{"paper_id":"377bfc70-e6bd-47e1-854f-def33586d87f","body_text":"1 \n \nEconomic evaluation of non-pharmacological interventions for fatigue in patients with long-term \nmedical conditions \nAuthors:  \nMon Mon Yee1 \nChristopher Burton1 \nJoanna Leaviss1 \nJessica E. Forsyth1 \nGeorge Daly1 \nSarah Davis1 \n \nCorresponding Author: Mon Mon Yee1 \nWord count: 3961 \n \nKEYWORDS  cost-effectiveness, fatigue, chronic conditions, non-pharmacological interventions \n \n  \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \nNOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.\n\n2 \n \nABSTRACT \nBackground Persistent fatigue is a frequent symptom in chronic medical conditions.  Systematic \nreviews of non-pharmacological interventions for fatigue have identified interventions that are effective \nat reducing fatigue, but there is limited published evidence on the cost -effectiveness of these \ninterventions.  \nObjective: To identify non-pharmacological fatigue interventions that have the potential to be cost-\neffective and would warrant further investigation in future cost-effectiveness studies.  \nDesign: Decision-analytic modelling with quality-of-life outcomes mapped from a systematic review \nand network meta-analysis of fatigue outcomes and intervention costs estimated from staff time. \nSetting: UK National Health Service (NHS)  \nParticipants: People with persistent fatigue associated with a chronic medical condition \nInterventions: Non-pharmacological fatigue interventions versus usual care \nPrimary and secondary outcome measures: Net monetary benefit  from a UK NHS and Personal \nSocial Services (PSS) perspective; quality-adjusted life years (QALYs) gained; intervention costs  \nvalued at 2022/23 prices; costs and benefits discounted at 3.5% per annum.  \nResults: In the base case analysis,  expected costs from the probabilistic analysis  for individual and \ngroup interventions were: £267 and £157 for physical activity promotion , £810 and £485 for CBT-\nFatigue, and £462 and £214 for mindfulness. The expected QALYs gained were similar for mindfulness \nand physical activity promotion (0.061 and 0.060 respectively), but lower for CBT-Fatigue (0.045). All \ninterventions provided positive incremental net monetary benefit  (INMB) versus usual care when \nvaluing a QALY at £20,000. However, since group interventions are less  costly than individual ones, \nand we assumed equivalent clinical benefit, they are expected to provide greater INMB. These findings \nremained robust across different scenarios , except for CB T-Fatigue (individual) which had negative \nINMB in some scenarios. \nConclusions: CBT-Fatigue, physical activity promotion and mindfulness interventions all \ndemonstrated the potential to be cost -effective versus usual care. Future research is recommended to \ncompare the cost-effectiveness of these interventions across a broad population with different chronic \nconditions.  \n \n \n  \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n3 \n \nSTRENGTHS AND LIMITATIONS OF THIS STUDY \n• Fatigue outcomes were estimated from a robust systematic review and network analysis that \npooled data across studies conducted across multiple chronic conditions, with multiple sclerosis \nbeing most common. \n• The treatment effects for interventions delivered to groups and individuals are assumed to be \nsimilar, but this assumption was only supported by an analysis exploring separate treatment \neffects for group and individual CBT-Fatigue interventions.  \n• The results for group physical activity promotion and individual mindfulness intervention s \nshould be treated with caution owing to there only being data beyond end of treatment for \nindividual physical activity promotion and group mindfulness interventions. \n• Healthcare costs were restricted to staffing costs for delivering the intervention and therefore \ndo not capture an y impact on resource use outside of the intervention  or any non -staff \nintervention cost such as access to a specific digital tool. \n \n \n  \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n4 \n \nINTRODUCTION \nPersistent fatigue is common in long-term medical conditions.1 Chronically-ill patients usually describe \nfatigue as “more than ordinary tiredness” that could have an impact on their quality of life. 2 Fatigue is \noften overlooked and can persist even after the underlying disease has been fully controlled. 3 The \nexperience of fatigue appears to show similarities across different conditions.3 Currently, there are no \nlicensed pharmacological treatments for fatigue in chronic conditions. Several non -pharmacological \ninterventions have been developed in order to address fatigue in such conditions, targeting physical or \npsychological aspects, or a combination of both. \nThis research has two main aims: to explore the potential cost -effectiveness of non-pharmacological \nfatigue interventions and to identify the interventions that would warrant further investigation in future \ncost-effectiveness studies. \nMETHODS \nThis work was conducted as a part of an evidence synthesis examining both the clinical and cost -\neffectiveness of non-pharmacological interventions for fatigue in long -term medical conditions. 4 The \nreview of clinical effectiveness studies including the network meta-analysis (NMA) and the review of \ncost-effectiveness studies are reported elsewhere. 