MS Fatigue Post High‐Efficacy Treatment

Authors: C. Sales, R. Patel, T. Wijeratne, A. Bruestle, N. Cherbuin

Abstract

Background

Multiple sclerosis (MS) is a major cause of disability, particularly among young adults, with fatigue affecting up to 95% of patients. Despite the availability of disease‐modifying therapies (DMTs), their impact on MS‐related fatigue (MSF) remains uncertain. This study investigates changes in fatigue levels among individuals with relapsing–remitting MS (RRMS) following the initiation of highly effective therapies (HETs).

Methods

A systematic search of MEDLINE, PubMed, Scopus, and PsycINFO identified studies involving adults with RRMS treated with HETs and assessed using validated fatigue measures at baseline and follow‐up. Data on demographics, MS duration, DMT type, and fatigue scores were extracted. Meta‐analyses using a random‐effects model calculated standardized mean differences (SMD) in fatigue scores post‐treatment.

Results

Eighteen studies comprising 4138 RRMS patients and 3806 person‐years of follow‐up were included. The overall SMD was −0.34 (95% CI −0.47 to −0.21), indicating a small but significant reduction in fatigue. Continuous treatments (e.g., natalizumab, ocrelizumab, fingolimod) showed significant improvements, while immune reconstituting therapies (e.g., alemtuzumab) did not. Among fatigue domains, only physical fatigue showed a significant reduction, particularly with natalizumab (SMD = −1.25; 95% CI −2.43 to −0.06) and when assessed using the fatigue scale for motor and cognitive functions scale (SMD = −1.52; 95% CI −2.87 to −0.17).

Conclusion

MS‐related fatigue levels, especially in its physical aspect, decrease after the initiation of HET. This finding reinforced the role of neuroinflammation in driving fatigue and highlighted the need for domain‐specific research and treatment strategies.

Introduction

Multiple sclerosis (MS) is a leading cause of disability worldwide [1]. It affects young individuals in the peak of their productive years, and this most often compromises their working capacity. A study across 16 European countries found that work capacity in people with MS (pwMS) drops from 82% at an expanded disability status scale (EDSS) score of 0%–8% at EDSS 9, with MS‐related fatigue (MS‐F) as a key driver [2]. Beyond employment, MS‐F has been found to be the most disabling MS‐related complication [3, 4].

The prevalence of MS‐F ranges from 55% to 77% [2, 5, 6, 7] in registry‐based studies, and up to 95% in epidemiological studies [8]. Although there are several validated MS‐F measures such as the fatigue severity score (FSS), modified fatigue impact scale (MFIS), and fatigue scale for motor and cognitive functions (FSMC), they are not routinely included as outcome measures in MS clinical trials. With research becoming increasingly patient‐centred, the use of patient‐reported outcomes (PROs) has grown. However, PROs have their own set of disadvantages and tend to be more subject to recall, language, timing, and fatigue biases [9]. Nonetheless, the use of PROs is now being encouraged as it provides an opportunity to estimate the impact of a medical condition from the patient's perspective, which is not necessarily appraised by clinicians [10].

In the last two decades, research on new and repurposed treatments for MS has been extensive. These treatments, including immunomodulators, sphingosine‐1 (S1P) inhibitors, B and T cell therapies, alpha integrin inhibitors, and immune reconstituting medications, have greatly reduced the risk of relapse and disease progression. Of those, alemtuzumab, cladribine, natalizumab, ocrelizumab, ofatumumab, and S1P modulators (fingolimod, ozanimod, ponesimod) are considered highly effective therapies (HET) based on the recommendation of the Multiple Sclerosis Therapy Consensus Group [11]. While these treatments are highly effective in reducing the risk of relapse, there is ambiguity as to their impact on MS‐F. For example, a systematic review including seven randomized controlled trials, which measured the impact of disease‐modifying treatments (DMT) in improving MS‐reported fatigue, reported only two studies identifying HET (ponesimod and teriflunomide) as effective in reducing fatigue levels [12]. However, it did not cover the most utilized HETs such as ocrelizumab, natalizumab, cladribine, and fingolimod because of a lack of randomized clinical trials (RCTs) which measured MSF as an outcome. More recent reports of MSF from RCTs have now demonstrated the advantages of HET compared to low to moderate interventions [13, 14]. However, most of the studies of other HETs still come from observational studies. Therefore, a detailed synthesis of the effects of these treatments is needed because recent observational studies focusing on these therapies have sometimes demonstrated an associated improvement in fatigue [15, 16, 17], while others have not [18, 19]. This study therefore aimed to synthesize the evidence on the degree to which the initiation of HET treatment is associated with changes in the level of fatigue in pwRRMS. Another aim was to determine whether potential changes in fatigue levels differ across functional domains (global, physical, cognitive, social, and psychosocial).

