Safety of exercise training in multiple sclerosis: An updated systematic review and meta-analysis

Abstract

Background:

A review of the safety profile of exercise training in multiple sclerosis (MS) has not been conducted since 2013.

Objective:

We undertook a systematic review and meta-analysis of randomised controlled trials (RCTs) of exercise training published since 2013 and quantified estimated population risks of clinical relapse, adverse events (AE) and serious adverse event (SAE).

Methods:

Articles reporting safety outcomes from comparisons of exercise training with non-exercise among persons with MS were identified. The risk of bias was established from study’s internal validity assessed using Physiotherapy Evidence Database (PEDro). Rates and estimated mean population relative risks (RRs; 95% confidence interval (CI)) of safety outcomes were calculated, and random-effects meta-analysis estimated the mean RR.

Results:

Forty-six interventions from 40 RCTs (N = 1780) yielded 46, 40 and 39 effects for relapse, AE, adverse effects and SAE, respectively. The mean population RRs ((95% CI), p-value) for relapse, AE and SAE were 0.95 ((0.61, 1.48), p = 0.82), 1.40 ((0.90, 2.19), p = 0.14) and 1.05 ((0.62, 1.80), p = 0.85), respectively. No significant heterogeneity is observed for any outcome.

Conclusion:

In studies that reported safety outcomes, there was no higher risk of relapse, AE, adverse effects or SAE for exercise training than the comparator. Exercise training may be promoted as safe and beneficial to persons with MS.

Background

Exercise training represents a rehabilitation-based approach for reversing multiple sclerosis (MS) dysfunction and managing symptoms and should be promoted among persons with MS throughout the disease trajectory.^1,2 Systematic reviews and meta-analyses report exercise training can improve physical fitness and walking mobility,^3,4 balance, ⁵ cognition, ⁶ fatigue, ⁷ depressive symptoms^8–10 and quality of life ⁷ in MS. Exercise has effects on the hippocampus and other brain structures,^11,12 sleep quality^13,14 and cardiovascular and metabolic comorbidity.^15,16 Upwards of 80% of people with MS do not engage in sufficient amounts of exercise necessary for health-related benefits. ¹⁷

The clinical guidelines for exercise prescription in MS^7,18,19 are endorsed internationally.^1,20 Yet, we know very little about the safety profile of exercise among persons with MS.

Relapse-risk, adverse events (AEs) and serious adverse events (SAEs)^21,22 represent metrics for gauging the safety profile of exercise in MS clinical studies, and we described those terms in Table 1. Three reviews have considered the safety of exercise training for MS.^23–25 The first and most analytical review included 26 randomised controlled trials (RCTs) published until November 2013. ²⁵ That review reported that persons with MS who received exercise training conditions had lower rates of relapse than control conditions (27% lower relapse rate). Notably, the rate of other AEs was higher for exercise, as represented by a 67% higher risk of AE from exercise than control. Those rates were not based on meta-analytic procedures.

Table 1.

Items for extraction and definition of terms used in this systematic review.

Study details	Extractions format and definitions
Author	First author surname, gender
Year	Year published
Title	Title of publication
Country	Corresponding author affiliation
Funding source	From text
Conflict of interest declared	Reported as yes, no, other
Clinical trial registry ID	From text
Methods
Study design	RCT, cluster RCT, other
Study aims	From text
Years of study recruitment	Years
Retention strategy employed	Described where relevant as; remuneration or other method
Participants
Age	Mean (standard deviation), median (range)
Gender	Male, female, other/unspecified
Race/ethnicity	Described where possible from text
Socio-economic status	Employment status, income, level of education or similar
MS phenotype	MS phenotype of participants; relapse remitting, primary progressive, secondary progressive, progressive relapsing Interpreted for reporting from overall participants as; RR, relapsing; all, relapsing and progressive; progressive only
Years diagnosed with MS	Mean (standard deviation), median (range) Interpreted for reporting as; <5 years, 5–10 years or >10 years
Disability level a	Described where possible using Expanded Disability Status Scale or Patient-Determined Disease Steps. Means (standard deviation) or median (range) will be reported. Other descriptors of disability will be reported Interpreted for reporting from overall participants; mild, moderate, severe, mild and moderate, moderate and severe Mild disability is usually categorised as EDSS: <4.5 or PDDS: 0–3, moderate disability is usually categorised as EDSS: 4.5–6 or PDDS: 4–5 and severe disability is usually categorised as EDSS: 6.5–9.5 or PDDS: 7–8
Inclusion/exclusion criteria	From text
Intervention
Frequency	Sessions/week
Intensity (aerobic exercise)	Described where relevant as; light, moderate, vigorous Light intensity exercise is usually between 9 and 11 on the Borg 6 to 20 RPE scale or 1–2 on the Borg 1–10 RPE scale Alternatively, light intensity exercise is 30%–39% VO2R or HRR16 Vigorous intensity exercise is usually between 14 and 17 on Borg’s 6–20 RPE scale. Alternatively, vigorous intensity exercise is 60%–89% VO2R or HRR26
Type a	Described where possible as; aerobic, resistance, flexibility, balance, neuromotor, combined, aerobic interval training or other Aerobic exercise training is a type of exercise in which the body’s large muscles move rhythmically for sustained periods. 27 Minimal guidelines for aerobic exercise are two 30-minute sessions per week11,23
	Resistance exercise training refers to activities where muscles work or hold against an applied force or weight to improve muscular fitness, traditional resistance training incorporates progressions and rest intervals.28,29 Minimal guidelines for resistance exercise are two sessions per week comprising 5–10 exercises.11,23 Flexibility exercise training considers the activities that are designed to preserve or extend range of motion 27 Balance training refers to activities designed to increase lower body strength and reduce the likelihood of falls 27 Neuromotor or multicomponent exercise training combines different motor skills (e.g. balance, coordination, gait, agility and proprioceptive training);28,30 this is not combined exercise training Combined exercise is a combination of different exercise types within an intervention (e.g. aerobic exercise and resistance exercise) Aerobic interval involves varying the exercise intensity at fixed time interval during a single exercise session26
Session time	Session minutes/day
Exercise prescription	Described where possible as; modality of exercise, equipment, sets and repetition and rest periods. Detail of progression through programme will be identified
Meeting minimum guidelines dose a	Identified from the frequency of aerobic exercise (2/week) and resistance exercise (2/week) sessions, the intensity and time of aerobic exercise (moderate intensity, 30 minutes)11,24 and the intensity of resistance exercise (one to four sets of 10–15 repetitions at 10–15 repetitions maximum)11,24,31
Programme duration	Number of weeks
Facilitator qualifications and training	Described where possible according to clinical qualification and/or studying qualification
Mode of delivery a	Described where possible as; supervised, independent or remotely supervised Supervised programmes are in-person and supervised by a researcher trained in exercise rehabilitation, an allied healthcare professional or students trained in exercise rehabilitation on allied healthcare. We will extract data on the setting where the exercise training is supervised. Independent programmes are completed in the participants community or home, and a researcher or health professional does not supervise the intervention in real-time. Information may be provided via mail or asynchronously via telehealth. Participants may provide feedback on intervention adherence to the researchers/health professionals. Synchronous communication is limited between the researcher/health professional team and the participant Remotely supervised programmes are completed in the participants community or home, asynchronous telecommunication to provide supervision and programming or intended advice is an important study construct. We will extract data on the setting where the exercise training is supervised
Description of comparator	Control condition will be categorised. We will extract data on the instruction provided to control participants, example categories include ‘usual activity’, ‘usual activity + social programme’ and ‘education’
Primary outcomes of interest	Only events occurring during the intervention period will be considered
Relapse	Relapse is an acute onset of new or worsening neurological symptoms, lasting over 24 hours32 Will be reported on using terms ‘relapse’ or a combination of words pertaining to ‘increase symptoms’, ‘symptom exacerbation’ From text, distinction of increased symptoms indicating a relapse will be determined from the text
Adverse event a (including adverse effects)	An adverse event is an unfavourable outcome that occurs during or after the intervention, 33 we consider AE to have a causal relationship, or not, to the intervention Will be reported on terms ‘adverse event’, ‘adverse effect’ or ‘injury’, ‘illness’, ‘falls’, ‘joint pain’, ‘upper respiratory tract infection’, ‘sprains’, ‘strains’, ‘muscle pain’ and ‘symptom exacerbation’ Described where possible as; musculoskeletal, respiratory illness, fall, cardiovascular and other From text, distinction between adverse event and adverse effect will be determined from text. We will identify the presence of causal language, for example, ‘engagement in intervention led to . . .’ or ‘event was unrelated to participation in the intervention’ to assist in our identification of adverse effects
Adverse effect	A adverse effect is an AE where a causal relationship is related to the intervention
Serious adverse event a (including serious adverse effects)	An SAE is an unfavourable outcome that results in death or is life-threatening, requires hospital admission or results in significant or permanent disability that occurs during or after the intervention, 34 we consider SAEs with a causal relationship, or not, to the intervention Will be reported on terms ‘serious adverse event’, ‘heart attack’, ‘myocardial infarction’, ‘stroke’, ‘pulmonary embolism’, ‘fracture’ and ‘dislocation’ to assist in our identification of adverse effects Described where possible as; musculoskeletal, respiratory illness, fall and cardiovascular From text, distinction between serious adverse event and serious adverse effect will be determined from text. We will identify the presence of causal language, for example, ‘engagement in intervention led to . . .’
Serious adverse effect	A serious adverse effect is an SAE where a causal relationship is related to the intervention
Retention rates	Retention is the completion of outcome measurements following the intervention Will be reported on number completed first postintervention follow-up data collection/number recruited
Intervention adherence rate	Adherence is the extent to which the participant follows the intervention corresponding with the agreed recommendations of the study, 35 we consider adherence as attendance to exercise sessions Will be reported on number of attended exercise sessions for the intervention From text: terms of attendance to exercise sessions for the intervention, for example, ‘attendance’, ‘journal’, ‘diary’ aspects reported will include ‘Frequency’, ‘intensity’, ‘modality’ and ‘duration’
Intervention compliance rate	Compliance is the extent to which the participant exercise behaviour matches the agreed recommendations of the study, 35 we consider compliance as the completion of the prescribed exercise programme Will be reported on compliance and completion of the prescribed programme From text: terms of completion of the exercise prescription, for example, ‘completed’, ‘dose’, ‘sets’, ‘repetitions’ and ‘prescription’
Risk of bias
PEDro 36
Inclusion criteria and source	Not scored – extracted as above
Random allocation	Yes/no
Concealed allocation	Yes/no
Baseline comparability	Yes/no
Subject blinding	Yes/no
Therapist blinding	Yes/No
Assessor blinding	Yes/No
Completeness of follow-up	Yes/No
Intention to treat analysis	Yes/No
Between group statistical comparisons	Yes/No
Point measures and variability	Yes/No

