Aim

In the current study, which is part of a large international survey of people with MS and the first baseline data collected as part of a longitudinal dataset, we investigate the associations between self-initiated PA or exercise, and HRQOL, relapse rate, disease activity and disability, as well as demographic variables.

Participants and data collection

Participants were recruited through online MS societies and platforms including websites, Facebook pages, twitter, blogs and forums where the survey was advertised. All media were specifically for people with MS, some actively promoted healthy lifestyle and most were based in English speaking countries including the US, Australia, New Zealand and the UK. The survey was in English and included a participant information and consent page which required a response before entering the survey. Participants were aged over 18 years and reported a physician confirmed diagnosis of MS. Contact details were recorded for future follow-ups as this is part of a longitudinal study.

Tools

The online survey comprised several validated questionnaires as well as researcher devised or amended items, and took approximately 40 minutes in total to complete with the option to come back to it at a later time if it was not completed at once. All data were self-reported.

Data analysis

Data were analyzed using SPSS version 22.0 (IBM Corporation). Continuous data were reported using mean (95% CI) or median (IQR) and categorical data using number (N) and percentage (%). Varying denominators are due to varying item completion. Analyses were only performed for those reporting a physician confirmed diagnosis of MS. The IPAQ produced a continuous variable of “MET minutes”, which was very positively skewed and could not be transformed, and a categorical variable. We used the categorical variable for analysis.

Pearson’s Chi Square was used to compare categorical variables, or Fisher’s exact test for two by two contingency tables with standardized adjusted residuals indicating over- or underrepresentation of groups. Analysis of variance was used to compare continuous variables across three or more groups with planned comparisons for post hoc testing. A Welch correction was carried out if the homogeneity of variances assumption was violated; this was the case for most variables except age, the MSQOL-54 ES and PHC. Alpha was set at 0.05 and two tailed tests of significance were used in all instances. These analyses for demographic variables were repeated for each of the three disability groups separately to control for disability level.

Regression analyses were carried out to identify independent predictors of outcomes. No regression analysis was carried out to predict disability as inverse causality seemed probable. Multiple regression (enter method) was used to identify predictors of PHC, MHC, SFS, OQOLS, and ES. It was ensured that data satisfied the assumptions of normality, linearity, and homoscedasticity by plotting the studentized residuals against the unstandardized predicted values of the dependent variable and assessing the spread. The MHC score had a skewed distribution of standardized residuals and a sensitivity analysis was undertaken. Independence of residuals was assessed by checking the Durbin Watson statistic (values between 1 and 3 were accepted). Variance inflation factor <5 was used as the criterion for absence of multicollinearity. Only correlations of <.70 between predictor variables were accepted as inspected from the correlation matrices. Data were checked for standardized residual values outside the range of -3.0 to 3.0 and outliers were inspected. There were no influential cases with a Cook value over 1. As data for 12 month relapse rate among those with relapsing-remitting MS were over-dispersed, negative binomial logistic regression was used after verifying that the assumptions were met in regards to multicollinearity (cut-off was >0.70). For disease activity (decreasing/stable versus increasing), binary logistic regression (enter method) was planned but not undertaken as the bivariate analysis showed no significant results. Parameter estimates and estimated marginal means (EMM) were reported as well as percentage change by dividing the difference between EMM of low level of PA and high level of PA, by the EMM of low level of PA. All percentages reported were adjusted for missing data or item non-completion.

Demographic variables

Physical activity and demographics

Physical activity and outcome variables

Quality of life

All subscales and composite scores of the MSQOL-54 were significantly related to PA and post-hoc tests showed that all subscales and composite scores were significantly different between low vs. moderate, low vs. high, and moderate vs. high levels of physical activity (all <.001 except between moderate and high activity on the mental health composite (p =.012)) (Figure 1).

Regression analyses showed that the model including age, gender, disability and PA significantly predicted PHC, MHC, ES, SFS and OQOLS scores, explaining between 10.6 and 43.5% of the variance (Table 4). Higher disability level was significantly associated with lower HRQOL scores, controlling for PA, age and gender. MHC, ES, SFS, and OQOLS scores increased with age and while male gender predicted a higher score on the PHC and ES it predicted a lower score on the OQOLS (Table 4).

