Fitness, physical activity, and exercise in multiple sclerosis: a systematic review on current evidence for interactions with disease activity and progression

The databases MEDLINE (via PubMed), EMBASE, CINAHL, and SPORTDiscus were used for electronic literature search from inception until January 31, 2021. The search strategy included Medical Subject Headings (MeSH) and text words of the defined MS population, surrogate terms of exercise, physical activity, and fitness, as well as clinical measures of disease activity and progression. The categories were combined through Boolean operators (‘‘AND’’, ‘‘OR’’) (Table S1 for complete search string). Titles and abstracts were screened by two independent reviewers (S.P., P.K.). Only peer-reviewed articles published in English language were included. Covidence Review Software, recommended by the Cochrane Collaboration, was used for the review process.

Two independent reviewers (S.P., P.K.) screened the studies for eligibility and methodological quality. The Cochrane Risk of Bias (RoB) tool was used to assess the risk of various bias domains (i.e., selection, reporting, performance, detection, and attrition) for included RCTs (Table S2) [29]. Each domain was judged as ‘high RoB’, ‘unclear RoB’, or ‘low RoB’. Decisions on the studies´ evidence level classification (classification I to IV) were based on standards published for therapeutic trials by the AAN (Table 2) [30]. According to the strength of evidence for respective outcomes in each research area (i.e., exercise, physical activity, fitness) based on the level of confidence in evidence (high confidence, moderate confidence, low confidence, very low confidence) [30], evidence-based recommendations will be provided, classified as highly likely to be effective, ineffective or harmful (level A), likely to be effective, ineffective or harmful (level B), possibly likely to be effective, ineffective or harmful (level C), or data is insufficient or too conflicting to conclude an effect (level U) (Table 2).

In cases of disagreement considering eligibility, methodological quality, or classification of evidence level, conflicts were resolved by consensus. If no consensus could be achieved, a third reviewer (N.J.) was consulted who was blinded to the judgments of the first two reviewers. Inter-rater correlation coefficients are 89.4% and 94.4% for AAN and Cochrane RoB, respectively.

All relevant study characteristics are synthesized and presented in Tables 3, 4 and 5.

A detailed description of study characteristics from all included studies can be found in Table 3 (RCTs, CTs), Table 4 (cross-sectional studies), and Table 5 (cohort studies, randomized parallel group trials). Due to the relatively short duration of existing exercise studies compared to clinical phase III trials, ranging from 4 to 26 weeks (mean duration of 13.4 weeks), only short-term effects were assessed in the included studies.

In the following sections, the studies´ evidence level classification is summarized to conclude the strength of evidence (level of confidence) and to give evidence-based recommendations (Table 2A and B) on (i) the effect of structured exercise interventions on clinical measures of disease activity and progression, and on (ii) interactions of the physical activity and fitness level with clinical measures of disease activity and progression.

In total, 19 clinical exercise studies (five of class II evidence, ten of class III evidence, four of class IV evidence) investigated the effect of an exercise intervention on clinical measures of disease activity and progression. Regarding RoB assessment, the five class II evidence RCTs achieved a mean score of 4.1/6 for ‘low RoB’, 0.57/6 for ‘unclear RoB’ and 0.72/6 for ‘high RoB’, whereas the remaining eight class III evidence RCTs achieved a mean score of 2.38/6 for ‘low RoB’, 2.75/6 for ‘unclear RoB’ and 0.87/6 for ‘high RoB’. Of the included (R)CTs, six studies applied aerobic exercise [12, 31‐35], three resistance exercises [11, 36, 37], five combined aerobic and resistance exercises [38‐42], and one study applied a mind–body exercise [43]. Of the included longitudinal cohort and parallel group studies, one applied combined resistance and aerobic exercise [44], two combined resistance and balance exercises [45, 46], and one both climbing and mind–body exercise [47].

Two class II, six class III, and three class IV studies [33, 34, 36, 38, 40‐46] report non-significant effects of exercise on EDSS, while two class III studies and one class IV study reported a decreased EDSS after exercise [35, 37, 47]. Two class II studies and one class III study reported exercise-induced improvements in the MSFC score after resistance exercise, combined endurance and resistance exercise, and interactive exergaming [11, 31, 41]. Considering MRI measures, changes in the percentage of brain volume were not observed in two class II studies [11, 12], whereas another class III study reported an increase in the left pallidum volume after exercise [32]. Although one of the class II studies reported an increase in cortical thickness [11], no change was observed in the other study [12]. Of note, none of the two studies reported an exercise-induced change in lesion load. The annualized relapse rate was investigated only in one class II study by Langeskov-Christensen et al. [12] who revealed a lower rate in the exercise group.

