A systematic review and meta-analysis of the effects of non-pharmacological interventions on quality of life in adults with multiple sclerosis

This systematic review and meta-analysis utilized the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) recommendations [13]. A systematic search reviewed all peer reviewed articles in English from inception of the database until March 4, 2022. Scopus, Web of Science, Cumulative Index of Nursing and Allied Health Literature (CINAHL Plus), The Cochrane Library, Medline and Embase were searched. Articles were selected based on broad keywords identified within the title and abstract of the publications. The following keywords and Boolean search criteria were implemented (‘multiple sclerosis’ OR ‘encephalomyelitis disseminate’ OR ‘demyelinating’) AND (‘Health Status Questionnaire’ OR ‘SF-36’ OR ‘Multiple Sclerosis Quality of Life-54’ OR ‘MSQOL-54’ OR ‘Multiple sclerosis quality of life inventory’ OR ‘MSQLI’). The full search strings are available in Additional file 1 for each database. The search was limited to three validated QoL measurement tools, the Multiple Sclerosis quality of life inventory (MSQLI), MSQOL-54 and SF-36. The search was not expanded to other validated tools as they were judged to differ widely in aspects such as their complexity, aspects of QoL measured, completion time and recall period. Moreover, some of the commonly used MS specific QoL tools measure aspects that are not covered in others making them not directly comparable instruments [9].

All articles were independently evaluated by two reviewers to assess eligibility and disagreements were discussed, as required, with a third reviewer to decide. We only included articles which performed studies on adult patients with MS and did not focus on those with additional comorbidities (Table 1). If patients within the study had other comorbidities the study was still included unless the particular co-morbidity was an inclusion criterion to participate. Studies evaluating the impact of acute clinical care, such as the evaluation of nursing practices, were not included. We define ‘non-pharmacological’ as any intervention used to improve quality of life without a pharmaceutical or surgical modality (dietary supplements, vitamins and nutraceuticals were excluded from the definition of a pharmaceutical agent). We categorized non-pharmacological interventions into five broad categories; physical activity, behavioral or psychological, tissue manipulation, nutraceuticals/supplement, diet, and other non-pharmacological interventions.

For the meta-analysis, we only included studies with a comparator arm including no intervention, usual/standard care (such as the continuation of a previous intervention), placebo, or a minor non-active intervention (such as education materials). Studies with multiple arms not containing a control group were reviewed and synthesized separately from the meta-analysis portion.

This review only included randomized control studies. For both the intervention and control arm, studies had to measure QoL at baseline and at another time point. Data on pre- and post-QoL scores such as mean, standard deviation (SD), standard error (SE), confidence intervals, or P-values had to be provided. Cross-over designs were not included due to (i) the potential issue of a carry-over effect which could confound the result and (ii) the difficulty in accounting for these long-term effects [14].

The following information was extracted from the studies: study author, year of publication, country-region of the study, whether the study was conducted in a single center or multicenter, sample size, participants’ MS type, inclusion criteria for the Kurtzke Expanded Disability Status Scale (EDSS), mean age and range of participants, percentage of female participants, a description of the intervention and control, frequency and duration of the intervention, QoL measurement tool used, a statement summarizing the main results related to QoL changes, and metrics associated with the results (pre- and post-scores, mean difference, SD, SE, etc.). If QoL was measured at multiple time points after baseline only the last reading was extracted. The mean difference between two timepoints was computed as the arithmetic difference, and a pooled SD was calculated using the initial and follow-up time points assuming normal distribution. The pooled SD was calculated by summing the variances of each time point and taking the square root.

There were two main outcomes of interest related to the measurement tool post intervention (SF-36 and MSQOL-54); the reported change in the physical composite score (PCS) and the reported change in the mental composite score (MCS). The PCS is a representation of role limitations due to physical health, including bodily pain, energy/fatigue, sexual function, social function, and health distress, and the MCS is a representation of role limitations due to mental health, including overall quality of life, emotional well-being, social function and vitality [15].

Studies included in the meta-analysis portion of the review were assessed using the standardized mean difference between the intervention and control arm as the chosen effect size. For those studies not reporting a PCS score, the physical functioning score was used as it contributes the most weight to the scoring of the PCS [15, 16]. Similarly for the MCS score, the mental health score or emotional well-being subscale was used in cases where it was not available [15, 16].

