixed training (i.e. muscle power training mixed with endurance training), or 'other' training (e.g. yoga, tai‐chi). To assess the safety of exercise therapy we counted the number of reported MS relapses in the people receiving exercise therapy and in people in a non‐exercise group and did not find a significant difference.

Quality of the evidence

Even though these results are promising, it is worth noting some methods used in the trials may have affected the reliability of the results. For example, most trials included a low number of participants and did not primarily aim to reduce fatigue (but, for instance, aimed to improve walking capability) with the assessment of fatigue being a secondary measure. However, in contrast, exercise therapy may also be less feasible for people with MS who are severely fatigued. In addition, the reporting and definition of MS relapses was in general poor, and lacked consistency. Future, high‐quality research is warranted to elucidate the feasibility, effects, and working mechanisms of exercise therapy. Future studies may benefit from a uniform definition of fatigue, and subsequently be designed to measure fatigue specifically.

Summary of findings

Background

General consensus is that multiple sclerosis (MS) is induced by an anomalous immune response leading to a disseminated distribution of demyelination across the central nervous system (CNS). Globally (112 countries included), an estimated 1.3 million people have MS (WHO 2008). The prevalence of MS has increased over the past years, predominantly due to increased survival rates (Koch‐Henriksen 2010).

MS is characterized by a wide spectrum of symptoms including cognitive decline, muscle weakness, spasticity, and excessive fatigue. Physical and mental deconditioning due to MS or comorbidities, or both, may play a pivotal role in the development or persistence of fatigue. Hence, improving physical and mental health by means of exercise therapy may provide a valuable intervention in the management of fatigue in people with MS. Exercise therapy has already been suggested to be beneficial for reducing fatigue in, for example, chronic fatigue syndrome (White 2011), stroke (Zedlitz 2012), and following cancer (Ahlberg 2003).

Description of the condition

Fatigue is reported in over 80% of the MS population (Zajicek 2010), and is often perceived as one of the most debilitating symptoms (Wynia 2008). The Multiple Sclerosis Council for Clinical Practice Guidelines defines fatigue as "a subjective lack of physical and/or mental energy that is perceived by the individual or caregiver to interfere with usual and desired activities" (MS Council 1998). This definition implies that fatigue can have mental and/or physical origin and is subjective in nature. Due to its subjective nature, fatigue can only be measured using self reported (or caregiver‐reported) questionnaires. Without interference, fatigue is sustained over time and has a considerable impact on health‐related quality of life (HRQoL) (Olascoaga 2010). The underlying mechanisms of fatigue are unknown; however, these mechanisms may be both primary and/or secondary related to MS. In summary, fatigue in people with MS is a disabling, subjective, and multi‐dimensional symptom with multiple causal and perpetuating pathways (Chaudhuri 2004).

Description of the intervention

In accordance with Rietberg 2005, exercise therapy was defined as "a series of movements with the aim of training or developing the body by a routine practice or as a physical training to promote good physical health" (Webster's New World Dictionary 1982). However, it is now common belief that exercise therapy may also benefit mental health in both healthy and diseased people (Cooney 2013). With respect to this definition, supervised or non‐supervised interventions of all duration, frequency, and intensity were considered. The control group included no exercise training (e.g. wait list) or another intervention.

How the intervention might work

The underlying mechanisms regarding the effects of exercise therapy on fatigue in MS are still poorly understood. Exercise therapy may increase the energy reserves available for physical work and alleviate the detrimental effects of physical inactivity (Andreasen 2011). In addition, exercise therapy and/or physical activity in itself may enhance neurobiological processes that could promote neuroprotection and neuroplasticity and reduce long‐term disability (White 2008a; White 2008b). The effects of hypothalamus‐pituitary‐adrenal (HPA) axis deregulation (Gottschalk 2005), long‐term potentiation, and long‐term depression as adaptive mechanisms (Cooke 2006) have gained increased attention. Exercise therapy has the ability to induce changes in these neurobiological processes; processes also closely related to chronic fatigue (Chaudhuri 2004). Compared to pharmacological and non‐pharmacological interventions such as energy conservation management and cognitive behavioural therapy, exercise may affect both primary mechanisms (e.g. neuroprotection or hormonal function), as well as secondary factors related to fatigue (e.g. inactivity or co‐morbidity). In addition, exercise therapy may be a relatively simple, easily accessible, non‐invasive intervention to reduce fatigue. Petajan and colleagues were one of the first to show that exercise therapy may be beneficial for people with MS and could be performed safely (Petajan 1996).

Why it is important to do this review

A previous Cochrane review entitled 'Exercise therapy for multiple sclerosis' investigated systematically the literature for evidence of the effectiveness of exercise therapy for people with MS, in terms of activities of daily living and HRQoL (Rietberg 2005). This systematic review and best‐evidence synthesis found strong evidence in favour of exercise therapy compared to no exercise therapy, in terms of muscle power function, exercise tolerance functions, and mobility‐related activities. However, it found no evidence for an effect of exercise therapy on fatigue when compared to no exercise therapy. Due to the prevalence, disabling character, and unexplained aetiology of fatigue in MS, the number of trials investigating exercise therapy and included fatigue as an outcome, has grown exponentially in the 2000s. In addition, a number of studies have suggested that non‐pharmacological interventions such as exercise therapy may have beneficial effects in the treatment of fatigue in other diseases, such as chronic fatigue syndrome (White 2011), and cancer (Ahlberg 2003). Hence, the review authors found it feasible to study the effects of exercise therapy on fatigue specifically, in an updated yet differentiated version of the original review (Rietberg 2005).

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Randomized controlled clinical trials (RCTs) including single‐blind, unblinded, and cross‐over trials, comparing exercise therapy to a no‐exercise control condition or another intervention.

Types of participants

Participants aged 18 years or over, with the clinical confirmed diagnosis of MS according to applicable diagnostic criteria (McDonald 2001; Polman 2005; Polman 2011; Poser 1983; Schumacher 1965).

Types of interventions

We considered for inclusion all trials that fitted the authors' definition of exercise therapy, except those trials in which the exercise therapy was associated with learning to handle products and technology in daily living (International Classification of Functioning, Disability and Health (ICF) model; e115). In addition, based on the ICF model, we categorized exercise therapy as related to endurance (b455), muscle power (b730), task‐oriented (i.e. d450; walking), and mixed or other (WHO 2012). Task‐oriented training is considered different from the other categories based on the aim of the intervention, that is, to improve performance at a certain task such as walking, and not to improve, for instance, endurance.

There were no restrictions as to the duration, frequency, or intensity of exercise therapy. Neither were there restrictions as to the content of the control treatment, which could have been either a different type or intensity exercise therapy as well as a non‐exercise control (e.g. wait list) or other treatment.

Types of outcome measures

Primary outcomes

We assessed fatigue as the primary outcome at the end of the intervention period, and during follow‐up as measured by:

questionnaires that primarily assessed fatigue, such as: Fatigue Severity Scale (FSS; Krupp 1989), Modified Fatigue Impact Scale (MFIS; Fisk 1994), Multidimensional Fatigue Index (MFI; Smets 1995), Visual Analogue Scale for fatigue;
sub‐scales of questionnaires that measured fatigue or sub‐scales not primarily designed for the assessment of fatigue but used in such, for example: Short Form‐36 sub‐scale (SF‐36; e.g. vitality sub‐scale; Ware 1992), and Multiple Sclerosis Quality of Life 54 (MSQoL‐54; e.g. physical functioning sub‐scale; Vickrey 1995). We only used these sub‐scales if it was specifically noted that these were included to assess fatigue.

Secondary outcomes

Safety of exercise therapy in people with MS, in terms of number of MS relapses and number of reported falls, during the intervention and follow‐up period. We defined an MS relapse as newly developed or recently worsened symptoms of neurological dysfunction, with objective confirmation, lasting more than 24 hours. However, we also considered less stringent criteria to identify a MS relapse (i.e. without objective confirmation).

Search methods for identification of studies

We applied no language restrictions to the search.

Electronic searches

The Trials Search Co‐ordinator searched the Cochrane Multiple Sclerosis and Rare Diseases of the Central Nervous System Group's Specialised Register, which is updated regularly and contains trials identified from the following.

The Cochrane Central Register of Controlled Trials (CENTRAL) (2014, Issue 10).
MEDLINE (PubMed) (1966 to 3 October 2014).
EMBASE (1974 to 3 October 2014).
CINAHL (EBSCOhost) (1981 to 3 October 2014).
LILACS (BIREME) (1982 to 3 October 2014).
PEDro (1999 to 3 October 2014);
Clinical trial registries (www.clinicaltrials.gov).
World Health Organization (WHO) International Clinical Trials Registry Portal (apps.who.int/trialsearch/).

Information on the Group's Trials Register and details of search strategies used to identify trials can be found in the 'Specialised Register' section within the Cochrane Multiple Sclerosis and Rare Diseases of the Central Nervous System Group's module.

Appendix 1 lists the keywords used in the search strategy.

Two review authors searched PsycINFO (1887 to 3 October 2014) (Appendix 2), and an independent medical librarian searched SPORTDiscus (1985 to 12 July 2012) (Appendix 3).

Searching other resources

To identify other relevant trial data we:

contacted authors of published trials when reported data were incomplete;
screened reference lists of review articles and primary trials found;
contacted authors of unpublished manuscripts to ask if they were willing to disclose their unpublished data; and
contacted experts in the field to identify further published or unpublished trials.

Data collection and analysis

Selection of studies

Two review authors (MH, IP) independently selected trials for inclusion using pre‐determined inclusion criteria. First, we screened titles and abstracts of all citations. Second, we acquired the full‐text of the remaining citations, and read each one to determine eligibility. In all cases, we resolved any disagreements about trial inclusions by consensus among review authors and consulted a third review author (GK) if disagreements persisted.

Data extraction and management

For each included trial, one review author (MH) and a research assistant (ME, see Acknowledgements) documented the following information in the pre‐defined form using Review Manager 5 (RevMan 2014), and Section 7.3 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a).

Trial design.
Participant characteristics (number, age, type of MS, Expanded Disability Status Scale (EDSS) score, gender, and disease duration).
Inclusion and exclusion criteria.
Brief description and type of the experimental intervention(s).
Brief description of the control intervention.
Outcomes and visits reported.

For quantitative analysis, one review author (MH) and a research assistant (ME) independently extracted trial outcome data. We entered all of the extracted information in both printed data extraction forms and Review Manager 5 (RevMan 2014).

Assessment of risk of bias in included studies

Two review authors or research assistants (MH and IP or ME) independently assessed the risk of bias and methodological quality for all included trials using the PEDro scale (Maher 2003), and provided a domain‐based clinical judgement based on the Cochrane 'Risk of bias' tool (Section 8.5 of the Cochrane Handbook for Systematic Reviews of Interventions; Higgins 2011b). The PEDro scale consists of 11 items assessing the randomization procedure, baseline characteristics of the trial sample, blinding, drop‐out rate, and results presented. Each item is rated as 0 (no) or 1 (yes). The first item (description of inclusion criteria) is excluded from the final PEDro score as it is related to the external validity of the trial and not to the internal validity and methodological quality of the trial. Hence, the PEDro scale entails 10 items leading to a score ranging from 0 to 10. The Cochrane 'Risk of bias' tool strongly resembles the items from the PEDro scale, however it demands a judgement‐based approach rather than an actual rating. Hence, they complemented each other in the assessment of methodological quality and risk of bias. The PEDro items are as follows.