Methods and Materials

This systematic review was conducted following the guidelines of the Meta‐analysis of Observational Studies in Epidemiology (MOOSE) and was registered in the PROSPERO database (CRD42025520020). Searches of electronic databases including MEDLINE, PubMed, Scopus, and PsycINFO were conducted to identify the relevant studies, and this was conducted on the 22nd of April 2024 to retrieve cross‐sectional and longitudinal studies.

Search String

The search string was developed using the PICOS approach (Population, Intervention, Comparison, Outcome and Study design). The following search string was (“multiple sclerosis” OR “relapsing–remitting” OR “relapsing‐remitting MS” OR RRMS) AND (fatig* OR asthenia OR neurasthenia OR “brain fog” OR malaise OR lassitude OR tiredness OR exhaustion) AND (treatment OR therap* OR regimen OR intervention OR medic*) AND (“disease‐modifying” OR “disease modifying” OR “highly effective” OR Natalizumab OR Tysabri OR Antegren OR Fingolimod OR Gilenya OR “FTY 720” OR Alemtuzumab OR Lemtrada OR Ozanimod OR Zeposia OR Ponesimod OR Ponvory OR Siponimod OR Mayzent OR Cladribine OR Mavenclad OR Rituximab OR Mabthera OR Rituxan OR Ocrelizumab OR Ocrevus OR Ofatumumab OR Kesimpta OR Daclizumab OR Zynbryta OR Mitoxantrone OR Pralifat OR Ublituximab OR Briumvi) AND (outcome* OR self‐report* OR scale* OR score* OR measure* OR “PRO” OR questionnaire* OR survey). Filters in PubMed were applied to exclude studies involving non‐human subjects and those not in English. The literature search was conducted without any time limitations.

Inclusion and Exclusion Criteria

Inclusion studies of (1) human samples, (2) adults aged 18 years and above, (3) relapsing–remitting MS patients on highly effective DMTs, (4) studies with a validated scoring system for measuring fatigue assessed at baseline and at least 3 months and up to 2 years after commencement of intervention.

Exclusion criteria (1) < 18 years of age, (2) progressive forms of MS, (3) mild to moderate effective DMTs, (4) MS mimics (other demyelinating disorders of the central nervous system such as clinically isolated syndrome, radiologically isolated syndrome, neuromyelitis optica, myelin oligodendrocyte disorders), (5) hematopoietic stem cell transplantation, (6) no fatigue measurement, and (7) review articles, systematic reviews, meta‐analysis, case reports, and case series.

Screening

Duplicate citations were deleted from the search results, and the remaining entries were initially screened by title by a single author (C.S.). Subsequently, all abstracts were independently double screened by using the predetermined inclusion/exclusion criteria, with any discrepancies resolved through consensus. Finally, the full‐text and supplementary materials of the remaining articles were double‐screened against the inclusion/exclusion criteria by two authors (C.S. and R.P.), and relevant data were extracted. If data were missing, authors were contacted via email to request the necessary information for inclusion in the review.

Data Extraction

Data from individual articles were independently extracted by two operators and any disagreement was resolved by consensus, or if necessary through a third assessor. The following information were (1) first author, (2) year of publication (3) countries were data was collected (4) study design (5) sample size (6) duration of MS (7) EDSS (8) percentage of females and (9) the specific measure of fatigue used.

