Objective Fatigue is a major disabling symptom in many chronic diseases including multiple sclerosis (MS), but treatment options are limited.Here, we tested the effectiveness of a self-guided , interactive, online fatigue management programme (ELEVIDA) based on principles of cognitive behavioural therapy (CBT) and related psychotherapeutic approaches (eg, mindfulness) for reducing fatigue in MS.
Methods Patients with MS and self-reported fatigue were recruited via the website of the German MS Society and assigned via an automated randomisation generator (1:1, no blocking or stratification) to a 12-week online intervention (ELEVIDA, n=139, 82% female, mean age 40.8, median patient determined disease steps (PDDS) 3.0) or a waitlist control group (n=136, 79% female, mean age 41.9, median PDDS 3.0). The primary outcome was the Chalder Fatigue Scale. Outcomes were assessed at baseline, at week 12 (postintervention) and at follow-up (week 24).
Results Compared with the control group, significantly greater reductions in Chalder Fatigue Scale scores were seen in the ELEVIDA group at week 12 (primary endpoint, intention-to-treat analysis: between-group mean difference 2.74 points; 95% CI 1.16 to 4.32; p=0.0007; effect size d=0.53), with effects sustained at week 24 (intention-to-treat analysis: between-group mean difference 2.19 points; 95% CI 0.57 to 3.82; p=0.0080).
Conclusions Our trial provides evidence for the effectiveness of a self-guided , internet-based intervention to reduce fatigue in MS. Interventions such as ELEVIDA may be a suitable low barrier, cost-effective treatment option for MS fatigue.
Trial registration number ISRCTN registry (number ISRCTN25692173).
- multiple sclerosis
Statistics from Altmetric.com
Fatigue is a common and disabling symptom in many neurological disorders such as stroke,1 traumatic brain injury,2 Parkinson’s disease3 and others. In patients with multiple sclerosis (MS), fatigue affects up to 70% of patients4 and more than three-quarters of these describe it as their most disabling problem.5 Importantly, fatigue is one of the most frequent causes for MS-related retirement6 and there is an association with depression7 and cognitive impairment.8
Despite the immediate clinical importance, treatment options for MS fatigue are limited (for a systematic review, see Khan et al9). Meta-analyses found no conclusive evidence for pharmacological treatment.10 In contrast, the potential benefit of non-pharmacological interventions has been supported by several meta-analyses.10–13 Cognitive behavioural therapy (CBT)-based approaches have been shown to be particularly effective in reducing MS-related fatigue (eg, van den Akker et al 14) which has also been confirmed by meta-analyses.10 15
However, there is a paucity of large randomised controlled trials (RCTs) in this area. Moreover, standard behavioural treatments may not be suitable for all patients with MS. For example, mobility impairments may interfere with patients’ ability to regularly travel to the therapist’s office. In addition, MS-related symptoms such as cognitive dysfunction or fatigue itself could make it difficult for patients to complete standard length, weekly sessions with a therapist. Finally, most of these approaches require well-trained professionals and supervision. Cost constraints may prevent implementation on a large scale or in rural areas, where therapists might not be available. Given the specific needs and requirements of patients with MS, internet-based approaches could be particularly well suited for delivering behavioural interventions to this patient population. Indeed, two small pilot RCTs in the UK and New Zealand have shown promising results for an internet-based version of a CBT programme for MS fatigue with either added telephone or email support.16 17 We thus conducted a large RCT of a novel, internet-based treatment approach and examined its effectiveness for MS-related fatigue.
Study design and population
This was a parallel group, two arm, RCT of a 12 week, internet-based, CBT intervention (ELEVIDA). Patients were automatically randomised to ELEVIDA or waitlist control group after completing baseline outcome measures (week 0). Follow-up of outcomes was obtained at week 12 (directly after the end of the intervention) and week 24 (long-term follow-up).