A random effect model was selected to accommodate the likely heterogeneity in the populations included in the studies. The meta-analysis was stratified by the type of non-pharmacological intervention and done for both the PCS and MCS separately utilizing RevMan 5.4.1 [17]. A pooled effect for each non-pharmacological subgroup was calculated with 95% confidence intervals and a corresponding overall effect across all studies was also performed. An I² statistic was calculated for each non-pharmacological intervention subgroup as well as overall to assess heterogeneity. I² statistics of 0–40%, 30–60%, 50–90% and 75–100% were considered as; might not be important, moderate heterogeneity, substantial heterogeneity and considerable heterogeneity, respectively [18]. A sensitivity analysis was incorporated by excluding a study from each subgroup and examining its effect on the I² statistic and SMD. Publication bias was analyzed and discussed by visual inspection of funnel plots. For those studies for which the mean was not within the 95% CI of the overall effect as per the forest plot, a t-test was performed using STATA 15.1, with a null hypothesis that the SMD in that study was equal to the overall/pooled standardized mean difference observed across all studies [19].

Studies included were evaluated using the Joanna Briggs Institute (JBI) critical appraisal checklist for randomized control trials (RCTs), a tool used to measure the methodological quality and risk of bias of a study [20, 21]. The completed JBI checklist is available in Additional file 2. Studies were deemed of insufficient quality and excluded if greater than 6 questions on the checklist were answered ‘No’.

A total of 2830 articles were identified from Scopus, Web of Science, CINAHL Plus, The Cochrane Library, Medline and Embase. Of these 1657 were identified as duplicate reports via Mendeley’s duplicate identification tool and through manual review. An additional 990 studies were excluded by two reviewers after reviewing the abstracts and titles. Full text review was conducted for the remaining 183 studies and of these 41 were deemed eligible to include in the current review (Fig. 1).

For the quantitative review, there were 30 applicable studies identified published between 2002 and 2021. The duration of therapy ranged from 3 days to 6 months with 80% of studies being exposed to the non-pharmacological intervention for greater than 2 months. Sample sizes ranged from 11 to 169 participants and the total number of participants across all studies included was 2089. The most common study setting was Iran, 11, then there were 6 studies from America, 3 from the United Kingdom, 3 from Italy and the remaining 7 studies were done in Germany, Iceland, Netherlands, Denmark, Finland and Turkey (Fig. 2). 66% of studies used MSQOL-54, an MS-specific instrument to measure HRQoL and the rest utilized SF-36, a generic HRQoL measurement tool. The average age of participants was 42.1 years old and the average percentage of females in a study was 78% (range 54–100%). Most studies, 77%, either did not preclude participation based on their EDSS score or had inclusion criteria less than 5.5 indicating the patient’s ability to walk without aid or rest for 100 m [22]. The remaining 23% of studies had inclusion criteria with EDSS scores > 5.5 indicating the needs for assistance and a more severe disability [22]. Further details of each included study as well as a summary of their main results related to the QoL outcome are presented in Table 2.

A full review of the methodological quality and risk of bias for the 30 RCTs included in the meta-analysis is summarized in Fig. 3, further details are available in Additional file 2. Although all studies are RCTs, it was unclear if true random assignment was utilized for 11 of the studies [23‐32]. The control arms and intervention arms were similar at baseline for 27 out of 30 studies. Differences between the intervention arm and control arm were in the majority (80%) of studies only attributable to the intervention assignment and not other factors such as baseline characteristics or follow-up frequency. For 11 studies there were differences between arms in the number of withdrawals and incomplete outcome data, and these differences were not sufficiently described [27, 33‐42]. However, 23 studies did conduct an intention-to-treat analysis which examined the effects based on the initial allocation of a participant. Namjooyan et al. [40] did not calculate SD correctly and this was recalculated based on the confidence interval presented. No studies were excluded on the basis of low methodological quality or a high risk of bias.

