Drug treatments for fibromyalgia: systematic review and network meta-analysis of randomised trials

Objective

To compare the effectiveness of drug treatments for the management of fibromyalgia.

Design

Systematic review and network meta-analysis of randomised trials.

Data sources

Medline, Embase, AMED (Allied and Complementary Medicine Database), PsycInfo, CINAHL, and Cochrane CENTRAL (Cochrane Central Register of Controlled Trials), from inception to 13 January 2025.

Eligibility criteria for selecting studies

Parallel, crossover, and cluster randomised trials that enrolled adults (aged ≥18 years) with fibromyalgia, according to any published diagnostic criteria or as diagnosed by a physician. Trials compared commercially available, investigational, or discontinued drug treatments with another drug, the same drug at a different dose or regimen, placebo, or standard care. Treatments also included select vitamins and supplements, such as melatonin, vitamin D, vitamin B6, and omega 3 fatty acids.

Results

151 eligible trials (32 592 participants) were included in this systematic review. Compared with placebo or usual care, low certainty evidence suggested that sodium oxybate, selective noradrenaline (norepinephrine) reuptake inhibitors (NRIs), serotonin-noradrenaline reuptake inhibitors (SNRIs), and gabapentinoids may increase the likelihood of patient reported meaningful improvement or recovery (risk difference 105 to 283 more per 1000 patients). Based on scores from the fibromyalgia impact questionnaire, low certainty evidence further suggested that sodium oxybate (mean difference –8.81, 95% confidence interval (CI) –11.78 to –5.84) and selective serotonin reuptake inhibitors (SSRIs) (mean difference –8.02, 95% CI –13.40 to –2.64) may reduce fibromyalgia symptoms, whereas tricyclic antidepressants, sodium oxybate, and gabapentinoids may improve sleep quality. No drug showed pooled effects on pain, physical health or mental health that met or exceeded a threshold for clinical importance. For sodium oxybate, selective NRIs, SNRIs, and cyclobenzaprine, responder analyses indicated that a potentially important proportion of patients achieved a reduction in pain of at least 30% (risk difference 111 to 202 more per 1000 patients). In addition, no drug increased the risk of discontinuing treatment or serious adverse events.

Conclusions

This systematic review and network meta-analysis provides low certainty evidence that sodium oxybate, selective NRIs, SNRIs, gabapentinoids, tricyclic antidepressants, and SSRIs may provide typically small benefits in select outcomes for people with fibromyalgia.

Systematic review registration

PROSPERO CRD42012003291.

Keywords: Pain, Pharmacology, Systematic Review

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Fibromyalgia is a chronic syndrome characterised by widespread pain, fatigue, functional limitations, cognitive impairment, and sleep disturbances, with an unclear cause

• Previous reviews suggested that serotonin-noradrenaline (norepinephrine) reuptake inhibitors (SNRIs), tricyclic antidepressants, and gabapentinoids may provide benefit

• The full range of patient important outcomes, the importance of treatment effects, the certainty of evidence, and trial trustworthiness have not been consistently evaluated

WHAT THIS STUDY ADDS

• In this systematic review and network meta-analysis, no drug showed pooled effects that met or exceeded an important benefit for any core outcome with moderate or high certainty evidence

• Low certainty evidence suggested that sodium oxybate, selective noradrenaline reuptake inhibitors (NRIs), SNRIs, and gabapentinoids may increase the likelihood of meaningful improvement or recovery

• Low certainty evidence also suggested that sodium oxybate and selective serotonin reuptake inhibitors (SSRIs) may reduce fibromyalgia symptoms, and tricyclic antidepressants, sodium oxybate, and gabapentinoids may improve sleep quality

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE, OR POLICY

• Clinicians and patients should be aware that drug benefits for fibromyalgia are typically modest and uncertain

• Common off-label options, such as medical cannabis and low dose naltrexone, lacked evidence of benefit

• Future trials may benefit from comparing drugs against non-pharmacological care, adopting contemporary diagnostic criteria, extending follow-up beyond a year, and minimising and appropriately accounting for attrition

Eligibility criteria

We included parallel, crossover, and cluster randomised trials that enrolled adults (aged ≥18 years) with fibromyalgia, according to any published diagnostic criteria or as diagnosed by a physician. Eligible trials compared commercially available, investigational, or discontinued drug treatments with another drug, the same drug at a different dose or regimen, placebo, or standard care. We also included select vitamins and supplements that have been recognised by the World Health Organization Model List of Essential Medicines or major regulators (FDA, European Medicines Agency, and Health Canada), such as melatonin, vitamin D, vitamin B6, and omega 3 fatty acids.⁴⁰ Although we did not restrict eligibility according to diagnostic criteria, we recorded the criteria used in each trial for subgroup analyses. We also did not restrict eligibility by language or date of publication.