Quality Assessment

Studsy quality was evaluated independently by two authors (C.S. and R.P.) using the Newcastle Ottawa Scale [20] which assessed study quality based on three (1) the selection of participants, (2) the comparability of groups, and (3) the assessment/ascertainment of the outcome of interest.

Statistical Analysis

Statistical analyses were conducted using the R package (4.4.1) RStudio (version 2024.09.1+394). The Metafor package was used to compute effect sizes, perform meta‐regression analyses, and generate forest plots. Ggplot2 was used to create bubble plots.

Summary Measures

The standardized mean difference (SMD) was used as the effect size of interest and was computed using the following SMD=Xpost−Xpre/SDdiffwhere X post and X pre are the mean scores after and before treatment, respectively while SDdiff is the pooled standard deviation of the differences between the pre‐treatment and post‐treatment scores. The pooled standard deviations were calculated with the PooledSD=n1−1SD12+n2−1SD22n1+n2−2

If standard deviation values were not available, standard deviation was approximated by (1) dividing the range by 4, (2) dividing the interquartile range by 1.35, or (3) using standard deviation values from other studies [21]. The standard errors of the mean and confidence intervals were transformed into standard deviations using the approach described by Higgins and Green [22].

Risk of Bias

The ROBINS‐I (Risk Of Bias In Non‐randomized Studies—of Interventions) tool was used to assess the risk of bias [23] across seven confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes, and selection of the reported result. The overall risk of bias was assessed as low, moderate, serious, or critical based on the cumulative assessment across all domains.

Meta‐Analysis

Due to the variability in sampling and methodologies expected across the studies, a random effects model using the restricted maximum likelihood estimator was applied in all analyses to estimate the average effect size distribution. Heterogeneity among the studies was evaluated using Cochran's Q statistic (with a p‐value of less than 0.01 indicating significant heterogeneity) and the I ^2^ statistic (with values of 25%, 50%, and 75% representing low, moderate, and high heterogeneity, respectively). The leave‐one‐out method was used to identify studies that significantly contributed to heterogeneity.

Results

Literature Search

The PRISMA flow diagram (Figure 1) outlines the study selection process for systematic review. Initially, 1940 studies were identified, with 276 duplicates removed, leaving 1664 studies for title screening. After excluding 1507 irrelevant titles, 157 studies underwent abstract screening, resulting in the removal of 119 irrelevant studies. A total of 19 studies met the inclusion criteria and were included in the study. Of these, 18 provided sufficient data for quantitative synthesis and were included in the final meta‐analysis.

The characteristics of studies included are presented in Table 1. These were published between the years 2005 and 2023, and reported on populations from North and South America, Asia‐Pacific, and Europe, with a total sample size of 4138 patients (weighted mean age [WMA] of 40.32 years, mean EDSS = 2.65, 71.79% females), collectively accounting for 3806 person‐years of follow‐up (PYF). The studies covered a range of DMTs including Alemtuzumab, Natalizumab, Fingolimod, Rituximab, Ponesimod, Mitoxantrone, and Ocrelizumab. Fatigue was assessed using PRO measures, including the Modified Fatigue Impact Scale (MFIS), Fatigue Severity Scale (FSS), Fatigue Scale for Motor and Cognitive Functions (FSMC), Fatigue Impact Scale (FIS), Neuro‐Quality of Life Fatigue (Neuro QOL Fatigue), and Fatigue Symptoms and Impacts Questionnaire (FSIQ). Table S1 summarizes the characteristics of each questionnaire.

Quality Assessment

Four studies (22.2%) were rated as having good methodological quality. Most studies' quality was rated as poor due to the inadequate comparability of cohorts (Table S2).

Risk of Bias Assessment

Potential study bias assessed with the ROBINS‐I tool was found to be high in most studies due to confounding, selection of participants, and measurement of outcomes. Overall, the majority of studies were rated as having a moderate risk of bias, with only a few studies rated as having a low risk (Table S3).