A decision framework for appropriate control conditions for clinical trials examining behavioural interventions has recently been published.18 According to this decision framework, participation risk for our patients was low and there is a paucity of evidence from large clinical trials in MS fatigue in general and for the efficacy of online-based therapeutic options for MS-related fatigue in particular. Thus, a waitlist control group was deemed appropriate for the current stage of development for ELEVIDA.18 Participants randomised to the waitlist control condition were provided with access codes to ELEVIDA after completing the week 24 assessments.
The trial was registered with the ISRCTN registry prior to patient enrolment (registration number ISRCTN25692173).
Patients were recruited by advertisements published on the website of the German MS patient organisation (Deutsche Multiple Sklerose Gesellschaft DMSG), both by the local DMSG chapter and nationally. In addition, information about the study was sent out via the e-newsletter of the Institut für Neuroimmunologie und Multiple Sklerose, and leaflets were distributed at the MS outpatient centre, University Medical Centre Hamburg-Eppendorf.
Patients were eligible if they had a diagnosis of MSi , were at least 18 years of age, reported fatigue at screening (as indicated by a score of 43 or higher on the Fatigue Scale of Motor and Cognition (FSMC)),19 reported no major neurological or psychiatric comorbidities (dementia, stroke, autism, or psychosis, although comorbid depression was allowed) and no MS relapse in the last 4 weeks.
Randomisation and masking
After completing the last page of the online questionnaire, patients were randomly allocated (1:1, no blocking or stratification) to ELEVIDA or to the waitlist control group using a fully automated computer algorithm (concealed allocation). For the randomisation, the EFS Survey platform used a random number generator built into the system to assign patients to one of the two groups (ELEVIDA or waitlist control). To complete assessments 12 weeks and 24 weeks after randomisation, participants were invited via standardised emails at predefined times (three reminders within 4 weeks each) by one of the study investigators (JP). All outcome measures were collected via an automated online interface.
The ELEVIDA programme was jointly developed by a multidisciplinary team of physicians, psychologists, psychotherapists and IT experts. Details about the development process can be found in the online Supplementary materials. The programme was made available at no cost for patients in this trial. In ELEVIDA, content is based on cognitive behavioural therapy (CBT) strategies and is conveyed chiefly via the technique of a ‘simulated dialogue’. Programme modules are composed of an introduction and a summary and include homework tasks. Patients are advised to access the programme once to twice per week. Participants are invited to respond continuously to narrative text passages provided by the programme using a multiple-choice format. Depending on patients’ responses, the programme tailors subsequently offered information to match the individual needs (eg, preference for elaborated explanations, additional exercises and shorter texts).
Supplementary file 1
All clinical and demographic descriptors were assessed by self-report from the patients via online questionnaires. The disability ratings were derived from the Patient Determined Disease Steps (PDDS) instrument.20 It has been shown that the PDDS correlates highly with Expanded Disability Status Scale scores rated by experienced neurologists, objective tests of walking ability such as the 6 min walking test and real world mobility as assessed by accelerometer measures.21
The prespecified primary outcome measure was the Chalder Fatigue Scale. This questionnaire assesses severity of physical and mental fatigue and is not disease specific. The scale contains 11 items covering physical fatigue (items 1–7) and mental fatigue (items 8–11). Two previous cross-sectional studies in autoimmune disorders (systemic lupus erythematosus and rheumatoid arthritis) estimated the minimally important difference for improvement on the Chalder Fatigue Scale between 0.7 and 1.4.22 ,23 The Chalder Fatigue Scale has good psychometric properties,24 has been validated in patients with MS25 and is sensitive to change.25 This is particularly important as other fatigue scales frequently used in MS such as the MFIS and the FSS have not been shown to be sensitive.26 Moreover, the Chalder Scale is a pure measure of fatigue severity in contrast to other frequently used fatigue scales in MS such as the MFIS, which measures impact of fatigue (rather than severity) and the FSS (which is a combination of severity and impact) as demonstrated in a recent validation study.25 The prespecified primary endpoint analysis listed in the registry entry was group differences from baseline to post-treatment (ie, 12 weeks after randomisation) and 3 months later (ie, 24 weeks after randomisation).