RCT: randomised controlled trial; MS: multiple sclerosis; RR: relative risk; EDSS: Expanded Disability Status Scale; PDDS: Patient-Determined Disease Steps; RPE: rate of perceived exertion; VO₂R or HRR–VO₂ Reserve or Heart Rate Reserve; AE: adverse event; SAE: serious adverse event; PEDro: Physiotherapy Evidence Database.

^aConsidered for subgroup analyses.

To date, the focus on intervention characteristics, such as exercise type, delivery style (e.g. supervised, independent or remotely supervised), participant disability level or the prescription of exercise consistent with minimal exercise guidelines for persons with MS,^7,18 has not been considered in terms of safety.

This systematic review builds on our previous review. ²⁵ By considering publications from 2013 onward, we identify the number of relapses, AE, adverse effects, SAE and serious adverse effects in exercise training studies in MS and quantify the relative risk (RR) of relapses, AE and SAE for exercise conditions compared with non-exercise and non-pharmacological comparison conditions. Secondary aims include exploration of potential sources of variability in overall RR, including (1) exercise types, (2) exercise delivery styles, (3) disability levels (based on established disability cut-points ³⁷ ) and (4) prescription of exercise training based on guidelines for persons with mild-to-moderate MS.^7,18

Methods

Our review adhered to the Preferred Reporting Items for Systematic review and Meta-Analysis Protocol statement. ³⁸ An original review by the research team ²⁵ was a basis for this study. The protocol ³⁹ for the current review was registered with the international prospective register of systematic reviews (PROSPERO 2020 CRD42020190544). The Participants, Interventions, Comparisons and Outcomes (PICO) strategy ⁴⁰ was adapted to suit the scope of the review as follows: the terms were persons with Multiple Sclerosis (participants) and Exercise training (intervention) (refer to search strategy). The comparison was a non-exercise, non-pharmacological comparator, and the outcomes included relapse, AE, adverse effects, SAE and serious adverse effects.

Eligibility criteria

The inclusion criteria were as follows: RCT reporting safety outcomes overall and per condition during the study; adults (aged 18 years or above) with MS and an intervention type which meets the definition of exercise training. ³³ We consider the definition of exercise as a subset of physical activity. Exercise is physical activity that is planned, structured, repetitive and purposive towards the improvement or maintenance of physical fitness. ³³ All non-active/non-exercise/non-pharmacological control conditions were considered, for example, ‘usual activity’.

The following exclusion criteria were applied: studies not written in English, non-human participants, physical interventions which are primarily sedentary (e.g. manual therapy, breathing exercises, pelvic floor exercises), physical intervention requiring external support to facilitate movement (e.g. hippotherapy or robot assisted exercise), studies where safety outcomes were not reported or unclearly reported (in cases of unclear reporting we contacted study authors a minimum of two times to collect safety data) and studies which were previously included in the 2014 publication. ²⁵

Search strategy and information sources

We searched for publications of RCTs of exercise training in persons with MS using eight databases: Ovid MEDLINE All; Ovid Embase; ProQuest PsycINFO; Cochrane Central Register of Controlled Trials (CENTRAL), Ebsco CINAHL, Scopus, Web of Science Core Collection (CC) and PEDro – Physiotherapy Evidence. We conducted our first search on 17 March 2020 and a final search on 27 April 2022. The search was completed from 2013 to the date of the search. This aligns with our original review search, which ended in November 2013. Supplemental Material 1 comprises the search strategy per database. We also searched the reference lists of included studies and relevant systematic reviews for studies meeting our inclusion criteria, searched clinical trial registries and performed forward citation searches on our included studies.