5 The NMA of randomised controlled trials (RCTs) \nexamined outcomes at three time points: end of treatment (EOT), short -term (ST) which was defined \nas up to 3 months after EOT, and long-term (LT) defined as more than 3 months after EOT. There was \na large diversity of non-pharmacological interventions reported across the RCTs included in the clinical \neffectiveness review. These were grouped into intervention categories to allow evidence to be pooled \nacross studies which were considered to have interventions that were sufficiently similar. A large \nnumber of intervention s were identified and  included in the NMA , and many of these failed to \ndemonstrate treatment effects that persisted beyond EOT . We therefore chose to focus the de novo \neconomic analysis on those interventions which had the strongest evidence for clinical effectiveness. \nConsequently, we selected interventions for inclusion in the de novo  economic evaluation  which \nreported fatigue outcomes at LT follow-up and which demonstrated a statistically significant difference \nin treatment effect compared to usual care in the NMA for either the ST or LT follow-up. The expected \nhealth outcomes and expected costs were estimated relative to usual care in order to align with the \napproach taken in clinical effectiveness review and NMA.  The cost-effectiveness analyses were \nconducted from the perspective of UK National Health Service ( NHS) and personal social services \n(PSS). Costs are reported in UK pound sterling, valued at 2022/23 prices and costs and benefits were \ndiscounted at 3.5% per annum. The data sources and assumptions included in the economic model were \ninformed by discussion with clinical experts and patient and public involvement experts who were able \nto draw on both their own lived experience and findings from focus groups conducted with people with \nfatigue associated with long-term conditions. \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n5 \n \nHealth outcome measures \nOnly a minority of studies included in the NMA reported a measure of health utility directly. Therefore, \nwe applied a mapping algorithm to estimate utility values from the  fatigue outcomes estimated by the \nevidence synthesis for each intervention category over which effectiveness estimates had been pooled.  \n \nIdentification and selection of mapping algorithms \nA targeted literature review was conducted to identify papers which describe statistical mapping \nmethods to convert fatigue -specific patient reported outcome measures used in chronic  health \nconditions to health state utility values derived from generic preference-based measures (PBM), such \nas the EQ-5D and SF-6D. The detailed methods are provided in the supplementary appendix. \nOur electronic database searches identified 96 papers. A study by Goodwin et al.6 that mapped from the \nFatigue Severity Scale (FSS) to health state utility values using EQ -5D-3L, SF-6D and MSIS-8D was \nidentified from both electronic database  and HERC database searches. After removing duplicates and \nreviewing the papers, two additional studies 7-8 were found mapping CIS-F to EQ -5D-5L and fatigue \nmeasured on a visual analogue scale ( V AS) to EQ -5D-3L, respectively . These three papers are \nsummarised in the supplementary appendix (Table S1). No further relevant studies were identified from \nour citation search. Of these three papers, Goodwin et al. was selected in preference to the remaining \npapers because it was the only included study that mapped from FSS, which is frequently used to \nmeasure fatigue severity, to PBMs. The studies by Eriksson et al.7 and Bloem et al.8 were considered \nless appropriate than the study by Goodwin et al. because they included other patient characteristics \n(e.g. disease severity measures, comorbidities, depression and anxiety scores) in their regressions which \nmay not be available in the studies included in our evidence synthesis . Goodwin et al. used five \nregression models to map from either FSS total score or FSS item score to each PBM (EQ-5D-3L, SF-\n6D and MSIS -8D) using both ordinary least squares (OLS) and censored least adjusted deviation  \n(CLAD) specifications. The regression model mapping from total FSS to SF -6D was reported as \nperforming better than the regression model mapping to EQ-5D. The MSIS-8D mapping algorithm was \nconsidered less useful for estimating utilities in populations with chronic conditions other than MS . \nAlthough the algorithm mapping to SF-6D was derived in an MS population, we felt this was broadly \napplicable to other populations of patients with chronic conditions as both the FSS and the SF -6D are \ngeneric tools that have been validated for use across a range of health care conditions.9 After considering \nthe mapping studies identified in this review, we chose to use the regression model mapping from FSS \nto SF-6D provided by Goodwin et al. in our de novo economic model. \nQuality-adjusted life-years \nThe QALY gains for each intervention and usual care were estimated using an area under the curve \napproach based on the average time points for fatigue outcomes reported in the studies contributing to \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n6 \n \nthe NMA. As the absolute utility values were not reported consistently across studies, we have used \ndata on the difference in fatigue scores between arms, which were then converted to absolute FSS scores \nand mapped to SF-6D values to estimate the QALYs gained relative to the usual care.  \nFor each intervention category, the studies contributing to that intervention category in the NMA were \nused to estimate the timing for EOT and the two follow -up points (ST and LT) used when estimating \nQALYs. For studies that have reported fatigue scores for two different LT follow-ups, the base-case \nNMA included the longest follow -up data but a sensitivity analysis was conducted incorporating the \nearliest LT follow-up point (i.e. closest to 3 months after EOT). This was done to ensure that any waning \nof treatment effect over time did not bias the analysis against studies that had a greater duration of \nfollow-up. We explored the impact of using the data from this NMA sensitivity analysis in the economic \nanalysis (scenario analysis) and adjusted the timing of the LT follow-up accordingly. The details of this \nNMA scenario analysis can be found in Table S9 of the supplementary appendix. \nIt was assumed that the baseline FSS score would be the same across interventions and the usual care  \ncomparator arm and this was based on an estimate from Goodwin et al.6 (43.73, 95%CI 14.13 to 73.33). \nThe absolute FSS score for usual care was kept fixed at its baseline value for all follow-up time points. \nFor the intervention arms, the standardised mean differences (SMDs) of fatigue scores provided by the \nNMA across all conditions for each intervention relative to the usual care were converted to a mean \ndifference on the FSS scale using the standard deviation (SD) of baseline FSS score of 15.1 reported by \nGoodwin et al. in the base case. The SD of FSS scores derived from studies included within the NMA \nwas used in the scenario analysis (see supplementary appendix for further details) . This difference in \nfatigue scores between intervention and usual care was used to estimate the absolute FSS score after \nbaseline for each intervention at each follow-up point (absolute FSS score = difference in FSS score + \nbaseline FSS score) . It did not seem reasonable to assume a sudden return to baseline fatigue scores \nafter the last follow-up. Therefore, it was assumed that interventions would experience a linear decline \nin treatment effect  between the last follow -up and 24 months after baseline  in the base case.  This \nassumption was informed by LT follow-up data from two studies which showed the potential for \ntreatment effects to be maintained  beyond one year.10-11 No further treatment effect was assumed to \npersist beyond that point. If data were missing at the ST follow-up point, a linear change in FSS between \nthe EOT and LT follow-up points was assumed.  This assumption was tested in the scenario analysis , \nwhere the treatment effect was assumed to be zero when data was unavailable. QALYs were discounted \nusing a discount rate of 3.5% per annum, as recommended by NICE.12 The time horizon was 24 months \nfor the base-case analysis and 15 and 48 months in the scenarios exploring pessimistic and optimistic \nassumptions, respectively, regarding the persistence of treatment effects. \nResource use and costs \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n7 \n \nThe studies included in the NMA were examined to determine the resources required to deliver the \ninterventions included within each intervention category for the evidence synthesis. Expected costs \nwere estimated separately for interventions delivered to groups versus those delivered to individuals \nas the latter are usually cheaper on a cost per patient basis. Information was extracted on the number \nof sessions, duration of sessions and the health-care professionals involved in delivering or facilitating \nthe interventions. For interventions delivered to groups, information was also extracted on the number \nof individuals who started the intervention, the group size and the number of groups (assumed to be \none if not stated) to allow an estimation of the average cost per patient. Unit costs were taken from the \nPersonal Social Services Research Unit ( PSSRU) Unit Costs of Health and Social Care 2023 13 with \nthe exception of cognitive behavioural therapy (CBT), which was only reported in a previous edition \nof the PSSRU Unit Costs ( 2017),14 and therefore, this cost was uplifted to 202 3 prices using the \nHospital and Community Health Service Pay and Prices index based on PSSRU 2023. Unit costs used \nto estimate the expected intervention costs are summarised in the supplementary appendix (Table S2). \nDetailed assumptions used in the costing analysis on a study-by-study basis are provided in Tables S3-\nS8 of the supplementary file . Median costs across each intervention category were used in the \neconomic analysis for the deterministic analysis . Interventions were typically shorter than 1 year so \ndiscounting was not applied. \nThe costing analysis includes only intervention costs and does not estimate its impact on other \nhealthcare resource use such as reduced  primary or secondary care attendances resulting from patients \nexperiencing lower fatigue or increased use from  potential adverse effects. Studies identified in the \ncost-effectiveness review which assessed the overall impact of an intervention on healthcare costs have \nbeen reported elsewhere.5 \nCost-effectiveness \nThe cost -effectiveness analysis was not taken on a study -by-study basis. Instead we focused on  \nintervention categories that have been considered sufficiently similar to be analysed together within the \nNMA. The same effectiveness was applied across group and individual interventions in the base case \nas these were pooled within the NMA. A scenario analysis was conducted to test the validity of assuming \nequivalent treatment effects for group and individual interventions, using data for CBT-Fatigue, where \nsufficient data were reported to estimate treatment effects separately (see Table S10 in the \nsupplementary appendix for detailed NMA results) . The net monetary benefit (NMB) was estimated \nbased on the willingness to pay threshold of £20,000 per QALY gained. The cost-effectiveness estimates \nwere generated using both deterministic and probabilistic versions of the model. The deterministic \nmodel applied the point estimates of the parameters, whereas the probabilistic model used Monte Carlo \nsampling across 5,000 iterations to generate distributions of expected health outcomes and costs for \neach treatment group. Uncertainty around treatment effect was handled using Convergence Diagnostics \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n8 \n \nand Analysis (CODA) samples while baseline FSS score and regression coefficients were sampled from \na normal distribution  and the intervention costs from gamma distribution s. When the variance -\ncovariance matrices for the mapping coefficients were not reported in the paper, it was assumed that the \ncoefficient for FSS was independent of the intercept parameter. As standard errors (SEs) of costs were \nunavailable, they were estimated using the lowest and highest expected costs of each intervention type \nacross relevant studies. When the intervention cost was limited to a single study, the SE was assumed \nto be equal to 25% of the expected cost. \nScenario analyses were also conducted exploring the impact of (i) optimistic and pessimistic durations \nfor treatment effect decline, (ii) alternative baseline FSS score and SD for FSS based on studies from \nthe NMA EOT network , (iii)  different SMDs for FSS between individual and group CBT -Fatigue \ninterventions, (iv) alternative SMDs using the shorter follow-up point in the LT network, (v) assuming \nthat SMD of mindfulness at the ST follow-up timepoint is zero  by removing the linear change \nassumption between EOT and LT, and (vi) lower and upper costs. \n \nRESULTS \nBased on the NMA results, physical activity promotion, CBT-based fatigue interventions (CBT-Fatigue) \nand mindfulness were found to be eligible for our economic analysis as they had RCTs reporting fatigue \noutcome at more than 3 months after the EOT and they had a statistically significant SMD in fatigue \nscores versus usual care  at either ST or LT follow -up. We chose not to include remote ischaemic \nconditioning in the economic analysis, as although this did achieve a statistically significant difference \nin fatigue versus usual care in the LT NMA, this was based on a single study in stroke patients and the \nintervention was potentially specific to the population and so less likely to be generalisable to patients \nexperiencing fatigue associated with other chronic conditions.  \nHealth outcomes \nQuality-adjusted life-years \nThe mapped SF-6D utility values assumed in the base-case analysis for interventions and the usual care \nare shown in Figure 1 and those used in the scenario analyses are presented in Figure S1 -S4 of the \nsupplementary file. Physical activity promotion was estimated to have the highest utility values at ST \nand LT follow -up timepoints in both base case and sensitivity analyses.  Both probabilistic and \ndeterministic base cases indicated that mindfulness and physical activity promotion interventions were \nassociated with higher QALYs gained compared to CBT-Fatigue, with mindfulness slightly exceeding \nphysical activity promotion (see Table 1). \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n9 \n \nFigure 1  Utility values of interventions across different timepoints \n \nCBT, cognitive behavioural therapy \n \nResource use and costs \nA cost was derived from at least one arm of twenty-one studies that included a total of twenty-three \nintervention arms that were classified as CBT -Fatigue, physical activity promotion  or mindfulness. \nCosts could not be estimated for five studies (Nguyen et al.15, Kucharski et al.,16 Okkersen et al.17, \nPottgen et al .18 and Katz et al .19) because no information was provided in the publications on the \nduration of sessions with healthcare providers  or the interventions included only web-based tools for \nwhich no cost data were available. The range of costs for each intervention is presented in Figure 2 as \nbox-and-whisker plots. As summarised in Table 1, the mean costs estimated by the probabilistic model \nwere similar to the median costs from the deterministic, suggesting that group physical activity \npromotion incurred the lowest costs, whereas individual CBT -Fatigue had the highest  costs due to \nhigher staff contact time in terms of frequency and duration of