Random allocation; the precise method of random allocation does not need to be specified. Procedures like coin‐toss and dice roll are considered random as well.
Allocation concealment; the person who determined if a person was eligible for inclusion in the trial was unaware, when this decision was made, of which group the person would be allocated to.
Groups were similar at baseline regarding the most prognostic factors; given the subject of the present review, these factors were considered disease severity, baseline fatigue, and depression.
Blinding of all participants; blinding means the participant did not know which group he/she had been allocated to.
Blinding of all therapists who administered the therapy; blinding means the therapist did not know which group the participant had been allocated to.
Blinding of all assessors; blinding means the assessor did not know which group the participant had been allocated to. In trials in which key outcomes are self reported (e.g. visual analogue scale, fatigue questionnaire), the assessor is considered to be blind if the participant was blind. It is highly unlikely that participants were unaware of treatment in terms of exercise therapy. Hence, this item was scored as zero for all included trials.
Measures of at least one key outcome, in the present review this is fatigue, were obtained from more than 85% of the participants initially allocated to groups.
All participants for whom outcome measures were available received the treatment or control condition as allocated or, where this was not the case, data for at least one key outcome were analysed by 'intention to treat'.
The results of between‐group statistical comparisons are reported for at least one key outcome, in the present review this is fatigue.
The trial provides both point measures and measures of variability for at least one key outcome, in the present review this is fatigue.

Measures of treatment effect

Depending on the nature of the outcomes, we used the following effect measures including 95% confidence intervals (CI).

Dichotomous data: odds ratio (OR), risk ratio (RR) or risk difference (RD); number needed to treat (NNT).
Continuous data: mean difference (MD) or standardized mean difference (SMD).
Time‐to‐event data: hazard ratio (HR).

Unit of analysis issues

Cross‐over trials

We included cross‐over trials as if it was a parallel‐group trial of exercise therapy versus a control condition, but only if it was reported that there were no significant differences in fatigue between baseline, and at the end of the wash‐out phase (i.e. reflecting no carry‐over effects). In case of significant differences in fatigue between baseline and the end of the wash‐out phase, we used only the results of the first period before cross‐over (see Section 16.4 of the Cochrane Handbook for Systematic Reviews of Interventions; Higgins 2011c).

Trials with multiple treatment groups

When included trials compared two exercise therapy interventions or two different training intensities, we considered the intervention that was the primary focus of the trial as the experimental intervention. In case of two or more exercise interventions and one non‐exercise control group, we combined the exercise interventions using the method described in Table 16.4 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011c). For subsequent sub‐group analyses, in which the exercise modalities were compared distinctively, these interventions were included as separate trials in the meta‐analysis and compared only with n_control/n_{interventions} of the control sample. The same approach was used in case trials used multiple fatigue outcome measures.

Dealing with missing data

If necessary, we contacted authors to obtain missing data. This led to additional data in two cases (Smedal 2011; Velikonja 2010). We used no statistical methods to impute missing data.

Assessment of heterogeneity

We examined heterogeneity across trials by inspecting the distribution of point estimates for the effect measure and the overlap in their CIs on the forest plot. We used the I² statistic to check the statistical consistency, defined as the ratio between‐trial variation compared to the overall variation (see Section 9.5.2 of the Cochrane Handbook for Systematic Reviews of Interventions; Deeks 2011). A value greater than 50% may indicate significant heterogeneity.

Assessment of reporting biases

To estimate the influence of unpublished papers and small studies on the overall effect (i.e. publication bias), we visually inspected the funnel plot. In the absence of small‐study effects, results from small trials with large standard errors will scatter widely at the bottom of a funnel plot while the spread narrows with increasing sample size and the plot will resemble a symmetrical inverted funnel. Conversely, an asymmetrical funnel plot is suggestive of small‐study effects. Small‐study effects can result from a combination of lower methodological quality of small trials, publication bias, or other reporting bias (Egger 1997; Nuesch 2010).

Data synthesis

Meta‐analysis

We used included trials that provided sufficient data for pooling for meta‐analysis. Even though the used outcome measures, as well as the type and duration of intervention, might differ, we adapted the pooled SMD as an overall measure of the effect of exercise therapy on fatigue. To check for heterogeneity, we used the Chi² test in conjunction with the I² statistic. In case of homogeneity (I² < 50%), we used a fixed‐effect model. In the situation of heterogeneity (Ι² ≥ 50%), we applied a random‐effects model to take into account the between‐trial variance. In accordance with Cohen, we classified effects sizes into small (less than 0.2), moderate (0.2 to 0.8), and large (greater than 0.8) (Cohen 1977). If possible, we then re‐expressed the SMD into the original continuous fatigue outcome, by multiplying the SMD with the standard deviation of the control group of the best‐powered (i.e. largest group) trial.

GRADE

We evaluated the overall quality of the body of evidence using the GRADE approach as recommended by The Cochrane Collaboration (Higgins 2011a). The quality of the evidence for a specific outcome is based upon performance against five principal domains: 1. limitations in design, suggestive of a high likelihood of bias; 2. inconsistency of results; 3. indirectness of the evidence, for instance if exercise therapy for fatigue was only studied in ambulant participants; 4. imprecision of results, for instance in trials with a small number of participants where CIs were wide; and 5. a high probability of publication bias. The quality starts at high when high‐quality RCTs provide results for the outcome, and can be reduced by a level each time if one of the five principal domains is not met. We used the following levels to measure the quality of the body of evidence.

High quality evidence: There are consistent findings among at least 75% of RCTs, with no limitations of the trial design; consistent, direct, and precise data, and no known or suspected publication biases. Further research is unlikely to change either the estimate of effect or our confidence in the results.

Moderate quality evidence: One of the five domains is not met. Further research is likely to have an important impact on our confidence in the estimate of effect, and may change the estimate.

Low quality evidence: Two of the five domains are not met. Further research is very likely to have an important impact on our confidence in the estimate of effect, and is likely to change the estimate.

Very low quality evidence: Three of the five domains are not met. We are very uncertain about the results.

No evidence: No RCTs were identified that addressed this outcome.

Subgroup analysis and investigation of heterogeneity

In case of sufficient power, we carried out different sub‐group analyses, for example, to compare the effect of exercise therapy for different exercise modalities, different control conditions, or different fatigue outcomes.

Sensitivity analysis

We undertook a sensitivity analysis to assess the influence of methodological quality on the intervention effect (PEDro > 5). Sensitivity analyses are sometimes confused with sub‐group analyses. Although some sensitivity analyses involve restricting the analysis to a subset of the totality of trials, the two methods differ in two ways. First, sensitivity analyses do not attempt to estimate the effect in the group of trials removed from the analysis, whereas in sub‐group analyses, estimates are produced for each sub‐group. Second, in sensitivity analyses, informal comparisons are made between different ways of estimating the same thing, whereas in sub‐group analyses, formal statistical comparisons are made across the sub‐groups (Higgins 2011a).

Results

Description of studies

See Characteristics of included studies; Characteristics of excluded studies.

Results of the search

See Figure 1.

The initial search, which included fatigue as a search component, yielded 213 citations (i.e. multiple sclerosis AND exercise therapy AND fatigue; Appendix 1). However, using the 'AND fatigue' section, independent of the keywords chosen, resulted in unacceptable insensitivity. That is, key trials that should have been part of the search results were not. Hence, we performed a second search without the fatigue section (i.e. multiple sclerosis AND exercise therapy; Appendix 4). The latter search was then updated in October 2014. After removal of duplicates, two review authors (MH and IP) independently screened 839 abstracts and titles for inclusion and exclusion criteria. The main reasons for exclusion were the trial design, no exercise therapy, or no fatigue outcome. We retrieved 85 full‐text articles of which we excluded an additional 40 (see Characteristics of excluded studies table) or moved to studies awaiting classification (see Characteristics of studies awaiting classification table). The final number of trials included in this review was 45, all written in English and published after 1995. We found no additional references based on reference screening of related reviews and included trials.

Included studies

The 45 trials that we included studied 69 different exercise therapy interventions. We categorized each exercise therapy intervention into the earlier described, ICF based (WHO 2012) modalities: endurance (23 interventions, b455), muscle power (nine interventions, b730), task‐oriented (five interventions, i.e. d450; walking), mixed (15 interventions), or other (17 interventions). See Characteristics of included studies table.

The mean (± SD) number of participants across all included 45 trials was 34 ± 38 for the intervention group and 16 ± 17 for the control group. In total, 1531 participants were involved in some type of exercise intervention while 719 participants served as non‐exercise controls. Furthermore, two trials included only people with a pre‐defined level of fatigue (Dettmers 2009; Hebert 2011). Three trials used fatigue as their primary measurement of outcome (Hebert 2011; Kargarfard 2012; Plow 2009). Out of the 45 included trials, 36 trials provided sufficient information on the outcome of fatigue to be included in a meta‐analysis. We contacted the corresponding authors of 11 trials to obtain missing information, which led to two additional trials included for the meta‐analysis. Based on the number and characteristics of the included trials, it was feasible to conduct a sub‐group analysis for the different fatigue outcomes used and the different types of exercise.

Excluded studies

See Characteristics of excluded studies; Figure 1.

We excluded 38 trials. The primary reason for excluding 26 trials was the lack of fatigue outcome measures included in the study design (Barrett 2009; Bjarnadottir 2007; Broekmans 2010; Broekmans 2011; Carter 2013a; Carter 2013b; Cattaneo 2007; Claerbout 2012; Dalgas 2009; DeBolt 2004; Fimland 2010; Gosselink 2000; Hilgers 2013; Hojjatollah 2012; Keser 2013; Marandi 2013a; Marandi 2013b; McAuley 2007; Miller 2011; Mutluay 2007; Nilsagard 2013; Patti 2003; Paul 2014; Rodrigues 2008; Romberg 2005; Solari 1999). Five trials did not apply a randomized controlled design (Bayraktar 2013; Castellano 2008; Heesen 2003; Keser 2011; Rasova 2006). The interventions in four trials could not be considered exercise therapy (Grossman 2010; Guerra 2014; Jackson 2008; Stephens 2001). Other reasons included: no original data (Sherman 2004, data from Oken 2004), and unable to translate (Shanazari 2013).

Risk of bias in included studies

Two review authors or research assistants (MH and ME or IVP) independently assessed the risk of bias for the included trials using the PEDro scale (see Table 4) and provided a domain‐based clinical judgement using the Cochrane 'Risk of bias' tool (see Characteristics of included studies table; Figure 2). The initial agreement was 84% (Cohen's Kappa = 0.674) of the 450 scored items on the PEDro scale. Following a consensus meeting, no disagreements persisted. The asymmetrical shape (lack of narrowing at the top of the plot) of the funnel plot does suggest presence of publication bias or small‐study effects (Figure 3).

