Immediate and Sustained Effects of Interventions for Changing Physical Activity in People with Multiple Sclerosis: Meta-analysis of Randomized Controlled Trials

Abstract

Objectives:

To examine the immediate and sustained effects of interventions for changing physical activity behavior in people with multiple sclerosis (MS), and to explore factors that might moderate intervention effects on physical activity behavior (eg, intervention type and duration, type of physical activity measurement, intensity of theory integration [degree of theory used in study design], and study quality).

Data Sources:

Systematic searches were conducted in 4 databases, including MEDLINE, CINAHL, PsychINFO, and Google Scholar, in October 2017 and October 2018. Updated searches were conducted in September 2019 with 2 additional databases (Embase and Scopus) and enhanced search terms.

Study Selection:

Studies were included that (1) incorporated a randomized controlled trial design of interventions that targeted change in physical activity behavior in adults with MS, namely, exercise training and behavioral intervention (alone and combined); (2) included self-reported and/or device-measured physical activity as an outcome; and (3) contained pre- and post-intervention assessments.

Data Extraction:

Data were extracted for immediate (pre- to post-intervention) and sustained (pre-intervention to follow-up) physical activity outcomes and study characteristics. Weighted mean effect sizes were expressed as standardized mean differences (SMD). Heterogeneity between each categorical moderator was compared using Q between statistics.

Data Synthesis:

The mean SMD was 0.56 for immediate changes (n = 24) and 0.53 for sustained changes (n = 7) of physical activity outcomes. Self-reported physical activity measures yielded larger effects (SMD, 0.64; n = 22) than those of device-measured physical activity (0.26; n=7). There appeared to be larger immediate effects of behavioral interventions (SMD, 0.71; n=9) than exercise training (SMD, 0.53; n=7) and combined interventions (0.37; n=8).

Conclusions:

Current evidence demonstrates that interventions are efficacious for increasing and potentially sustaining physical activity behavior in adults with MS. The effects appear to be optimized based on the delivery of behavioral interventions alone, and these interventions may be capable of supporting long-term behavior change.

Multiple sclerosis (MS) is a prevalent, immune-mediated disease of the central nervous system that results in demyelination and transection of axons in the brain, brain stem, optic nerves, and spinal cord.¹ The extent and location of damage within the central nervous system result in heterogeneous outcomes, including walking and cognitive dysfunction and symptomatic fatigue, pain, and depression.² MS and its consequences can further compromise participation in activities of daily living and health-related quality of life.^3,4

Participation in physical activity, including exercise training, has gained acceptance as a nonpharmaceutical, behavioral approach for managing or alleviating many consequences of MS.^5,6 This is based on meta-analyses indicating beneficial effects of physical activity on walking,⁷ fatigue,⁸ depression,^9,10 and indices of quality of life. ¹¹ There may be additional benefits of physical activity for modifying the disease itself¹² as well as reducing the rate of relapses ¹³ and cardiovascular comorbidities. ¹⁴

Nevertheless, persons with MS engage in less physical activity than those from the general population.^15–17 This is supported by a meta-analysis indicating a lower level of physical activity participation in people with MS than adults without conditions or disorders (mean effect size, −1.00).¹⁷ One study has demonstrated that fewer than 20% of people with MS meet the recommended guideline for moderate-to-vigorous physical activity.¹⁶ These data indicate that the majority of people with MS do not engage in sufficient levels of physical activity necessary for accruing benefits of this health behavior.

The persistent problem of physical inactivity in persons with MS has been addressed through the delivery of exercise training programs or behavioral interventions (alone and combined).¹⁸ The behavioral interventions, in particular, might enhance the effects of interventions on long-term physical activity behavior (ie, sustainability).^19–21 These interventions typically teach behavior change techniques, such as self-monitoring, goal-setting, and feedback, that align with thoery.²² To date, meta-analyses have reported that behavioral interventions could increase physical activity behavior,^23.24 but the immediate and, in particular, sustained effects were estimated from a small number of carefully selected studies. This hindered subanalyses of potentially influential intervention characteristics. Accordingly, there are several unknown features regarding the effects of interventions on changes in physical activity. For example, are behavioral interventions alone sufficient to elicit immediate and sustained changes in physical activity behavior, or do behavioral interventions need to be supplemented with exercise training? Does the degree of integrating a behavioral change theory within an intervention (ie, intensity of theory integration) moderate changes in physical activity behavior? The idea of theory integration involves the degree of theory wied in the intervention design, delivery, and evaluation, and can range from sparse through extensive.²⁵ To that end, performing a meta-analysis with carefully selected moderators can help interventionists identify optimal strategies for increasing and sustaining physical activity behavior in people with MS.

We undertook a meta-analysis of randomized controlled trials (RCTs) that provided a quantitative synthesis of the immediate and sustained effects of exercise training programs and behavioral interventions (alone and combined) for changing physical activity in people with MS. The meta-analysis focused on 2 separate time-phases, namely post-intervention (immediate effects) and follow-up (sustained effects). We further explored factors that might moderate intervention effects on physical activity behavior, including intervention characteristics (type and duration), type of physical activity measurement, intensity of theory integration, and study quality.

Methods

This meta-analysis followed the Preferred Items for Systematic Reviews and Meta-Analyses statement.²⁶ Using systematic review procedures, intervention studies that reported changes of physical activity behaviors in people with MS were identified, reviewed, and synthesized from 6 electronic databases: MEDLNE, CINAHL, PsychNFO, Embase, Scopus, and Google Scholar. The protocol registration was not performed before commencement.

Search strategy

We conducted initial searches of 4 electronic databases (ie, MEDLINE, CINAHL, PsychINFO, and Google Scholar) on October 3, 2017, from the period of July 1963 (inception date) through September 2017. The searches were updated on October 31, 2018 (until September 2018). Three categories of search terms were used, namely, interventions (eg, exercise, physical activity, or behavicr), outcomes (eg, health behavior or physical activity), and disability (multiple sclerosis or MS). We conducted an updated search on September 19, 2019 (until September 2019) to identify papers and reports that could have potentially been published during the preparation of this article for submission. The sean:hes were enhanced with 2 additional databases (ie, Embase and Scopus) and a broader list of searching terms: interventions (eg, exercise, physical activity, behavior therapy, or health promotion), outcomes (eg, health behavior, physical activity, or accelerometry/actigraphy), and disability (multiple sclerosis or MS). Examples of the MEDLINE searches are shown in supplemental appendix Sl (available online only at http://www.archives-pmr.org/ ).

Eligibility criteria

Studies were deemed eligible based on the following criteria: (1) included adults with MS (≥18y); (2) incorporated RCT designs of interventions that targeted change in physical activity behavior (structured exercise training, behavioral intervention, or both); (3) included self-reported and/or device-measured physical activity; (4) contained a pre- and post-assessment period; and (5) published in English in peer-reviewed journals.

