Economic Evaluations of Educational, Physical, and Psychological Treatments for Fibromyalgia

A Systematic Review With Meta-Analysis

Cristina Maria Nunes Cabral; Gisela Cristiane Miyamoto; Katherinne Ferro Moura Franco; Judith Ekkina Bosmans


Pain. 2021;162(9):2331-2345. 

In This Article


Study Design and Registration

The protocol of this systematic review was prospectively registered in the International Prospective Register of Systematic Reviews—PROSPERO (CRD42019119972).

Inclusion Criteria

We included economic evaluations conducted alongside randomized controlled trials or performed using modelling methods that evaluated educational, physical, and psychological interventions for patients with fibromyalgia. Studies conducted in adult patients (aged 18 years or older) were included. Diagnosis should be based on published classification and diagnostic criteria for fibromyalgia of the American College of Rheumatology (1990 and/or 2010/2011 and/or 2016 criteria[51]) or on rheumatologist diagnosis. All nonpharmacological interventions to manage fibromyalgia delivered in any setting (inpatient, outpatient, community, or home-based[2]) were included, such as exercise therapy, behavioural therapy, educational intervention, manual therapy, and acupuncture, among others. The comparator was no treatment, usual care, and other types of nonpharmacological or pharmacological interventions.


The primary outcomes were healthcare and societal costs, and QALYs. Healthcare costs included use of all medical care services, comprising the study intervention, any other services received for health problems, and follow-up care to the intervention or the underlying condition. Societal costs considered, in addition to healthcare costs, the care provided by family or friends, costs paid by patients and their family themselves, and productivity losses.[38] A QALY is a measure of health that combines both length of life and health-related quality of life.[38,47] The secondary outcome was any important clinical outcome, eg, impact of fibromyalgia and pain intensity, among others.

Literature Search

The search was performed in the following databases: PUBMED, EMBASE, CINAHL, Cochrane Central Register of Controlled Trials, Physiotherapy Evidence Database (PEDro), PsycINFO, EconLit, National Health Service Economic Evaluation Database, and Health Technology Assessment Database. The search terms were based on search strategies used in previously conducted systematic reviews regarding fibromyalgia[6,19] and economic evaluations.[31,53] The complete search strategy is presented in Appendix 1 (available as supplemental digital content at All databases were searched from the inception to April 1, 2020, except the National Health Service Economic Evaluation Database (updated until March 31, 2015) and Health Technology Assessment Database (updated until March 31, 2018). In addition, clinical trial databases from the US National Library of Medicine[45] and from the World Health Organization International Clinical Trials Registry Platform[52] were searched for study protocols. Reference lists of included studies and relevant systematic reviews were hand searched to identify additional studies. There were no language restrictions, and only full-text articles were included.

Data Collection and Extraction

Two independent reviewers (C.M.N.C. and G.C.M.) selected eligible studies, first based on the title and abstract, and then the full-text article. Disagreements between reviewers were solved first by discussion and then by arbitration of a third reviewer (J.E.B.), if disagreement persisted. All publications related to the eligible studies, such as published protocols or clinical studies, were used for data extraction. Two independent reviewers (C.M.N.C. and G.C.M.) extracted data using a customized data extraction sheet on bibliometric properties (authors, year of publication, and language), demographic and clinical characteristics of the participants (eg, age, sex, sample size, diagnostic criteria, and pain duration), the content of the interventions according to the Template for Intervention Description and Replication guide,[17] design aspects of the economic evaluation (country, currency, type and perspective of the economic evaluation, time horizon, willingness to pay threshold, reference year, and discounting), outcome measurement (clinical outcomes, instruments/questionnaires, duration of follow-ups, and type of healthcare and societal costs), and study results (costs and effects stratified by the group).[31] Disagreements between the reviewers during data extraction were also solved first by discussion and then by arbitration of a third reviewer (J.E.B.), if disagreement persisted.