There were 16 physical activity intervention studies in the meta-analysis, and activities involved rehabilitations programs [43, 44], yoga [23, 36, 45], aerobic and strength exercises [26‐28, 32‐34, 46], aquatic exercise [47], balance and eye-movement exercises [35], treadmill training and cycling programs [24, 25, 36] (Fig. 4). The physical component of QoL saw an overall positive effect with an SMD of 0.40 (95% CI 0.14, 0.66), however, substantial heterogeneity was observed (I² = 69%, P < 0.001). Hebert et al. [35] which utilized a specifically created regiment of balance and eye-movement exercises for people with MS (BEEMS) for 4 months showed a larger SMD of 1.71 (95% CI 1.22, 2.20). Kargarfard et al. [47] which utilized a supervised aquatic exercise program for 8 weeks also showed a larger SMD of 2.38 (1.21, 3.55). Performing a sensitivity analysis by excluding Hebert et al. still showed a significant pooled effect of physical activity interventions on the PCS, SMD 0.24 (95% CI 0.06, 0.42) with an (I² = 32%, P = 0.09). An overall positive effect was also seen on the mental component of QoL with an SMD of 0.31 (95% CI 0.08, 0.55), however, substantial heterogeneity was observed (I² = 62%, P < 0.001). Hebert et al. [35] and Kargarfard et al. [47] showed a larger SMD of 1.63 (95% CI 1.15, 2.12) and 1.96 (95% CI 0.88, 3.04) respectively. A sensitivity analysis excluding Hebert et al. still showed a significant pooled effect of physical activity interventions on the MCS, SMD to 0.18 (95% CI 0.04, 0.32) with an (I² = 0%, P = 0.52). The changes in the I² statistics indicate that most of the heterogeneity was likely due to Hebert et al. for both PCS and MCS.

There were 5 behavioral and psychological interventions studies in the meta-analysis and activities included brain training programs [37], a short social cognitive treatment [48], self-care programs and lifestyle change classes [38, 39, 49]. The physical component of QoL did not see a significant overall effect with an SMD of 0.22 (95% CI − 0.19, 0.62) with substantial heterogeneity observed (I² = 81%, P < 0.001). Momenabadi et al. [49] which investigated training sessions educating participants on health-promoting self-care behaviors lasting 4 months showed a significant effect on PCS, SMD of 1.19 (95% CI 0.71, 1.67). A sensitivity analysis excluding Momenabadi et al. still showed no significant pooled effect of behavioral/psychological interventions on the PCS, SMD 0.01 (95% CI − 0.17, 0.19) with an (I² = 0%, P = 0.94). A significant overall effect was also not seen on the mental component of QoL with an SMD of 0.38 (95% CI − 0.04, 0.79) with substantial heterogeneity observed (I² = 81%, P ≤ 0.001). Momenabadi et al. also showed a large significant effect on MCS, SMD of 1.35 (95% CI 0.86, 1.83). A sensitivity analysis excluding Momenabadi et al. still showed no significant pooled effect of behavioral/psychological interventions on the MCS, SMD 0.13 (95% CI − 0.05, 0.31) with an (I² = 0%, P = 0.91). The changes in the I² statistics indicate that almost all of the heterogeneity was likely due to Momenabadi et al. for both PCS and MCS.

There were 5 nutraceutical and supplements studies in the meta-analysis and included vitamin D [29], Korean ginseng tablets [50], folic acid tablets & vitamin B12 injections [30], grape seed extract capsules [51] and a traditional formulation containing cinnamon, ajwain and Iranian boragom [40]. The interventions ranged in duration from 1 to 3 months. An overall positive effect was seen on the physical component of QoL with an SMD of 0.78 (95% CI 0.34, 1.22), however, substantial heterogeneity was observed (I² = 71%, P = 0.007). Ashtari et al. [29] was the only study within the subgroup that did not show a significant effect on PCS, SMD of 0.25 (95% CI − 0.15, 0.66). Performing a sensitivity analysis by excluding Namjooyan et al. [40] still showed a significant pooled effect of nutraceutical/supplement interventions on the PCS, SMD 0.54 (95% CI 0.28, 0.80) with an (I² = 9%, P < 0.001). The change in the I² statistic indicates that most of the heterogeneity was likely due to Namjooyan et al. An overall positive effect was also seen on the mental component of QoL with an SMD of 0.86 (95% CI 0.11, 1.62), however, considerable heterogeneity was observed (I² = 90%, P < 0.001). A sensitivity analysis was done but no single study contributed significantly to the heterogeneity in MCS. Korean ginseng, the traditional formulation containing cinnamon, ajwain and Iranian boragom, and grape seed extract capsules all showed a significant effect on the mental health component of QoL; SMD 0.62 (95% CI 0.062, 1.17), SMD 2.15 (95% CI 1.45, 2.85), and SMD 1.45 (95% CI 0.88, 1.97), respectively.