Studies were excluded based on the following criteria: (1) non-research publications (ie, conference presentations, study protocol, dissertations); (2) therapeutic and/or pharmaceutical interventions that required the assistance of a licensed therapist or devices (eg, robotic/body-weight support gait training, constrain-induced movement therapy, functional electrical stimulation); and (3) insufficient information for calculating effect size (ES) regarding physical activity outcome measures. Self-reported outcomes that included a broad aspect of health and well-being were only included when the subcategory of physical activity score was reported separately (eg, Health-Promoting Lifestyle Profile Il). Device-measured physical activity outcomes that were obtained within the intervention itself (ie, session attendance) were not included in this meta-analysis, as we were interested in behavior change rather than compliance.

Screening process and data extraction

The search terms were developed through interactions with a librarian and further resulted in electronic search strings per database. We refined the search criteria for generating a broader search and better representing the literature, thereby yielding a more reliable estimate of the overall effect of interventions on physical activity change. The primary analyst (Y.K.) conducted searches of the electronic databases and narrowed the search results based on the eligibility criteria of this review (eg, human subject research, RCTs). After retrieving the studies, two analysts (primary and secondary [B.L.]) independently performed the screening process. This process included the following steps: (1) removed duplicate studies; (2) screened all studies at the abstract level; (3) reviewed the remaining studies in full-text level; (4) assessed the methodological quality of included studies; and (5) evaluated the intensity of theory integration of intervention design (if applicable). Disagreements were resolved by a senior author (R.M.). The senior author independently assessed a study and determined the final decision for inclusion or exclusion of the study in die meta-analysis and of theory integration.

Data extraction processes were performed by the primary analyst, and the extracted data were cross-checked for accuracy by the secondary analyst. Data were organized into 2 spreadsheets: (1) participant and intervention characteristics using the Population, Intervention Comparator, Outcome, Timing, Setting framework and (2) physical activity outcomes. The participant characteristics included age, sex, type of MS, disease severity, and time since diagnosis. The intervention characteristics consisted of program and training prescription (frequency, intensity, time, and type), duration of the intervention (weeks from pre- to post-assessment) and follow-up (weeks from post- to follow-up assessment), and name of behavior change theories applied to frame and deliver the intervention. The physical activity outcome included name and type of measure (self-reported, device-measured), sample size (pre- to post-assessment, pre-assessment to follow-up), mean and SD of pre- and post-assessment, and any follow-up data for both intervention and control goups.

Moderator variables

Intervention type

Intervention type was categorized into 3 levels: behavioral intervention, exercise training, and combined. Behavioral interventions weæ operationally defined as based on inclusion of behavior change techniques informed by theory for changing behavior.²² Exercise training studies were defined as the delivery of structured and planned physical training. Studies that used both behavioral intervention and exercise training categorized as combined. We explored the intervention type as a moderator of possible variability in the effect of intervention types on change in physical activity.

Intervention duration

Intervention duration was categorized into 2 levels: interventions of 12 weeks or less (≤12wk) and more than 12 weeks (>12wk). The levels were determined based on moderator analyses of a previous meta-analysis.²⁴

Type of physical activity measurement

Type of physical activity measurement was categorized into 2 levels: self-reported (questionnaire) and device-measured physical activity (eg, accelerometry). We explored the type of physical activity measurement as a moderator to see whether or not the magnitude of physical activity behavior change differs based on the presumption of larger effects with self-reported than device-measured physical activity.

Intensity of theory integration

Theory integration informs the degree of theory included in the study design, delivery, and evaluation (ie, sparse through extensive). The intensity of theory integration was evaluated using a modified version of the Theory Coding Scheme. ^25,27 The modified version differs from the original in that it emphasizes the identification of theory, constructs, and specific methods related to theory measurement and evaluation in the intervention design, and excludes non-relevant items (eg, evaluation of how the study results might be used for refining theories). The coding scheme was previously applied to evaluate the magnitude of the theory application of physical activity interventions among breast cancer survivors.²⁵ The coding consisted of 8 items, rated as present or absent of application, with a range of scores between O and 8. The intensity of theory application was classified and interpreted as followings: level 1 (sparse) if 3 or fewer items were satisfied, level 2 (moderate) if 4 to 5 items were satisfied, or level 3 (extensive) if 6 or more items were satisfied. Higher scores reflect greater inclusion of theory in the design, delivery, and evaluation of a study and its effects on physical activity behavior change. We explored the intensity of theory integration as a moderator based on the presumption that more intense inclusion of theory in informing the intervention would yield larger changes in physical activity.

Study quality

The methodological quality of each study was assessed using the Physiotherapy Evidence Database (PEDro) scale.²⁸ The PEDro scale has a maximum possible score of 10 points. Two items (blinding of therapists/trainers and blinding of subjects) were considered as not applicable in rehabilitation research when comparing an intervention group with a nontraining control group. Therefore, these 2 items of each study were credited. ^29,30 A higher score indicates better methodological quality. The methodological quality of each study was then categorized into 2 levels using the Spinal Cord Injury Rehabilitation Evidence system. ³¹ This is a 5-level system that distinguishes between studies of differing quality and incorporates the types of research designs commonly used in rehabilitation research (supplemental appendix S2, available online only at http://www.archives-pmr.ory). This system has been applied in several systematic reviews and meta-analyses in the field of MS reseamh.^30,32.34 The level of the evidence within RCTs was interpreted as level 1 if the PEDro score was greater than 6 and level 2 when the score was 6 or lower. ³¹ We added study quality as a moderator to explore whether or not the quality of studies influence the effects of the interventions For example, low-quality studies may yield larger effects with high chances of bias than high-quality studies.

Data analysis

The mean ES was computed using a random effect model based on the assumption that the samples of the selected studies represent the population and that the true effects differ between studies³⁵ and adjusted by sample size using the Comprehensive Meta-analysis Software.^a The weighted mean ESs were expressed as standardized mean differences (SMD), which is often referred to as Cohen’s d. SMD was calculated based on the mean change of physical activity outcomes from before and after the intervention minus the mean change of the control group. The mean changes between the 2 groups were then divided by pooled change score SD of intervention and control groups. A positive SMD indicated an improvement in physical activity behavior after intervention (favors intervention), whereas a negative SMD represented a worsening of physical activity behavior in the intervention group compared to controls (favors control). The ES was interpreted according to Cohen’s benchmarks³⁶: small, SMD of 0.20; medium, SMD of 0.50; and large, SMD of 0.80. We further computed a 95% confidence interval around the mean ES. Confidence intervals excluding zero were considered statistically significant The interpretation of analyses accounted for statistical significance as well as clinical meaningfu1ness based on the guideline of ½. SD (ie, Cohen’s d of 0.5); this guideline of ½ SD has been deemed as a universal threshold for judging effects as clinically meaningful.³⁷