Outcome variable	Predicting variable	Adjusted R2	Beta	SE	p
Physical health composite	Intercept	.435 P < .001	61.209	1.832	<.001
	Age		.037	.040	.355
	Male gender		2.393	.974	.014
	Moderate level of physical activity		8.331	.944	<.001
	High level of physical activity		12.263	.998	<.001
	Moderate level of disability		-22.172	.898	<.001
	High level of disability		-26.950	1.451	<.001
Mental health composite	Intercept	.106 P < .001	51.818	2.125	<.001
	Age		.341	.045	<.001
	Male gender		.068	1.168	.954
	Moderate level of physical activity		6.369	1.097	<.001
	High level of physical activity		8.175	1.176	<.001
	Moderate level of disability		-10.699	1.047	<.001
	High level of disability		-9.457	1.653	<.001
Energy subscale	Intercept	.167 P < .001	31.844	2.149	<.001
	Age		.224	.046	<.001
	Male gender		2.783	1.185	.019
	Moderate level of physical activity		8.590	1.114	<.001
	High level of physical activity		13.927	1.189	<.001
	Moderate level of disability		-12.872	1.063	<.001
	High level of disability		-9.838	1.672	<.001
Social function subscale	Intercept	.292 P < .001	67.104	2.135	<.001
	Age		.141	.045	.002
	Male gender		.784	1.179	.506
	Moderate level of physical activity		8.304	1.109	<.001
	High level of physical activity		10.544	1.184	<.001
	Moderate level of disability		-18.749	.594	<.001
	High level of disability		-29.595	1.058	<.001
Overall quality of life subscale	Intercept	.196 P < .001	63.904	1.770	<.001
	Age		.109	.038	.004
	Male gender		-2.432	.977	.013
	Moderate level of physical activity		6.086	.917	<.001
	High level of physical activity		9.247	.976	<.001
	Moderate level of disability		-11.700	.873	<.001
	High level of disability		-15.867	1.372	<.001

Female gender, low level of activity and low level of disability were used as reference categories in regression analyses.

Bold indicates significant finding at <0.05 level.

PA significantly increased all HRQOL measures, controlling for age, gender and disability level (Table 5).

Gender	Level of disability	Level of physical activity	Physical health composite	Mental health composite	Energy subscale	Social function subscale	Overall QOL subscale
			EMM	EMM	EMM	EMM	EMM
			SEa	SEb	SEc	SEd	SEe
Male	High	High	50.6	66.1	48.9	55.3	59.8
			1.7	2.0	2.0	2.0	1.6
		Moderate	46.6	64.3	43.6	53.1	56.6
			1.7	1.9	2.0	2.0	1.6
		Low	38.3	57.9	35.0	44.7	50.6
			1.5	1.7	1.7	1.7	1.4
	Moderate	High	55.3	64.9	45.9	66.1	64.0
			1.2	1.5	1.5	1.5	1.2
		Moderate	51.4	63.1	40.6	63.9	60.8
			1.2	1.4	1.4	1.4	1.2
		Low	43.1	56.7	32.0	55.6	54.7
			1.1	1.3	1.3	1.3	1.1
	Low	High	77.5	75.6	58.8	84.9	75.7
			1.1	1.3	1.3	1.3	1.1
		Moderate	73.6	73.8	53.4	82.6	72.5
			1.1	1.3	1.3	1.3	1.1
		Low	65.3	67.4	44.9	74.3	66.4
			1.2	1.4	1.4	1.4	1.2
Female	High	High	48.2	66.0	46.2	54.5	62.2
			1.5	1.8	1.8	1.8	1.5
		Moderate	44.2	64.2	40.8	52.3	59.1
			1.5	1.7	1.7	1.7	1.4
		Low	35.9	57.9	32.2	44.0	53.0
			1.3	1.4	1.4	1.4	1.2
	Moderate	High	53.0	64.8	43.1	65.4	66.4
			1.0	1.2	1.2	1.2	1.0
		Moderate	49.0	63.0	37.8	63.1	63.2
			0.9	1.1	1.1	1.1	0.9
		Low	40.7	56.6	29.2	54.8	57.2
			0.8	0.9	0.9	0.9	0.8
	Low	High	75.1	75.5	56.0	84.1	78.1
			0.8	0.9	0.9	0.9	0.8
		Moderate	71.2	73.7	50.7	81.9	74.9
			0.8	0.9	0.9	0.9	0.7
		Low	62.9	67.3	42.1	73.6	68.9
			0.8	1.0	1.0	1.0	0.8

^aCovariate age appearing in the model is evaluated at the following value: 45.0.

^bCovariate age appearing in the model is evaluated at the following value: 45.5.

^cCovariate age appearing in the model is evaluated at the following value: 45.7.

^dCovariate age appearing in the model is evaluated at the following value: 45.6.

^eCovariate age appearing in the model is evaluated at the following value: 45.6.

Keeping age, gender and disability at a constant (the average), increasing from low to moderate and to high PA increased predicted physical health scores from 47.7 to 56.0 to 59.9 (25.6% change from low to high), mental health scores from 60.6 to 67.0 to 68.8 (13.5% change from low to high), energy score from 35.9 to 44.5 to 49.8 (38.7% change from low to high), social function score from 57.8 to 66.1 to 68.4 (18.3% change from low to high), and OQOLS from 58.5 to 64.5 to 67.7 (15.7% change from low to high).

Summary of findings

This study shows that self-initiated PA, or exercise, is associated with better HRQOL outcomes, regardless of disability level, in a very large, international sample of people diagnosed with MS. Increasing levels of PA were associated with increased levels of energy, social function, physical, mental, and overall QOL as measured by the MSQOL-54 while controlling for age, gender and disability level. Of note, the increases in HRQOL scores, between low to moderate levels of PA, were more pronounced than those between moderate to high levels of PA, indicating that PwMS who are currently participating in low levels of PA may benefit more from increasing their level of PA than those with moderate level increasing to a high level of PA.