Prespecified secondary outcome measures were MS-specific motor and cognitive fatigue measured by the Fatigue Scale for Motor and Cognitive Functions (FSMC),19 anxiety and depression measured by the Hospital Anxiety and Depression Scale (HADS-A and HADS-D),27 health-related quality of life (QoL) quantified by the Hamburg Quality of Life Questionnaire for MS (HAQUAMS),28 29 and self-reported cognitive difficulties as assessed by the Multiple Sclerosis Neuropsychological Screening Questionnaire (MSNQ).30 Activities of daily living as assessed by the Frenchay Activity Index (FAI)31 was prespecified as an exploratory outcome. Additional questionnaires were obtained as exploratory measures (assessing illness perception, coping and personality traits and responses to symptoms, see trial registration) and these will be reported separately.
Sample size calculation was based on the previously published face-to-face/telephone-delivered CBT intervention trial 32 and performed using G*Power V.3.1 software. Target sample size in the trial registration was 204 participants (ie, 102 per group). Sample size was based on an estimated standardised effect size (Cohen’s d) of 0.35 (given that a fully automated programme would likely have smaller effects than face-to-face therapy) with significance level α set at 0.05 (one tailed) and a power of 0.80.
Continuous clinical characteristics and baseline scores are summarised by means and SD. Categorical clinical characteristics are described by frequencies and percentages. Treatment effects were estimated and tested using an intention-to-treat approach as primary analysis based on multiple imputations with baseline fatigue (Chalder) and sex as predictors. The number of imputations was set to 100, which is in line with the recent recommendations.33 The primary analysis model was an analysis of covariance (ANCOVA) with Chalder fatigue score at week 12 as dependent variable and intervention group as factors and baseline Chalder fatigue score as covariate. The same analysis was used for secondary endpoints.
We imputed missing scores at week 12 in the primary analysis and weeks 12 and 24 for follow-up analyses making a missing-at-random assumption. The sensitivity of our results to the missing at random assumption was explored by missing-not-at-random models including control-based pattern imputation and a tipping-point analysis shifting the expected values of the dropouts in the intervention group. We also reran the primary analysis in all patients with neurologist-confirmed diagnosis (n=120) as well as the subgroup who had accessed the programme at least once (ie, ‘modified’ intention-to-treat (ITT)) as sensitivity analyses. Additional sensitivity analyses included ANCOVA (change from baseline in the Chalder scale as dependent variable) with last observation carried forward (LOCF) adjusting for baseline Chalder scale scores. Moreover, we computed the ANCOVA using the complete cases sample (per protocol analysis) as well as a mixed effects model repeated measures (MMRM) analysis including all available follow-up data as further sensitivity analyses.
Treatment group differences were estimated by least-squares means and are reported with 95% CI and two-sided p values testing the null hypothesis of no treatment difference. A p value smaller than 0.05 is considered statistically significant in the primary analysis. Secondary analyses are of an exploratory nature and were not adjusted for multiple testing.
We conducted several post hoc analyses of the data to estimate clinical relevance of the achieved effects. Following FDA guidelines for patient-reported outcome measures, we analysed data on minimal clinically important difference (MCID) in patients’ QoL. Here, we used the Hamburg QoL Questionnaire (HAQUAMS), a validated MS-specific QoL instrument with established thresholds for clinically relevant change (threshold of 0.36 on the fatigue subscale).28 As recommended by the FDA (UCM193282), we used the anchor-based MCID34 and measured the treatment effect against the definitions put forth by Kieser and Hauschke.35 We also analysed activities of daily living as a measure of ELEVIDA’s impact on patients’ every day life as measured by the FAI.31
Adverse events were not prespecified but defined post hoc as an increase in fatigue by more than three points or an increase in levels of depression (HADS-D) from below to above the established threshold for clinically relevant symptoms (>8). Adverse events are described as frequencies and percentages by intervention group.