When a study used 2 intervention groups (2 different types of exercise) and 1 control group, we created a mean ES in that study by averaging the ESs estimated from these groups. This process included the following steps: (1) assign subgroups (ie, exercise 1 and exercise 2) within the study; (2) compute data (ie, sample size, mean, SD) of each subgroup; (3) compute data of control group more than once; (4) create 2 ESs estimated from each subgroup (ie, the mean ESs of exercise 1 vs control and of exercise B vs control); and (5) average the ESs estimated from these groups and then create a mean ES in that study. The variance of the averaged ESs was corrected by the Comprehensive Meta-analysis Software to account for the doubled sample size of the control group. ^35,38 When a study had multiple physical activity outcomes within a study, based on the same participants, we created a mean ES per study by averaging the ESs estimated from these outcomes. This process included the following steps: (1) compute data of each outcome; (2) create multiple ESs estimated from each outcome (eg, accelerometry, Godin Leisure-Time Exercise Questionnaire); and (3) average the ESs estimated from these outcomes and then create a mean ES in that study. The multiple ESs from the same study are assumed to be not independent among the different groups and outcomes.^35,38

The analyses were conducted separately for the 2 time phases: (1) inimediate (pre-post) and (2) sustained (pre-follow-up) physical activity outcomes. We further performed moderator analyses of each time phase for the physical activity outcomes. The moderators were the intervention type (behavioral intervention vs exercise training vs combined), the intervention duration (≤12wk vs > 12wk), the type of physical activity measurement (self-reported vs device-measured), the intensity of theory integration (level 1 vs level 2 vs level 3), and study quality (level 1 vs level 2).

To estimate heterogeneity in ESs, we used an I² statistic. We report the Q statistic as a test of heterogeneity with the caveat that it may have low power to detect heterogeneity and is dependent on the number of studies.³⁹ I² represents the percentage of variability in observed results caused by heterogeneity rather than sampling error.³⁹ I² values of 25%, 50%, and 75% represent low, medium, and high heterogeneity, respectively.⁴⁰ When substantial heterogeneity was observed (determined by I² value >50%), we further explored the reasons for heterogeneity using the visual inspection forest plots and study characteristics. To estimate heterogeneity in ESs between each categorical moderator, we used the Q between statistic (Q^B). Statistically significant Q^B indicates that we can reject the null hypothesis of no moderating effect in physical activity outcome (ie, effect size variability can be explained by the categorical moderator variables). ³⁵

Results

Study description

The results of the study selection process are provided in a Preferred Items for Systematic Reviews md Meta-Analyses flowchart in figure 1. The search strategy returned 2851 studies from the electronic databases. After removing duplicates, 2479 studies were screened at the abstract level based on the eligibility criteria, and 229 studies were retained. The remaining studies were reviewed at the full-text level. This yielded a total of 24 studies^41–64 that met the eligibility criteria and provided sufficient data on physical activity to compute an ES (ie, sample size, means, SDS). Of those, 9 studies were located through additional searches: 4 from the external searches (October 2017) and 5 from die updated searches (September 2019).

Fig 1.

Preferred Items for Systematic Reviews and Meta-Analyses flowchart.

Among the 24 studies in the immediate analyses, 9 studies (38%) included at least 1 follow-up point. Of those, 2 studies^61,62 did not provide follow-up data of groups because the waitlist controls received intervention upon the completion of the post-intervention measures. Therefore, only 7 studies^{42–44,54,56,59,63} were included in the sustained analyses. The detailed characteristics of the studies included in the quantitative synthesis are presented in table 1 using the Population, Intervention, Comparator, Outcome, Timing, Setting framework.

Table 1.

Characteristics of the studies included in the quantitative synthesis (n=24) reported using the PICOTS framework