1. Risk of bias.

Study ID	Fatigue scale	ITT	1	2	3	4	5	6	7	8	9	10	Total score
Ahmadi 2013	FSS	no	1	0	1	0	0	0	1	1	1	1	6
Aydin 2014	FSS	no	1	0	1	0	0	0	1	0	1	1	5
Bansi 2013	FSMC	no	1	1	1	0	0	0	1	1	1	1	7
Brichetto 2013	MFIS	no	1	0	1	0	0	0	1	1	1	1	6
Briken 2014	MFIS	no	1	1	0	0	0	0	1	0	1	1	5
Burschka 2014	FSMC	no	0	0	0	0	0	0	0	0	1	1	2
Cakt 2010	FSS	no	1	1	1	0	0	0	0	0	1	1	5
Carter 2014	MFIS	yes	1	1	1	0	0	0	1	1	1	1	7
Castro‐Sanchez 2012	MFIS, FSS	no	1	1	1	0	0	0	1	0	1	1	6
Collett 2011	FSS	yes	1	1	1	0	0	0	1	1	1	1	7
Coote 2014	MFIS	no	1	1	1	0	0	0	0	1	1	1	6
Dalgas 2010	FSS	no	1	1	1	0	0	0	0	1	1	1	6
Dettmers 2009	MFIS	no	1	1	1	0	0	0	0	1	0	0	4
Dodd 2011	MFIS	yes	1	1	1	0	0	0	1	1	1	1	7
Frevel 2014	MFIS, FSS	no	1	0	1	0	0	0	1	0	1	1	5
Fry 2007	FSS	no	1	0	1	0	0	0	1	1	0	1	5
Gandolfi 2014	FSS	no	1	1	1	0	0	0	0	1	1	1	6
Garrett 2013	MFIS	no	1	1	1	0	0	0	0	1	1	1	6
Geddes 2009	FSS	no	1	0	1	0	0	0	0	1	1	1	5
Hayes 2011	FSS	no	1	0	1	0	0	0	1	1	1	1	6
Hebert 2011	MFIS	yes	1	1	1	0	0	0	1	1	1	1	7
Hogan 2014	MFIS	no	1	1	0	0	0	0	0	1	1	1	5
Kargarfard 2012	MFIS	yes	1	1	1	0	0	0	0	1	1	1	6
Klefbeck 2003	FSS	no	1	0	1	0	0	0	1	0	1	1	5
Learmonth 2011	FSS	yes	1	1	1	0	0	0	0	1	1	1	6
McCullagh 2008	MFIS	no	0	0	1	0	0	0	0	1	1	1	4
Mori 2011	FSS	no	1	0	1	0	0	0	1	1	0	0	4
Mostert 2002	FSS	no	1	0	1	0	0	0	0	0	0	1	3
Negahban 2013	FSS	no	1	0	1	0	0	0	1	1	1	1	6
Oken 2004	MFI (general)	no	1	0	1	0	0	0	0	1	0	1	4
Petajan 1996	POMS fatigue, FSS	no	1	0	1	0	0	0	0	1	0	1	4
Plow 2009	MFIS	no	1	0	0	0	0	0	0	1	0	0	2
Rampello 2007	MFIS	no	1	0	1	0	0	0	0	1	1	1	5
Sabapathy 2010	MFIS	no	1	0	1	0	0	0	0	1	1	1	5
Sangelaji 2014	FSS	no	1	0	1	0	0	0	0	1	0	0	3
Schulz 2004	MFIS	no	1	0	0	0	0	0	1	1	1	1	5
Smedal 2011	FSS	yes	1	0	1	0	0	0	1	1	1	1	6
Skjerbaek 2014	FSMC	no	1	1	1	0	0	0	1	1	1	1	7
Straudi 2014	FSS	yes	1	0	1	0	0	0	1	1	1	1	6
Surakka 2004	FSS	no	1	0	1	0	0	0	0	1	1	1	5
Sutherland 2001	POMS fatigue	no	1	0	1	0	0	0	1	1	1	1	6
Tarakci 2013	FSS	no	1	1	1	0	0	0	1	1	1	1	7
van den Berg 2006	FSS	no	1	1	1	0	0	0	0	1	1	1	7
Velikonja 2010	MFIS	no	1	0	0	0	0	0	1	0	0	1	4
Wier 2011	FSS	no	1	0	1	0	0	0	1	1	1	0	5
% of trials			96%	42%	87%	0%	0%	0%	53%	80%	80%	89%

Effect of exercise therapy for fatigue in multiple sclerosis ‐ overall analysis
Patient or population: people with multiple sclerosis  Intervention: exercise therapy ‐ overall analysis
Outcomes	*Illustrative comparative risks (95% CI)**		No of participants  (trials)	Quality of the evidence  (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Overall analysis
Fatigue	No risk assumed	The mean fatigue outcome in the intervention groups was  0.35 standard deviations lower   (0.57 to 0.13 lower)	1603  (36 trials)	⊕⊕⊕⊝  moderate¹	Indirectness (‐1)
Exercise versus no‐exercise control	No risk assumed	The mean fatigue outcome in the intervention groups was 0.58 standard deviations lower (0.81 to 0.34 lower) compared to a no‐exercise control group	1325  (27 trials)	⊕⊕⊕⊝  moderate¹	Indirectness (‐1)
Exercise versus exercise	No risk assumed	The mean fatigue outcome in the intervention groups was 0.28 standard deviations higher (0 to 0.56 higher) compared to an exercise control group	278  (9 trials)	⊕⊕⊝⊝  low^1,2	Indirectness (‐1) Imprecision (‐1)
The argumentation for downgrading the grades of evidence is provided in the footnotes. CI:* confidence interval.
GRADE Working Group grades of evidence  High quality: Further research is very unlikely to change our confidence in the estimate of effect.  Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.  Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.  Very low quality: We are very uncertain about the estimate.

Study	Time (i.e. duration of intervention)	Fatigue scale	Effect	Other outcomes	Effect
Ahmadi 2013 Compared aerobic training vs. control	8 weeks	FSS	‐	BBS Walk time Walk distance BDI BAI	+ ‐ + ns ns
Ahmadi 2013 Compared yoga vs. control	8 weeks	FSS	‐	BBS Walk time Walk distance BDI BAI	+ ns + ‐ ‐
Brichetto 2013 Compared Nintendo® Wii® balance training vs. control	4 weeks	MFIS	ns	BBS Open‐eye stabilometry Closed‐eye stabilometry	+ ‐ ‐
Briken 2014 Compared arm‐ergometry vs. control	10 weeks	MFIS	‐	VO_2peak 6MWT SDMT VLMT TAP (alertness) TAP (shift of attention) LPS RWT IDS ‐ SR30	ns + ns + ns ‐ ns ns ‐
Briken 2014 Compared rowing vs. control	10 weeks	MFIS	ns	VO_2peak 6MWT SDMT VLMT TAP (alertness) TAP (shift of attention) LPS RWT IDS ‐ SR30	ns ns ns + ns ns ns ns ns
Briken 2014 Compared bicycling vs. control	10 weeks	MFIS	ns	VO_2peak 6MWT SDMT VLMT TAP (alertness) TAP (shift of attention) LPS RWT IDS ‐ SR30	+ + ns + ‐ ‐ ns ns ‐
Burschka 2014 Compared Tai‐Chi yoga vs. control * no change of fatigue in experimental group, increase in fatigue in control group	24 weeks	FSMC*	‐	Balance Co‐ordination CES‐D QLS	+ + ‐ +
Cakt 2010 Compared progressive resistance training vs. control	8 weeks	FSS	‐	Duration of exercise W_max TUG DGI FR FES 10MWT BDI SF‐36	+ + ‐ + + ‐ ‐ ‐ unk
Cakt 2010 Compared home‐based exercise vs. control	8 weeks	FSS	ns	Duration of exercise W_max TUG DGI FR FES 10MWT BDI SF‐36	ns ns ns ns ns ns ns ns unk
Carter 2014 Compared a pragmatic exercise intervention vs. control	12 weeks	MFIS	‐	GLTEQ Accelerometer MSQoL‐54 MSFC 6MWT EDSS	+ + + ns ns ns
Castro‐Sanchez 2012 Compared Ai‐Chi aquatic programme vs. control	20 weeks	FSS MFIS ‐ physical ‐ cognitive ‐ psychosocial	ns ‐ ns ns	Pain MPQ‐PRI MPQ‐PPI RMDQ Spasm  MSIS‐29 ‐ physical ‐ psychological BDI BI	‐ ‐ ns ‐ ‐ ‐ ‐ ‐ ns
Dalgas 2010 Compared progressive resistance training vs. control	12 weeks	FSS MFI‐20 ‐ General fatigue ‐ Physical fatigue ‐ Reduced activity ‐ Reduced motivation ‐ Mental fatigue	‐ ‐ ns ns ns ns	MDI SF‐36 ‐ PCS ‐ MCS MVC (knee extensor) FS (%)	‐ ‐ ns + +
Dodd 2011 Compared progressive resistance training vs. control	10 weeks	MFIS ‐ physical ‐ cognitive ‐ psychosocial	‐ ‐ ns ns	MSIS‐88 muscle stiffness MSIS‐88 muscle spasms 2MWT Walking speed Leg press endurance (repetitions) Reversed leg press endurance (repetitions) 1RM leg press (kg) 1RM reversed leg press (kg) WHOQoL‐BREF overall quality of life WHOQoL‐BREF overall health WHOQoL‐BREF physical health	ns ns ns ns ns + ns ns ns ns +
Fry 2007 Compared inspiratory muscle training vs. control	10 weeks	No interaction effects reported
Garrett 2013 Compared physiotherapist‐led exercise vs. control	10 weeks	MFIS ‐ physical ‐ cognitive	‐ ‐ ns	MSIS‐29 physical component MSIS‐29 cognitive component 6MWT	‐ ‐ +
Garrett 2013 Compared fitness instructor‐led exercise vs. control	10 weeks	MFIS ‐ physical ‐ cognitive	‐ ‐ ns	MSIS‐29 physical component MSIS‐29 cognitive component 6MWT	‐ ‐ +
Garrett 2013 Compared yoga vs. control	10 weeks	MFIS ‐ physical ‐ cognitive	‐ ‐ ns	MSIS‐29 physical component MSIS‐29 cognitive component 6MWT	ns ‐ ns
Geddes 2009 Compared home walking vs. control	12 weeks	FSS	ns	6MWT PCI RPE	ns ns ns
Hebert 2011 Compared vestibular rehabilitation vs. no exercise control	6 weeks	MFIS	‐	SOT DHI 6MWT	+ ‐ ns
Hebert 2011 Compared exercise control vs. no exercise control	6 weeks	MFIS	ns	SOT DHI 6MWT	ns ns ns
Hogan 2014 Compared group physiotherapy vs. control	10 weeks	MFIS	ns	BBS 6MWT MSIS29v2	+ ns ns
Hogan 2014 Compared individual physiotherapy vs. control	10 weeks	MFIS	ns	BBS 6MWT MSIS29v2	+ ns ns
Hogan 2014 Compared yoga vs. control	10 weeks	MFIS	ns	BBS 6MWT MSIS29v2	+ ns ns
Kargarfard 2012 Compared aquatic training vs. control	8 weeks	MFIS ‐ physical ‐ psychosocial ‐ cognitive	‐ ‐ ‐ ‐	MSQoL‐54 ‐ Physical ‐ Mental	+ +
Klefbeck 2003 Compared inspiratory muscle training vs. control	10 weeks	No interaction effects reported
Learmonth 2011 Compared leisure centre‐based exercise group vs. control	12 weeks	FSS	ns	T25FW 6MWT BBS TUG QPW BMI PF ABC HADS LMSQoL	ns ns ns ns ns ns + ns ns ns
McCullagh 2008 Compared group circuit training vs. control	12 weeks	No interaction effects reported
Mori 2011 Compared transcranial magnetic stimulation (TMS) vs. control	2 weeks	No interaction effects reported
Mori 2011 Compared exercise control vs. control	2 weeks	No interaction effects reported
Mostert 2002 Compared short‐term exercise vs. control	4 weeks	No interaction effects reported
Negahban 2013 Compared exercise therapy vs. control	5 weeks	FSS	‐	VAS scale for pain MAS BBS TUG 10MWT 2MWT MSQoL‐54	ns ‐ + ‐ ‐ + ns
Negahban 2013 Compared massage + exercise therapy vs. control	5 weeks	FSS	‐	VAS scale for pain MAS BBS TUG 10MWT 2MWT MSQoL‐54	‐ ns + ‐ ‐ + ns
Oken 2004 Compared Iyengar yoga classes plus home programme vs. control	24 weeks	No interaction effects reported
Oken 2004 Compared weekly bicycle exercise classes along with home exercise vs. control	24 weeks	No interaction effects reported
Petajan 1996 Compared aerobic training vs. control	15 weeks	FSS POMS ‐ fatigue	ns ns	EDSS ISS VO_2max PWC HR_max Upper extremity strength Lower extremity strength POMS SIP	ns ns + + ns + + ns ns
Sangelaji 2014 Compared combination exercise therapy vs. control	10 weeks	FSS	‐	EDSS BBS 6MWT MSQoL	ns + + +
Schulz 2004 Compared aerobic interval training vs. control	8 weeks	MFIS ‐ physical ‐ cognitive ‐ social	ns ns ns ns	W_max VO_2max HR_max W endurance Lactate change Immune and neurotrophic factors IL‐6 (rest) IL‐6 (AUC) sIL‐6R (rest) sIL‐6R (AUC) BDNF (rest) BDNF (AUC) NGF (rest) NGF (AUC) HAQUAMS ‐ fatigue/thinking ‐ mobility lower ‐ mobility upper ‐ social function ‐ mood	ns ns ns ns ‐ ns ns ns ‐ ns ns ns ns ‐ ns ns ns ‐ ‐
Skjerbaek 2014 Compared endurance training vs. control	4 weeks	FSMC	ns	VO_2peak HR_peak 9HPT Hand grip power Box and blocks 6‐minute wheelchair MDI MSIS‐29	ns ns ns ns ns ns ns ns
Straudi 2014 Compared task‐oriented circuit training vs. control	2 weeks	FSS	ns	10MWT 6MWT TUG DGI MSWS‐12 MSIS‐29 ‐ physical ‐ psychosocial ‐ psychological	ns + ns ns + + + ‐
Surakka 2004 Compared inpatient rehabilitation plus home‐based exercise vs. control	26 weeks	FSS	ns	Leg flexor/extensor torque Motor fatigue Ambulatory fatigue index	ns ns ns
Sutherland 2001 Compared aerobic aquatic training vs. control	10 weeks	No interaction effects reported
Tarakci 2013 Compared group exercise programme vs. control	12 weeks	FSS	‐	BBS 10MWT  10SCT R Hip flexors MAS  L Hip flexors MAS R Hamstring MAS  L Hamstring MAS R Achilles MAS  L Achilles MAS MusiQoL	+ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ +
van den Berg 2006 Compared treadmill exercise vs. control (cross‐over)	4 weeks	No interaction effects reported