All statistical analyses were carried out using SAS V.9.4. Patient-level data for primary and secondary outcome measures are available for download (online Supplementary file 2).
Supplementary file 2
Patients were enrolled from 11 July 2014 to 28 November 2014 (duration of the recruitment period: 140 days). The Consolidated Standards of Reporting Trials flow chart is displayed in figure 1; patient characteristics at baseline are provided in tables 1 and 2.
As shown in figure 1, the overall dropout rate at postintervention was 19% (n=51).
During the study period, participants randomised to ELEVIDA accessed the programme on average 14.5 times (SD 13.0) and had logged activity within the programme for on average 16.3 different days (SD 17.1) during the intervention phase. A total of n=15 (11%) participants in the intervention group never accessed the programme.
Overall, fatigue levels as measured by the Chalder scale decreased in both groups, but this decline was more pronounced in the ELEVIDA group (see figure 2 and table 3). The intention-to-treat primary analysis (multiple imputations with baseline Chalder score and sex as predictors) yielded a statistically significant treatment effect favouring ELEVIDA at week 12 with a mean difference on the Chalder scale of −2.74 points (95% CI −1.16 to −4.32; p=0.0007; d=0.53).
Numerous sensitivity analyses were conducted to confirm these results. Treatment effects were almost identical when sex was not used as a predictor in the multiple imputations model (mean difference, −2.76 points; 95% CI −1.17 to −4.36, p=0.0007). We also ran the primary ITT analysis in the subgroup of 120 patients who had a neurologist-confirmed MS diagnosis. This analysis yielded a group difference of −2.50 points (95% CI −0.07 to −4.93) and was still statistically significant (p=0.043), despite the considerably smaller sample size. When patients randomised to the treatment group who never accessed the programme were excluded from the analysis (ie, a ‘modified ITT analysis’), results indicated a group difference of −3.33 points (95% CI −4.92 to −1.74, p<0.0001). Statistically significant treatment effects at week 12 were also confirmed in sensitivity analyses using LOCF (mean difference −2.04; 95% CI −0.61 to −3.48, p=0.0055) and the per protocol analysis (mean difference −3.39; 95% CI −1.77 to −5.01, p<0.0001).
We further investigated the sensitivity of our primary analysis to the missing at random assumption. Applying a multiple imputation procedure with control-based pattern imputation resulted in a significant intervention effect of −2.53 (95% CI−4.14 to −0.3, p=0.0020), that is, only slightly smaller than the primary analysis. Furthermore, we carried out a tipping-point analysis investigating when the observed statistically significant treatment difference would turn into a non-significant result. This was done by gradually shifting the expected value of the dropouts in the intervention group. We found that the result would tip with an increase of 3.92 points on the Chalder Fatigue Scale in dropouts. Since it appears fairly unrealistic that the response of the dropouts would differ by such a large amount from the responses seen in patients who completed the trial, the analysis adds further support to our primary analysis. Taken together, the primary endpoint was met in the primary as well as all sensitivity analyses.
Effects on fatigue levels at week 12 were confirmed using the secondary endpoint FSMC as well as its subscales for motor fatigue and cognitive fatigue favouring ELEVIDA (see table 3). There were also significant reductions on anxiety as measure by the HADS-A favouring ELEVIDA. No significant effects were seen on depressive symptoms (HADS-D) or perceived cognitive problems (MSNQ). Domain-specific QoL was significantly increased for three of the HAQUAMS subscales: fatigue, thinking and mobility lower extremities, favouring ELEVIDA. No significant effects were detected on QoL domains mobilitiy lower extremities, mood and communication. Levels of significance and CIs for all secondary endpoints are presented in table 3.