		Population				Timing
Study	Croup	Sample Size: n Age (y), Mean ± SD Sex (M, F): n MS Type (RR), n (%) Disability Level (EDSS or PDDS), Mean ± SD MS Duration (y), Mean ± SD	Intervention and Comparator	Physical Activity Outcome	Length of INT (wk)	length of Follow-up (Wk)
Bombardier et al41	INT	n=44 Age: 47.1±8.9 Sex: 5, 39 RR: 32 (76) EDSS: ≤5.5 MS duration: 9.4±7.1	Behavioral coaching using MI • l initial face-to-face session at a clinic (60 min) • 7 scheduled telephone calls (30 min/session)• 1 final face-to-face session at a clinic (60 min)	7-dsy Physical Activity Recall	12	-
	COM	n=48 Age: 49.7±7.9 Sex: 8, 40 RR: 36 (75) EDSS: 5.5 or less  MS duration: 11.7±7.9	Waitlist control
Carter etal42	INT	n=16 Age: 33.5±6.5 Sex: 2,14 RR: ns EDSS: 3.0±1.1 MS duration: ns	Exercise training (flexibility, aerobic, strengthening, balance) Incorporating behavioral techniques into the exercise sessions using the TTM • 2 times/wk 1-on-l, supervised training (60 min/session) • Aerobic exercise targeted 50%–69% of APMHR and RPE 11–13• 1 time/wk unsupervised, home exercise	GLTEQ	10	12
	COM	n=14 Age: 40.9±8.7 Sex: 2, 12 RR: ns EDSS: 3.1±1.7 MS duration: ns	Usual care
Coote et al43	INT	n = 33 Age: 43.3±9.9 Sex: 4, 29 RR: 27 (82) EDSS: 3.3±0.7 MS duration: 6.7±5.7	Exercise training (aerobic, strengthening) • 6 supervised, group exercise and coaching sessions (75–90 min/session) • Aerobic exercise targeted moderate intensity • Strengthening exercise targeted lset of 10–15 reps and then progressed to 2 sets of 8–12 reps Behavioral coaching based on SCT • 4 times via telephone	GLTEQ and Sense wear armband	10	12, 24
	COM	n=32 Age: 41.9±9.3 Sex: 6, 26 RR: 27 (84) EDSS: 3.3±0.7 MS duration: 7.0±6.l	Attention control Exercise training (aerobic, strengthening) Education sessions unrelated to PA behavior (eg, diet, sleep, temperature and hydration, immunizations)
Dlugonski et al44	INT	n=22 Age: 48.5±10.1 Sex: 4, 18 RR: 22 (100) PDDS: 1.0 (0–6) (median, range) MS duration: 10.3±9.2	Behavioral intervention; contents (text-based and video files) are developed based on SCT • New contents and video coaching calls (5–10 min) were available 4 time for 1st month, 2 times for 2nd month, and 1 time for 3rd month • Participants were encouraged to wear a pedometer for recording daily steps and self-monitoring purpose • Home; internet	GLTEQ	12	12
	COM	n=23 Age: 44.8±9.1 Sex: 2, 21 RR: 23 (100) PDDS: 1.0 (0–6) (median, range) MS duration: 8.5±6.2	Waitlist control
Duff et al45	INT	n=15 Age: 45.7±9.4 Sex: 3, 12 RR: 14 (93) PDDS: 2.1 ±1.8 MS duration: ns	Exercise training (PiLates) and massage after each PiLates session • Pilates: 50 min, 2 times/wk • Massage: 60 min, 2 times/wk • Clinic, group exercise	Accelerometer	12	-
	COM	n=15 Age: 45.1±7.4 Sex: 4,11 RR: 11 (73) PDDS: 2.3±2.3 MS duration: ns	Massage: 60 min, 2 times/wk
Ennis et al46	INT	n=31 Age: 45±9 Sex: ll, 20 RR: 16 (50) EDSS: 0–3 (22%); 3.5–6 (69%); 6.5–7 (9%) MS duration: 7±5	Multidisciplinary health promotion education based on self-efficacy belief (exercise and physical activity, fatigue and stress management, nutritional awareness) • 180 min, 1 time/week • Hospital; group	HPLP II (physical activity subscale)	8	-
	COM	n=30 Age: 46±8 Sex: 11,19 RR: 12 (40) EDSS: 0–3 (23%); 3.5–6 (74%); 6.5–7 (3%) MS duration: 8±6	Waitlist control
Hayes et al47	INT	n=33 Age: 43.3±9.9 Sex: 4, 29 RR: 27 (82) EDSS: 3.3±0.7 MS duration: 6.7±5.7	Same as Coote et al43	GLTEQ	10	-
	COM	n=32 Age: 41.9±9.3 Sex: 6, 26 RR: 27 (84) EDSS: 3.3±0.7 MS duration: 7.0±6.1		Same as Coote et al43
Learmonth et al48	INT	n=20 Age: 51.4±8.06 Sex: 5, 15 RR: ns EDSS: 6.14±0.36 MS duration: 13.4±6.4	Exercise training (aerobic, strengthening, balance) using a circuit training approach • 60 min, 2 times/wk Leisure center; supervised, group	Phone FITT	12	-
	COM	n=12 Age: 51.8±8.0 Sex: 4, 8 RR: ns EDSS: 5.82±0.51 MS duration: 12.6±8.1	Usual care
Learmonth et al49	INT	n=29 Age: 48.7±10.4 Sex: 1, 28 RR: 26 (90) EDSS: 1.25 (2.5) (median, IQR) MS duration: 14.8±8.5	Exercise training (aerobic and strengthening) • Aerobic exercise (walking) targeted moderator intensity, 10–30 min, 2 times/wk • Pedometer for monitoring and tracking • Strengthening exercise targeted 1–2 sets of 10–15 repetitions for 10 exercises, 2 times/wk • Home; DVD Behavioral coaching to monitor progression and discuss newsletter contents, which is developed based on SCT • 6 times for 12 wk • Internet; one-on-one coaching	GLTEQ and accelerometer	16	-
	COM	n=28 Age: 48.2±9.1 Sex: 1, 27 RR: 25 (90) EDSS: 2 (3) (median, IQR) MS duration: 13.0±7.7	Waitlist controL
McAuley et al50	INT	n=24 Age: 59.62±1.43 Sex: 6,18 RR: 16 (66.7) EDSS: ns MS duration: 18.10±9.42	Exerdse training (flexibility, strengthening, balance) • 3 times/wk • Strengthening training targeted 1–2 sets of 8–10 repetitions (RPE 10–12) and then progress to 2 sets of 10–12 repetitions (RPE 13–15) • Home; DVD	GLTEQ	24	-
	COM	n=24 Age: 59.78±1.50 Sex: 6, 18 RR: 16 (66.7) EDSS: ns MS duration: 59.78±l.50	Attention control • Watching Healthy Aging documentaiy (85 min) DVD, home
Mostert and Kesselring51	INT	n=13 Age: 45.23±8.66 Sex: 3,10 RR: 4 (30.8) EDSS: 4.6±1.2 MS duration: 11.2±8.5	Exercise training (aerobic) • 30 min, 5 times per 2 wk using a bicycle ergometer • Rehabilitation center; supervised	BAECKE (wortc, sport, leisure)	4	-
	COM	n=l3 Age: 43.92±13.90 Sex: 2, 11 RR: 5 (38.5) EDSS: 4.5±1.9 MS duration: 12.6±8.1	Usual care
Metl et al52	INT	n=23 Age: 46.1±10.4 Sex: 2, 21 RR: 23 (100) PDDS: 2.0±1.8 MS duration: 8.l±6.5	Same as Dlugonski et al44	GLTEG	12	-
	COM	n=25 Age: 45.6±9.2 Sex: 2, 22 RR: 25 (100) PDDS: 2.1±1.9 MS duration: 7.3±6.2	Waitlist control
Motl et al53	INT	n=23 Age: 52.3±10.3 Sex: 2, 21 RR: 20 (87.0) EDSS: 3.5 (1.5) (median, IQR) MS duration: 14.4±10.4	Behavioral intervention; contents (test-based and video files) are develop based on SCT • New contents and video coaching rails were available 7 times for 1st 2 mo, 4 times For 2nd 2 mo, and 2 times for 3rd 2 mo • Graphical goal tracking • Home; internet	GLTEQ and accelerometer	24	-
	COM	n=24 Age: 51.4±7.4 Sex: 5, 19 RR: 21 (87.5) EDSS: 3.5 (2.0) (median, IQR) MS duration: 21.1±8.7	Waitlist control
Paul et al54	INT	n=45 Age: 55.6±10.2 Sex: 13, 32 RR: 15 (33) EDSS: 5.0 (median) MS duration: 10 (12) (median, IQR)	Exercise training (aerobic, strengthening, balance) • 2 times/wit • Exercise (videos, text, and audio descriptions) and educational contents (disease-specific advice) were delivered via a website, Alterations of exercise based on level and comments • Home, internet	Accelerometer	24	12
	COM	n=45 Age: 56.5±9.1 Sex: 8, 37 RR: 15 (33) EDSS: 6.0 (median) MS duration: 15 (13) (median, IQR)	Attention control • Printed sheet of exercise program • 2 times/wk, home
Pilutti et aL55	INT	n=4l Age: 48.4±9.1 Sex: 11, 30 RR: 31 (75.6) EDSS: 2.0 (4.0) (median, IQR) MS duration: 10.6±7.1	Same as Motl et al53	GLTEQ and accelerometer	24	-
	COM	n=41 Age: 49.5±9.2 Sex: 9, 32 RR: 34 (82.9) EDSS: 3.0 (3.0) (median, IQR) MS duration: 13.0±9.l	Waitlist control
Plow et al56	INT	n=14 Age: 47±9 Sex: 0,14 RR: 14 (100) PDDS; 1.79+1.72 MS duration: 8±7	Behavioral intervention; contents (customized pamphlets) were developed based on SCT and TTM • 1 pamphlet for eveiy 3 weeks • Home; mailout Exercise training • 2 times in-person, to prescribe an individualized home exercise program	GLTEQ and PADS	12	12
	COM	n=16 Age: 48±10 Sex: 0, 16 RR: 16 (100) PDDS: 2.69±2.06 MS duration: 10±7	Waitlist control
Rice et al57	INT	n=9 Age:: 53.3±11.1 Sex: 3, 6 RR: 3 (33.3) EDSS/PDDS: ns MS duration: 13.2±8.9	1 time wheelchair skill/technique training using multimedia Behavior coaching based on SCT (home; telephone; 1 time/wk)	Accelerometer	12
	COM	n=5 Age; 54±0.4 Sex: 1,4 RR: 2 (40) EDSS/PDDS: ns MS duration: 17.6±8.5	Waitlist control
Sandroff et al58	INT	n=37 Age 48.8±8.3 Sex: 10, 27 RR: 28 (75.7) EDSS: 0–2 (48.6%); 3–6 (51.4%); 10.7 (S.8) MS duration: ns	Waitlist control	IPAQ	24
	COM	n=39 Aga: 50.3±8.4 Sex: 9, 30 RR: 32 (82.1) EDSS: 0–2 (46.2%); 3–6 (53.8%); 13.4 (9.4) MS duration: ns	Waitlist control	IPAQ	24
Stuifbergen et al 59	INT and COM	n=56 (INT) n=57 (CON) Age: 45.79±10.9 Sex: 0,113 RR: 62 (55) EBSS/PDDS; ns MS duration: 10.76±6.92	Multidisciplinary health promotion education (90 min 1 time) • 90 min, 1 time/wk (dink, group) • Bimonthly phene calls during follow-up period (home, tsbphone) Usual cars control group	HPLP II PA subscale	8	12 and 24
Suh et al60	INT	n=34 Age: 50.1 ±8.1 Sex: ns RR: 33 (97.1) PDDS: 2.0±1,8 MS duration: 11.6±7.1	Behavior coaching based on SCT • Printed newsletters (mail and e-mail) • 1 time/wk, one-on-one telephone coaching • Pedometer and logbook for self-monitoring and motivation purpose	GLTEQ	6	-
	COM	n =34 Age: 48.0±9.4 Sex: ns RR: 33 (97.1) PDDS: 2.2±1.8 MS duration: 12.7±8.8	Attention control • Received educational materials unrelated to physical activity (stress management, nutrition, allergies) • 1 time/wk telephone call to check whether or not the participant received newsletters
Tallner et al61	INT	n=59 Age: 40.9±10.4 Sex: 15, 44 RR: 52 (88.1) EDSS: 2.8±0.8 MS duration: 9.8±9.2	Exercise training (aerobic, strengthening) • Moderate to high intensity (RPE 11–16) • Home-based, supervised via internet Aerobic training (walking, cycling, jogging, swimming) • 10–60 min, 1 time/wk Strengthening training • 2 times/wk, 2–3 sets per exercise Waitlist control	BAECKE (sport)	12	12
	COM	n=67 Age: 40.7±9.5 Sex: 17, 50 RR: 57 (85.1) EDSS: 2.7±0.8 MS duration: 9.2±7.2
Thomas et al62	INT	n=15 Age: 50.9±8.08 Sex: 1,14 RR: 12 (80) EDS5/PDDS: ns MS duration: ns	Gaming intervention using Nintendo Wii • 2 supervised, face-to-face sessions for familiarization at a hospital (wk 1 and 2) • 3 home visit to set up the equipment and risk assessment (wk 3, 7, 16) • 3 telephone/e-mail for monitoring and ongoing support (wk 5, 12, 20) Incorporating behavioral techniques throughout intervention using MI	GLTEQ	24	24
	COM	n=15 Age: 47.6±9.26 Sex: 2,13 RR: 9 (60) EDSS/PDDS: ns MS duration: ns	Waitlist control
Turner et al63	INT	n=31 Age: 52.7±11.6 Sex: ns RR: 19 (65.5) EDSS/PDDS: ns MS duration: 11.33±9.00	Behavioral coaching using MI • 1 time/wk (90 min for 1st session, 30–60 min for remaining sessions) • Home, telephone Exercise training using DVD • Encouraged to perform ≥45 min of high-intensity exercise, 1–2 times/wk	GLTEQ	12	12
	COM	n=33 Age: 53.6±13.1 Sex: ns RR: 23 (69.7) EDSS/PDDS: ns MS duration: 11.85±10.41	Attention control • Self-directed education using DVD information (facilitating motivation, ability to matched peer model to promote self-efficacy, examples) • Home, mail out
Wens et al64	INT	n = 11 Age: 47±9.9 Sex: 5, 6 RR: S (72.7) EDSS: 2.7±1.0 MS duration: ns	Exercise training (aerobic, strengthening), 5 PASIPD time/2 wk High intensity continuous aerobic training (cycle ergometer) • 5 1 min intervals (wk 1–5) with 80%−90”% MHR • 5 2 min intervals (wk 6–12) with 90%.−100% MHR Moderate to high-intensity resistance training • Progressed from 1 set of 10 repetitions to 2 sets of Z0 repetitions for 5 exercises	PASIPD	12
	INT	n=12 Age: 43±10.4 Sex: 5, 7 RR: 10 (83.3) EDSS: 2.3±1.0 MS duration: ns	Exercise training (aerobic, strengthening), 5 times/2 wk High-intensity interval training (cycle and treadmill walking/running) • 1 6 min interval (wk 1–5) with 80%−90% MHR • 2 10 min intervals (wk 6–12) with 80%-90% MHR Moderate- to high-intensity resistance training • Same as high-intensity continuous training group
	COM	n=11 Age: 47±9.9 Sex: 2, 9 RR: 8 (72.7) EDSS: 2.5±1.0 MS duration: ns	Usual care