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Fatigue	36	1603	Std. Mean Difference (IV, Random, 95% CI)	‐0.35 [‐0.57, ‐0.13]
1.1 Exercise versus non‐exercise control	27	1325	Std. Mean Difference (IV, Random, 95% CI)	‐0.58 [‐0.81, ‐0.34]
1.2 Exercise versus exercise	9	278	Std. Mean Difference (IV, Random, 95% CI)	0.28 [0.00, 0.56]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Fatigue	26	1304	Std. Mean Difference (IV, Random, 95% CI)	‐0.53 [‐0.73, ‐0.33]
1.1 Modified Fatigue Impact Scale (MFIS)	8	688	Std. Mean Difference (IV, Random, 95% CI)	‐0.40 [‐0.58, ‐0.22]
1.2 Fatigue Severity Scale (FSS)	13	449	Std. Mean Difference (IV, Random, 95% CI)	‐0.56 [‐0.95, ‐0.17]
1.3 Other	5	167	Std. Mean Difference (IV, Random, 95% CI)	‐0.54 [‐1.01, ‐0.07]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Fatigue	26	1299	Std. Mean Difference (IV, Random, 95% CI)	‐0.52 [‐0.72, ‐0.32]
1.1 Endurance	11	266	Std. Mean Difference (IV, Random, 95% CI)	‐0.46 [‐0.78, ‐0.15]
1.2 Muscle power	4	207	Std. Mean Difference (IV, Random, 95% CI)	0.03 [‐0.65, 0.71]
1.3 Task‐oriented	2	36	Std. Mean Difference (IV, Random, 95% CI)	‐0.34 [‐1.02, 0.33]
1.4 Mixed	6	495	Std. Mean Difference (IV, Random, 95% CI)	‐0.73 [‐1.23, ‐0.23]
1.5 Other	9	295	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [1.00, ‐0.29]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Fatigue	26	1299	Std. Mean Difference (IV, Fixed, 95% CI)	‐0.50 [‐0.62, ‐0.37]
1.1 Endurance	11	266	Std. Mean Difference (IV, Fixed, 95% CI)	‐0.43 [‐0.69, ‐0.17]
1.2 Muscle power	4	207	Std. Mean Difference (IV, Fixed, 95% CI)	‐0.19 [‐0.53, 0.15]
1.3 Task‐oriented	2	36	Std. Mean Difference (IV, Fixed, 95% CI)	‐0.34 [‐1.02, 0.33]
1.4 Mixed	6	495	Std. Mean Difference (IV, Fixed, 95% CI)	‐0.63 [‐0.83, ‐0.43]
1.5 Other	9	295	Std. Mean Difference (IV, Fixed, 95% CI)	‐0.54 [‐0.79, ‐0.29]

Methods	Design: random assignment to aerobic treadmill training (AT), yoga (Y) or a wait‐list control (NEX) Setting: physiotherapy clinic, Iran Categorization: AT = endurance, Y = other
Participants	n = 31; AT = 10, Y = 11, NEX = 10 Inclusion criteria: physician diagnosis MS, self reported EDSS 1 >< 4.0, able to walk a treadmill > 5 minutes at stable speed, no physical activity in the last 3 months Exclusion criteria: cardiovascular disease, liver or kidney failure, lung disease, diabetes, thyroid disorder, gout or orthopaedic limitations, pregnant, addicted (i.e. smoking) Type MS: ? Disease duration (yr) ± SD: AT = 5.6 ± 3.3, Y = 4.7 ± 5.6, NEX = 5.0 ± 3.1 Mean age (yr) ± SD: AT = 36.8 ± 9.2, Y = 32.3 ± 8.7, NEX = 36.7 ± 9.3 % Female: 100% Mean EDSS ± SD: AT = 2.4 ± 1.2, Y = 2.0 ± 1.1, NEX = 2.3 ± 1.3
Interventions	AT: 8 weeks, 3x per week, 30 minutes physiotherapist‐led treadmill training at 40‐75% of predicted maximal heart rate + 20 minutes of stretching. Comfortable walking speed as starting point, progression directed by participant Y: 8 weeks, 3x per week, 60‐70 minutes Hatha yoga led by a physiotherapist and neurologist. Hatha yoga included stretching, postures, breathing, and meditation NEX: wait list control
Outcomes	FSS, BDI, BAI, BBS, 10MWT, 2MWT (walking speed, walking endurance)
Notes	Drop‐outs No drop‐outs reported Measurements Baseline, 8 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomization procedure not described
Allocation concealment (selection bias)	Unclear risk	Allocation procedure unknown
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	No drop‐outs reported
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Low risk	No other sources of bias identified

Methods	Design: random assignment to supervised hospital‐based callisthenic exercise (EX) OR home‐based callisthenic exercise (Home‐EX) Setting: outpatient clinic, Iran Categorization: EX = muscle power, Home‐EX = muscle power
Participants	n = 40; EX = 20, Home‐EX = 20 Inclusion criteria: RRMS, EDSS < 4.5, aged 18‐50 years Exclusion criteria: acute exacerbation; Ashworth spasticity score > 2.0; thyroid disorder; history of serious psychiatric disorders, alcohol abuse, or other chronic diseases Type MS: RRMS Disease duration (yr) ± SD: EX = 6.43 ± 2.78, Home‐EX = 7.40 ± 3.43 Mean age (yr) ± SD: EX = 32.62 ± 3.15, Home‐EX = 33.00 ± 4.06 % Female (n/n group): EX = 9/16, Home‐EX = 11/20 Mean EDSS ± SD: EX = 3.6 ± 1.3, Home‐EX = 3.4 ± 2.1
Interventions	EX: 12 weeks, 5x per week, 60 minutes hospital‐based exercise supervised by a physiatrist. 3 out 5 session/week were callisthenic exercise (60 minutes), 2 out of 5 sessions were relaxation exercises (20 minutes). Callisthenic exercises were focused on the large muscles and were applied rhythmically and in combination with breathing exercises. The callisthenic training sessions consisted of 15 minutes' warming up, 20 minutes' intensive training, 10 minutes' cooling down, and 15 minutes' relaxation Home‐EX: same as EX yet participants were asked to perform the exercise at home and their exercise schedule was followed by telephone every day
Outcomes	FSS, 10MWT, BBS, MusiQoL, HADS
Notes	Drop‐outs EX: 2 failure to adapt to exercise, 2 transportation problems Home‐EX: no drop‐outs Measurements Baseline, 12 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated numbers
Allocation concealment (selection bias)	Low risk	Concealed allocation procedure
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	High risk	Drop‐out rate of 10% in the exercise condition, 0 drop‐outs reported in the control condition
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Low risk	No other sources of bias identified

Methods	Design: random assignment to Ergometer Land Group (ELG) or Ergometer Water Group (EWG) Setting: Department of Sports Therapy, Rehabilitationsklinik Valens, Valens, Switzerland Categorization: ELG = endurance, EWG = endurance
Participants	n = 60; ELG = 30, EWG = 30 Inclusion criteria: EDSS 1.0‐6.5 Exclusion criteria: incontinent, persistent infections, cardiovascular and pulmonary diseases, or immunosuppressive therapy Type MS: ? Disease duration (yr) ± SD: ? Mean age (yr) ± SD: ELG = 52 ± 14.2, EWG = 50 ± 14.2 % Female (n/n group): ELG = 18/28, EWG = 17/25 Mean EDSS ± SD: ELG = 4.7 ± 1.5, EWG = 4.6 ± 1.4
Interventions	ELG: 3 weeks, 5x per week; 30‐minute physiotherapist‐led overland ergometry training at 60% of peak oxygen uptake/70% maximal heart rate EWG: 3 weeks, 5x per week; 30‐minute physiotherapist‐led aquatic training at 70% maximal heart rate corrected for water temperature (‐7 beats per minute). Water temperature at 28°C
Outcomes	FSMC, neurotrophin, cytokines
Notes	Drop‐outs ELG: 2 immunosuppressive medication during training EWG: 3 immunosuppressive medication during training, 3 unable to comply with daily time schedule Measurements Baseline, 3 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomization procedure not described
Allocation concealment (selection bias)	Unclear risk	Allocation procedure unknown
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	Unclear risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Overall drop‐out rate < 15%, without indications of differences in reasons
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Low risk	No other sources of bias identified