For interventions involving a change in diet, only one study was included in the meta-analysis and involved a modified Mediterranean diet, lasting 6 months [41]. A positive effect was seen on the physical component of QoL with an SMD of 0.42 (95% CI 0.09, 0.75). A significant effect was not seen on the mental component of QoL with an SMD of 0.28 (95% CI − 0.04, 0.61).

There was one study included as part of the tissue manipulation category and involved reflexology, lasting 3 months [52]. A positive effect was seen on both the physical component of QoL with an SMD of 0.79 (95% CI 0.26, 1.32) and the mental component of QoL with an SMD of 0.90 (95% CI 0.36, 1.43).

Other interventions included a naturopathic medicine regimen [31], MS education [31], and hippotherapy [42]. A positive effect was seen on the physical component of QoL with an SMD of 0.44 (95% CI 0.07, 0.81). Only the Vermohlen et al. study [42], which utilized hippotherapy once a week for 3 months, showed a significant improvement in PCS, SMD of 0.62 (95% CI 0.08, 1.16). A significant overall effect was not seen on the mental component QoL with an SMD of 0.33 (95% CI − 0.06, 0.71). However, the Vermohlen et al. study showed an improvement in MCS, SMD of 0.63 (95% CI 0.09, 1.17).

For the effect of non-pharmacological intervention on the physical health component of QoL there were 4 studies for which the SMD was outside the 95% CI of the pooled overall effect (Fig. 4) [35, 40, 47, 49]. For each of these studies a t-test was performed with a null hypothesis that the SMD was no different than the overall pooled SMD of 0.44 (SD: 4.08). Hebert et al. and Kargarfard et al. both showed strong evidence against the null hypothesis (P-value = 0.0042 and 0.0439 respectively). Namjooyan et al. and Momenabadi et al. showed weak evidence against the null hypothesis and it is possible that the difference in SMD was due to chance (P-value = 0.0692 and 0.4323, respectively).

For the effect of non-pharmacological intervention on the mental health component of QoL there were 5 studies for which the SMD was outside the 95% CI of the pooled overall effect (Fig. 4) [35, 40, 47, 49, 51]. For each of these studies, a t-test was performed with a null hypothesis that the SMD was no different than the overall pooled SMD of 0.42 (SD: 4.20). Hebert et al. showed strong evidence against the null hypothesis (P-value = 0.0085). Siahpoosh et al. (P-value = 0.0955), Namjooyan et al. (P-value = 0.0525), Momenabadi et al. (P-value = 0.3718) and Kargarfard et al. (P-value = 0.1455) showed weaker evidence against the null hypothesis and it is possible that the differences in SMD were due to chance.

Summary of studies not included in meta-analysis—qualitative synthesis: there were 11 studies not included in the quantitative synthesis which are described in detail in Additional file 3. Seven of these studies compared physical regiments [53‐59], two studies compared behavioral/psychological regimens [60, 61], one study (Weinstock-Guttman et al.) compared supplements [62] and one study examined the effects of transcranial random noise stimulation (tRNS) (Salemi et al.) [63]. Nine studies showed an improvement in some aspects of QoL, with the exceptions of Pilutti et al. [58] and Plow et al. [61]. Of these, five studies did not show a statistically significant difference in the effect between the groups. Of note Khalil et al. [55] considered the impact of a virtual reality exercise program two times per week and 1 balance exercise session at home or traditional balance exercises at home without virtual reality over a period of 6 weeks. The results showed that the arm incorporating virtual reality had a significantly larger effect on both PCS and MCS than traditional balance exercises (P-value < 0.05)[62]. Solari et al. [59] considered the impact of an inpatient rehabilitation program consisting of daily exercises two times per day lasting 45 min or a home exercise program for 3 weeks. The inpatient rehabilitation group improved significantly more in the MCS and this was sustained at 3 and 9 weeks (P-value = 0.001) [66]. Impellizzeri et al. [60] considered the impact of a conventional cognitive rehabilitation (CCR) and neurologic music therapy (NMT) or just CCR 6 times per week for 8 weeks. Results showed a significantly greater improvement in mental health in the CCR and NMT group as compared to the only CCR group (P-value < 0.001) [67]. Salemi et al. [63] considered the impact of tRNS applied 15 min daily for 2 weeks or a sham control group [70]. The results showed a significant increase in the tRNS group in both the ‘change in health’ and role limitation due to physical problems’ subscales, P-value = 0.006 and 0.001, respectively [63].