Abbreviations: APHMR, age-predicted maximum heart rate; BAECKE, Baecke Physical Activity Questionnaire; COM, comparator; EDSS, Expanded Disability Status Scale; FITT, frequency, intensity, time, and type; GLTEQ, Godin Leisure-Time Exercise Questionnaire; HPLP II, Health-Promoting Lifestyle Profile II; INT, intervention; IPAQ, International Physical Activity Questionnaire; IQR, interquartile range (Q3-Q1); MHR, maximum heart rate; MI, motivational interviewing; ns, not specified; PA, physical activity; PADS, Physical Activity and Disability Survey; PASIPD, Physical Activity Scale for Individuals with Physical Disabilities; PDDS, Patient Determined Disease Steps; PICOTS, Population, Intervention, Comparator, Outcome, Timing, Setting; RPE, ratings of perceived exertion; RR, relapsing-remftting; SCI, spinal cord injury; 5CT, social cognitive theoiy; TTM, transtheeretical modeL

Participant characteristics

Summary characteristics of all studies included in the quantitative synthesis are provided in table 2. Overall, 1373 people with MS were included in the studies with a mean age of 48±2 years. The samples predominately consisted of (1018 of 1245, 82%) and relapsing-remiti.ng MS (985 of 1321, 75%); 2 studies did not clarify sex^60,63 or type of MS.^42,48 Studies reported disease severity using the Expanded Disability Status Scale or Patient Determined Disease Steps. The study participants generally with mild-to-moderate disability (eg, Expanded Disability Status Scale scores between 0 and 6.5). Only 1 study included individuals with MS who used a wheelchair.⁵⁷ The mean of MS among the participants was 12±2 years. Several studies did not report either disease severity of participants,^50,59,62,63 or MS duration.^42,45,62,64

Table 2.