sessions. \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n10 \n \nFigure 2  Distribution of costs by type of intervention \n \nPhy. activity, physical activity; CBT, cognitive behavioural therapy \n \nCost-effectiveness \nThe incremental costs and QALYs and the incremental net monetary benefit (INMB) for each \nintervention versus usual care, using deterministic and probabilistic models, are presented in Table 1. It \ncan be seen that for all interventions, the INMB versus usual care is positive, which means that the \nmonetary value of the QALYs gained is greater than the intervention cost, indicating that the \nintervention would be considered cost-effective when valuing a QALY at £20,000. In both deterministic \nand probabilistic analyses, physical activity promotion (group intervention) is estimated to provide the \nhighest INMB compared to usual care , followed by group mindfulness, and then individual physical \nactivity promotion. The higher cost of CBT-Fatigue interventions mean that they have the lowest INMB \nversus usual care  with individual CBT -Fatigue having the lowest INMB . The results of scenario \nanalyses (see Table S11 of the supplementary file) are similar to the base case  for the majority of the \ninterventions, with the exception being CBT-Fatigue (individual intervention). Due to the higher cost \nof this intervention, it has a negative INMB, indicating an incremental cost-effectiveness ratio (ICER) \ngreater than £20,000 per QALY versus usual care under certain conditions: (i) pessimistic assumption \nof the treatment effect decline, (ii) alternative baseline FSS score and (iii) higher intervention costs.  \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n11 \n \nTable 1   Base-case cost-effectiveness results \nIntervention type \nGroup or \nindividual \nintervention \nNumber \nof study \narms \nMedian \ncost (£) \nMean \nQALYs \n(discounted) \nIncremental \nCosts \nIncremental \nQALYs \nNMB at \n£20,000 \nthreshold \nINMB \nagainst UC \nBase-case (deterministic) \nUsual care - - £0 1.227 - - £24,531 -  \nPhysical activity promotion Individual 3 £267 1.287 £267 0.060 £25,465 £934 \nCBT-Fatigue Individual 10 £817 1.271 £817 0.045 £24,606 £75 \nMindfulness Individual 2 £465 1.287 £465 0.060 £25,272 £741 \nPhysical activity promotion Group 2 £156 1.287 £156 0.060 £25,576 £1,045 \nCBT-Fatigue Group 5 £484 1.271 £484 0.045 £24,939 £408 \nMindfulness Group 1 £215 1.287 £215 0.060 £25,522 £991 \nBase-case (probabilistic) \nUsual care - - £0 1.223 - - £24,463 - \nPhysical activity promotion Individual 3 £267 1.283 £267 0.060 £25,397 £934 \nCBT-Fatigue Individual 10 £810 1.268 £810 0.045 £24,553 £90 \nMindfulness Individual 2 £462 1.284 £462 0.061 £25,222 £760 \nPhysical activity promotion Group 2 £157 1.283 £157 0.060 £25,508 £1,045 \nCBT-Fatigue Group 5 £485 1.268 £485 0.045 £24,878 £415 \nMindfulness Group 1 £214 1.284 £214 0.061 £25,470 £1,007 \nQALY , quality-adjusted life years; NMB, net monetary benefit; INMB, incremental net monetary benefit; UC, usual care; CBT, cognitive behavioural therapy \n \n \n \n \n \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n12 \n \nDISCUSSION \nThere have been a number of trial-based economic evaluations of non -pharmacological interventions \nfor fatigue in specific  medical conditions. 11,20-22 To our knowledge, no study has assessed the cost-\neffectiveness of such interventions across multiple conditions using decision analytic modelling . Our \nde novo economic assessment evaluated the cost-effectiveness of interventions that had been found, in \na systematic review and NMA of interventions across long-term medical conditions, to demonstrate a \nstatistically significant improvement in fatigue beyond the end of treatment . The base-case model \nsuggests that physical activity promotion, CBT-Fatigue and mindfulness could all be cost-effective at \nthe willingness-to-pay threshold of £20,000 per QALY gained, regardless of whether they are delivered \nindividually or in groups. These findings were robust under the scenario analyses conducted with the \nexception of CBT-Fatigue delivered to individuals, which had the lowest INMB in the base -case and \nwas therefore less robust to scenarios with more pessimistic assumptions.  \nStrengths and limitations \nOne of the key strengths of our analysis is that it  was based on the pooled results of fatigue outcomes \nacross multiple long-term conditions , derived from  a comprehensive systematic review and  NMA. \nStudies on multiple sclerosis accounted for around half of the included studies in each network. Both \nbase case and scenario analyses are probabilistic, capturing the model parameter uncertainty. In addition, \nthe probabilistic model explicitly incorporated the correlation between efficacy estimates (fatigue \noutcomes) using the CODA samples generated by the NMA. \nOur economic assessment is also subject to some limitations. V ery few studies reported the UK \nestimates of resource use other than those associated with delivering the intervention and therefore, the \nestimated incremental costs might be overestimated or underestimated  if the interventions result in \ndecrease or increase in other resource use . Additionally, a detailed cost breakdown for internet or app \nbased tools (e.g., MS Invigor8 20, ELEVIDA18 Tailorbuilder23) were unavailable; therefore, the