Data management and extraction

Screening and data extraction were undertaken using Covidence (https://www.covidence.org) and Microsoft Excel, respectively. Articles were screened, and data were extracted per our protocol. ³⁹ Briefly, two authors independently screened studies at the title/abstract and full-text stages, and any disagreement resolved through third author discussion. Data were also extracted independently by two authors, with disagreement resolved through discussion. Where data were not reported in manuscripts, authors were contacted a minimum of two times. Where data could not be sought, data were recorded as not reported. Details of data extracted from each study are available in Table 1.

The risk of bias was determined via the Physiotherapy Evidence Database (PEDro) scale. ³⁴ The PEDro scale is an 11-item scale that relates the external validity or generalisability of the sample (item 1) and the internal validity (items 2–11) of a study. We focused on internal validity (range = 0–10). Thus, a score of 10 indicates the study satisfies the maximum items for internal validity. We considered a score above six representative of high internal validity ²⁵ and hence high methodological quality.

Outcomes

The primary outcomes of interest were (1) Relapse, (2) Adverse Events (AEs), (3) Adverse Effects, (4) Serious Adverse Events (SAEs) and (5) Serious Adverse Effects. Definitions of each outcome are provided in Table 1. Where data were not reported and could not be sought through author contact, data were recorded as not reported.

Analysis

Analyses were conducted using SPSS (IBM SPSS v26; MeanES and MetaReg macros) and Stata (Stata Corp v14; meta).

Calculation of RR

Overall, RR was calculated for each of our primary outcomes (for exercise compared with the comparator). The unit of analysis for each outcome was the total number of relapses, AEs or SAEs experienced in each group of each study.

For example, the overall rate of relapse, defined as the total number of reported relapses in the exercise and comparator (i.e. number of relapses in each condition by the total participants in each condition), was pooled across all studies, respectively. We calculated the overall RR of relapse using standard risk procedures (i.e. ratio of rates) defined as the ratio of participants in a treatment group who experience an illness, condition or event (i.e. relapse) to those in a comparator who experience the same illness, condition or event. ⁴¹ We calculated RR by dividing the overall rate of relapse, AEs, adverse effects or SAEs for exercise by the overall rate of relapse, AEs or SAEs, respectively, in the comparator. RR > 1.0 indicated no difference in risk between exercise and comparator; RR > 1.0 indicated higher risk for exercise training and RR < 1.0 indicated a lower risk for exercise training. We only reported data from studies that directly report on the presence or absence of relapse or events. RR, its standard error and 95% confidence intervals (CIs) were calculated according to standard procedures. ⁴² Where zeros created problems with the computation of RR or standard errors, 0.5 was added to all cells (e.g. exercise relapse, exercise non-relapse, control relapse and control non-relapse).^43,44

Data synthesis and analysis

Using an SPSS MeanES macro, ⁴⁵ a random-effects model was used to aggregate the mean RR for each outcome. Based on established methods, ⁴⁶ Cochrane Q and I ² were calculated to evaluate heterogeneity and consistency, respectively. Sampling error was calculated using Cochrane’s Q according to established methods; heterogeneity was indicated if sampling error accounted for <75% of the observed variance and when p < 0.05 for Q ⁴⁷ . I ² values of 25%, 50% and 75% indicate small, moderate and large amounts of heterogeneity, respectively. ⁴⁸

Results

Study selection

We identified 39 eligible reports via the database search and one other via forward citation search (Figure 1). Of those 40 RCTs, 46 exercise interventions were included in the analysis. Sufficient data were included to perform meta-analysis and report on the primary outcome of (1) relapse (k = 46); (2) AEs (k = 40) and (3) SAEs (k = 39). We identified 34 interventions with adverse effects. As adverse effects are AE where a causal relationship is related to the intervention we cannot undertake meta-analyses comparison, as our comparator was usual care (i.e. not an intervention). Serious adverse effects were not reported in any study.

Figure 1.

PRISMA flow diagram of updated review.

*Authors contacted twice for full texts.

Study characteristics

Table 2 provides characteristics of included RCTs. The studies were published between 2013 and 2022. The majority of the studies were conducted in Iran (n = 8, 20%), followed by the United States (n = 7, 18%), Belgium (n = 4, 10%), Turkey and Italy (n = 3 per country), Australia, Denmark and the United Kingdom (n = 2 per country), and Brazil, Ireland, the Netherlands, Norway, Slovenia and Spain (n = 1 per country). In 23 (56%) studies, the first author was female. Funding was disclosed for 27 (68%) studies.

Table 2.

All included studies.