Methods	Design: random assignment to Nintendo^® Wii^® balance training (EX) or control (NEX) Setting: Italian Multiple Sclerosis Foundation, Genoa, Italy Categorization: EX = other
Participants	n = 36; EX = 18, NEX = 18 Inclusion criteria: no relapse < 6 months, EDSS ≤ 6.0, Ambulation Index ≤ 4 Exclusion criteria: psychiatric disorder, blurred vision, severe cognitive impairment Type MS: ? Disease duration (yr) ± SD: EX = 11.2 ± 6.4, NEX = 12.3 ± 7.2 Mean age (yr) ± SD: EX = 40.7 ± 11.5, NEX = 43.2 ± 10.6 % Female (n/n group): EX = 10/18, NEX = 12/18 Mean EDSS ± SD: EX = 3.9 ± 1.6, NEX = 4.3 ± 1.6
Interventions	EX: 4 weeks, 3x per week; 1 hour Nintendo^® Wii^® balance board training. Special focus on soccer heading, skiing, table tilt, snowboarding, tight rope walking, and zazen (relaxation) NEX: 4 weeks, 3x per week; 1 hour static and dynamic lower body exercise using single and double leg stances with or without an equilibrium board
Outcomes	MFIS, BBS, open and closed eye stabilometry
Notes	Drop‐outs No drop‐outs reported Measurements Baseline, 4 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Biased‐coin randomization procedure
Allocation concealment (selection bias)	Unclear risk	Allocation concealment procedure unknown
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	No drop‐outs reported
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Low risk	No other sources of bias identified

Methods	Design: random assignment to arm‐cycling (ARM), rowing (ROW), bicycling (BIC), and a control (NEX) Setting: MS outpatient clinic, Germany Categorization: ARM = endurance, ROW = endurance, BIC = endurance
Participants	n = 47; ARM = 12, ROW = 12, BIC = 12, NEX = 11 Inclusion criteria: definite secondary‐progressive or primary‐progressive MS, EDSS 4.0‐6.0 Exclusion criteria: contraindications for exercise, start immunomodulatory treatment < 6 months, steroid treatment < 4 weeks, MS relapse < 12 months, abnormal liver or kidney function, immunodeficiency, other serious medical illness or psychiatric, developmental or neurological disorder Type MS: SPMS and PPMS Disease duration (yr) ± SD: ARM = 17.1 ± 7.2, ROW = 14.1 ± 6.1, BIC = 13.3 ± 5.4, NEX = 18.9 ± 9.8 Mean age (yr) ± SD: ARM = 49.1 ± 8.5, ROW = 50.9 ± 9.2, BIC = 48.8 ± 6.8, NEX = 50.4 ± 7.6 % Female (n/n group): ARM = 5/10, ROW = 7/11 , BIC = 6/11 , NEX = 6/10 Mean EDSS ± SD: ARM = 5.2 ± 0.9, ROW = 4.7 ± 0.8, BIC = 5.0 ± 0.8, NEX = 4.9 ± 0.9
Interventions	ARM, ROW, and BIC: 8‐10 weeks, 2 or 3 times per week, 15‐45 minutes aerobic training led by licensed physiotherapists at the individual determined aerobic threshold (AT), 120%AT and 130%AT. Aim was to perform 20 training sessions in which the time spent at the higher levels was increased. Only the mode of exercise (i.e. ARM, ROW, or BIC) different between groups NEX: wait list control
Outcomes	MFIS, VO_2peak, 6MWT, SDMT, VLMT, TAP, LPS, RWT, IDS ‐ SR30
Notes	Drop‐outs ARM: 1, logistics + 1 did not receive allocated intervention ROW: 1, unrelated injury BIC: 1, fatigued NEX: 1, lost to follow‐up Measurements Baseline, 8‐10 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Automated biased coin procedure
Allocation concealment (selection bias)	Low risk	Randomization after determining eligibility
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Overall drop‐out rate was 10%, equally balanced across groups
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Low risk	No other sources of bias identified

Methods	Design: Random assignment to Tai Chi (EX) or treatment as usual (NEX) Setting: Department of neurology, Bayreuth, Germany Categorization: EX = other
Participants	n = 38; EX = 21, NEX = 17 Inclusion criteria: diagnosis of any type of MS, able to walk without a walking aid, EDSS < 5, relapse‐free < 4 weeks Exclusion criteria: severe cognitive impairment Type MS: RRMS, SPMS, and clinically isolated syndrome (n = 1) Disease duration (yr) ± SD: EX = 6.0 ± 4.7, NEX = 7.8 ± 6.8 Mean age (yr) ± SD: EX = 42.6 ± 9.4, NEX = 43.6 ± 8.0 %Female (n/n group): EX = 10/15, NEX = 12/17 Median EDSS (range): EX = 2 (1‐4), NEX = 4 (1‐4.5)
Interventions	EX: 6 months, 1x per week, 90 minutes of centre‐based, structured, compact Tai Chi. The Tai Chi programme was based on the Yang‐style 10‐form. Exercises were structured so that during each session of the course, the same essential elements were repeated NEX: treatment as usual
Outcomes	FSMC, balance, co‐ordination, CES‐D, QLS
Notes	Drop‐outs EX: 5 time constraints, 1 health problems Measurements Baseline, 6 months
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Assignment to EX or NEX was based on participants' weekday preference
Allocation concealment (selection bias)	High risk	Unconcealed allocation
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	High risk	Drop‐out rate in EX group was 29%, no drop‐outs reported in the control condition. In the EX group, adherence varied between 15 and 44 out of 50 classes offered
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Unclear risk	There was an apparent (yet non‐significant) difference in the disease severity and baseline fatigue scores between the EX and NEX group

Methods	Design: random assignment to progressive resistance training group (PRT), home‐based resistance training (EX), or control (NEX) Setting: rehabilitation clinic, Ankara, Turkey Categorization: PRT = muscle power, EX = muscle power
Participants	n = 45; PRT = 15, EX = 15, NEX = 15 Inclusion criteria: definite RRMS or SPMS, EDSS ≤ 6.0 stand upright > 4 seconds, no steroid or immunosuppressive therapy (or both) < 1 month Exclusion criteria: severe MS, acute exacerbation, active physical therapy or regular exercise training programme < 1 month, unable to cycle a static bike, visual involvement or diplopia, high level of spasticity, persistent severe fatigue or depression Type MS: RRMS or SPMS Disease duration (yr) ± SD: PRT = 9.2 ± 5.0, EX = 6.2 ± 2.2, NEX = 6.6 ± 2.4 Mean age (yr) ± SD: PRT = 36.4 ± 10.5, EX = 43.0 ± 10.2, NEX = 35.5 ± 10.9 %Female (n/n group): PRT = 9/14, EX = 8/10, NEX = 6/9 Mean EDSS ± SD: ?
Interventions	PRT: 2x per week for 8 weeks; 30 minutes' cycling progressive resistance training, 5 minutes' walking/stretching and 20‐25 minutes' balance exercises supervised by a physiatrist EX: 2x per week for 8 weeks; home‐based exercise programme consisting of 5 minutes' walking/stretching and 20‐25 minutes' balance exercises NEX: wait list
Outcomes	FSS, number of relapses, duration of exercise, W_max, TUG, DGI, FR, FES, 10MWT, BDI, SF‐36
Notes	Drop‐outs PRT: 1 acute exacerbation EX : 2 work‐related, 2 acute exacerbation, 1 unknown NEX: 3 acute exacerbation, 3 unknown Measurements Baseline, 8 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated random numbers were used
Allocation concealment (selection bias)	Unclear risk	Unclear who performed the randomization procedure
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	High risk	Overall drop‐out rate was 27%, with a relative higher drop‐out rate in the control (40%) versus the exercise group (combined 20%). However, no indications for differences in reasons between groups
Selective reporting (reporting bias)	Low risk	All introduced outcomes were reported
Other bias	High risk	The PRT intervention included group‐based components

Methods	Design: random assignment to pragmatic exercise intervention + usual care (EX) or usual care (NEX) Setting: MS clinic, UK Categorization: EX = mixed
Participants	n = 120; EX = 60, NEX = 60 Inclusion criteria: clinical diagnosis MS, EDSS 1‐6.5, able to walk 10 metres, aged 18‐65 years, clinically stable, and stable medical treatment < 3 months Exclusion criteria: structured exercise ≥ 3x per week, ≥ 30 minutes; living ≥ 20 minutes travelling from study centre, co‐morbid conditions precluding exercise participation Type MS: RRMS, SPMS, and PPMS Disease duration (yr) ± SD: EX = 8.4 ± 7.4, NEX = 9.2 ± 7.9 Mean age (yr) ± SD: EX = 45.7 ± 9.1, NEX = 46.8 ± 8.4 % Female (n/n group): EX = 43/60, NEX = 43/60 Mean EDSS ± SD: EX = 3.8 ± 1.5, NEX = 3.8 ± 1.5
Interventions	EX: 12 weeks, 3x per week; partly home based with more supervised sessions in the first 6 weeks of the intervention and vice‐versa in the second 6 weeks. Short bouts (5 x 3 minutes) of low to moderate intensity aerobic exercise at 50‐69% of predicted maximal heart rate or 12‐14 on the BORG scale. Participants were encouraged during the intervention period to increase the duration of the bouts or reduce the resting time in between bouts. When appropriate, participants could also perform 6 different resistance exercises using body resistance, light weights, or Therabands. Generally, 1‐3 sets of 20 repetitions based on the participant's level of disability, strength, and stage of the programme. In case of balance or control problems, also balance board and exercise ball work was included. The supervised exercise sessions included also cognitive behavioural techniques to promote long‐term physical activity behaviour NEX: usual care
Outcomes	MFIS, GLTEQ, Accelerometer, MSQoL‐54, MSFC, 6MWT, EDSS
Notes	Drop‐outs EX: 2 relapse, 2 ill health, 1 poor adherence, 6 unable to contact NEX: 1 relapse, 2 ill health, 1 work commitments, 1 no reason give, 5 unable to contact Measurements Baseline, 12 weeks, 6 months' follow‐up
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Distant randomization service; minimization according to gender and EDSS. Allocation not disclosed until after baseline measurement
Allocation concealment (selection bias)	Low risk	Independent and off‐site randomization service
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Drop‐out rate in the combined exercise group was 17%, in the control group 18%. No indication for differences in reasons
Selective reporting (reporting bias)	Low risk	Study protocol published
Other bias	Low risk	No other sources of bias identified