Clinical relevance of observed treatment effects
In order to address clinical relevance, we conducted additional post hoc analyses on secondary outcomes. Anchor-based thresholds for minimally clinically important change (MCID) are the recommended way of judging clinical significance. MCID estimates are established for patients’ QoL on the HAQUAMS scale, a predefined secondary endpoint of our trial. The threshold for fatigue impact on QoL measured by the HAQUAMS is 0.36 on the fatigue subscale.28 Here, ITT analyses indicated a ‘probably clinically significant effect’ using published interpretation guidelines.35 Finally, we analysed the impact on patients’ daily lives using the exploratory endpoint activities of daily living (FAI). Here, a significant increase was observed in the ELEVIDA group compared with the control group (FAI, mean difference, −1.97; 95% CI −0.58 to −3.35, p=0.0053).
Treatment effects remained statistically significant at 24 weeks (primary analysis multiple imputations: mean group difference on the Chalder scale of −2.19 points; 95% CI −0.57 to −3.82, p=0.0080). Sensitivity analyses using LOCF (−1.53 points; 95% CI −0.09 to −2.96, p=0.0376) and per protocol analyses (−2.87 points; 95% CI −1.20 to −4.54, p=0.0008) were also statistically significant. Finally, long-term stability was confirmed using MMRM analysis (mean difference, −2.79 points; 95% CI −1.15 to −4.44, p=0.0010). Group differences also remained significant at 24 weeks for the secondary endpoints FSMC and both of its subscales and HAQUAMS subscales fatigue and thinking as well as the activities of daily living (FAI) (see table 3). However, treatment effects were no longer significant at 24 week follow-up for anxiety (HADS-A) and the HAQUAMS subscale mobility lower extremities (see table 3).
For this trial, there were no predefined safety measures/adverse events. However, we used several measures post hoc as potential adverse events. First, we examined the number of patients who showed a reliable increase in fatigue (as determined by the primary endpoint Chalder) of three points or more from baseline to postintervention (week 12). Out of all patients completing week 12, this was seen in eight patients in the ELEVIDA group and 23 patients in the control group. An increase from below to above the clinical threshold for depression (HADS-D >8) was seen in five patients in the ELEVIDA group and six patients in the control group. For anxiety (HADS-A >8), this was observed in six patients in the ELEVIDA group and 11 patients in the control group. One patient randomised to the control group died during the trial for unknown reasons. He was 47 years of age and of moderate disability when he entered the study. We were not able to obtain more information about his death.
To our knowledge, this trial is the largest of any published behavioural or pharmacological fatigue treatment trial in MS to date. Beyond meeting the primary endpoint, we showed persistence of treatment benefits up to 6 months postrandomisation, suggesting that patients in the intervention group may have acquired new skills during the programme that helped them to manage their fatigue symptoms even after they had completed the programme. Moreover, we demonstrated a positive effect on activities of daily living and clinically relevant improvements in QoL for the treatment group, which was maintained at 6 months, underscoring the robustness of our findings as well as supporting their clinical relevance.
In the meta analysis by Asano and Finlayson,10 exercise interventions achieved an effect size of d=0.57 (10 trials) and education programmes (including CBT) yielded d=0.54 (eight trials). Another meta-analysis including all MS exercise studies that had measures of fatigue included as primary or secondary endpoints, the standardised effect size was d=0.45 (18 trials)12 and the most recent meta-analysis reports a standardised mean difference of 0.53 for exercise (when analysing 26 trials with non-exercise control conditions).36 ELEVIDA in the present trial achieved an effect size of d=0.53. Thus, the effects of ELEVIDA are comparable to those achieved by highly structured and supervised exercise interventions and behavioural interventions delivered in person, although self-guided programmes have much less requirements in terms of resources and infrastructure for delivery.
Behavioural interventions (including ELEVIDA) thus have a better evidence base for treating MS fatigue than pharmacological treatments with amantadine and modafinil, where the meta-analysis found non-significant effect (d=0.07; seven trials).10 There was also inconclusive evidence for amantadine, modafinil and pemoline in a more recent meta-analysis.37 However, it should be noted that there are individual placebo-controlled trials of other drugs including aspirin38 and vitamin D,39 which might in the future also have a sufficient evidence base for clinical use, possibly in combination with efficacious behavioural interventions for MS fatigue.