Study details First author, year, gender, country	PEDro score	Participants	Exercise prescription	Reporting
		Allocated/completing, mean age (SD), female: male, nationality/race/ethnicity, phenotype (s), mean years diagnosed (SD), MS disability level	Duration, exercise type, intensity, frequency/week for session length, supervision (supervisor qualification), achieving MS guidelines	Consort	Relapses	A Events	A Effects	S A Events
Ahmadi et al. 49 Female Iran	8	Intervention 1 10/10, 36.8 (9.17) years, 10:0, NR, NR, 5–10 years, Mild Intervention 2 11/11, 32.3 (8.7) years, 11:0, NR, NR, 5–10 years, Mild Control 10/10, 36.7 (9.32) years, 10:0, NR, NR, 5–10 years, Mild	Intervention 1 8 weeks, Aerobic (constant treadmill training), 40%–75% age predicted HRmax, 3/week for 30 minutes, supervised (NR), NA Intervention 2 8 weeks, Neuromotor (yoga), NR, 3/week for 60/70 minutes, supervised (physiotherapist), NA	No	Yes	No	No	No
Arntzen et al. 50 Female Norway	8	Intervention 40/39, 52.2 (12.9) years, 27:12, NR, All, > 10 years, Mild Control 40/39, 48.0 (8.8) years, 29:11, NR, All, > 10 years, Mild	6 weeks, Combined (Core stability and Dual tasks), NR, 2 or 3/week for 30–60 minutes, supervised (physiotherapist) progressing to independent, Not achieving guidelines	Yes	Yes	Yes	Yes	Yes
Azimzadeh et al. 51 Female Iran	8	Intervention 18/16, NR, 18:0, All Iranian, RR, 5–10 years, moderate Control 18/18, NR, 18:0, All Iranian, RR, 5–10 years, moderate	12 weeks, Other (Tai chi), NR, 2/week for 60 minutes, supervised (Tai chi instructor), NA	No	Yes	No	No	No
Backus et al. 52 Female The United States	8	Intervention 12/6, 56.2 (10.0), 3:3, 4 Black 2 White, All, NR, Severe Control 9/6, 54.7 (11.6), 4:2, 3 Black 3 White, All, NR, Severe	12 weeks, Aerobic (constant FES cycling), NR, 3/ week for 30–60 minutes, supervised (NR), NA	Yes	Yes	Yes	Yes	Yes
Baquet et al. 53 Female Germany	9	Intervention 34/29, 38.2 (9.6), 21:13, NR, RR, 5–10 years, Mild Control 34/29, 39.6 (9.7), 25:9, NR, RR, 5–10 years, Mild	12 weeks, Aerobic (Interval bicycle ergometer), varying across 10 different powers, 2–3/week for 15–150 minutes, supervised (physiotherapist), NA	Yes	Yes	Yes	Yes	Yes
Callesen et al. 54 Male Denmark	9	Intervention 1 23/17, median 52 (IQR 38–64) years, 16:7, NR, All, > 10 years, Moderate Intervention 2 28/24, median 52 (IQR 31–75), 23:5, NR, All, 10 years, Moderate Control 20/18, median 56 (IQR 30–73), 16:4, NR, All, 10 years, Moderate	Intervention 1 10 weeks, Resistance (resistance machine, lower limb), 8 RM, 2/week for 60 minutes, supervised (physiotherapist), NA Intervention 2 10 weeks, Neuromotor (Balance and motor control), Varying BOS, 2/week for 60 minutes, supervised (physiotherapist), NA	Yes	Yes	Yes	Yes	Yes
Cameron et al. 55 Male The United States	8	Intervention 47/37, 57.5 (13.3), 34/13, 43 Caucasian 33 African American 1 Hispanic, All, NR, Mild to severe Control 49/38, 53.5 (10.6), 36/13, 41 Caucasian 2 African American 1 Asian 3 Hispanic 2 other, All, NR, Mild to severe	8 weeks, Neuromotor (balance), NR, 1/week for 60 minutes, supervised (NR)	Yes	Yes	Yes	Yes	Yes
Felippe et al. 36 Male Brazil	8	Intervention 14/13, median 35 (IQR16.5) years, 3:10, NR, RR, 5–10 years, Mild Control 14/14, median 38 (IQR 15.3) years, 5:9, NR, RR, 5–10 years, Mild	24 weeks, Other (cognitive motor exercise), RPE 11–14, 2/week for 60 minutes, supervised (physiotherapist), NA	Yes	Yes	Yes	Yes	Yes
Carter et al. 56 Male The United Kingdom	9	Intervention 60/54, 45.7 (9.1), 43:17, 90% White, All, 5–10 years, Mild Control 60/53, 46 (8.4), 43:17, 95% White, All, 5–10 years, Mild	12 weeks, Combined (Aerobic and Resistance), aerobic; 50%–69% Agre predicted HRmax/12%14 RPE, resistance; 20 RM, 3/week for 30–60 minutes, supervised progressing to independent, guidelines achieved.	Yes	Yes	Yes	Yes	Yes
Faramarzi et al. 57 Male Iran	7	Intervention 48/43, NR, 43:0, NR, NR, NR, NR Control 48/46, NR, 46:0, NR, NR, NR, NR	12 weeks, Combined (Aerobic, resistance, Pilates, stretching), low intensity, 3/week for 90–110 minutes, supervised (physiotherapist/exercise physiologist), Not achieving guidelines	Yes	Yes	No	No	No
Feys et al. 58 Male Belgium	6	Intervention 21/18, 36.6 (8.5), 20:1, NR, NR, 5–10 years, Mild Control 21/17, 44.4 (8.5), 18:3, NR, NR, 5–10 years, Mild	12 weeks, Aerobic (outdoor walking-running), Up to 5–10 km/hour higher than initial capacity, 3/week for NR time, remotely supervised, NA	Yes	Yes	Yes	Yes	Yes
Fleming et al. 59 Male Ireland	7	Intervention 39/29, 46.7 (10.0), 36:3, NR, NR, NR, Mild to moderate Control 41/34, 47.4 (10.2), 33:8, NR, NR, NR, Mild to moderate	8 weeks, Neuromotor (Pilates), NR, 3/week for NR time, Independent, NA	Yes	Yes	Yes	Yes	Yes
Gheitasi et al. 60 Male Iran	8	Intervention 15/15, 32.1 (6.3), 0:15, NR, NR, < 5 years, Moderate Control 15/15, 30.6 (5.3), 0:15, NR, NR, < 5 years, Moderate	12 week, Neuromotor (Pilates), NR, 3/week for 60 minutes, Supervised (NR), NA	Yes	Yes	Yes	Yes	Yes
Guillamó et al. 61 Female Spain	6	Intervention 1 8/8, 42 (NR), 6:2, NR, RR, NR, Mild Intervention 2 11/8, 42 (NR), 8:3, NR, RR, NR, Mild Control 10/9, 42 (NR), 7:3, NR, RR, NR, Mild	Intervention 1 40 week, Combined (constant; Aerobic, resistance, balance, stretching), aerobic; 50%–85% HRmax, 6/week for 40 minutes, Supervised and concurrently independent, Guidelines not achieved Intervention 2 40 week, Combined (Interval; Aerobic, resistance, balance, stretching), 11–12 RPE and 17–18 RPE, 6/week for 40–90 minutes, Supervised and concurrently independent, Achieving guidelines	Yes	Yes	Yes	Yes	Yes
Author Gunn et al. 62 Female The United Kingdom	9	Intervention 30/25, 58.7 (10.8), 20:10, NR, All, NR, Severe Control 26/24, 60 (8.5), 17:9, NR, All, NR, Severe	13 week, Other (gait and functional training), 3/5 on ‘wobbly-scale’, NR/week for 120 minutes, supervised (physiotherapist) and concurrently independent	Yes	Yes	Yes	Yes	Yes
Hansen et al. 63 Female Belgium	9	Intervention 23/16, 46.0 (11.0), 10:6, All Caucasian, All, NR, Mild Control 13/11, 48.0 (10.0), 6:5, All Caucasian, All, NR, Mild	24 weeks, Combined (Aerobic walking or cycling and Resistance whole body), aerobic; 12–14 RPE; resistance 15 RM, 5/fortnight for NR minutes, Supervised (NR), Achieving guidelines	Yes	Yes	Yes	Yes	Yes
Heine et al. 64 Male The Netherlands	8	Intervention 43/37, 43.1 (9.8), 32:11, NR, All, NR, Mild Control 46/36, 48.2 (9.2), 33:13, NR, All, NR, Mild	16 week, Aerobic (Interval cycle ergometry), 3:1:1 minute at 40%:60%:80% peak power, 2/week for 30 minutes, Supervised (physiotherapist), NA	Yes	Yes	Yes	Yes	Yes
Heinrich et al. 65 Female Germany	7	Intervention 25/17, 51.9 (7.9), 17:8, NR, NR, > 10 years, Moderate Control 27/22, 50.3 (6.9), 15:12, NR, NR, > 10 years, Moderate	12 weeks, Aerobic (FES cycling), 12–14 RPE. 2–3 week for NR minutes, supervised (physiotherapist) progressing to independent, NA	Yes	Yes	Yes	Yes	Yes
Hoang et al. 66 Male Australia	9	Intervention 28/23, 53.4 (10.7), 21:7, NR, All, > 10 years, moderate Control 22/21, 51.4 (12.8), 17:5, NR, All, > 10 years, moderate	12 weeks, Other (exergames), NR, 2/week for 30 minutes. Independent, NA	Yes	Yes	Yes	Yes	Yes
Kooshiar et al. 67 Male Iran	8	Intervention 20/18, NR, 20:0, All Iranian, All, < 5 years, Mild Control 20/19, NR, 20:0, All Iranian, All, < 5 years, Mild	8 weeks, Other (aquatic-aerobic, -stretching, -resistance), NR, 3/week for 45 minutes, Supervised (physiotherapists), NA	Yes	Yes	Yes	Yes	Yes