Quality Assessment

Two independent reviewers (C.M.N.C. and G.C.M.) assessed the risk of bias of the eligible studies. Disagreements between the reviewers were solved first by discussion and then by arbitration of a third reviewer (J.E.B.), if disagreement persisted. The reporting quality of the economic evaluations was assessed using the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) statement,[18] which provides recommendations to improve the reporting of health economic evaluations. The CHEERS checklist has 24 items, and each item was scored as yes, no, or not applicable. Items 4 (target population and subgroups), 7 (comparators), 8 (time horizon), and 11 (measurement of effectiveness) were scored as yes even if the rationale for the choice of participants or comparators, the appropriateness of the time horizon, and the use of a single study as a sufficient source of clinical effectiveness was not presented.[31] Item 9 (discount rate) was scored as yes if the authors clearly stated the discount rate in the study and not applicable if the authors did not state the time horizon but used a time horizon of 12 months or less. Modelling studies with yes for at least 17 of 24 items (71%), and randomized controlled trials with yes for at least 14 of 21 (67%) possible items (items 15, 16, and 18 are specific for modelling studies[31]) were considered as having sufficient quality.

The risk of bias of the randomized controlled trials was assessed using the Cochrane risk-of-bias tool (RoB 2).[16] The RoB 2 is structured in 5 domains: bias arising from the randomization process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in measurement of the outcome, and bias in selection of the reported result. For each domain, the RoB 2 comprises a series of signalling questions and a judgement about the risk of bias of the domain facilitated by an algorithm. The criterion to reach an overall risk of bias judgement for a specific outcome is the following: low risk of bias if the trial is judged to be at low risk of bias for all domains, some concerns if the trial is judged to raise some concerns for at least one domain, but not to be at high risk of bias for any domain, and high risk of bias if the trial is judged to be at high risk of bias in at least one domain or the trial is judged to have some concerns for multiple domains in a way that substantially lowers confidence in the result. Although the RoB 2 should be assessed for each outcome separately, we reported this for the randomized controlled trial on the whole because there were no differences in the RoB 2 assessment between self-reported outcomes.

Data Synthesis

Costs were converted to American Dollars ($) and Euros (€, using values from the 19 countries of the Euro area) using the purchasing power parity for the reference year of the study.[33] Subsequently, costs were inflated to 2019 prices using the consumer price indices for American Dollars[44] and Euros.[41] If the index year of the study was not reported in the included study, the year of publication was used. Mean differences between costs and effects and 95% confidence intervals (CIs) of a study were calculated, when not presented by the study itself, using the Review Manager 5.3 based on the means and standard deviations stratified by groups if available. If only the mean values of each group were available, the mean difference was calculated by subtracting the mean in the intervention group from the mean in the comparator group. The ICERs for all eligible studies were recalculated to reflect 2019 Dollars and Euros by dividing the difference in costs by the difference in effects. An intervention was considered dominant over the comparator when it was more effective and less costly than the comparator treatment, and dominated by the comparator if it was less effective and more costly than the comparator treatment.[9] If an intervention was more expensive and more effective than the comparator treatment, the intervention was deemed cost-effective in comparison with the comparator treatment if the incremental cost-utility ratio was below the defined willingness to pay threshold. The threshold we used was equal to the one used by the United Kingdom's National Institute for Health and Clinical Excellence (NICE) of £20,000 per QALY gained (ie, $29,412 or €20,765).[3,4] For clinical outcomes other than QALYs, no formal willingness to pay thresholds exist.

Costs and effects were pooled in a meta-analysis for studies in which interventions and comparators, time horizon and perspective of the economic evaluation were similar. Assuming that the intervention effects across the studies were not identical, we used a random-effects model and assessed heterogeneity based on the I2 statistics.[16] When substantial heterogeneity was present (I2 > 50%), we scrutinised the studies for differences in sample characteristics, details of the economic evaluation, instruments used for outcome data collection and risk of bias. After that, if we could not find an explanation for the heterogeneity, we performed meta-analysis and presented the results with caution.