Summary characteristic of all studies and studies with a follow-up period included in the quantitative synthesis

	All Studies (n = 24)		Studies With Follow-up (n = 7)
Characteristics	No.	%	No.	%
Total no. of participants
Intervention	901	50	279	49
Control	901	50	286	51
Average no. of participants (mean ± SD)
Intervention	24±11		21±12
Control	24±13		22±13
Type of Intervention
Exercise training	7	29	1	14
Behavioral intervention	9	38	2	29
Combined	8	33	4	57
Length of intervention, wk
Median (range)	12 (4–24)		-
Length of follow-up, wk
Median (range)	-		12 (12–24)
Physical activity outcome
Self-reported	22	67	7	70
Device-measured	7	21	2	20
Both	4	12	1	10
Study quality
PEDro score, median (range)	9 (6–10)		9 (7–10)
Level 1 (>6)	22	84	7	100
Level 2 (≤6)	2	16	0	0
Intensity of theory integration
Median (range)	4 (3–7)		4.5 (3–6)
Level 1 (≥3)	5	29	2	33
Level 2 (4–5)	6	35	2	33
Level 3 (≤6)	6	35	2	33

Intervention characteristics and settings

The interventions consisted of exercise training (7 of 24, 29%),^{45,48,50,51,54,61,64} behavioral interventions (9 of 24, 38%),^{41,44.46,52,53,55,58–60} and combined (8 of 24, 33%).^{42,43,47,49,56,57,62,63} Exercise modality mainly included aerobic (cg, walking, cycling), strengthening (eg, weight lifting), and/or balance training (cg, standing still, walking with objects). Exercise training was either supervised and delivered in a laboratory, clinic, or community center (n = 4, 57%) or unsupervised and delivered via DVD or internet in participants’ homes (n = 3, 43% ). Behavioral and combined interventions were generally framed and delivered based on 1 or more behavior change theories. The behavior change theories that were predominantly used included social cognitive theory (n = 12, 76%),^{43,44,47,49,52,53,55–58,60,62} motivational interviewing (n = 3, 18% ), ^41,62,63 transtheoretical model (n=2, 12%),^42,56 and self-efficacy theory (n=2, 12%).^46,59 Three studies (18%) applied more than 1 theory.^56,59,62

Comparator

Interventions typically included nonactive comparison groups (ie, usual waitlist control). Only 4 studies included active comparison groups (ie, attention control) and delivered either education sessions or materials unrelated to physical activity behavior.^{43,47,50,54,60,63}

Physical activity outcomes

The studies included a variety of physical activity measures. The device-measurement of physical activity included either a waist-worn accelerometer or sense wear armband (7 of 21, 33%).^{43,45,49,53–55,57} The self-report measures of physical activity include: Godin Leisure-Time Exercise Questionnaire (13 of 21, 70% )^{42–44,47,49,50,52,53,55,56,60,62,63}; Baecke Physical Activity Questionnaire (n=2, 10%)^51,61; Health-Promoting Lifestyle Profile II (n=2, 10%)^46,59; International Physical Activity Questionnaire (n= 1, 5%)⁵⁸; Physical Activity and Disability Survey (n=l, 5%)⁵⁶; 7-day Physical Activity Recall (n=l, 5%)⁴¹;Phone-Frequency Intensity Time Type Questionnaire (n= 1, 5%)⁴⁸; Physical Activity Scale for Individuals with Physical Disabilities (n= 1, 5%).⁶⁴ Five studies included more than 1 measure of physical activity.^{43,49,53,55,56}

Timing

The average duration of interventions was 12.4±5.8 weeks (median, 12wk) and ranged between 4 and 24 weeks. The average duration of follow-up was 15.4±5.9 weeks (median, 12wk) and ranged between 12 and 24 weeks. All 7 studies included a 12-week follow-up of physical activity outcomes; 2 studies included an additional 24-week follow-up assessment.

Intensity of theory integration

Both behavioral (n=9) and combined interventions (n=8) were framed and delivered based on 1 or more behavior change theories. Using the modified version of the Theory Coding Scheme,^25,27 5 studies were classified as level I (sparse),^{42,53,55,58,63} 6 as level 2 (moderate),^{41,44,47,56,57,62} and 6 as level 3 (extensive).^{43,46,49,52,59,60} Studies with exercise training alone did not apply a behavior change theory, and these studies were excluded from the moderator analyses. The intensity of theory integration of studies that applied behavior change theory included in the quantitative synthesis is provided in figure 2 and table 3.

Fig 2.

Intensity of theory integration. Abbreviation: n/a, not applicable.

Table 3.

Moderators of all studies included in the quantitative analyses

Study	Study Quality (Level)	Theory	Theory Intensity (Level)	Training Type	Intervention Length (wk)	Follow-up Length (wk)	Instrument Type (Device-Measured/Subjective)	Physical Activity Outcome
Bombardier et al41	1	MI	Moderate	Behavior	12	-	Subjective	7-Day PAR
Carter et al42	1	TTM	Sparse	Combined	10	12	Subjective	GLTEQ
Coote et al43	1	SCT	Extensive	Combined	10	12	Subjective Device-measured	GLTEQ SenseWear armband
Dlugonski et al44	1	SCT	Moderate	Behavior	12	12	Subjective	GLTEQ
Duff et al45	1	-	-	Exercise	12	-	Device-measured	Accelerometer
Ennis et al46	1	Self-efficacy	Extensive	Behavior	8	-	Subjective	HPLP II
Hayes et al47	1	SCT	Moderate	Combined	10	-	Subjective	GLTEQ
Learmonth et al48	1	-	-	Exercise	12	-	Subjective	Phone FITT
Learmonth et al49	1	SCT	Extensive	Combined	16	-	Subjective Device-measured	GLTEQ Accelerometer
McAuley et al50	1	-	-	Exercise	24	-	Subjective	GLTEQ
Mostert and Kesselring51	2	-	-	Exercise	4	-	Subjective	BAECKE
Motl et al52	1	SCT	Extensive	Behavior	12	-	Subjective	GLTEQ
Motl et al63	1	SCT	Sparse	Behavior	24	-	Subjective Device-measured	GLTEQ Accelerometer
Paul et al54	1	-	-	Exercise	24	12	Device-measured	Accelerometer
Pilutti et al55	1	SCT	Sparse	Behavior	24	-	Subjective Device-measured	GLTEQ Accelerometer
Plow et al56	1	SCT and TTM	Moderate	Combined	12	12	Subjective	GLTEQ PADS
Rice et al57	2	SCT	Moderate	Combined	12	-	Device-measured	Accelerometer
Sandroff et al58	1	SCT	Sparse	Behavior	24	-	Subjective	IPAQ
Stuifbergen et al59	1	HBM, self-efficacy, Pender’s model of health promotion	Extensive	Behavior	8	12	Subjective	HPLP II
Suh et al60	1	SCT	Extensive	Behavior	6	-	Subjective	GLTEQ
Tallner et al61	1	-	-	Exercise	12	-	Subjective	BAECKE
Thomas et al62	1	MI, SCT, cognitive behavioral, self-determination	Moderate	Combined	24	-	Subjective	GLTEQ
Turner et al63	1	MI	Sparse	Combined	12	12	Subjective	GLTEQ
Wens et aL64	1	-	-	Exercise	12	-	Subjective	PASIPD

Abbreviations: BAECKE, Baecke Physical Activity Questionnaire; FITT, frequency, intensity, time, and type; GLTEQ, Godin Leisure-Time Exercise Questionnaire; HBM, health belief model; HPLP II, Health-Promoting Lifestyle Profile II; IPAQ, International Physical Activity Questionnaire; MI, motivational interviewing; PAR, physical activity recall; SCT, social cognitive theoiy; TTM, transtheoretical model.