Learmonth et al. 68 Female The United States	8	Intervention 29/25, 48.7 (10.4), 28:1, 20 Caucasian 17 African American, RR, > 10 years, Mild Control 28/26, 48.2 (9.1), 27:1, 18 Caucasian 8 African American 1 American Indian 1 Latino, RR, > 10 years, Mild	16 weeks, Combined (Aerobic walking and resistance), ⩾100 steps/min and 12–14, 4/week for 30 minutes, Independent, Guidelines achieved.	Yes	Yes	Yes	Yes	Yes
Louie et al. 69 Female Australia	8	Intervention 12/9, 48.9 (9.7), 6:6, NR, All, > 10 years, Moderate Control 11/8, 48.3 (14.1), 7:4, NR, All, > 10 years, Moderate	12 week, Combined (core stability, balance, stretching, resistance and aerobic circuit exercise), NR, 2/week for 60 minutes, Supervised (physiotherapist) progressing to independent, Not achieving guidelines	Yes	Yes	Yes	Yes	Yes
Lutz et al. 48 Female Germany	7	Intervention 8/8, 52.4 (10.4), 7:1, NR, All, > 10 years, Mild Control 6/6, 56 (7.4), 6:0, NR, All, > 10 years, Mild	6 weeks, Other (participant choice of aerobic, strength, neuromotor), NR, Supervised (Sport scientist), NA	Yes	Yes	Yes	Yes	Yes
Magnani et al. 70 Female The United States	6	Intervention 13/11, 47.8 (10.8). 6:7, NR, NR, NR, NR Control 12/10, 40.7 (13.0), 7:5, NR, NR, NR, NR	24 weeks, Combined (Aerobic, resistance, stretching), aerobic; 50%–80% max work rate, resistance; 15%–30% of max load, 3/week for 60 minutes, Supervised (NR), Guidelines achieved	No	Yes	Yes	No	Yes
McAuley et al. 71 Male The United States	8	Intervention 24/22, 59.6 (1.4), 18:6, NR, All, > 10 years, Mild and moderate Control 24/24, 59.8 (1.5), 18:6, NR, All, > 10 years, Mild and moderate	26 weeks, Combined (Aerobic, strength, stretching), 10–15 RPE. 3/week for 30–50 minutes, Remotely supervised, Not achieving guidelines	Yes	Yes	Yes	No	Yes
Moradi et al. 72 Female Iran	8	Intervention 8/8, 34.4 (11.1), 0:8, NR, All, 5–10 years, Mild Control 10/10, 33.1 (7.0), 0:10, NR, All, 5–10 years, Mild	8 week, Resistance (Weight machines upper and lower limb), NR, 5/week for 30 minutes, Independent, Not achieving guidelines	No	Yes	No	No	No
Negaresh et al. 73 Male Iran	9	Intervention 36/34, NR, 22:12, NR, NR, 5–10 years, Mild Control 30/27, NR, 18:9, NR, NR, 5–10 years, Mild	8 week, Aerobic (Upper and lower limb cycle ergometry), 60%–75% peak work rate, 3/week for 42–66 minutes, Supervised (Exercise physiologists), NA	Yes	Yes	Yes	Yes	Yes
Ozkul et al. 74 Female Turkey	8	Intervention 17/17, 35.9 (9.7), 13:4, NR, NR, 5–10 years, Mild and moderate Control 17/17, 36.8 (9), 13:4, NR, NR, 5–10 years, Mild	8 week, Combined (Aerobic and Pilates), aerobic; <70%–80%–70% HRmax, 3/week for 30 minutes, Supervised (physiotherapist), Not achieving guidelines.	Yes	Yes	Yes	Yes	Yes
Ozkul et al. 75 Female Turkey	8	Intervention 12/10, 46.0 (NR), 6:4, NR, All, > 10 years, Mild Control 11/10, 21.5 (NR), 6:4, NR, All, > 10 years, Mild	6 weeks, Neuromotor (task-oriented functional training), NR, 3/week for 30 minutes Supervised (physiotherapist). NA	Yes	Yes	Yes	Yes	Yes
Pau et al. 76 Male Italy	6	Intervention 11/11, 47.4 (10.8), 5:6, NR, NR, NR, Mild Control 11/11, 44.5 (13.5), 5:6, NR, NR, NR, Mild	24 weeks, Combined (Aerobic and resistance), aerobic up to 80% of HRmax, resistance 15% of 1 RM–30% of 1 RM, 2/week for 60 minutes, Supervised (strength and conditioning coaches), Achieving guidelines	Yes	Yes	No	No	No
Riemenschneider et al. 77 Male Denmark	7	Intervention 42/38, 37.3 (10.1), 29:13, NR, NR, > 10 years, Mild Control 42/39, 37.4 (9.7), 34:8, NR, NR, > 10 years, Mild	48 weeks, Aerobic (Constant aerobic, Interval aerobic), constant aerobic 60%–80% HRmax. Interval aerobic 65%–95% HRmax, 2/week for 3–60 minutes, Supervised (NR), NA	Yes	Yes	Yes	Yes	Yes
Sangelaji et al. 78 Male Iran	6	Intervention 42/39, 33.1 (7.7), 24:15, NR, NR, NR, NR Control 30/22, 32.1 (6.4), 15:7, NR, NR, NR, NR	10 weeks, Combined (Aerobic, resistance, balance, stretching), aerobic 65%–75% HRmax, 3/week for 40 minutes, Supervised (NR), Guidelines achieved	No	Yes	Yes	No	Yes
Savšek et al. 79 Female Slovenia	8	Intervention 14/12, 39.7 (6.7), 11:3, NR, RR, 5–10 years, Mild Control 14/13, 42.3 (5.7), 11:3, NR, RR, 5–10 years, Mild	12 weeks, Aerobic (standing aerobic exercise), 50%–70% HRres, 2/week for 60 minutes, Supervised (aerobics trainer), NA.	Yes	Yes	Yes	Yes	Yes
Sosnoff et al. 80 Male The United States	8	Intervention 13/10, 60.1 (6.3), 10:3, NR, All, > 10 years, Moderate Control 14/12, 60.1 (6.0), 11:3, NR, All, > 10 years, Moderate	12 weeks, Combined (balance, resistance and stretching), NR, 2/week for 45–60 minutes, Supervised progressing to independent, not achieving guidelines	Yes	Yes	Yes	Yes	Yes
Sosnoff et al. 81 Male The United States	7	Intervention 1 11/10, 62.3 (7.5), 8:2, NR, All, > 10 years, Moderate Intervention 2 8/8, 59.3 (6.5), 6:2, All, > 10 years, Moderate Control 9/8, 63.3 (11.2), 8:2, NR, All, > 10 years, Mild	Intervention 1 12 weeks, Other (Falls preventing strength and balance exercise), NR, 2/week for 45–60 minutes, Supervised progressing to independent, NA Intervention 2 12 weeks, Other (Falls preventing strength and balance exercise and education), NR, 2/week for 45–60 minutes, Supervised progressing to independent, NA	Yes	Yes	Yes	Yes	Yes
Straudi et al. 82 Female Italy	9	Intervention 12/12, 49.9 (7.5), 7:5, NR, All, > 10 years, Moderate Control 12/12, 55.3 (13.8), 10:2, NR, All, > 10 years, Moderate	14 week, Combined (Aerobic, Resistance and Neuromotor and stretch), NR, 5 (week 1–2) and 3 (weeks 3+), for 120 or 60 minutes, Supervised (physiotherapist), Not achieving guidelines	Yes	Yes	Yes	Yes	Yes
Straudi et al. 83 Female Italy	8	Intervention 18/13, 49.7 (13.6), 11:7, NR, All, > 10 years, Moderate Control 18/16, 52.6 (12.6), 12:6, NR, All, > 10 years, Moderate	12 weeks, Neuromotor (task-oriented functional training), NR, 3/week for 60 minutes, Supervised (NR) progressing to independent, NA	Yes	Yes	Yes	Yes	Yes
Wens et al. 84 Female Belgium	7	Intervention 1 12/12, 47.0 (3.0), 6:5, NR, All, NR, Mild Intervention 2 11/11, 43.0 (3.0), 7:5, NR, All, NR, Mild Control 11/11, 47.0 (3.0), 9:2, NR, All, NR, Mild	Intervention 1 12 weeks, Combined (interval aerobic bicycle ergometer, resistance upper limb and lower limb), aerobic; 80%–100% of HRmax, resistance; 10–20 RM, 5/fortnight for 45–75 minutes, Supervised (physiotherapist), Achieving guidelines Intervention 2 12 weeks, Combined (continuous aerobic bicycle ergometer, resistance upper limb and lower limb), aerobic; 80%–90% HRmax, resistance; 10–20 RM, 5/fortnight for 45–75 minutes, Supervised (physiotherapist), Achieving guidelines	Yes	Yes	Yes	No	Yes
Wens et al. 16 Female Belgium	5	Intervention 15/15, 42.0 (3.0), 9:6, NR, All, NR, Mild Control 7/7, 44.0 (2.0), 5:2, NR, All, NR, Mild	24 weeks, Combined (aerobic bicycle ergometer, resistance upper limb and lower limb), 12–14 RPE, 5/fortnight, Supervised (physiotherapist), Achieving guidelines.	Yes	Yes	Yes	No	Yes
Young et al. 85 Female Turkey	8	Intervention 1 27/21, 49.7 (9.4), 22:5, 12 White 15 Black, RR, All, > 10 years, Mild Intervention 2 26/21, 48.4 (10.0), 20:6, 15 White 9 Black 2 other, RR, All, > 10 years, Mild Control 28/19, 47.3 (10.3), 24:4, 17 White 11 Black, RR, All, > 10 years, Mild	Intervention 1 12 weeks, Neuromotor (movement to music), NR, 3/week for 60 minutes, Supervised (dance instructor), NA Intervention 2 12 weeks, Neuromotor (Yoga), NR, 3/week for 60 minutes, Supervised (yoga instructor), NA	Yes	Yes	Yes	Yes	Yes