Recommendations: Based on results from two class II and six class III studies, there is moderate confidence (level B) that the EDSS score remains unchanged over the exercise intervention period. Moderate confidence (level B) from two class II studies indicates that exercise does not decrease disease activity indicated by the change in lesion load. Due to results from two class II and one class III study, there is moderate confidence (level B) that exercise improves the MSFC in pwMS. Moreover, there is low confidence (level C) from one class II study that exercise decreases the annualized relapse rate. While results from two class II studies show moderate confidence (level B) that exercise has no effect on brain volume in pwMS, there is very low confidence (level U) for the effect of exercise to increase cortical thickness.

Therefore, it is likely that exercise neither decreases nor stabilizes the EDSS and lesion load, whereas the MSFC score is likely to be improved. The brain volume is not likely to be changed by exercise, meaning that exercise neither increases nor maintains brain volume. Further, it is possible that exercise decreases the annualized relapse rate. Due to inconsistency, current data is insufficient to determine the effect of exercise on cortical thickness.

In total, eight cross-sectional studies (two of class III evidence, six of class IV evidence) and four cohort studies with follow-up measurement time points (all of class IV evidence) investigated the effect of the physical activity level on clinical measures of disease activity and progression. Physical activity was assessed subjectively by a questionnaire (GLTEQ or IPAQ) in three studies [48‐50], while the remaining studies used devices (i.e., Fitbit^®, ActiGraph^®, Axivity AX3^®, Actiwatch^®) to report daily or weekly step counts in most of the cases.

Two class III studies [51, 52] and eight class IV studies [48‐50, 53‐57] report significant negative associations between the physical activity and the EDSS. With respect to MRI outcomes, Kalron et al. (class IV) reported an increased hippocampal volume in active compared to inactive pwMS, whereas all other brain regions were not affected [48]. Positive associations between the level of physical activity and volume of whole brain gray and white matter as well as deep gray matter structures such as the thalamus, caudatus, putamen, and hippocampus were observed by Klaren et al. [23] (class IV).

Recommendations: Due to the low evidence classification of ten studies (two class III, eight class IV) reporting negative associations between the physical activity level and the EDSS, there is very low confidence (level U) that a high level of physical activity slows down disease progression quantified by EDSS. Again, due to the low evidence classification of two studies (class IV), there is very low confidence (level U) that a high level of physical activity increases or maintains (region-specific) brain volume in pwMS.

Therefore, current data is insufficient to determine the effect of physical activity on EDSS and (region-specific) brain volume in pwMS.

In total, 12 cross-sectional studies (three of class III evidence, nine of class IV evidence) investigated the effect of the fitness level on clinical measures of disease activity and progression. CRF was quantified as VO_2peak across all studies, except for two studies that quantified the VO_2max [58, 59]. Strength was assessed in four studies as maximal strength with isokinetics [60], dynamometer [61, 62], or the one-repetition maximum [63].

One class III study [60] and six class IV studies [21, 62‐66] report a negative association between the EDSS and VO_2peak or strength measures, respectively, while the two class IV studies that assessed VO_2max do not report an association [58, 59]. Considering MRI measures, increased gray matter volume in midline cortical structures (class III) and deep gray matter structures (class IV) was found in pwMS with a higher CRF [22, 67]. Fritz et al. [61] revealed a positive association between a summed strength score and the corticospinal area. Only one class III study assessed lesion load volume and reported a negative association with the CRF of pwMS [67].

Recommendations: Due to the low evidence classification of seven studies (one class III, six class IV) reporting negative associations between the fitness level and EDSS, there is very low confidence (level U) that a high fitness level slows down disease progression assessed by EDSS. Due to the low number and low evidence classification of studies investigating the association between the fitness level and MRI-based outcomes, there is very low confidence (level U) that the (region-specific) brain volume and the lesion load are increased (brain volume), decreased (lesion load), or maintained by a high fitness level.

Therefore, current data are insufficient to determine the effect of a high fitness level on the EDSS, (region-specific) brain volume and lesion load in pwMS.