Overall Analysis

The meta‐analysis of 18 longitudinal studies showed an overall improvement in fatigue scores after treatment with highly effective DMTs. The overall standardized mean differences (SMD) pre‐ and post‐initiation of HET was −0.34 (−0.47; −0.21), indicating a small but significant reduction in fatigue scores after treatment (Figure 2). The funnel plot is nearly symmetric, indicating low publication bias (Figure S1). The Egger's test revealed no significant study effect bias (t‐test = 0.22, p‐value = 0.8275, bias estimate = 0.0470, SE = 0.2124).

Subgroup Analysis

Stratified analysis of continuous (natalizumab, ocrelizumab, rituximab, fingolimod, ponesimod, and mitoxantrone) and Immune Reconstituting Treatments (IRT) showed that continuous treatments were associated with a statistically significant decrease in fatigue (SMD = −0.34 [−0.48; −0.20]), while the IRT, which all used Alemtuzumab, were not (Figure 2).

Further analysis of studies according to the type of patient‐reported outcome and treatment duration was also performed. Among various measures of fatigue, only FSMC, FSS, and MFIS 5 showed a statistically significant difference in SMD scores at −0.86 (95% CI −1.58 to −0.14), −0.30 (95% CI −0.4 to −0.19), and −0.53 (95% CI −0.75 to −0.30), respectively (Figure 3a). An effect of the length of treatment was also detected. The SMD in fatigue scores after 6–12 months of treatment was −0.28 (95% CI −0.43 to −0.13), while that of 18–24 months was at −0.48 (95% CI −0.68 to −0.29), both indicating statistically significant differences, although treatment effect was slightly greater in the latter (Figure 3b). A meta‐analysis of studies where patients previously treated with beta‐interferon and/or glatiramer acetate switched to HET showed a statistically significant difference at both 6 to 12 (SMD = −0.31, 95% CI −0.39 to −0.23) and 18 to 24 (SMD = −0.88, 95% CI −0.88 to −0.13) months of treatment with a larger effect for the latter (Figure 4). Assessment of the effect of each drug according to mechanism subclass was also performed, showing that B cell treatments (ocrelizumab, rituximab) have the greatest effect (SMD = −0.530, SD = 0.524) in terms of fatigue improvement, while S1P inhibitors (Fingolimod, Ponesimod) were the least effective (SMD = −0.180, SD = 0.140) (Figure 5).

Meta‐Regression Analysis

The possible moderating effects of age, EDSS, and duration of treatment in the associations between specific types of continuous treatment and fatigue were investigated by meta‐regression and showed no significant effect of these variables (Figure 6). Analysis showed that there was no statistically significant relationship between age and standardized mean differences in fatigue scores before and after treatment (regression coefficient = −0.0211, p value = 0.8495). The same trend was observed for both EDSS (regression coefficient = −0.0614, p value = 0.9059) and duration of MS (regression coefficient = −0.0278, p value = 0.6049) (Table S4). Similarly, no effect of the duration of MS (< 10 and ≥ 10 years), age group (< 40 and ≥ 40), or EDSS (< 3 and ≥ 3) was detected (Figure 7). Greater reduction in fatigue scores was noted among those with longer durations of MS (≥ 10 years), older ages (≥ 40 years old), and those with EDSS of ≥ 3, although the mean differences between the groups were not statistically significant (p value for duration of MS = 0.12, p value for age = 0.52, p value for EDSS = 0.62).

Assessment of Fatigue Domains

The effect of HET on various fatigue domains such as physical, cognitive, and psychological was also evaluated and detailed in Figure 8. There was no statistically significant difference in the overall effect with physical fatigue measures (SMD = −0.98, 95% CI −2.02 to 0.96), although subgroup analysis according to DMT indicates that natalizumab resulted in a statistically significant reduction in SMD with an overall effect of −1.25 (95% CI −2.43 to 00.06). The overall effect on cognitive (SMD = −0.29, 95% CI −1.37 to 0.79) and psychological domains (SMD = −0.29, 95% CI −1.37 to 0.79) was not significant irrespective of DMT. Subgroup analysis according to the type of PRO was also performed, which did not show any statistically significant differences in the overall effect except for FSMC motor with SMD of −1.52 (95% CI −2.87 to −0.17) (Figure S2).