The funnel plot of the studies reporting a PCS outcome is presented in Additional file 4. There is asymmetry in the plot implying that there is the possibility of publication bias [71]. The gap in the bottom left corner of the plot indicates that studies with a larger SE which found no significant effect on PCS may have gone unpublished [64]. The funnel plot of the studies reporting an MCS outcome is also presented in Additional file 4. This plot is more symmetrical around the mean SMD compared to the PCS outcome funnel plot, implying that there may have been some studies which found a strong negative effect but went unpublished [64].

Across the 30 studies included in the meta-analysis there was a moderate effect for both the PCS and MCS on QoL for physical interventions. Although these results had a high risk of heterogeneity, most was contributed by a single study [Hebert et al.] and when this was removed a moderate effect was still observed for both PCS and MCS. Aquatic exercise [47] and balance and eye movement exercises (BEEMS) [35] showed a larger effect on PCS and MCS than the pooled effect within the physical intervention category (Fig. 4). When a t-test was performed comparing the SMD of individual studies and the overall pooled SMD in PCS, BEEMS and aquatic exercise both showed a statistically significant result indicating high probability that the difference was not due to chance. BEEMS also showed a significantly different effect on the MCS as compared to the overall pooled effect. BEEMS is a promising finding as previous studies have also shown a positive effect on patients with MS in aspects such as fatigue, posture control and disability [65‐68]. Dizziness and instability are significant symptoms in patients with MS [69] and perhaps improving these over the course of the 4-month BEEMS program contributed greatly to the larger increase in MCS/PCS observed as compared to physical interventions which focused exclusively on either aerobic or resistance training.

Behavioral and psychological interventions did not show an overall pooled effect on PCS or MCS. Although there was a substantial amount of heterogeneity, it was mostly contributed by one study, Momenabadi et al. [49]. Conversely, the results of a review conducted by Gil-Gonzalez et al. [10] did comment on an effect of these types of interventions on QoL. This was a surprising finding, however, although some of the studies did report a significant effect on MCS in the intervention arm, not all studies included in the review by Gil-Gonzalez [10] included a control arm and this may have contributed to this apparent discrepancy.

Nutraceuticals and supplements overall showed an improvement in the PCS and MCS even after accounting for heterogeneity between studies, although the source of heterogeneity was unable to be determined for the effects on MCS. Therefore, it is likely that this subgroup is a good option to increase PCS and a possible one to improve MCS in patients with MS. Caution is recommended when generalizing these findings as all studies in this subgroup were done in a single country, Iran. The results in the diet intervention group, involving a modified Mediterranean diet, showed a moderate effect on PCS and no effect on MCS. This result needs to be interpreted with caution as the control arm was a traditional Iranian diet [41] which may differ substantially from other diets around the world such as the typical western diet. The tissue manipulation category involving reflexology demonstrated a benefit to both PCS and MCS. However, the tissue manipulation subgroup only included one single country study with a small sample size [52]. Thus, the conclusions of the study may not be widely generalizable and require further investigation.

The remaining studies not falling into any other categories were naturopathic medicine regimen [31], MS education [31], and hippotherapy [42]. The results were pooled, however, the interventions varied substantially and thus do not warrant discussion as a summary estimate. Hippotherapy showed a positive improvement in PCS, SMD of 0.62 (95% CI 0.08, 1.16) and although this is only one study a previous non-RCT also showed a positive effect of hippotherapy on aspects such as pain, muscle tension and balance [70]. Thus, it would be worthwhile for future studies to further explore the efficacy of hippotherapy on MS and QoL.