We excluded pseudorandomised trials, trials that compared pharmacological with non-pharmacological interventions without a placebo, sham, usual care, or alternative drug treatment arm, and conference abstracts. When all arms of a trial received the same background treatment, whether pharmacological or non-pharmacological, we classified that shared treatment as usual care and extracted the contrast between randomised pharmacological interventions. This approach isolates the incremental effect of the drug under study, because the background treatment applies equally across arms.

We also excluded randomised trials with <20 participants allocated to each arm. Small trials are unlikely to meaningfully contribute to meta-analyses, and are more likely to include unrepresentative samples and arms that are prognostically imbalanced at baseline, and are at higher risk of publication bias.⁴¹ Although any such threshold is arbitrary, 20 participants for each arm represents a pragmatic balance between including too small trials and excluding potentially informative trials.

Search strategy

In collaboration with an experienced research librarian, we searched Medline, Embase, CENTRAL (Cochrane Central Register of Controlled Trials), CINAHL, AMED (Allied and Complementary Medicine Database), and PsycInfo, from inception to 13 January 2025, for eligible randomised trials. Our search combined concepts related to fibromyalgia with a filter for randomised trials,⁴² without any restrictions on language of publication or other search filters. We removed duplicate search records with the Bond University Institute for Evidence Based Healthcare Systematic Review Accelerator deduplicator tool.⁴³ Online supplement 1 shows our search strategy. We supplemented our search by reviewing the reference lists of relevant systematic reviews,^{20 24 26} reviewing references of eligible studies, and soliciting two experts (Mary-Ann Fitzcharles and Amir Minerbi) for trials that may have been missed by our search.

Study selection

After training and calibration exercises to ensure sufficient agreement, pairs of reviewers worked independently and in duplicate to screen the titles and abstracts of the search records and subsequently the full texts of records considered potentially eligible at the title and abstract screening stage. Screening was conducted with Covidence (https://www.covidence.org), an online systematic review software. Non-English-language search records were reviewed either using DeepL machine translation software (www.deepl.com/en/translator) or by reviewers fluent in the language of publication. Reviewers resolved disagreements by discussion or, if necessary, by the adjudication of a third reviewer. One reviewer screened references against the Retraction Watch database.⁴⁴ Our intention was to exclude trials that were retracted because of concerns about the integrity of the trial’s findings or with major errors, but we did not identify any eligible trials that were later retracted.

Data extraction

After training and calibration exercises, pairs of reviewers worked independently and in duplicate to collect data from eligible trials with a data collection form in an Excel spreadsheet (Microsoft Office Excel 2019) that was pilot tested. Reviewers resolved disagreements by discussion, and a senior reviewer reviewed and confirmed all extractions. We used artificial intelligence tools (Claude and ChatGPT) to verify and format the extracted data. The authors reviewed, edited, and confirmed all AI generated outputs. For publications in languages other than English, we used DeepL translation software (www.deepl.com/en/translator) to perform extractions, and extractions were verified by a fluent speaker.

We collected data on trial characteristics (design, country, funding source, and diagnostic criteria), patient characteristics (age, sex, comorbid anxiety and depression, and time since fibromyalgia diagnosis or duration of fibromyalgia symptoms), characteristics of interventions and comparators (eg, drug, dose, and duration), and patient important outcomes, informed by core outcome sets for fibromyalgia and chronic pain, and discussions with healthcare providers and patients.^{33 45–47} These outcomes were patient reported improvement or recovery, pain intensity, physical health, mental health, fibromyalgia symptoms, sleep quality, tolerability, and safety.

For outcomes that could be captured with different instruments or measures, we prioritised those most frequently reported among eligible randomised trials, supported by strong psychometric properties, and of clear relevance to patients and clinicians.⁴⁸ For patient reported recovery or improvement, we prioritised data on self-reported recovery or, if unavailable, responses of “very much better” or “much better” on the Patient Global Impression of Change scale, a single item measure of perceived overall health change.⁴⁸ When studies used alternative response options, we classified responses indicating greater than moderate improvement as evidence of recovery or meaningful improvement.

When trials reported more than one measure of pain intensity, we applied a prespecified hierarchy: pain at rest over pain during activity; average pain over least or worst pain; pain assessed across multiple days (preferably seven days) over shorter or instantaneous assessments; contemporaneous ratings over recalled pain (eg, daily ratings averaged over seven days over retrospective seven day recall); numeric rating scales over visual analogue scales; paper based over electronic formats; and unidimensional pain instruments over multidimensional instruments with embedded pain domains. We excluded measures of pain interference, pain frequency, and number of tender points.