SD: standard deviation; MS: multiple sclerosis; NR: not reported; NA: not applicable; RR: relative risk; FES: functional electrical, stimulation; IQR: interquartile range; RM: maximum load of one repetition; BOS: base of support; A Events: adverse events; A Effects: adverse effects, S A Events: serious adverse events.

Methodological quality

Methodological quality judgement identified PEDro scores ranging between 5 and 9 (mean score 7.7 ± 1.0) on the 10-point internal validity scale (refer to Supplemental Tables). Almost all studies (n = 39, 98%) achieved a score of at least 6 out of 10, indicating high quality (Table 2). No study secured participant blinding, and all but one study failed to secure therapist blinding. ⁶²

Participant characteristics

The mean number of participants per control group or exercise training group was 21 (range 6–60) and 22 (range 8–60), respectively. Over half (n = 21, 53%) of studies reported a mean age over 45 years. The mean age of participants in the studies ranged between 32 and 63 years. There were 679 females and 279 males in the exercise interventions and 597 females and 225 males in the control conditions. Data on participant nationality, race or ethnicity were provided in eight studies.

Participants had primarily mild MS disability in 22 (55%) studies,^{16,36,48–50,53,56,58,61,63,64,68,72–77,79,84,85} primarily moderate MS disability in nine (23%) studies^{51,54,60,65,66,69,81–83} and primarily severe MS disability in 3 (8%) studies.^52,55,62 Participants had mild and moderate MS disability in three (8%) studies,^59,70,71 and two studies did not report disability level.^57,78 Participants primarily had a relapsing remitting MS phenotype in 6 (15%) studies^{36,51,53,61,68,79} and all MS phenotypes in 20 (50%) studies;^{48,50,52,54,56,62–64,66,67,69,71,72,74,80–84} MS phenotype was not reported in 13 (33%) studies.^{16,49,57–60,65,70,73,75,77,78,85} Participants had been diagnosed with MS for less than 5 years in two studies,^60,67 between 5 and 10 years in nine (23%) studies^{36,49,51,53,56,58,74,79,81} and over 10 years in 14 (35%) studies;^{48,50,54,65,66,68,69,71,75,77,80–82,85} 13 (33%) studies did not report disease duration.^{16,52,55,57,59,61–64,70,76,78,84}

Exercise intervention characteristics

Exercise intervention length ranged between 2 and 48 weeks (mean 14 weeks), with the number of sessions per week between one and seven (mean 3 sessions per week). Prescribed intensity varied across studies. Nine aerobic,^{49,52,53,58,64,65,73,77,79} 2 resistance,^54,72 16 combined^{16,50,56,57,61,63,68–71,74,76,78,80,82,84} and 7 neuromotor^{54,55,59,60,75,83,85} exercise interventions were included. Eight other exercise interventions did not meet these criteria types of exercise and were classified as ‘other’.^{36,48,49,51,62,66,67,81} Details on each intervention are listed in Table 2. Eight of the combined exercise interventions^{16,56,61,63,68,70,76,78,84} met the 2013 MS exercise guidelines^7,18 (see Table 1). The exercise sessions lasted between 30 and 100 minutes. Twenty-five interventions were supervised in-person,^{16,36,48,49,51–55,57,60,63,64,67,70,73–79,82,84,85} three interventions were completed independently by participants, with minimal researcher input,^59,66,72 three interventions included remote supervision,^58,68,71 three interventions included both supervised and independent exercise throughout,^56,61,62 and seven interventions began supervised and progressed to independent exercise.^{50,61,65,69,80,81,83}

Safety criteria

Rate of events

Our systematic search for relevant studies indicated that many (n = 50) studies were excluded for unclear or missing safety reporting (Supplemental reference list). Among the included studies, the number and description of relapses, AE, adverse effects and SAE are presented in Table 3. Across all studies, 40 (100%), 35 (88%), 30 (75%) and 34 (85%) reported relapses, AE, adverse effects and SAE, respectively. The number of relapses per study in the exercise and control conditions ranged from 0 to 6 and 0 to 9, respectively, across the studies, with an overall prevalence rate of 14% and 43%, respectively. The number of AEs in the exercise and control conditions ranged from 0 to 7 and 0 to 5 per study, respectively, and yielded overall prevalence rates of 38% and 22%, respectively. The number of SAEs in the exercise and control conditions was 0 to 4 and 0 to 5, respectively, and yielded prevalence rates of 9% and 10%, respectively. Adverse effects were only applicable to the exercise condition (i.e. not quantifiable against control for effect size and RR) and the number of adverse effects ranged from 0 to 6, with a prevalence rate of 29%.