The promotion of an active lifestyle is a major issue in the treatment of a broad range of diseases that are tightly linked to metabolic and immune-mediated disarrangement, including autoimmune diseases such as MS [19, 20]. With regard to MS, a recent article highlights the importance of physical activity and its promotion [24], since current evidence confirms that pwMS are still physically less active than the healthy population, although guidelines have been developed [17, 18]. Despite many limitations that are associated with the EDSS, i.e., a high intra- and inter-rater variability, non-linearity, and a limited responsiveness [10], the EDSS still represents the most frequently used clinical measure for disease progression and was assessed in ten out of the twelve included studies. All of the studies show a negative association between the physical activity level and the EDSS which means that pwMS with a higher EDSS are less physically active. However, this seems plausible in a way that neuromuscular functioning decreases with disease progression [68], leading to the need of a walking aid when the score is around six. Further, it should be considered that there is an increase in fatigue with a higher level of disability, which in turn may decrease the motivation to be physically active or to engage in exercise [69]. Positive associations have also been observed between the physical activity level and MRI-based outcomes. Physically more active pwMS showed increased volumes of gray matter including various subcortical brain regions of which the hippocampus region seems particularly sensitive [23, 48]. This is in line with evidence from preclinical animal models and human studies that support the beneficial effect of aerobic exercise on the hippocampus region [70, 71]. Although these findings are important and provide a clear rationale to determine effects of changes in physical activity on clinical measures of disease progression and activity longitudinally, it remains elusive why only a limited number of studies integrated follow-up measurements [49, 53, 57, 72]. A pilot study of Bahmani et al. [49] showed that vigorous physical activity, assessed by questionnaire, decreased during the first 2 years of disease, although not being associated with changes in the EDSS. Since this study covers a sensitive time frame of the disease which is proposed being a window of opportunity in MS exercise therapy [26], these findings are of high importance. Another longitudinal study tracked the physical activity of pwMS with a median EDSS score of 4.0 for 1 year [53]. Participants with a clinically meaningful increase in disability during this period showed a reduced daily step count and, more interestingly, those with a baseline daily step count below the cohort median had higher odds of clinically meaningful disability worsening within this year. These results may be indicative of an important role of physical activity to prevent disease progression.

Taken together, the findings from cross-sectional studies are interesting and provide a rationale to investigate the effect of structured physical activity on clinical measures of disease activity and progression in well-designed randomized controlled studies. This is highly important, since only relying on cross-sectional studies does not move research in this field forward.

An increase in physical activity or engagement in exercise programs results in higher fitness, usually quantified as CRF or strength. A recent study showed that CRF is positively associated with total brain volume and local gray matter volumes such as the right hippocampal gyrus in healthy adults [73]. Another study revealed a positive association between an increased CRF and a reduced brain atrophy in early-stage Alzheimer disease compared to healthy adults [74]. Therefore, an increase in CRF might be of high relevance for pwMS. Indeed, pwMS with higher levels of VO_2peak/max or higher muscle strength showed increased gray matter volume including various subcortical brain regions [22, 67]. Prakash et al. [67] further demonstrated a negative association between CRF and lesion load volume. Interestingly, a recent RCT reported that increased CRF in pwMS following 24 weeks of aerobic exercise was associated with an increase in the gray matter parenchymal fraction [12], thereby underscoring findings from cross-sectional studies that assume potential neuroprotective effects [22, 67]. Corresponding to findings from studies assessing physical activity, 9 out of the 12 included studies used the EDSS for correlation analyses with the fitness level. As already discussed in the Physical activity section, the consistent negative associations between the EDSS score and the fitness level observed in existing cross-sectional studies may be reasonable, since a moderate to high fitness level is the result of an active lifestyle which in turn has been shown to be associated with decreased disease severity. So again, despite these consistent associations, well-designed studies assessing both changes in fitness and disease progression/activity outcomes are necessary to provide evidence of higher quality.

In contrast to the results from cross-sectional studies, exercise studies predominantly showed non-significant effects on the EDSS, even when studies lasted 24 weeks or longer [40, 41, 43]. Studies that reported improvements suffer from methodological issues such as no randomization or matching of groups [35], no inclusion of a passive control group and focus on EDSS sub-domains [47], or a rather low sample size [37, 47]. However, low sample sizes are also observed in studies that report no changes in the EDSS [38, 40, 43]. Of note, only 8 of the included 13 RCTs tested 15 or more pwMS per group (the other 5 RCTs did not include more than 10 participants per group [37‐40, 43]) which underlines the small power of existing RCTs. The relatively short intervention period, ranging from 5 to 26 weeks, may hamper the interpretation of significant results. Dalgas et al. recommend conducting clinical exercise studies lasting more than 1 year [6]. A recent study revealed that assessing short-term disability progression over 3–6 months to estimate treatment effects may overestimate the accumulation of permanent disability by 24–30% [75]. In this regard, most of the outcome measures assessed in the reviewed studies (i.e., EDSS, relapse rate, percentage brain volume change) may not be sensitive enough to quantify changes over a relatively short time period of up to 6 months. Therefore, longer intervention periods with follow-up measurements at regular intervals are reasonable to elucidate (long-term) exercise-induced effects on clinical measures in pwMS. This would enable researchers to quantify other important outcome parameters which may be affected by exercise, e.g., the time and magnitude of recovery after relapses in relapsing–remitting MS. Since RCTs usually aim to determine significant improvements in the intervention group compared to the control group, it should be considered that this is not always reasonable for measures such as the EDSS, brain volume or lesion load. Especially in the context of progressive disease, it would be a therapeutic success if these clinical parameters would remain stable over the exercise intervention period which optimally lasts for two or more years.