An improvement of 5 points on the SF-36, the tool from which the MSQOL-54 is derived, is generally seen as clinically significant [30, 31]. The estimated increases of more than 25% on the physical health composite (12.2 points), and estimated increases of around 15% on the mental health composite (8.2 points), social function subscale (10.6 points) and overall QOL subscale (9.2 points) between low and high levels of PA would therefore be clinically significant and of interest to many PwMS. The estimated increase of almost 40% on the energy subscale (13.9 points) is especially noteworthy. Fatigue, related to perceived energy, is a debilitating symptom for most PwMS and is positively affected by PA [32]. The benefits of PA on HRQOL measures are in line with most randomised controlled trials (RCTs) and other studies assessing the association between QOL and PA [15, 16, 33–37], with one meta-analysis showing the effect of PA on QOL superior to medication[14]. However, at least two RCTs showed no difference in QOL in people with MS after participating in exercise programs [12, 13]. Although most studies included only people with mild to moderate disability, a small pilot study including PwMS with high disability showed increases on MSQOL subscales of energy (60%), physical health composite (30.3%), mental health composite (22.4%), and social function subscale (12.6%) (but a decreased score of 2.6% on overall QOL subscale) after a 12 week treadmill training program [38]. The effect of PA on QOL could be mediated through improvements in social participation, fatigue, mood, self-efficacy, perceived stress and/or body functions [5, 39–43]. If the type of exercise involves a team or group, it can be expected that PA has a positive effect on perceived social function, whereas being able to complete a training program may improve perceived self-efficacy. However, since QOL is an umbrella term encompassing many of these aspects, it may be difficult to disentangle the relationship between the variables.

Increased level of disability was associated with decreased levels of PA, consistent with the literature [7, 8]. Increasing disability and symptoms may prevent PwMS from participating in PA. However, this may be a two-way causal relationship as some evidence is emerging that PA has a disease-modifying effect for PwMS [44], and may have beneficial effects through changes in anti-inflammatory and neurotropic factors in PwMS [45–47] but more research is needed. PwMS generally have lower rates of PA compared to the general population [7, 48], even those with mild disability, [49] although evidence is emerging that PA is beneficial for PwMS regardless of level of disability [38, 50]. Several RCTs have found PA to favorably affect level of disability [12, 51], and a lack of PA may also negatively affect disease progression and characteristics in PwMS [9, 10]. The current study design does not allow separating the effects that PA and disability have on each other but future longitudinal data of this cohort may provide further insights.

Due to these expected interactions, regression analyses controlled for disability levels, as disability is expected to affect PA levels as well as the outcome measures. For the subsample of people with relapsing-remitting MS, disease activity was not associated with level of PA and although increased PA was associated with decreased 12 month relapse rate, it was not a significant predictor of relapse rate in the regression analysis when controlling for age, gender and disability level. A recent literature review regarding the effects of PA on disease progression concluded that the evidence from intervention studies was not strong enough to draw definite conclusions but MRI data and self-reported data from PwMS show a possible disease-modifying effect [44]. A longitudinal survey of PwMS found that those with higher baseline activity levels compared to those who had lower levels at baseline showed less functional decline over time suggesting that PA may favorably affect disease progression [43]. Although a survey of PwMS reported that 15% of respondents indicated that aerobic exercise worsened their MS symptoms, conversely 16% reported improvement [52]. There is no other evidence that PA has a deleterious effect on MS [6, 53], and a recent study has reported that PA is not related to relapses [54]. Our data also indicated that for PwMS, with low to moderate disability level, increased PA was generally related to a healthier BMI, which is expected to result in better physical health overall [9].

Limitations

The sample in this study consisted of volunteers from English speaking backgrounds with access to the internet and with the ability to complete this survey online. Most participants were female and between the ages of 38 and 53 years, and thus our population may be less representative of the heterogeneous spectrum of PwMS. It may therefore not be generalizable to all PwMS, despite the number, and variety of backgrounds of participants. All measures were self-reported and it is therefore not possible to assess accuracy or reliability. Due to some incomplete data in the duration items of the IPAQ, a small number of people who indicated to participate in activities once or more per week may have been incorrectly classified in the low level PA group. Although there were 235 severely disabled PwMS who completed PA related items, only 26 and 16 reported moderate and high levels of PA, respectively, which reduced the power to detect significant differences in this group.

IPAQ, the instrument widely used for studying PA in PwMS, has been shown by some to overestimate the intensity and duration of physical activity in the general population [55] and it has been reported that PwMS have difficulties recalling the duration of activities in retrospect as required for the IPAQ [22]. Also, the level of exercise relative to someone’s ability and fitness was not taken into account, rather the IPAQ measures an objective standard of intensity of exercise and this may therefore not accurately measure activity efforts in people with high levels of disability. Further, certain types of PA may be more beneficial than others, and effects may differ between types of MS; this study did not differentiate between these, but future planned research may be able to elucidate this.