Methods	Design: random assignment to Ai‐Chi aquatic exercise programme (EX) or control group (NEX) Setting: Multiple Sclerosis Association of Almeria, Almeria, Spain Categorization: EX = other
Participants	n = 73; EX = 36, NEX = 37 Inclusion criteria: diagnosis MS, aged 18‐75 years, VAS pain score > 4 for at least two months, EDSS ≤ 7.5 Exclusion criteria: treatment with other complementary and alternative medicine ≤ 3 months, relapse requiring hospitalization or steroid treatment ≤ 2 months Type MS: ? Disease duration (yr) ± SD: EX = 10.7 ± 9.1, NEX = 11.9 ± 8.7 Age (yr): EX = 46 ± 9.97, NEX = 50 ± 12.31 % Female (n/n group): EX = 26/36, NEX = 24/37 Mean EDSS ± SD: EX = 6.3 ± 0.8, NEX = 5.9 ± 0.9
Interventions	EX: Ai‐Chi exercise programme (20 weeks, 2x per week, 60 minutes per session) in shoulder‐depth 36°C water; consisting of 16 different slow and broad movements of the arms, legs, and torso to work on balance, strength, relaxation, flexibility, and breathing. Each physiotherapist‐led session began and ended with 10 minutes of relaxation. During the course of each session relaxing tai‐chi music was played NEX: 20 weeks, 2x per week, 60 minutes per session led by a physiotherapist identical to the 10‐minute relaxation (breathing, contraction ‐ relaxation) periods of the intervention group yet in a therapy room (26°C), on an exercise mat without ambient music
Outcomes	FSS, MFIS, number of relapses, pain (VAS), MPQ‐PRI, MPQ‐PPI, RMDQ, spasm (VAS), MSIS‐29, BDI, BI
Notes	Drop‐outs NEX: 2 relapsed Measurements Baseline, 20 weeks, 24 weeks, 30 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated random list
Allocation concealment (selection bias)	Low risk	Participants were randomly assigned by a blinded researcher
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Drop‐out rate was 3%
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Unclear risk	Large proportion of participants with unknown disease course despite 'a definite diagnosis of MS' as inclusion criterion

Methods	Design: random assignment to continuous (CON), intermittent (INT), or mixed endurance training (CB) Setting: Oxford Brookes University and at 4 community leisure centres, London, UK Categorization: CON = endurance, INT = endurance, CB = endurance
Participants	n = 61; CON = 20, INT = 18, CB = 20 Inclusion criteria: none Exclusion criteria: medical condition precluding safe exercise, unable to walk 2 minutes, unable to sit 60 seconds on cycling ergometer and pedal 60 seconds with no resistance Type MS: RRMS, SPMS, PPMS and unknown Disease duration (yr) ± SD: CON = 15 ± 8, INT = 11 ± 7, CB = 12 ± 11 Mean age (yr) ± SD: CON = 52 ± 8, INT = 50 ± 10, CB = 55 ± 10 % Female (n/n group): CON = 16/20, INT = 14/18, CB = 9/17 Mean EDSS ± SD: ?
Interventions	CON: 12 weeks, 2x per week; 20 minutes supervised continuous endurance training at 45% of maximal power INT: 12 weeks, 2x per week; 20 minutes of supervised intermittent endurance training consisting of 30 seconds at 90% maximal power, 30 seconds rest CB: 12 weeks, 2x per week; 20 minutes of supervised intermittent (10 minutes) endurance training followed by continuous endurance training (10 minutes)
Outcomes	FSS, 2MWT, TUG, leg extensor power, BI, SF‐36
Notes	Drop‐outs CON: no drop‐outs INT: 1 loss of consciousness during intervention, 1 exacerbation of symptoms, 1 knee pain during cycling, 1 leg pain during cycling CB: 1 exacerbation of knee injury, 1 leg pain during cycling Measurements Baseline, 6 weeks, 12 weeks (post‐intervention), 24 weeks (follow‐up)
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomization list
Allocation concealment (selection bias)	Low risk	Central randomization by statistician
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	High risk	Overall drop‐out rate 10%, unequally balanced in favour of the continuous exercise protocol versus the intermittent and combined protocol
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Low risk	No other sources of bias identified

Methods	Design: random assignment to endurance training (EX) and control treatment (NEX) Setting: inpatient rehabilitation unit, Schmieder Konstanz clinic, Konstanz, Germany Categorization: EX = endurance
Participants	n = 30; EX = 16 (see drop‐outs), NEX = 15 Inclusion criteria: mild to moderate MS (EDSS < 4.5), complained of fatigue, walking distance < 2500 metres Exclusion criteria: serious leg weakness (degree of paresis < 4), ataxia at rest, spasticity at rest, relapse < 3 months, corticosteroid treatment < 3 months, prominent cognitive deficits, major depression, insufficient motivation for additional training Type MS: RRMS, SPMS, PPMS Disease duration (yr) ± SD: EX = 8.0 ± 5.9, NEX = 6.1 ± 4.3 Mean age (yr) ± SD: EX = 45.8 ± 7.9, NEX = 39.7 ± 9.1 % Female (n/n group): EX = 10/15, NEX = 11/15 Mean EDSS ± SD: EX = 2.6 ± 1.2, NEX = 2.8 ± 0.7
Interventions	EX: 3 weeks, 3x per week, 45 minutes per session. Warming up, mild strength training, repetitive endurance exercise followed by relaxation and feedback. All sessions were supervised and playful elements were introduced to camouflage training difficulties NEX: 3 weeks, 3x per week, 45 minutes per session. Warming up, sensory training, stretching, balance, co‐ordination training, and periods of relaxation
Outcomes	MFIS, FSMC, maximal walking distance, relative walking ability, BDI, HAQUAMS
Notes	Drop‐outs EX: 1 training too demanding; replaced by a newly recruited 16th participant Measurements Baseline, 3 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Random order list
Allocation concealment (selection bias)	Unclear risk	Allocation concealment procedure unknown
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	High risk	6% drop‐out rate in the exercise condition. However, only 19 of 30 forms completed at discharge (63%)
Selective reporting (reporting bias)	High risk	No point and variance measures reported
Other bias	Low risk	No other sources of bias identified

Methods	Design: random assignment to group progressive resistance training programme (EX) or usual care in combination with a social programme (NEX); stratified by ambulation index score Setting: MS outpatient clinic, Victoria, Australia Categorization: EX = muscle power
Participants	n = 76; EX = 39, NEX = 37 Inclusion criteria: aged ≥ 18 years, confirmed RRMS, Ambulation index score of 2, 3, or 4, medical clearance to participate Exclusion criteria: acute exacerbation < 2 months, benign or progressive/relapsing types of MS, serious unstable medical condition, any concurrent condition, progressive resistance training < 6 months Type MS: RRMS Disease duration (yr) ± SD: ? Mean age (yr) ± SD: EX = 47.7 ± 10.8, NEX = 50.4 ± 9.6 % Female (n/n group): EX = 26/36, NEX = 26/35 Mean EDSS ± SD: ?
Interventions	EX: group PRT programme (10 weeks, 2x per week). Exercise targeted the key lower limb muscles for supporting body weight and for generating and absorbing power during walking (leg press, knee extension, calf raise, leg curl, and reverse leg press). Training intensity was 10‐12 repetitions at 10‐12 repetition maximum. Load was increased if 2 sets of 12 repetitions were completed. 2 minutes rest between each set NEX: usual care + social programme 1 hour per week
Outcomes	MFIS, MSSS‐88, adverse events, number of relapses, 2MWT, 1RM seated leg press, 1RM reversed leg press, WHOQoL‐BREF
Notes	Drop‐outs EX: 3 withdrew after allocation (2 experienced relapse but did attend assessment) NEX: 2 withdrew after allocation, 3 experienced relapse, 1 did not attend at follow‐up Measurements Baseline, 10 weeks, 22 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Separate randomization procedure for each stratum using permuted blocks
Allocation concealment (selection bias)	Low risk	Sealed, opaque envelopes made by research co‐ordinator
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Overall drop‐out rate 12%. Not equally balanced across the exercise (8%) and control group (16%); however, no indication for differences in reasons taken into account the 2 participants who did have a relapse but attended the assessment
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Low risk	No other sources of bias identified

Methods	Design: internet‐based home training (IBT) versus hippotherapy (HT) Setting: regional MS groups, Germany Categorization: IBT = other, HT = other
Participants	n = 18; IBT = 9, HT = 9 Inclusion criteria: definite MS, EDSS 2‐6, aged 18‐60 years, clinically stable < 4 weeks Exclusion criteria: internal or orthopaedic diseases unrelated to MS, allergy or aversion to horses, previous hippotherapy post‐diagnosis Type MS: RRMS and SPMS Disease duration (yr) ± SD: IBT = 16.1 ± 11.3, HT = 22.3 ± 8.3 Mean age (yr) ± SD: IBT = 44.3 ± 8.1, HT = 46.9 ± 7.6 % Female (n/n group): IBT = 7/9, HT = 8/9 Mean EDSS ± SD: IBT = 3.8 ± 1.5, HT = 3.8 ± 1.1
Interventions	IBT: 12 weeks, 2x per week, 45 minutes' balance and strength e‐training; exercise performed on unstable surface in eyes open or closed condition. 5‐8 different exercise, 2‐3 sets, 8‐15 repetitions. BORG 11‐14. Perceived exertion monitored by physical therapist HT: 12 weeks, 12x per week, 30 minutes' balance exercises by means of horse riding exercises and riding patterns/movements
Outcomes	MFIS, FSS, BBS, DGI, isometric muscle strength, TUG, 2MWT, HAQUAMS
Notes	Drop‐outs IBT: 1 unable to work with computer HT: 1 MS relapse Measurements Baseline, 12 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Drawing lots
Allocation concealment (selection bias)	Low risk	Opaque envelopes were used
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Overall drop‐out rate 11% equally balanced across groups
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Low risk	No other sources of bias identified

Methods	Design: random assignment to home‐based inspiratory muscle training (EX) or control (NEX) Setting: Physical Therapy Department, University of Michigan‐Flint, Flint, Michigan, USA Categorization: EX = other
Participants	n = 46; EX = 23, NEX = 23 Inclusion criteria: ambulatory, aged ≥ 18 years Exclusion criteria: acute respiratory infection, oral temperature > 100°F (37.8°C), smoking, cardiac of musculoskeletal condition unrelated to MS Type MS: RRMS, SPMS, PPMS or primary relapsing MS Disease duration (yr): ? Mean age (yr) ± SD: EX = 50 ± 9.1, NEX = 46.2 ± 9.4 % Female (n/n group): EX = 21/23, NEX = 17/23 Mean EDSS ± SD: EX = 3.96 ± 1.80, NEX = 3.36 ± 1.47
Interventions	EX: 10 weeks, daily; 3x 15 repetitions of home‐based Inspiratory muscle training using a Threshold Inspiratory Muscle Trainer. Resistance was adjusted to match a perceived exertion of 15, 16 on the BORG scale NEX: no intervention; informative telephone call 4, 8, and 10 weeks
Outcomes	FSS, 6MWT, PCI, RPE
Notes	Drop‐outs 2 due to illness unrelated to a neurological exacerbation, 3 no longer wished to participate Measurements Baseline, 10 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Randomized by date of enrolment
Allocation concealment (selection bias)	High risk	Researcher could foresee date of enrolment
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Unclear risk	Overall drop‐out rate 11%. No information provided on the distribution between groups
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Low risk	No other sources of bias identified

PERMALINK

Exercise therapy for fatigue in multiple sclerosis

Martin Heine

Ingrid van de Port

Marc B Rietberg

Erwin EH van Wegen

Gert Kwakkel

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

Plain language summary

Summary of findings

Background

Description of the condition

Description of the intervention

How the intervention might work

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Primary outcomes

Secondary outcomes

Search methods for identification of studies

Electronic searches

Searching other resources

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Unit of analysis issues

Cross‐over trials

Trials with multiple treatment groups

Dealing with missing data

Assessment of heterogeneity

Assessment of reporting biases

Data synthesis

Meta‐analysis

GRADE

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

Results

Description of studies

Results of the search

1.