Discussion

This meta‐analysis provides a comprehensive synthesis of the changes in MSF among pwRRMS treated with HET. The key finding is a small reduction in fatigue following HETs, which may still be clinically meaningful in chronic neurological conditions like MS. This reduction was most evident in the physical fatigue domain, while changes in cognitive and psychosocial fatigue were less consistent. Notably, continuous treatments such as ocrelizumab, natalizumab, and fingolimod were associated with greater improvements than immune reconstituting therapies like alemtuzumab.

Several mechanisms may explain these findings. First, HETs may exert a direct anti‐inflammatory effect that alleviates fatigue. Fatigue in MS is closely linked to neuroinflammation, with elevated levels of pro‐inflammatory cytokines such as TNF‐α and interleukins being associated with fatigue severity [36, 37]. By targeting these inflammatory pathways, HETs may reduce the systemic and central nervous system inflammation that contributes to fatigue. The stronger effects observed with continuous therapies may reflect their ability to maintain consistent immunomodulation, as opposed to the more episodic immune resetting seen with immune reconstitution therapies. This is particularly relevant in the context of physical fatigue, which is thought to be more directly influenced by inflammatory processes affecting motor pathways [38].

Second, differences in disease severity and baseline disability may influence treatment response. In this study, patients receiving continuous treatments had lower baseline EDSS scores (mean ~2.5) compared to those on IRTs (~3.4), suggesting that inflammation‐driven fatigue may be more responsive to HET in earlier disease stages. In contrast, patients with more advanced disease may experience fatigue that is increasingly driven by neurodegenerative mechanisms, such as axonal loss and cortical atrophy, which are less amenable to anti‐inflammatory therapies [39, 40]. This distinction aligns with the hypothesis that MS‐related fatigue evolves over time, with inflammatory mechanisms predominating early, and neurodegenerative processes becoming more prominent later [26].

Third, the observed improvements may partly reflect relief from fatigue‐inducing effects of prior treatments. Many patients in the included studies had switched to HETs from lower‐efficacy therapies such as beta‐interferons or glatiramer acetate, both of which are known to cause fatigue as a side effect [13, 41, 42]. The transition to HETs may therefore result in an immediate improvement in fatigue simply by removing these adverse effects. However, the sustained reduction in fatigue observed over 18–24 months suggests that the possible benefits of HETs may extend beyond this initial relief. This longer‐term improvement supports the notion that HETs may also address underlying disease mechanisms contributing to fatigue, rather than merely alleviating treatment‐related side effects.

Fourth, variability in fatigue measurement tools may have influenced the results. Only studies using certain scales—such as the FSMC, FSS, and MFIS‐5—showed statistically significant changes in fatigue scores. This raises the possibility that some instruments are more sensitive to treatment effects than others. For example, FSMC's motor subscale detected significant improvements, aligning with the observed reduction in physical fatigue. In contrast, broader or multidimensional tools like the full MFIS‐21 may dilute domain‐specific effects, making it harder to detect meaningful changes [43]. This highlights the importance of selecting appropriate, domain‐sensitive instruments when evaluating fatigue in MS.

Finally, most included studies were observational; non‐specific or expectancy effects could have influenced fatigue reporting. Patients' awareness of receiving a more potent therapy may have shaped their perceptions of fatigue, particularly in the absence of blinding. However, the consistency of findings across multiple cohorts, longer duration, and fatigue domains strengthens the case for a genuine treatment effect. Moreover, the lack of improvement in cognitive and psychosocial fatigue—domains more susceptible to subjective bias—suggests that the observed changes in physical fatigue are less likely to be explained by placebo effects alone. These findings support the growing consensus that early, aggressive treatment of MS is warranted, as physical fatigue appears to be closely linked to inflammatory activity and may be more responsive to HET.

Taken together, the most plausible explanation of the present findings is that HETs reduce MS‐related fatigue primarily through their anti‐inflammatory action, particularly in patients with lower disability and inflammation‐driven fatigue.