It is worth mentioning that only one study considered objectively assessed relapses to calculate the relapse rate which represents an important outcome in phase III trials [12]. According to a survey that collected data on the annualized relapse frequency in an American population, 44.1% of the participants reported less than one relapse in the preceeding 2 years, whereas 35.5% reported 1–2 and 20.2% more than two relapses, respectively [76]. A time frame of 6 months or shorter might therefore be insufficient to assess a meaningful exercise-induced effect on the (medically confirmed) relapse frequency. Compared to clinically evident relapses, T2-hyperintense and gadolinium-enhancing T1 lesion formation can occur subclinically, thereby representing markers that could be identified in shorter periods of time [10]. That makes the assessment of lesion formation potentially more suitable to detect disease activity in exercise studies that last 6 months or shorter. However, following 24 weeks of progressive resistance or aerobic exercise, no changes in lesion load were observed [11, 12]. These studies did also not reveal changes in brain volume, and the relative cortical thickness was increased after exercise in only one of the two studies [11]. Brain atrophy accumulates very slowly and when assessed in clinical trials to determine treatment efficiency of drugs, follow-up time points of brain atrophy quantification are usually several years [77, 78]. Although 6 months of aerobic exercise has been shown to increase gray and white matter volumes in an aged population [13], this time frame still might be too short to determine meaningful effects in pwMS. In this regard, confounding variables such as age and disease duration need to be considered when investigating short-term effects of exercise on brain atrophy measures in pwMS. No study observed an exercise-induced positive modulation of the hippocampal structures [12, 32], although evidence from preclinical animal models and human studies support a beneficial effect of aerobic exercise on the hippocampus region [70, 71]. Again, exercise studies lasting more than 1 year might reveal larger and clinically meaningful effect sizes. But rather than aiming to increase (region-specific) brain volume, maintaining the exisiting brain volume by counteracting neurodegeneration represents another meaningful outcome to consider. In this regard, Kjølhede et al. [11] revealed a trend in whole brain volume preservation following progressive resistance exercise for 24 weeks.

After 12 weeks of interactive step training and 24 weeks of progressive resistance exercise, respectively, the MSFC was improved [11, 31]. Only Romberg et al. [41] conducted a study which lasted longer (26 weeks), resulting in an improvement of the MSFC as well. When looking more into the different dimensions of the MSFC, both Romberg et al. [41] and Kjølhede et al. [11] observed improvements in the timed 25-foot walk test and the nine-hole peg test (9-HPT) in the exercise compared to the passive control group, whereas no changes were observed in the paced auditory serial addition test (PASAT). Hoang et al. [31] included the symbol digit modalities test (SDMT) and the 10-meter walk test instead of the PASAT and the 25-foot walk test. Despite the selection of different tools to assess the same clinical dimensions, the results are consistent with those of the aforementioned studies. Here, only the 9-HPT and 10-meter walk test improved in the exercise group compared to the control group, with results from the SDMT remaining unchanged. These results indicate that the cognitive dimension of the MSFC, assessed by the PASAT or SDMT, is less responsive to exercise over three to 6 months than the functioning of upper and lower extremities. This is in line with a recent meta-analysis that does not support the efficacy of exercise training on global or domain-specific cognitive performance in pwMS [79].

Since almost all drug-based disease-modifying therapies have been shown to be effective primarily in the early relapsing–remitting disease course [3], theoretical considerations came up addressing a potential window of opportunity for exercise as a supportive disease-modifying treatment early in the disease course [26]. In this regard, it is important to note that the mean disease duration of MS populations from included studies is 9.27 years for (R)CTs, 10.52 years for cross-sectional studies, and 13.27 years for cohort studies. Only Bahmani et al. [49] included newly diagnosed pwMS to record changes in the physical activity behavior over 2 years. Therefore, it may be reasonable to focus more on the disease duration of participants rather than discussing solely intensity, frequency, and duration of exercise regimens. Documentation of the medication at study onset as well as changes of medication during the study period is important to control for disease-modifying effects that are not attributable to exercise. From the included studies, however, only five RCTs and one cohort study report on the participants´ medication status [11, 12, 31, 37, 38, 49].