Included studies

Excluded studies

Risk of bias in included studies

1. Risk of bias.

2.

3.

Allocation

Blinding

Incomplete outcome data

Selective reporting

Other potential sources of bias

Effects of interventions

Summary of findings for the main comparison. Overall analysis for fatigue in multiple sclerosis.

Summary of findings 2. Per fatigue measure for fatigue in multiple sclerosis.

Summary of findings 3. Per exercise group for fatigue in multiple sclerosis.

Effect of exercise therapy on fatigue

2. Group‐by‐time effects (interaction) of included trials: exercise versus control.

3. Group‐by‐time effects (interaction) of included trials: exercise versus exercise.

1.1. Analysis.

2.1. Analysis.

Safety

Sub‐group analysis per fatigue outcome

3.1. Analysis.

Modified Fatigue Impact Scale

Fatigue Severity Scale

Other

Methods	Design: random assignment to robot‐assisted gait training (RAGT) or sensory integration balance training (SIBT) Setting: outpatient clinic, Neurological Rehabilitation Unit, Italy Categorization: RAGT = other, SIBT = other
Participants	n = 26; RAGT = 14, SIBT = 12 Inclusion criteria: aged 30‐60 years, EDSS 1.5‐6.5, MMSE > 23, ability to maintain standing position without aids for at least 1 minute and ability to walk independently for at least 15 metres, absence of concomitant neurological or orthopaedic conditions that may interfere with ambulation Exclusion criteria: any rehabilitation treatment < 1 month, MS relapse < 3 months, pharmacological treatment not well defined or changed during study, presence of paroxysmal vertigo, lower limb botulinum toxin injections < 12 weeks Type MS: RRMS or SPMS Disease duration (yr): RAGT = 13.5 ± 7.6, SIBT = 14.9 ± 8.68 Mean age (yr) ± SD: RAGT = 50.83 ± 8.42, SIBT = 50.1 ± 6.29 % Female (n/n group): RAGT = 7/12, SIBT = 9/10 Mean EDSS ± SD: RAGT = 3.96 ± 0.75, SIBT = 4.35 ± 0.67
Interventions	RAGT: 6 weeks, 2x per week, 50‐minute individual sessions of RAGT. The RAGT group was treated by means of the electromechanical Gati Trainer GT1 (Reha‐stim, Berlin, Germany). Participants received 40 minutes of RAGT followed by 10 passive lower limb joint mobilizations and stretching exercises. Each training consisted of 2 x 15‐minute sessions separated by 5 minutes of rest if required. The first session was performed at 20% supported body‐weight and 1.3 km/hour speed; the second sessions at 10% supported body weight and 1.6 km/hour speed SIBT: 6 weeks, 2x per week, 50‐minute individual sensory integration balance training. Each session consisted of exercises fitting to 3 different levels of difficulty and repeated under 3 different sensory conditions (free vision, wearing a mask, and wearing a helmet). Level I included tasks that induced external destabilization of the centre‐of‐body mass (CoP). Level II included exercises of self destabilization of the CoP. Level III consisted of exercises of external destabilization and exercise of self destabilization of CoP while standing on different types of compliant surfaces. During each treatment session, a total of 10 exercises (3 from level I, 3 from level II and 4 from level III) were repeated 2‐5 times with a 5‐minute period
Outcomes	FSS, spatiotemporal gait parameters, BBS, SOT, stabilometric assessment, MSQoL‐54, ABC
Notes	Drop‐outs RAGT: 2 difficulties in transportation SIBT: 2 medical complications Measurements Baseline, 6 weeks, 1‐month follow‐up
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	'Simple software‐generated randomisation scheme'
Allocation concealment (selection bias)	Low risk	Randomization by an independent blinded collaborator not involved in the treatment or care of participants
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Overall drop‐out rate 15.4% equally balanced across groups
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Low risk	No other sources of bias identified

Methods	Design: random assignment to physiotherapists‐led exercise (PT), yoga (YG), fitness‐instructor led exercise (FI), or control (NEX) Setting: Clinical Therapies Department, University of Limerick, Limerick, Ireland Categorization: PT = mixed, YG = other, FI = mixed
Participants	n = 314; PT = 80, YG = 77, FI = 86, NEX = 71 Inclusion criteria: 0‐2 on the GNDR mobility sub‐scale, aged > 18 years, diagnosis of MS Exclusion criteria: relapse/steroid treatment < 3 months, pregnant, serious co‐morbidity Type MS: RRMS, SPMS, PPMS, benign, or unknown Disease duration (yr): PT = 9.8 ± 7.0, YG = 11.6 ± 8.0, PI = 10.5 ± 6.9, NEX = 10.6 ± 8.2 Mean age (yr) ± SD: PT = 51.7 ± 10.0, YG = 49.6 ± 10.0, PI = 50.3 ± 10.0, NEX = 48.8 ± 11.0 % Female (n/n group): PT = 50/63, YG = 44/63, PI = 45/77, NEX = 43/49 Mean EDSS ± SD: ?
Interventions	PT: 10 weeks, 1x per week; 1‐hour circuit style ‐ body weight/free weight resisted group (6‐8 participants) training. Exercises consisted of, for example: sit to stand, bridging, shoulder flexion, walking, cycling. Classes were supervised by a trained physiotherapist. Difficulty of the exercises was aimed to fail at the 12th repetition. In addition, participants were advised to do aerobic training of choice, 2x per week at 65% HR_max YG: 10 weeks, 1x per week; 1‐hour sessions led by YG instructors. The YG sessions were not pre‐defined; but consisted of breathing exercise, relaxation, range of motion exercise, and dynamic weight‐bearing poses FI: 10 weeks, 1x per week; 1‐hour community fitness led by local fitness instructors. The content of the intervention was not pre‐defined NEX: no change in exercise habits for 10 weeks
Outcomes	MFIS, MSIS‐29v2, 6MWT
Notes	Drop‐outs PT: 7 health condition, of which 2 relapse, 1 personal factors, 2 environmental factor, 7 unknown YG: 5 health condition, of which 2 relapse, 3 personal factors, 1 environmental factor, 5 unknown FI: 7 health condition, of which 3 relapse, 5 personal factors, 7 unknown NEX: 11 health condition, of which 6 relapse, 6 personal factors, 1 environmental factor, 4 unknown Measurements Baseline, 12 weeks (post intervention), 24 weeks follow‐up for intervention groups only A secondary analysis to predict the response of each therapy on gait function was also published (Kehoe 2014). The results of a second study strand, in more severely disabled people with MS, has been included as a separate study (Hogan 2014)
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Block randomization per geographical region
Allocation concealment (selection bias)	Low risk	Randomization performed by central co‐ordinator
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Overall drop‐out rate 23% equally balanced across the different exercise conditions.
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Low risk	No other sources of bias identified

Methods	Design: random assignment to home walking programme (EX) and no training (NEX) Setting: Department of Physical Therapy, The George Washington University, Washington, USA Categorization: EX = task‐oriented
Participants	n = 15; EX = 9, NEX = 6 Inclusion criteria: aged 18‐65 years, diagnosis MS > 1 year, no exacerbation < 6 months, no participation in regular aerobic exercise < 6 months, EDSS ≤ 6.0 Exclusion criteria: cardiovascular, pulmonary, or orthopaedic conditions Type MS: ? Disease duration (yr): EX: 3‐30 (information provided for 5 participants), NEX = ? Age range (yr): EX = 40‐64, NEX = 22‐50 % Female (n/n group): EX = 6/8, NEX = 3/4 Mean EDSS ± SD: ?
Interventions	EX: individualized home walking programme (12 weeks, 3x per week). Training heart rate (THR) range was determined based on 6MWT using the Karvonen formula. Each session consisted of 5 minutes below THR, 15 minutes in THR, and cooling down below THR. Time within THR increased gradually during training programme NEX: no training, walking programme offered following completion of trial
Outcomes	FSS, 6MWT, PCI, RPE
Notes	Drop‐outs EX: 1 poor compliance and failure to show at post‐test assessment NEX: 2 poor compliance and failure to show at post‐test assessment Measurements Baseline, 12 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomization by coin toss
Allocation concealment (selection bias)	Unclear risk	Unclear who performed coin toss
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Overall drop‐out rate was 20%, equally balanced across groups
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	High risk	Small and unevenly distributed sample

Methods	Design: random assignment to standard exercise programme plus supplementary resistance training (EX) or standard exercise programme (STD) Setting: Department of Physical Therapy, University of Utah, Salt Lake City, USA Categorization: EX = mixed, STD = endurance
Participants	n = 22; EX = 11, STD = 11 Inclusion criteria: definite MS, no exacerbation < 3 months, aged 18‐65 years, ambulatory, impaired gait pattern, no lower extremity joint problems Exclusion criteria: participation in regular strength training Type MS: ? Disease duration (yr) ± SD: EX = 12.5 ± 11.2, STD = 11.8 ± 7.3 Mean age (yr) ± SD: EX = 48.0 ± 11.9, STD = 49.7 ± 10.98 % Female (n/n group): EX = 5/9, STD = 6/10 Mean EDSS ± SD: EX = 5.3 ± 1.0, STD = 5.2 ± 1.0
Interventions	EX: supplementary lower extremity eccentric ergometric resistance training (12 weeks, 3x per week). During the course of the training programme resistance was increased based on perceived exertion (BORG 7/20 ‐ BORG 13/20) STD: aerobic training, lower extremity stretching, upper extremity strength training, and balance exercises (12 weeks, 3x per week)
Outcomes	FSS, voluntary maximal isometric strength (flexion/extension of knee, hip, dorsi, and sum), TUG, 6MWT, stair test, 10MWT, BBS
Notes	Drop‐outs EX: 1 difficulty committing to 3x per week schedule STD: 1 during pre‐testing, 1 discontinued intervention Measurements Baseline, 12 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information on randomization procedure
Allocation concealment (selection bias)	Unclear risk	Allocation concealment procedure unknown
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Overall drop‐out rate was 14% equally balanced across the exercise groups
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Low risk	No other risk of bias identified