The lack of improvement in cognitive and psychosocial fatigue may reflect the influence of non‐inflammatory factors such as mood, personality traits, and environmental stressors, which are less amenable to pharmacological intervention [44, 45, 46]. These domains may require complementary strategies, including psychological support, cognitive rehabilitation, and lifestyle modifications. Moreover, emerging evidence suggests that impaired neurogenesis in limbic regions, such as the hippocampus—critical for mood regulation and memory—along with broader neurodegenerative changes, may contribute to persistent fatigue in more advanced stages of MS [47, 48].

Clinical Implications and Future Directions

These findings have several important clinical implications. The observed reduction in MSF, particularly in the physical domain, supports the early use of HETs in RRMS to preserve function and improve quality of life. Clinicians should consider fatigue profiles—especially physical fatigue—when selecting treatment strategies and recognize that continuous therapies may offer more consistent benefits than immune reconstituting treatments.

Measuring fatigue in clinical trials has historically not been a standard endpoint, which limits our ability to assess the impact of treatments on this symptom in double‐blind studies. This is particularly concerning given how profoundly fatigue affects pwMS' quality of life. Future research should prioritize randomized controlled trials that include fatigue as a primary outcome, stratify patients by disease stage and prior treatment, and employ domain‐specific fatigue measures. Long‐term follow‐up will be essential to determine whether improvements are sustained beyond 2 years, and how pharmacological and non‐pharmacological interventions might interact to address the multifaceted nature of MSF. In addition, differentiating between fatigue subtypes may offer valuable clinical physical fatigue could serve as a more sensitive indicator of ongoing inflammatory disease activity, whereas cognitive and mood‐related fatigue may better reflect underlying neurodegenerative processes and disease progression.

Strength and Limitations

This meta‐analysis has several strengths and limitations. A major strength is its comprehensive synthesis of real‐world data from 3806 person‐years of follow‐up across diverse settings, using validated PROs. The inclusion of domain‐specific analyses and subgroup comparisons by treatment type and duration adds nuance to the findings. However, most included studies were observational and lacked control groups, limiting causal inference. Potential confounding from prior treatments and variability in fatigue measurement tools may also have influenced results. Additionally, the reliance on PROs introduces the possibility of expectancy effects, although the consistency of findings across cohorts and time points strengthens the case for a genuine treatment effect. Lastly, the interpretation of these findings must be tempered by the methodological limitations of the included studies. Only four studies (22.2%) were rated as having good methodological quality, while the majority were assessed as poor due to inadequate comparability of cohorts. This raises concerns about potential confounding and selection bias, which may have influenced the observed treatment effects. Moreover, the ROBINS‐I assessment revealed a high potential for bias in most studies, particularly in domains related to confounding, participant selection, and outcome measurement. These biases could have led to overestimation or underestimation of treatment effects, especially in the absence of randomization and blinding.

Conclusion

In conclusion, HETs may offer clinically meaningful improvements in MS‐related fatigue, particularly in its physical dimension. This supports the hypothesis that fatigue in MS is, at least in part, driven by neuroinflammatory processes that are responsive to immunomodulatory treatment. However, the complexity of fatigue—spanning physical, cognitive, and psychosocial domains—underscores the need for a multidimensional approach to treatment and research.

Author Contributions

C. Sales: conceptualization, methodology, data curation, formal analysis, writing – original draft, visualization; R. Patel: methodology, data curation, writing – reviewing and editing; T. Wijeratne: conceptualization, supervision, writing – reviewing and editing; A. Bruestle: formal analysis, investigation, writing – reviewing and editing; N. Cherbuin: conceptualization, methodology, formal analysis, writing – reviewing and editing, supervision.

Ethics Statement

Approval from the institutional review board or ethics committee was not needed for this systematic review.

Consent

This study did not collect individuals' data, and participation consent was not required.

Conflicts of Interest

C.S. has received speaker honoraria from Novartis and Merck and has also received a scholarship from the Australian Government Research Training Program. R.P. has received honoraria from Roche, Novartis, Merck, and Biogen. None of the other authors has any conflicts of interest to disclose.

Supporting information