Study quality

The methodological quality assessment of all studies included in the quantitative synthesis is provided in table 3 and supplemental appendix S3 (available online only at http://www.archives-pmr.org/). The overall methodological quality was good (median, 9; range, 6–10) based on the PEDro scale. Using the Spinal Cord Injury Rehabilitation Evidence system, 22 studies were classified as level I^{41–50 52–56 58–64} and 2 studies were classified as level II.^51,57

Immediate effect of interventions on physical activity behavior

The immediate effects of interventions on physical activity behavior from the 24 studies are illustrated in figure 3 and table 4. Overall, there was a statistically significant increase in physical activity levels favoring intervention compared with control (P<.001); the SMD was moderate (0.56) in magnitude and surpassed the threshold for clinical meaningfulness. The test of heterogeneity was significant (Q=35.69, df=23, P=.044, I²=36%) and supported examination of moderator variables.

Fig 3.

Immediate, post-intervention effect on overall physical activity levels.

Table 4.

Moderators of the effects of the interventions on physical activity behavior

Categorical Moderator	QB	P Value	Level of Moderator	Study No.	SMD	SE	Lower CI	Upper CI	I2
Immediate effects
Overall	35.69 (Q)*	.04		24	0.56	0.07	0.42	0.70	36
Intervention type	3.89	.14	Exercise	7	0.53	0.15	0.24	0.82	40
			Behavioral	9	0.71	0.12	0.46	0.95	54
			Combined	8	0.38	0.11	0.15	0.60	0
Intervention duration	0.01	.93	≥12 wk	17	0.57	0.09	0.38	0.75	40
			<12 wk	7	0.55	0.13	0.30	0.80	32
Measurement type	6.21*	.01	Device-measured	7	0.26	0.12	0.04	0.49	8
			Self-reported	17	0.62	0.09	0.45	0.79	37
Intensity of theory integration	0.09	.96	Level 1 (sparse)	5	0.61	0.15	0.33	0.90	27
			Level 2 (moderate)	6	0.58	0.15	0.29	0.86	17
			Level 3 (extensive)	6	0.55	0.17	0.21	0.88	63
Study quality	0.56	.45	Level 1	22	0.57	0.08	0.42	0.72	40
			Level 2	2	0.31	0.34	−0.35	0.98	0
Immediate effects
Studies included follow-up	10.74 (Q)	.10		7	0.40	0.14	0.12	0.67	44
Intervention type			Exercise	1	0.05	0.24	−0.43	0.52	0
			Behavioral	2	0.64	0.41	−0.15	1.44	79
			Combined	4	0.39	0.17	0.05	0.73	19
Intervention duration			≥12 wk	6	0.47	0.16	0.16	0.77	42
			<12 wk	1	0.05	0.24	−0.43	0.52	0
Measurement type			Device-measured	2	0.13	0.18	−0.22	0.49	0
			Self-reported	5	0.53	0.19	0.17	0.89	49
Intensity of theory integration			Level 1 (sparse)	2	0.30	0.21	−0.12	0.72	0
			Level 2 (moderate)	2	1.05	0.25	0.57	1.53	0
			Level 3 (extensive)	2	0.25	0.16	−0.05	0.56	0
Study quality			Level 1	7	0.40	0.14	0.12	0.67	44
			Level 2	0
Sustained effects
Overall	8.83 (Q)	.18		7	0.53	0.13	0.27	0.79	32
Intervention type			Exercise	1	0.12	0.26	−0.38	0.62	0
			Behavioral	2	0.74	0.28	0.20	1.29	56
			Combined	4	0.56	0.16	0.24	0.88	0
Intervention duration			≥12 wk	1	0.12	0.26	−0.38	0.62	0
			<12 wk	6	0.62	0.12	0.37	0.86	10
Measurement type			Device-measured	2	0.16	0.20	−0.23	0.54	0
			Self-reported	5	0.68	0.13	0.44	0.93	0
Intensity of theory integration			Level 1 (sparse)	2	0.65	0.22	0.22	1.08	0
			Level 2 (moderate)	2	0.96	0.25	0.51	1.48	0
			Level 3 (extensive)	2	0.42	0.16	0.10	0.74	0
Study quality			Level 1	7	0.53	0.13	0.27	0.79	32
			Level 2	0

Abbreviations: CI, confidence interval; SE, standard error.

^*P<.05.

Among the studies included in the analysis of sustainability (7 of 24, 29%), there was a statistically significant immediate increase in physical activity levels favoring the intervention conditions compared with control conditions (P=.005); the SMD was small (0.40) in magnitude and did not surpass the threshold for clinical significance. The level of heterogeneity was not significant (Q=l0.74, df=6, P=.095, I² =44%).

Sustained effect of interventions on physical activity behavior

Seven studies^{42,44,54,56,59,63} were included in the sustainability analyses with 12-week follow-up points. Only 2 studies provided 24-week follow-up points, which were omitted from the analysis. Four studies with follow-up periods included combined interventions,^42,43,56,63 2 studies included behavioral intervention only,^44,59 and 1 provided only exercise training.⁵⁴

The sustained effects on overall physical activity levels are provided in figure 4 and table 4. There was a statistically significant increase in physical activity levels favoring intervention compared with control (P<.001); the SMD was moderate (0.53) and surpassed the threshold for clinical significance. The level of heterogeneity was not significant (Q=8.83, df=6, P=.183, I²=32%).

Fig 4.

Sustained, 12-week follow-up effects on overall physical activity levels.

Moderator effects of interventions on physical activity behavior

The moderator variables for understanding variability in the average ES are provided in table 3. The immediate effects of interventions on physical activity behavior by moderators (ie, intervention type and duration, measurement type of physical activity, intensity of theory integration, and study quality) are presented in table 4.

The Q_B statistic indicated that the immediate effects of the intervention on physical activity outcomes significantly differed based on the type of physical activity measurement (self-reported vs device-measured) (Q_B = 6.21, df=2, P = .013). The interventions with self-reported outcomes yielded a moderate effect (SMD, 0.62), whereas there was a small effect (SND, 0.26) for studies that included device-measured physical activity. Other Q_B values were not significant for the study or intervention characteristic moderators.