cost of \nsuch tools were excluded from the model, which might lead to an underestimation of costs for some \ninterventions. In terms of intervention costs per patient, interventions delivered to groups were found \nto be less costly than the same type of interventions delivered individually. For group interventions, \ngroup size determined the cost per patient. Where the group size was not reported in the studies, it was \nassumed that there was a single group and that may have underestimated the cost per patient if smaller \ngroups were used instead. \nThere is a broad range of costs  for each type of intervention  across included studies, indicating high \nlevel of heterogeneity  in expected cost estimates . Sensitivity analyses explored the impact of th e \nheterogeneity in intervention costs , and only CBT-Fatigue (individual) was found to have an ICER \ngreater than £20,000 per QALY versus usual care when the highest cost was applied. This finding aligns \nwith the results from the within-trial economic evaluations reported by Thomas et al.,21 Chong et al.,22 \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n13 \n \nand Hewlett et al.11 where CBT was dominated by usual care because of higher intervention costs and \nsmall negative incremental QALYs. Chong et al .22 compared physical activity promotion to CBT \ndirectly using a within -trial analysis of the RCT (reported by Bachmair et al.24) and their conclusion \nthat CBT was dominated by physical activity promotion is replicated in our de novo analysis informed \nby effectiveness data from multiple studies. Given that CBT-Fatigue may include elements of physical \nactivity promotion or mindfulness, these findings raise the question of whether there is “added value” \nfrom the additional cognitive components of CBT-Fatigue interventions, given that these are generally \nmore intensive to deliver.  \nThe expected QALYs were based on the assumption that the difference between interventions and usual \ncare would start to decline after the last follow -up timepoint and become zero at 24 months (from the \nbaseline). This assumption was tested using optimistic and pessimistic scenarios regarding the treatment \neffect dissipation . All interventions were less cost -effective when making a more pessimistic \nassumption about the persistence of the treatment effect but only CBT-Fatigue (individual) had an ICER \ngreater than £20,000 per QALY when assuming that the treatment effect declined rapidly within a year \nafter the intervention.  Another limitation is that the probabilistic model had to use SEs for mapping \ncoefficients when sampling from a  normal distribution instead of using variance -covariance matrix, \nwhich might overestimate the uncertainty of the mapping algorithm.  \nThe scenario analysis exploring the impact of allowing for different treatment effects for CBT-Fatigue \nfrom group and individual interventions had minimal impact on the cost-effectiveness estimate for CBT-\nFatigue versus usual care  due to prediction of similar treatment effects . However, it should be noted \nthat we were unable to conduct a similar analysis exploring the impact of allowing for different \ntreatment effects of group and individual interventions for the mindfulness and physical activity \npromotion. This was due to there being only being data beyond the end of treatment available for group \nmindfulness interventions and individual physical activity promotion interventions. As such the cost-\neffectiveness results for individual mindfulness interventions and group physical activity promotion \ninterventions should be interpreted with caution. \nAlthough our analysis is able to estimate which interventions have the highest INMB, any head-to-head \ncomparison should be interpreted with caution given the heterogeneity in intervention costs across the \nincluded studies as shown by the overlapping confidence intervals for the intervention costs (see figure \n2). However, across all the scenario analyses explored , CBT-Fatigue had lower QALY gains than \nmindfulness or physical activity promotion due to a smaller and non-statistically significant difference \nin fatigue scores at the ST follow-up. Although it should also be noted that no efficacy data were \navailable for mindfulness from the ST follow-up and the assumption of a linear change in fatigue \nbetween the EOT and the LT follow-up point for the mindfulness intervention could be optimistic . \nHowever, in our scenario analysis we found that mindfulness had greater QALY gains than CBT-Fatigue \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n14 \n \neven when assuming no clinical efficacy at the short-term follow-up point for mindfulness. The analysis \nfor CBT-Fatigue was also the only analysis in which there was data from multiple studies at both the \nST and LT follow-up points and in which the studies informing these endpoints were conducted across \nmore than one type of chronic condition (nine studies for multiple sclerosis, three for stroke, three for \nmusculoskeletal disease and one for inflammatory bowel disease). Therefore, the greater QALY gains \nfor mindfulness and physical activity promotion in this analysis should not be overinterpreted as \ndemonstrating clinical superiority for patients with chronic conditions , as the LT data for these \ninterventions are based on a smaller set of studies covering a limited set of conditions ( 2 studies for  \nphysical activity promotion in musculoskeletal conditions and 1 study for mindfulness in multiple \nsclerosis for mindfulness).   \n \nCONCLUSIONS \nThis de novo economic evaluation indicates that mindfulness interventions, physical activity promotion \ninterventions and CBT -fatigue interventions have the potential to be a cost-effective means for \nimproving quality of life in people experiencing fatigue associated with a chronic condition when \ncompared to usual care. If it is assumed that group and individual interventions have similar efficacy, \nas supported by our analysis of CBT-Fatigue interventions, then group interventions tend to be lower \ncost to deliver and are therefore more cost-effective than individual interventions. Whilst CBT-Fatigue \nhad higher costs and lower QALY gains than the other interventions, the clinical effectiveness estimates \nfor the other interventions are based on fewer studies conducted across a  narrower range of conditions. \nWe therefore recommend that future research is conducted to compare the cost-effectiveness of CBT-\nFatigue, physical activity promotion and mindfulness interventions across a broad population with \ndifferent chronic conditions.  \n \nAuthor affiliations  \n1Sheffield Centre for Health and Related Research (SCHARR), University of Sheffield, Sheffield, UK \n \nAcknowledgements The authors would like to thank Helen Dawes, Vincent Deary, Julia Newton, Kate \nFryer, Samantha McCormick and David Coyle for their advice on the data sources and assumptions \ninforming the economic analysis and the interpretation of the results. We would like to thank Kate Ren, \nProfessor of Statistical Health Technology Assessment at the University of Sheffield, who provided \nstatistical advice and guidance to JEF and GD when conducting the network meta-analysis and analysis \nof baseline fatigue scores. \n \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n15 \n \nAuthor Contributions MMY and SD conceptualised and designed the economic evaluation. JL, JEF, \nGD and CB conducted the systematic review and network meta -analysis that informed the economic \nanalysis and provided advice on the incorporation of clinical efficacy evidence within the analysis. All \nauthors contributed to development, analysis, writing and editing the manuscript. All authors read and \napproved the final version. \nFunding  This work was supported by National Institute for Health and Care Research (NIHR) with \nthe grant number NIHR154660. \nCompeting interests  None of the authors have any conflicts of interest to declare. \nPatient and public involvement  The research group included two members with relevant lived \nexperience and an academic expert in patient and public involvement who were involved in discussion \nregarding the data sources and assumptions for the economic modelling. Their contributions were \ninformed by 5 focus groups convened of people with fatigue associated with long term conditions. \nPatient consent for publication  Not applicable. \nEthics approval  Not applicable. \nProvenance and peer review  Not commissioned; externally peer reviewed. \nData availability statement  All data relevant to the study are included in the article or uploaded as \nsupplementary file. The economic model is available from the corresponding author upon reasonable \nrequest \nSupplementary material  This content has been supplied by the author(s).  \nORCID iDs \nMon Mon Yee  0009-0001-6921-7529 \nChristopher Burton  0000-0003-0233-2431 \nJoanna Leaviss  0000-0002-5632-6021 \nJessica E. Forsyth  0000-0002-5839-9160 \nGeorge Daly  0000-0002-2367-9584 \nSarah Davis  0000-0002-6609-4287 \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n16 \n \nREFERENCES \n1. Goërtz YMJ, Braamse AMJ, Spruit MA, et al. Fatigue in patients with chronic disease: results \nfrom the population -based Lifelines Cohort Study. Sci Rep  2021;11(1):20977. doi: \n10.1038/s41598-021-00337-z [published Online First: 20211025] \n2. Jaime-Lara RB, Koons BC, Matura LA, et al. A Qualitative Metasynthesis of the Experience of \nFatigue Across Five Chronic Conditions. J Pain Symptom Manage 2020;59(6):1320-43. doi: \n10.1016/j.jpainsymman.2019.12.358 [published Online First: 20191220] \n3. Whitehead LC, Unahi K, Burrell B, et al. The Experience of Fatigue Across Long-Term Conditions: \nA Qualitative Meta -Synthesis. J Pain Symptom Manage  2016;52(1):131-43.e1. doi: \n10.1016/j.jpainsymman.2016.02.013 [published Online First: 20160524] \n4. Leaviss J, Burton C, Booth A, et al. Effectiveness of Interventions For Fatigue in Long term \nconditions (EIFFEL) PROSPERO 2024 [Available from: \nhttps://www.crd.york.ac.uk/PROSPERO/view/CRD42023440141. \n5. Leaviss J, Forsyth J, Booth A, et al. Effectiveness of non -pharmacological Interventions For \nFatigue in Long term conditions (EIFFEL) - systematic review and network meta-analysis. \n[Submitted], 2025. \n6. Goodwin E, Hawton A, Green C. Using the Fatigue Severity Scale to inform healthcare decision-\nmaking in multiple sclerosis: mapping to three quality -adjusted life-year measures (EQ-\n5D-3L, SF -6D, MSIS -8D). Health and Quality of Life Outcomes  2019;17(1):136. doi: \n10.1186/s12955-019-1205-y \n7. Eriksson J, Kobelt G, Gannedahl M, et al. Association