Table 3.

Number and rate of relapses, adverse events and effects and serious adverse events.

Study (intervention)	Overall		Relapses		Adverse events			Adverse effects			Serious adverse events
	Con	Ex	n (%)		n (%)		Type	n (%)		Type	n (%)		Type
			Con	Ex	Con	Ex		Con	Ex		Con	Ex
Ahmadi et al. 49	10	10	0, 0%	0, 0%	NR	NR		NR	NR		NR	NR
Ahmadi et al. 49	10	11	0, 0%	0, 0%	NR	NR		NR	NR		NR	NR
Arntzen et al. 50	40	40	0, 0%	0, 0%	0, 0%	0, 0%		0, 0%	0, 0%		0, 0%	0, 0%
Azimzadeh et al. 51	18	18	0, 0%	0, 0%	NR	NR		NR	NR		NR	NR
Backus et al. 52	9	12	1, 11%	0, 0%	0, 0%	4, 22%	E: 3 Wounds, knee injury	0, 0%	0, 0%		0, 0%	0, 0%
Baquet et al. 53	34	34	0, 0%	1, 3%	2, 6%	1, 3%	C: Mental health E: Lumbago	0, 0%	0, 0%		0, 0%	0, 0%
Callesen et al. 54	20	28	0, 0%	0, 0%	0, 0%	0, 0%		0, 0%	0, 0%		0, 0%	0, 0%
Callesen et al. 54	20	23	0, 0%	0, 0%	0, 0%	4, 17%	E: Illness, 3 Falls	0, 0%	2, 9%	E: Training fatigue, low back pain	0, 0%	0, 0%
Cameron et al. 55	49	47	9, 18%	6, 13%	5, 10%	3, 6%	C: 3 Upper RT infections E: 3 Upper RT infections	0, 0%	2, 4%	E: Fall, muscle pain	5, 10%	4, 9%	C: Knee replacement, cervical disc disease, hiatal hernia, psychosis, sepsis E: 2 Knee replacement, fracture, influenza
Carter et al. 56	60	60	1, 2%	1, 2%	2, 3%	2, 3%	C: 2 Illness E: 2 Illness	0, 0%	0, 0%		0, 0%	0, 0%
Faramarzi et al. 57	48	48	1, 2%	0, 0%	NR	NR		NR	NR		NR	NR
Felippe et al. 27	14	14	0, 0%	0, 0%	0, 0%	1, 7%	E: Depression	0, 0%	0, 0%		0, 0%	0, 0%
Feys et al. 58	21	21	0, 0%	1, 5%	1, 5%	2, 10%	C: Discomfort E: Discomfort, psychosocial problems	0, 0%	6, 29%	E: 4 Muscle strain, 2 training fatigue	0, 0%	0, 0%
Fleming et al. 59	41	39	0, 0%	0, 0%	0, 0%	0, 0%		0, 0%	0, 0%		0, 0%	0, 0%
Gheitasi et al. 60	15	15	0, 0%	0, 0%	0, 0%	0, 0%		0, 0%	0, 0%		0, 0%	0, 0%
Guillamó et al. 61	10	11	0, 0%	0, 0%	0, 0%	0, 0%		0, 0%	0, 0%		0, 0%	0, 0%
Guillamó et al. 61	10	8	0, 0%	0, 0%	0, 0%	0, 0%		0, 0%	0, 0%		0, 0%	0, 0%
Gunn et al. 62	26	30	0, 0%	0, 0%	1, 4%	2, 7%	C: Anxiety E: 2 Anxiety	0, 0%	0, 0%		0, 0%	1, 3%	E: Fall with hospitalisation
Hansen et al. 63	13	23	0, 0%	0, 0%	1, 8%	7, 30%	C: Illness E: 7 Illness/low motivation	0, 0%	0, 0%		0, 0%	0, 0%
Heine et al. 64	46	43	1, 2%	0, 0%	1, 2%	2, 5%	C: Comorbidity E: Comorbidity, 2 low mood	0, 0%	1, 2%	E: Training fatigue	0, 0%	0, 0%
Heinrich et al. 65	27	25	0, 0%	1, 4%	0, 0%	1, 4%	E: Shoulder pain	0, 0%	0, 0%		0, 0%	2, 8%	E: 2 Hospitalisation
Hoang et al. 66	22	28	0, 0%	1, 4%	1, 5%	0, 0%	C: Health reasons E: Exercise intervention group	0, 0%	0, 0%		0, 0%	0, 0%
Kooshiar et al. 67	20	20	0, 0%	1, 5%	0, 0%	0, 0%		0, 0%	0, 0%		0, 0%	0, 0%
Learmonth et al. 68	28	29	0, 0%	0, 0%	0, 0%	6, 21%	E: Snowboarding injury, fall, 2 increased fatigue, vision problems, drop foot	0, 0%	0, 0%		0, 0%	0, 0%
Louie et al. 69	11	12	1, 9%	1, 8%	1, 9%	1, 8%	C: Tooth abscess E: Pain	0, 0%	0, 0%		1, 9%	1, 8%	C: Fractured vertebrae E: Fractured toes
Lutz et al. 48	6	8	0, 0%	0, 0%	0, 0%	0, 0%		0, 0%	0, 0%		0, 0%	0, 0%
Magnani et al. 70	12	13	2, 17%	1, 8%	0, 0%	0, 0%		NR	NR		0, 0%	1, 8%	E: Car accident
McAuley et al. 71	24	24	0, 0%	0, 0%	0, 0%	1, 4%	E: Illness	NR	NR		0, 0%	0, 0%
Moradi et al. 72	10	8	0, 0%	0, 0%	NR	NR		NR	NR		NR	NR
Negaresh et al. 73	30	36	0, 0%	1, 3%	0, 0%	0, 0%		0, 0%	0, 0%		0, 0%	0, 0%
Ozkul et al. 74	11	12	0, 0%	0, 0%	0, 0%	0, 0%		0, 0%	0, 0%		0, 0%	0, 0%
Ozkul et al. 75	17	17	0, 0%	0, 0%	0, 0%	0, 0%		0, 0%	0, 0%		0, 0%	0, 0%
Pau et al. 76	11	11	0, 0%	0, 0%	NR	NR		NR	NR		NR	NR
Riemenschneider et al. 77	42	42	0, 0%	4, 10%	0, 0%	2, 5%	E: 2 Spinal disc herniation	0, 0%	0, 0%		0, 0%	0, 0%
Sangelaji et al. 78	30	42	1, 3%	1, 2%	0, 0%	1, 2%	E: Muscle pain	NR	NR		0, 0%	0, 0%
Savšek et al. 79	14	14	6, 43%	2, 14%	0, 0%	0, 0%		0, 0%	0, 0%		0, 0%	0, 0%
Sosnoff et al. 80	14	13	0, 0%	1, 8%	0, 0%	0, 0%		0, 0%	0, 0%		1, 7%	0, 0%	C: Fall with hospitalisation
Sosnoff et al. 81	9	8	0, 0%	0, 0%	2, 22%	3, 38%	C: Surgery, fall E: 3 Falls	0, 0%	0, 0%		0, 0%	0, 0%
Sosnoff et al. 81	9	11	0, 0%	0, 0%	2, 22%	1, 9%	C: Surgery, fall E: Fall	0, 0%	0, 0%		0, 0%	0, 0%
Straudi et al. 82	12	12	0, 0%	0, 0%	0, 0%	0, 0%		0, 0%	0, 0%		0, 0%	0, 0%
Straudi et al. 83	18	18	2, 11%	1, 6%	0, 0%	0, 0%		0, 0%	0, 0%		0, 0%	0, 0%
Wens et al. 84	11	12	0, 0%	0, 0%	0, 0%	0, 0%		NR	NR		0, 0%	0, 0%
Wens et al. 84	11	11	0, 0%	0, 0%	0, 0%	0, 0%		NR	NR		0, 0%	0, 0%
Wens et al. 16	7	15	0, 0%	0, 0%	0, 0%	0, 0%		NR	NR		0, 0%	0, 0%
Young et al. 85	28	27	0, 0%	0, 0%	0, 0%	1, 4%	E: Muscle strain	0, 0%	0, 0%		0, 0%	1, 4%	E: Stress fracture
Young et al. 85	28	26	0, 0%	0, 0%	0, 0%	0, 0%		0, 0%	1, 4%	E: Muscle strain	0, 0%	1, 4%	E: Stroke