Regarding the intervention type as a moderator, the studies that delivered behavioral intervention alone produced a moderate effect (SMD, 0.71) with a medium level of heterogeneity (I² = 54%). The SMDs of the studies with exercise training alone and combined interventions were 0.53 and 0.38, respectively, with low levels of heterogeneity (I²=63%). Regarding intervention duration as a moderator, both levels (≤12 wk and >12wk) produced moderate effects with low levels of heterogeneity. Regarding the intensity of theory integration as a moderator, all levels produced medium effects, but the studies with extensive use of theory had a high level of heterogeneity Regarding the study quality, level 1 studies yielded a moderate effect (SMD, 0.57) with a low level of heterogeneity, whereas level 2 studies had a small effect (SND, 0.31) with a confidence interval, including zero.

Discussion

This study provided a comprehensive meta-analysis that quantified the immediate and sustained effects of interventions to increase physical activity behavior among people with MS. The cumulative evidence demonsfrated that the interventions had moderate effects on both immediate and sustained changes in physical activity behavior; those effects exceeded ½ SD as a threshold for clinical meaningfulness. The moderator analyses identified study features associated with the trend for larger physical activity changes, and these included self-reported physical activity measurement (vs device-measured physical activity) and behavioral interventions (vs exercise training and combined interventions), but not study quality, intervention duration, and intensity of theory integration. Such findings have important implications for designing and developing future RCTs that target physical activity behavior for people with MS.

When examining immediate changes in physical activity, the interventions (n =24) had moderate effects on physical activity behavior (SMD, 0.56). The magnitude of the effect is consistent with previous meta-analyses of behavioral interventions in people with MS (ES, 0.64)²⁴ and other neurological disorders, including MS (ES, 0.53).⁶⁵ The findings are further comparable with those reported by a previous meta-analysis of interventions that were delivered through technology in people with MS (ES, 0.59).⁶⁶ However, we note that the previous meta-analyses have focused on small, specific interventions, such as behavioral interventions or technology-based interventions. Overall, our findings indicate that people with MS who participate in an intervention can increase physical activity levels upon completing the intervention.

When examining sustained changes in physical activity, the interventions (n = 7) had moderate effects on physical activity behavior (SMD, 0.53) that were comparable with a previous meta-analysis (SND, 0.60).²³ The resultant sustained effects appeared slightly higher than those identified from pre- to post-intervention (ie, immediate effects) (n = 7; SMD, 0.40). This indicates that participants increased physical activity behavior pre- to post-intervention and sustained, and perhaps, built upon these changes throughout the follow-up period. Only 1 of the 7 studies did not report the application of behavioral change strategies. ⁵⁴ Collectively, the findings emphasize the importance of incorporating behavioral change techniques aligned with theory within interventions that target sustainable changes in physical activity behavior; such findings regarding sustainability should be confirmed by RCTs.

Regarding moderator analyses of the immediate intervention effects, we identified moderate effects when studies included self-reported measures of physical activity (SMD, 0.62), and small effects in studies that used device-measured physical activity (SMD, 0.26). This observation is comparable with previous research ^23,24 that demonstrated the differences between selfand device-measured physical activity. Such findings should be interpreted with caution as the studies in the present meta-analysis primarily included self-reported measures of physical activity.

Moderator analyses of intervention type demonsfrated that the behavioral interventions alone yielded the largest effect (SMD, 0.71), followed by the exercise training studies and the combined interventions There was a medium level of heterogeneity among the behavioral interventions (I² =54%). This appeared to be influenced by 2 studies with multidisciplinary health promotion education (ie, wellness interventions) that produced small effects. This is consistent with a previous meta-analysis of physical activity interventions in healthy adults that reported larger effects with interventions that used behavioral strategies (ES, 0.25) compared with other interventions that targetted general health education (ES, 0.17).⁶⁷ Overall, these findings indicate that behavioral interventions alone may be capable of addressing the long-standing problem of physical inactivity in MS.

Interestingly, interventions that provided both exercise training and behavioral coaching resulted in a smaller effect (SND, 0.38) than the behavioral interventions alone on immediate changes in physical activity behavior (SMD, 0.71). One likely explanation for this difference could be the dose of the behavioral coaching. The combined interventions consisted of 185 total sessions, and only 57 sessions (31%) focused solely on behavioral coaching. Exercise training accounted for 103 sessions (56%) and the other 25 sessions (13%) embedded behavioral coaching during exercise training. The behavioral interventions delivered only behavioral coaching content (84 sessions), and the combined interventions may have had a lower volume or dose of behavioral coaching. Of note, because of inadequate reporting, we were unable to discern the precise amount of time spent on behavioral coaching. Further research may explore optimal doses of behavioral coaching for physical activity change in MS.

Regarding the intensity of theory integation, we anticipated that a higher intensity would yield larger immediate changes in physical activity. However, we observed moderate and meaningful effects of interventions across all levels of theory integration. One explanation is that the classification method (a modified version of a behavior theory coding framework) may not have been sensitive to detect such differences, with the sparse amount of reported details concerning behavioral coaching. One previous meta-analysis encountered a similar issue when attempting to classify and analyze behavioral interventions based on the original Theory Coding Scheme.²³

Future directions

The findings of this meta-analysis are encouraging, but several knowledge gaps require further investigation. The most common follow-up duration was rather short (3mo), and this is critical as the impetus for designing physical activity interventions is that people will adopt and maintain an active lifestyle over a prolonged period for improving and managing health and function. The analysts observed that interventions often did not provide enough detail regarding behavior change techniques to allow replication of study procedures or identification of mechanisms that resulted in the observed changes in behavior. Of note, our findings indicated that few studies examined sustainable changes in physical activity behavior. Sustainability is a critical area that warrants further investigation, considering that people with MS experience numerous barriers (ie, personal and environmental) for sustained behavior change.

Study Limitations

This meta-analysis had limitations. The study participants had mostly mild-to-moderate mobility disability, which is not surprising as walking was the most commonly prescribed type of physical activity. However, this finding indicates a need for programs that are inclusive of a wider variety of movement capabilities, such as for people who use wheelchairs. Only 1 study included people with MS who used wheelchairs as a primary mobility aid. ⁵⁷ This limits the generalizability of our findings among people with mild-to-moderate disabilities. The findings of the sustainability analyses were statistically significant but should be interpreted with caution given tie relatively small number of studies that included follow-up; this further hindered the statistical comparisons among the categorical moderators. Ideally, the findings of moderator analyses should be compared and interpreted in the context of associations with other moderators.

Conclusions

This meta-analysis demonstrated that behavioral interventions alone are efficacious for increasing and perhaps sustaining physical activity behavior in adults with MS. These findings are encouraging and provide an initial foundation for future research exploring sustainability. We further identified several knowledge gaps that require additional research, including follow-up durations and the moderafing effects of the intervention and/or participant characteristics.

List of abbreviations:

ES

eftect size

MS

multiple sclerosis

PEDro

Physiotherapy Evidence Database

Q

test for heterogeneity in study effect size

Q_B

test for heterogeneity within categorical moderators

RCT

randomized controlled trial

SMD

standardized mean difference