Methods	Design: random assignment to vestibular rehabilitation (VES), exercise control (EX), or wait‐list control group (NEX); stratified by yes/no involvement of brain stem/cerebellar Setting: Department of Physical Medicine and Rehabilitation, and Department of Neurology, University of Colorado, Aurora, USA. Medical Campus Categorization: VES = other, EX = endurance
Participants	n = 38; VES = 12, EX = 13, NEX = 13 Inclusion criteria: definite MS, aged 18‐65 years, able to walk 100 metres with at most a single‐sided device, score ≥ 45/85 on the MFIS, score < 72 on SOT; limited standing balance Exclusion criteria: unable to walk, use of pharmacological agent that control or cause fatigue, change in MS medication < 3 months, MS‐related relapse < 6 months, other conditions that cause fatigue, impaired or limited upright postural control, participation in a vestibular or endurance exercise programme < 2 months Type MS: RRMS or SPMS Disease duration (yr) ± SD: VES = 6.5 ± 5.6, EX = 5.1 ± 3.2, NEX = 9.1 ± 7.3 Mean age (yr) ± SD: VES = 46.8 ± 10.5, EX = 42.6 ± 10.4, NEX = 50.2 ± 9.2 % Female (n/n group): VES = 9/12, EX = 11/13, NEX = 11/13 Mean EDSS ± SD: ?
Interventions	VES: standardized vestibular rehabilitation programme (6 weeks, 2x per week, 55 minutes). In addition, individual exercises were selected and constituted an individualized home training programme (postural control, eye movement). Also 5 minutes of fatigue management education was provided EX: endurance and stretching exercises (6 weeks, 2x per week); 5 minutes warming up, 2 x 15 minutes at 65‐75% HR_peak, 2‐5 minutes cooling down. In addition, an individualized home training programme (endurance and stretching) at a similar intensity as the supervised programme and 5 minutes of fatigue management education were provided NEX: wait list control
Outcomes	MFIS, SOT, 6MWT, DHI, BDI
Notes	Drop‐outs NEX: 1 refused continuation after wait‐list control allocation Measurements Baseline (weeks 1, 2, 4), intervention phase (weeks 6, 8, 10), follow‐up phase (weeks 12, 14); not all measures assessed at each occasion
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomization procedure not described
Allocation concealment (selection bias)	Low risk	Randomization performed by a clinician not involved in the trial
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Drop‐out rate was 3%
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Low risk	No other sources of bias identified

Methods	Design: random assignment to group physiotherapy (GP), individual physiotherapy (IP), yoga (Y), or control (C) Setting: Clinical Therapies Department, University of Limerick, Limerick, Ireland Categorization: GP = muscle power, IP = muscle power, Y = other
Participants	n (drop‐outs) = 146; GP = 66, IP = 45, Y = 16, C = 19 Inclusion criteria: confirmed diagnosis of MS, GNDS score of 3 or 4 Exclusion criteria: relapse or steroid treatment < 12 weeks, pregnant, aged < 18 years Type MS: RRMS, SPMS, PPMS, unknown Disease duration (yr) ± SD: GP = 18 ± 9, IP = 13 ± 8, Y = 15 ± 8, C = 10 ± 3 Mean age (yr) ± SD: GP = 57 ± 10, IP = 52 ± 11, Y = 58 ± 8, C = 49 ± 6 % Female (n/n group): GP = 30/48, IP = 20/35, Y = 8/13, C = 13/15 Mean EDSS ± SD: ?
Interventions	GP: 10 weeks, 1x per week, 1‐hour group physiotherapy. GP was a self paced circuit style class of exercises (sit to stand, squat, heel raises, step‐ups, side stepping, and tandem) that targeted strength and balance with the aim of increasing balance and mobility. The aim was to perform 1 set of 12 repetitions and according to the progression was increased to 3 sets of 12 repetitions IP: 10 weeks, 1x per week, 1‐hour individual physiotherapy with the same content as the GP Y: 10 weeks, 1x per week, 1‐hour yoga classes by yoga instructors of the Yoga Federation of Ireland. The content of each yoga class was kept in a diary C: usual care
Outcomes	MFIS, BBS, 6MWT, MSIS29v2
Notes	Drop‐outs GP: 1 relapse, 5 unwell at day of measurement, 3 unwell and discontinued intervention, 1 steroids for lower back pain, 2 moved to different study strand, 1 on holiday, 1 requested 1 : 1 treatment, 1 fall, 3 unknown IP: 2 relapse, 1 unwell at day of measurement, 1 inclement weather, 2 on holidays, 3 moved to different study strand, 1 unknown Y: 1 discontinued intervention by choice, 2 unknown C: 2 relapse, 1 rapidly progressing MS, 1 conflicting appointments Measurements Baseline, 12 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Block randomization per geographical region
Allocation concealment (selection bias)	Low risk	Randomization performed by central co‐ordinator
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	The overall drop‐out rate was 24%, equally balanced across exercise (24%) and control (21%)
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Unclear risk	Unequal group sizes

Methods	Design: random assignment to 8‐week aquatic exercise programme (EX) or no training/wait list (NEX) Setting: Faculty of Physical Education and Sport Sciences, University of Isfahan, Isfahan, Iran Categorization: EX = endurance
Participants	n = 32; EX = 16, NEX = 16 Inclusion criteria: definite MS, > 2 years after diagnosis, no relapse < 1 month, ability to participate regular exercise sessions Exclusion criteria: relapse during intervention, disease preventing participation Type MS: RRMS Disease duration (yr) ± SD: EX = 4.9 ± 2.3, NEX = 4.6 ± 1.9 Mean age (yr) ± SD: EX = 33.7 ± 8.6, NEX = 31.6 ± 7.7 % Female (n/n group): EX = 10/10, NEX = 11/11 Mean EDSS ± SD: EX = 2.9 ± 0.9, NEX = 3.0 ± 0.7
Interventions	EX: 8‐week aquatic exercise (3x per week, 60 minutes); 10‐minute warm‐up, 40‐minute exercise, 10‐minute cool‐down at 50‐75% heart rate reserve. Exercises included activities focused on joint mobility, flexor and extensor muscle strengths, balance movements, posture, functional activities, and intermittent walking NEX: no training/maintain current habits
Outcomes	MFIS, MSQoL‐54
Notes	Drop‐outs EX: 6, no reasons specified NEX: 5, no reasons specified Measurements Baseline, 4 weeks, 8 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Shuffled envelopes
Allocation concealment (selection bias)	Low risk	Sealed envelopes
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Unclear risk	Overall drop‐out rate 34%, equally balanced across groups. Reasons for drop‐out not specified
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	High risk	A large increase in fatigue in the control condition

Methods	Design: random assignment to 10 weeks' inspiratory muscle training (EX) or control group (NEX) Setting: 3 rehabilitation outpatient clinics in Stockholm, Sweden Categorization: EX = other
Participants	n = 15; EX = 7, NEX = 8 Inclusion criteria: progressive MS according to Poser criteria, EDSS 6.5‐9.5 Exclusion criteria: EDSS < 6.0, chronic obstructive airways, asthma, emphysema, cystic fibrosis, heart insufficiency, another diagnosis/disorder Type MS: progressive MS Disease duration (yr); range: EX = 12 (3‐19), NEX = 20 (12‐35) Mean age (yr); range: EX = 46 (37‐49), NEX = 52.5 (38‐61) % Female (n/n group): EX = 1/7, NEX = 5/8 Mean EDSS; range: EX = 7.5 (6.5‐8.0), NEX = 8.0 (6.5‐9.0)
Interventions	EX: inspiratory muscle training (10 weeks, 70 sessions). Each session consisted of 3x 10 loaded inspirations with 1‐minute rest in between at 40‐60% of maximal inspiratory pressure NEX: no additional training besides regular physiotherapy
Outcomes	FSS, maximal inspiratory ‐ expiratory pressure, FVC, FEV, VC, RPE after morning bathing/dressing
Notes	Drop‐outs No drop‐outs reported Measurements Baseline, 10 weeks, 14 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomization procedure not described
Allocation concealment (selection bias)	Unclear risk	Allocation concealment procedure unknown
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	No drop‐outs reported
Selective reporting (reporting bias)	High risk	Not all outcomes reported
Other bias	Low risk	No other sources of bias identified

Methods	Design: random assignment to leisure centre‐based exercise group (EX) or control group (NEX); separate randomization for 2 different treatment sites Setting: Multiple Sclerosis Service, Scotland, UK Categorization: EX = mixed
Participants	n = 32; EX = 20, NEX = 12 Inclusion criteria: confirmed diagnosis MS, EDSS 5‐6.5, stable rehabilitation and drug therapy < 1 month, cognitive score > 24 on the MMSE Exclusion criteria: exacerbation < 3 months, medical condition that may preclude participants from exercise intervention Type MS: ? Disease duration (yr) ± SD: EX = 13.4 ± 6.4, NEX = 12.6 ± 8.1 Mean age (yr) ± SD: EX = 51.4 ± 8.06, NEX = 51.8 ± 8.0 % Female (n/n group): EX = 15/20, NEX = 8/12 Mean EDSS ± SD: EX = 6.14 ± 0.36, NEX = 5.82 ± 0.51
Interventions	EX: leisure centre‐based exercise group (12 weeks, 2x per week). 10‐minute warm‐up of aerobic and stretching components; 30‐40 minutes of circuit exercises (8‐12) designed for aerobic endurance, resistance, and balance NEX: usual routine, avoid new exercise regimen for the 12‐week study period
Outcomes	FSS, T25FW, 6MWT, BBS, TUG, QPW, BMI, PF, ABC, HADS, LMSQoL
Notes	Drop‐outs EX: 1 family commitments, 1 participate in other study, 1 increased work commitments, 1 influenza‐like symptoms, 1 suspected trigeminal neuralgia NEX: 1 unable to commit time for assessments, 1 unable to attend due to weather conditions Measurements Baseline, 8 weeks, 12 weeks
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computerized randomization procedure
Allocation concealment (selection bias)	High risk	Randomization list was generated beforehand to include a potential 20 participants
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Overall drop‐out rate 22%, without indication of differences in reasons
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	Low risk	No other sources of bias identified

Methods	Design: random assignment to 12‐week group circuit training (EX) or control group (NEX) Setting: Physiotherapy Gym, Dublin, Ireland Categorization: EX = mixed
Participants	n = 30; EX = 17, NEX = 13 Inclusion criteria: definite MS, independently mobile without use of aids, able to attend twice weekly + exercise independently at home Exclusion criteria: relapse < 3 months; history of cardiac conditions that would limit exercise capacity; cognitive or psychosocial condition that would limit class participation Type MS: RRMS and SPMS Disease duration (yr) ± SD: EX = 5.4 ± 4.35, NEX = 5 ± 3.52 Mean age (yr) ± SD: EX = 40.5 ± 12.68, NEX = 33.58 ± 6.1 % Female (n/n group): EX = 14/17, NEX = 10/13 Mean EDSS ± SD: ?
Interventions	EX: leisure centre‐based group classes (12 weeks, 2x per week) 5‐minute warm‐up, 40‐minute exercises. 4 different stations (10 minutes each): treadmill walking/running, cycling, stair‐master training, arm‐strengthening exercises, volleyball, and outdoor walking. In addition, participants were required to do a single type home exercise programme of choice for 40‐60 minutes at RPE 11‐13 NEX: normal activity + physiotherapy once monthly
Outcomes	MFIS, FAMS, MSIS‐29, exercise capacity (HR_max, RPE)
Notes	Drop‐outs EX: 2 relapses, 1 inconvenient timing of classes, 1 pregnancy, 1 personal reasons NEX: 1 moved home Measurements Baseline, 3 months, 6 months
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomization by blinded withdrawal of labelled slips from box
Allocation concealment (selection bias)	Unclear risk	Only 2 slips were present in the box during the allocation of each participant
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Due to the nature of the intervention, neither participants nor personnel could be blinded
Blinding of outcome assessment (detection bias)   All outcomes	High risk	Fatigue was self reported and therefore outcome assessment was not blinded
Incomplete outcome data (attrition bias)   All outcomes	High risk	Overall drop‐out rate 20%, unequally balanced across exercise (29%), and control (8%)
Selective reporting (reporting bias)	Low risk	No selective reporting identified
Other bias	High risk	After 17 people had been allocated to the experimental group the remaining people were allocated to the control group to maintain equal group sizes