between Disability, Cognition, Fatigue, \nEQ-5D-3L Domains, and Utilities Estimated with Different Western European Value Sets \nin Patients with Multiple Sclerosis. Value Health  2019;22(2):231-38. doi: \n10.1016/j.jval.2018.08.002 [published Online First: 20181012] \n8. Bloem AEM, Mostard RLM, Stoot N, et al. Perceptions of fatigue in patients with idiopathic \npulmonary fibrosis or sarcoidosis. J Thorac Dis 2021;13(8):4872-84. doi: 10.21037/jtd-21-\n462 \n9. Knoop V , Mathot E, Louter F , et al. Measurement properties of instruments to measure the \nfatigue domain of vitality capacity in community -dwelling older people: an umbrella \nreview of systematic reviews and meta -analysis. Age and Ageing  \n2023;52(Supplement_4):iv26-iv43. doi: 10.1093/ageing/afad140 \n10. Gay MC, Cassedanne F , Barbot F , et al. Long-term effectiveness of a cognitive behavioural \ntherapy (CBT) in the management of fatigue in patients with relapsing remitting multiple \nsclerosis (RRMS): a multicentre, randomised, open-label, controlled trial versus standard \ncare. J Neurol Neurosurg Psychiatry 2024;95(2):158-66. doi: 10.1136/jnnp -2023-331537 \n[published Online First: 20240111] \n11. Hewlett S, Almeida C, Ambler N, et al. Group cognitive-behavioural programme to reduce the \nimpact of rheumatoid arthritis fatigue: the RAFT RCT with economic and qualitative \nevaluations. Health Technol Assess 2019;23(57):1-130. doi: 10.3310/hta23570 \n12. National Institute for Health and Care Excellence. NICE health technology evaluations: The \nmanual, 2022:1-200. \n13. Personal Social Services Research U. Unit Costs of Health and Social Care. 2023. \nhttps://www.pssru.ac.uk/unitcostsreport/ (accessed 06/03/24). \n14. Personal Social Services Research U. Unit Costs of Health and Social Care. 2017. \nhttps://www.pssru.ac.uk/project-pages/unit-costs/ (accessed 06/03/24). \n15. Nguyen S, Wong D, McKay A, et al. Cognitive behavioural therapy for post-stroke fatigue and \nsleep disturbance: a pilot randomised controlled trial with blind assessment. \nNeuropsychological Rehabilitation  2019;29(5):723-38. doi: \n10.1080/09602011.2017.1326945 \n16. Kucharski D, Lange E, Ross AB, et al. Moderate-to-high intensity exercise with person -\ncentered guidance influences fatigue in older adults with rheumatoid arthritis. \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint \n\n17 \n \nRheumatol Int 2019;39(9):1585-94. doi: 10.1007/s00296-019-04384-8 [published Online \nFirst: 20190720] \n17. Okkersen K, Jimenez -Moreno C, Wenninger S, et al. Cognitive behavioural therapy with \noptional graded exercise therapy in patients with severe fatigue with myotonic dystrophy \ntype 1: a multicentre, single -blind, randomised trial. The Lancet Neurology  \n2018;17(8):671-80. doi: https://doi.org/10.1016/S1474-4422(18)30203-5 \n18. Pöttgen J, Moss-Morris R, Wendebourg JM, et al. Randomised controlled trial of a self-guided \nonline fatigue intervention in multiple sclerosis. J Neurol Neurosurg Psychiatry  \n2018;89(9):970-76. doi: 10.1136/jnnp-2017-317463 [published Online First: 20180316] \n19. Katz P , Margaretten M, Gregorich S, et al. Physical Activity to Reduce Fatigue in Rheumatoid \nArthritis: A Randomized Controlled Trial. Arthritis Care Res (Hoboken)  2018;70(1):1-10. \ndoi: 10.1002/acr.23230 [published Online First: 20171206] \n20. Moss-Morris R, McCrone P , Yardley L, et al. A pilot randomised controlled trial of an Internet-\nbased cognitive behavioural therapy self -management programme (MS Invigor8) for \nmultiple sclerosis fatigue. Behaviour Research and Therapy  2012;50(6):415-21. doi: \nhttps://doi.org/10.1016/j.brat.2012.03.001 \n21. Thomas S, Thomas PW, Kersten P , et al. A pragmatic parallel arm multi -centre randomised \ncontrolled trial to assess the effectiveness and cost -effectiveness of a group -based \nfatigue management programme (FACETS) for people with multiple sclerosis. J Neurol \nNeurosurg Psychiatry  2013;84(10):1092-9. doi: 10.1136/jnnp -2012-303816 [published \nOnline First: 20130521] \n22. Chong HY , McNamee P , Bachmair EM, et al. Cost-effectiveness of cognitive behavioural and \npersonalized exercise interventions for reducing fatigue in inflammatory rheumatic \ndiseases. Rheumatology (Oxford)  2023;62(12):3819-27. doi: \n10.1093/rheumatology/kead157 \n23. Torkhani E, Dematte E, Slawinski J, et al. Improving Health of People With Multiple Sclerosis \nFrom a Multicenter Randomized Controlled Study in Parallel Groups: Preliminary Results \non the Efficacy of a Mindfulness Intervention and Intention Implementat ion Associated \nWith a Physical Activity Program. Frontiers in Psychology 2021;12 \n24. Bachmair EM, Martin K, Aucott L, et al. Remotely delivered cognitive behavioural and \npersonalised exercise interventions for fatigue severity and impact in inflammatory \nrheumatic diseases (LIFT): a multicentre, randomised, controlled, open -label, parallel-\ngroup trial. Lancet Rheumatol 2022;4(8):e534-e45. doi: 10.1016/s2665-9913(22)00156-4 \n[published Online First: 20220627] \n \n \n \n \n . CC-BY 4.0 International licenseIt is made available under a \nperpetuity. \n is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint \nThe copyright holder for thisthis version posted August 5, 2025. ; https://doi.org/10.1101/2025.08.01.25332798doi: medRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}