Several limitations have to be considered when interpreting our results. First, the dropout rate in the current trial was 19% and somewhat higher in the ELEVIDA group compared with the control group. This might indicate that the intervention is not equally suited for all patients or that additional tools to reduce attrition might be necessary. One possibility is that patients with more pronounced cognitive impairment might have had difficulties using the programme and thus dropped out. However, self-reported cognitive complaints (as assessed by the MSNQ) were not related to treatment response (ie, group×MSNQ interaction in an additional ITT analysis, p=0.72). Thus, we believe (moderate) cognitive impairment is unlikely to be a major barrier for use of such programmes. Attrition is a concern for any health intervention but this has to be weighted against the resources required to deliver the intervention and thus the reach and scalability of such an approach. Here, internet-based, self-guided tools such as ELEVIDA have a clear edge over traditional delivery methods (eg, face-to-face interventions by highly skilled therapists). Thus, their public health impact can be larger even if attrition is higher than for more traditional therapies.
Disease variables (including disease course and disability ratings) were obtained using self-report. However, MS diagnosis was confirmed by written reports from their neurologist in more than 90% of the selected patients, and a sensitivity analysis restricted to this subsample yielded very similar estimates of the treatment effect (which were also statistically significant despite the smaller sample size of this analysis). It should also be noted that we did not assess sleep disorders in our patients, which often co-occur or overlap with fatigue in MS. In a related matter, we did not assess whether or not participants started any new or additional treatments (pharmacological or non-pharmacological) after enrolment. In future studies, it may thus be informative to explore effects of comorbid symptoms or possible co-interventions on treatment outcome.
Finally, our trial had no active control group. Given limited evidence for effective internet-based fatigue management programmes in MS on one hand and the potential impact such easily scalable approaches would have in clinical care, we feel that the waitlist control group chosen here was appropriate for the current stage of development of our intervention.18 Moreover, there was no indication of a nocebo response in the control group. However, the choice of control group in trials of behavioural interventions have been shown to have a pronounced effect on the observed effect size and waitlist control conditions or treatment as usual typically yield the largest estimates.18 40 Thus, future trials evaluating ELEVIDA in MS should consider additional control conditions to separate specific and non-specific effects of such interventions, specifically a comparison with face-to-face CBT interventions.
In conclusion, ELEVIDA could offer an interim solution or low barrier option to help to reach patients in areas where therapists are not available or provide care to patients in settings where resources are limited.
↵i To verify that participants had a clinical diagnosis of MS, half of the patients were randomly selected by the EFS platform and were requested to send written confirmation of their diagnosis (letter signed by their neurologist or copies of their medical records by email, mail or fax). Of the selected patients, 90.2% provided the requested confirmed diagnosis. The minimum required fatigue score and the full list of excluded neurological and psychiatric comorbidities were unknown to the patients when they completed screening and eligibility assessment.
CH and SMG contributed equally.
Contributors JP, IKP, CH and SMG designed the study. JP, BM and CH contributed to the development of the intervention tool. JP, JMW, LF and SL obtained the data. TF analysed the data. JP, RMM, JMW, LF, SL, SK, BM, TF, IKP, CH and SMG interpreted the data. SMG wrote the paper. JP, RMM, JMW, LF, SK, BM, TF, IKP and CH revised the paper for intellectual content.
Funding This research was supported by a research grant from the Gemeinnützige Hertiestiftung (grant no. P1130079-Multiple Sklerose). SMG is supported by a Heisenberg Professorship from the Deutsche Forschungsgemeinschaft (DFG, GO1357/5-1 and 5-2).
Competing interests BM is an employee of GAIA AG, the developer, owner and distributor of ELEVIDA.
Patient consent Obtained.
Ethics approval Ethikkommission der Ärztekammer Hamburg (review number PV4772).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Patient level data on the primary and secondary endpoint are provided as supplementary material.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.