From an evolutionary perspective, the human physiology is inherently associated with a moderate to high level of physical activity which affects many interconnected or rather soluble cellular systems such as central nervous system (CNS) structures and the immune system [19, 80]. Since MS is a neuroinflammatory disease, mitigation of both peripheral/central inflammation and neurodegeneration is of high importance. In this context, exercise represents a proposed therapy approach without side effects [5, 6]. Three weeks of exercise during an inpatient rehabilitation have been shown to decrease the systematic inflammatory index and the neutrophil-to-lymphocyte ratio in pwMS [81], the latter being associated with disease-specific symptoms, the EDSS, and disease activity [82, 83]. Interestingly, only high-intensity interval training reached significant results compared to moderate exercise. The same dose-dependent effect was shown for the reduction of matrix metalloproteinase-2 in another study with a similar study design [8], assuming that a higher cardiorespiratory stimulus over 3 weeks beneficially modulates blood–brain barrier integrity and decreases translocation of inflammatory immune cells into the CNS. Corresponding to these findings, results from a mouse model of experimental MS revealed exercise-induced inhibition of a decreased tight junction protein expression in the CNS observed in non-exercising mice [16]. Another animal study revealed significantly reduced demyelination and infiltration of proinflammatory Th17 cells into the CNS, whereas anti-inflammatory regulatory CD4⁺ T cells were enriched [15]. Again, high-intensity exercise was superior to moderate exercise, thereby adding evidence that a high cardiovascular stimulus over a certain time period may be superior to reach anti-inflammatory and neuroprotective effects. The exercise-induced intermittent metabolic stress is proposed to enhance neuronal survival, resilience, and plasticity through ketone body-mediated signaling [71]. The increased expression of the brain-derived neurotrophic factor plays an important role in this context. Indeed, increases in serum levels of brain-derived neurotrophic factor, also observed in pwMS in response to exercise [9], may be the result of epigenetic alterations induced by ketone bodies [84].

Another potential mechanism that may underlie exercise-mediated benefits in pwMS is the elevated metabolic flux of tryptophan degradation toward the immunosuppressive and neuroprotective end product kynurenic acid. Animal [85] and human [86] studies have demonstrated that exercise increases the flux along the metabolic kynurenine pathway to kynurenic acid, thereby preventing a pathological accumulation of kynurenine and quinolinic acid as well as increasing the availability of anti-inflammatory mediators. However, a better understanding of exercise-induced kynurenine pathway alterations in pwMS is needed to conclude on its contribution to improvements in symptoms or possibly also in disease progression.

Despite the increasing number of publications in the field, this review identified a lack of high quality evidence that exercise exerts consistent beneficial effects on several clinical measures of disease progression and activity. This could be due to the low sample size, heterogeneity of the included study population, and short duration of existing studies. Indeed, due to the short duration (averageing 13.4 weeks) compared to clinical phase III trials, only short-term effects were assessed. This raises the question if the available studies are appropriate to address the research question whether exercise affects established clinical measures of disease activity and progression in pwMS, since longer observations periods are needed to validly evaluate exercise-induced effects the relapse frequency, brain volume changes, or the EDSS score as discussed above. Further, only one exercise study assessed clinical measures of disease progression and activity as a primary outcome [12] which substantiates the underpowering of existing RCTs. In addition, the outcomes mainly do not correspond to outcomes used in phase III clinical trials to quantify disease progression or activity such as the relapse rate, changes in Gadolinium-enhancing T1- or T2-hyperintense lesion load, or brain atrophy [10]. Surprisingly, the MSFC was not considered in any cross-sectional or cohort study despite the fact that this clinical measure is commonly used in clinical trials and shows profound correlation with other clinically relevant measures such as the EDSS, relapse rate, white matter lesion load, and brain volumetric measures [10]. Despite the abundance of cross-sectional studies indicating beneficial effects of a high level of physical activity or fitness level on clinical measures of MS, it needs be considered that there is no causality within these associations and that negative associations between the EDSS and the physical activity or fitness level are somehow reasonable as discussed respectively in the “Physical activity” and “Fitness” sections within the discussion.