The results of the qualitative review, including 11 studies, were complimentary to that of the meta-analysis and quantitative synthesis. The Gandolfi et al. [54] study which incorporated SIBT in one of the arms provides further evidence for the effectiveness and importance of balance training for people with MS. Khalil et al. and Munari et al. both incorporated virtual reality (VR) into the physical intervention and had positive outcomes on QoL [55‐57]. Pilutti et al. which investigated a stepper training program in one arm and a treadmill training in the other arm found no significant effects on any QoL outcome for either of these interventions [58]. It is possible, however, that no effect was observed due to the fact that all patients had progressive MS and were significantly disabled (EDSS score 6–8). This may indicate that perhaps exercise interventions may diminish in effectiveness in those with a more disabling disease status. Impellizzeri et al. added neurologic music therapy (NMT) to conventional cognitive rehabilitation (CCR) and this showed a greater improvement in mental health indicating the need to further explore NMT [60]. Another unique intervention was transcranial random noise stimulation (tRNS), a form of non-invasive brain stimulation, that showed statistically significant improvements in the ‘change in health’ and ‘role limitation due to physical problems’ subscales [63]. tRNS also needs to be further studied in MS patients with a larger sample size to more concretely understand its effects on QoL.

The systematic review was successful in identifying 1173 studies for screening which speaks to the wide and broad search criteria. This led to 41 RCTs of similar design to be included in this review which allowed for the ability to make robust and useful discoveries. There was minimal selection bias introduced by restricting the review to SF-36, MSQLI, MSQoL-54 as these are the most widely used and comparable measure of HR-QoL in MS. Non-pharmacological studies were of similar duration, with 77% being administered for 2 months or greater which allowed for a fair comparison to be made across interventions and provided further support to pool the results. Also, the studies included in the meta-analysis were overall representative of the global MS population, with studies covering 10 diverse countries (Fig. 2) [71] and being majority female in representation (range 54–100% female in studies compared to 69% globally). The mean age of a participant in this review at 42.1 years old with a SD of 7.7 was comparable to the global mean age of MS diagnosis of 32 years old [3]. There was likely some selection bias within the studies; this is because for 11 of the studies we are not certain if true randomization was used and for 15 studies it was unclear if allocation was concealed. For those studies administering a nutraceutical/supplement sufficient blinding of participants was maintained using a placebo [29, 40, 50, 51]. However, in 13 studies those assessing QoL were not blind to assignment, and this may have introduced detection bias overestimating the results [23, 25, 28, 33, 34, 39, 45, 48, 49, 51, 52, 72]. In addition, for many studies it was not feasible to blind participants to the interventions and this may have contributed significant bias and led to an overestimation of the true benefit. Confounding was not deemed to be an issue in this study as potential confounders such as age or EDSS score were controlled for via randomization. Finally, it is also possible based on visual inspection of the funnel plots that publication bias was present and thus studies that showed a negative effect of non-pharmacological interventions went unpublished. However, this may have been due to heterogeneity in the interventions, for example, the majority of nutraceutical/supplement studies showed a statistically significant effect on PCS whereas the majority of behavioral/psychological studies showed no statistically significant effect on PCS. Overall, the quality of the studies was good as all were RCTs with similar patient characteristics at baseline and the majority, 23, conducted an intention-to-treat analysis. No studies were excluded from this review on the basis of quality.

Studies were limited in size with a mean sample size of 67 patients, and this can affect the power of the studies and uncertainty of the effect size. Additionally, most studies were conducted at a single center making it difficult to generalize the outcomes to a wider region or country. There was also some heterogeneity in the disability status of patients across studies. This is noteworthy, as those with more disabling disease may be less likely to derive benefit from certain interventions and conversely those with less disabling disease may be less likely to derive benefit from interventions designed for those with a higher EDSS score. The duration of therapy and follow-up was overall quite short and given this is a chronic condition one cannot make claims that the benefits seen will be maintained over longer periods. Lastly, it should be noted that this review was not prospectively registered.