E: exercise intervention group; C: control group; NR: not reported; RT: respiratory tract; Con/C: control; Ex/E: exercise; NR: no results.

Meta-analyses of RRs

Overall, 125 effects were derived from 40 RCTs; 46, 40 and 39 effects were derived for relapse, AEs, adverse effects and SAEs, respectively. Forest plots for relapse, AE and SAE are illustrated in Figures 2–4, respectively. The number of relapses, AE, adverse effects and SAE per clinical and exercise delivery moderator is shown in Supplemental Tables 4–6, respectively.

Figure 2.

Forest plot of relapse.

Figure 3.

Forest plot of adverse events.

Figure 4.

Forest plot of serious adverse events.

Of 46 effects provided for relapse, 20 RR were < 1, suggesting lower risk of relapse for exercise training; 13 RR were > 1, suggesting higher risk of relapse for exercise training. However, the 95% CIs for all 46 effects encompassed 1; therefore, statistical significance cannot be inferred. The mean effect ∆ = 0.95 was not statistically significant (95% CI: 0.61, 1.48; z = −0.22, p > 0.82); the effect was not heterogeneous (Q₄₅ = 8.66, p = 1.00; I ² = 0%), and sampling error accounted for 100% of observed variance.

Of the 40 effects provided for AE, 12 RR were < 1, suggesting lower risk of AE for exercise training; 19 RR were > 1, suggesting higher risk of AE for exercise training. However, the 95% CIs for all 40 effects encompassed 1; therefore, statistical significance cannot be inferred. The mean effect ∆ = 1.40 was not statistically significant (95% CI: 0.90, 2.19; z = 1.48, p > 0.14); the effect was not heterogeneous (Q₃₉ = 13.58, p = 1.00; I ² = 0%), and sampling error accounted for 100% of observed variance.

Of 39 effects provided for SAE, 17 RR were < 1, suggesting lower risk of SAE for exercise training; 9 RR were > 1, suggesting higher risk of SAE for exercise training. However, the 95% CIs for all 39 effects encompassed 1; therefore, statistical significance cannot be inferred. The mean effect ∆ = 1.05 was not statistically significant (95% CI: 0.62, 1.80; z = 0.19, p > 0.85); the effect was not heterogeneous (Q₃₈ = 3.85, p = 1.00; I ² = 0%), and sampling error accounted for 100% of observed variance.

Discussion

This study quantified the rate of relapse, AE, adverse effects and SAE in RCTs of exercise training in persons with MS published between November 2013 and April 2022; since our original review. ²⁵ The exercise training in this review comprised studies of aerobic, strength or neuromotor exercise and included both supervised and unsupervised exercise. We identified that in studies where safety data are reported that exercise was safe for persons with MS based on the following: (1) there were no serious adverse effects reported from exercise training and (2) there were no higher risks of relapse, AEs or SAEs in participants who engaged in exercise training compared with participants in control conditions.

The results of our meta-analyses did not reveal any significant variability in risk of exercise training across the potentially important factors of exercise type, delivery style (e.g. supervised, independent or remotely supervised), participant disability level or the prescription of exercise consistent with minimal exercise guidelines for persons with MS.^7,18 This finding is important and supports the current evidence-based guidelines of exercise for all persons with MS across the disease trajectory. ¹ Together, the confirmed safety of exercise and available guidelines should provide confidence for clinicians, health advocacy organisations and governments interested in promoting exercise among all persons with MS.

When reviewing articles for inclusion in our study, many were excluded for not clearly reporting safety data. Based on the unclear safety reporting, our study might not capture all safety events occurring in RCTs of exercise training in MS. Our past systematic review ²⁵ and studies providing a narrative description of relevant safety events in MS exercise studies^23,24 have all advocated that researchers include transparent and consistent reporting of safety outcomes. Here, we direct the field towards clarity in the definitions and reporting of relapse, AE, adverse effects SAE and serious adverse effects and would encourage consideration of the applied definitions as described in Table 1.^21,22,86 We recommend that researchers enable strategies to address the dearth of safety-related outcomes in MS research.^25,87 Researchers should define these events in study protocols, report all events in study reporting, even if unanticipated or deemed unrelated to the exercise itself, and clearly report if no events occurred. ⁸⁷ To facilitate this, we recommend researchers follow a plan to monitor, record and vet all relapses and AE and SAE during exercise studies. Methodological quality was high for almost all included RCTs, indicating a low risk of study bias. Our inclusion criteria for reporting on safety outcomes may be associated with studies of higher methodological quality. The identified high study quality brings confidence to our results, and we acknowledge our recommendation advocating for exercise promotion as safe is associated with exercise interventions delivered with high-quality methodology.

One limitation of our study was the lack of heterogeneity across effects that did not permit an analysis of the influence of clinical characteristics or exercise prescription characteristics. Another limitation is that the RCTs often included screening for risk factors for AE with exercise training and those presenting with elevated risk profiles were excluded from the RCTs. This might obfuscate the actual safety profile and make our results only applicable under conditions wherein participants have low risk of contraindications. Since our search was completed we are aware of only two new articles reporting study results of exercise: telehealth supervised pilates or yoga training ⁸⁸ and in-person supervised strength-training, ⁸⁹ respectively, in MS using a randomised controlled design and reporting on safety outcomes.^89,90 Three AEs, three respiratory infections were reported in the intervention groups of the pilates or yoga-based study, with no events in the control group, while no AEs were reported in the strength-training study group. These results would not change our overall conclusions.

Conclusion

In exercise studies involving persons with MS where safety outcomes were reported, we identified that aerobic, strength or neuromotor exercise performed under both supervised and unsupervised settings was safe for persons with MS based on the following: (1) there were no serious adverse effects reported from exercise training and (2) there were no higher risks of relapse, AE or SAE in participants who took part in the exercise training interventions compared with participants who took part as control participants.