A mean difference in pain reflects an average effect. When treatment response varies across participants, a small mean difference can coexist with important improvement in a subgroup. This challenge affects the interpretation of continuous outcomes generally. To help with the interpretation of pain relief, we also extracted the proportion of participants who achieved at least a 30% reduction in pain intensity from baseline, a commonly reported threshold of response that closely aligns with the minimum important difference for pain on a numeric rating scale.

For physical and mental health, we collected data from the 36 item short form health survey (SF-36) physical and mental component summary scores or, when not reported, the SF-36 physical function and mental health subscales.⁴⁹ To assess the overall impact of symptoms of fibromyalgia, we extracted data from both the original and revised fibromyalgia impact questionnaire.⁵⁰ For sleep quality, we included all validated or ad hoc instruments that directly assessed sleep quality, excluding those focused on pain related sleep interference or energy levels.

We assessed overall acceptability by estimating patients' willingness to continue treatment. We extracted the proportion of participants who discontinued the study drug, placebo, or usual care for any reason, including for adverse events, poor tolerability, or lack of efficacy. When trials did not report these data, we extracted the number of participants who withdrew from the trial. Observational evidence suggests that many patients discontinue their drug treatments within a year because of poor tolerability or diminishing benefit.^{51 52} Whether a patient remains on a drug or discontinues treatment because of tolerability or lack of efficacy is therefore a critical question when starting treatment. All cause discontinuation answers this question directly. We also identified serious adverse events, relying on each trial's classification of severity.

For dichotomous outcomes, we extracted the number of patients and events for each arm and, for continuous outcomes, the number of patients, a measure of central tendency (mean or median), and a measure of variability (eg, standard deviation, standard error, 95% confidence interval). For continuous outcomes, we preferentially extracted change scores from baseline to deal with interpatient variability. When trials did not report change scores, we extracted outcomes at follow-up. In randomised studies, randomisation balances baseline values across arms, such that the difference in follow-up measures will, on average, equal the difference in change scores. We did not convert final scores to change scores, which requires assumptions on the correlation between baseline and follow-up scores.

For each outcome, we extracted the results at the longest point at which randomisation was maintained and before discontinuation of the drug or drugs being studied. We prioritised intention-to-treat analyses without imputation over analyses with imputed data, per protocol analyses, and as-treated analyses. Common imputation strategies, such as last observation carried forward and baseline observation carried forward, impose assumptions about the trajectory of missing outcomes that are difficult to verify and, when violated, may bias results.^{53 54} These methods also treat imputed values as though observed, thereby underestimating variance.⁵³ Likewise, conventional multiple imputation and mixed models for repeated measures generally rely on a missing-at-random assumption. This assumption may be questionable in fibromyalgia trials because adverse effects and inadequate effectiveness often drive attrition, potentially leaving missing outcomes related to patients’ unobserved health status.^{51 52 55} Therefore, consistent with established guidance, we extracted observed case data and dealt with bias from missing data by risk of bias assessment and GRADE (Grading of Recommendations Assessment, Development, and Evaluation) judgments about the certainty of the evidence.^{42 56}

Given the fluctuating course of fibromyalgia and the potential for long term effects of drugs, we extracted data only from the first phase of crossover trials, before washout and crossover. When available, we collected outcome data grouped by diagnostic criteria, sex, baseline anxiety or depression, and employment or education status, anticipating that these variables might modify the effects of treatment.

Risk of bias and trial trustworthiness

After training and calibration, two reviewers independently assessed the risk of bias for each outcome within each trial with a modified version of the revised Cochrane risk of bias tool for randomised trials (RoB 2).⁵⁷ Reviewers resolved disagreements by discussion or by consultation with a third reviewer, if necessary. RoB 2 evaluates five domains: bias in randomisation (eg, sequence generation and allocation concealment); bias from deviations from the intended intervention (eg, lack of masking); bias from missing outcome data (eg, attrition); bias in outcome measurement (eg, unmasked assessment); and selective reporting (eg, reporting influenced by results). In assessing deviations from the intended intervention, we focused on the effect of assignment rather than adherence because this approach reflects the effect most relevant to clinical practice. Our modified tool retained the same domains as the original tool, but with revised response options (ie, definitely low, probably low, probably high, and definitely high risk of bias) and guidance tailored to concerns relevant for our review (eg, removing guidance for assessing risk of bias of adhering to the intervention and listing important co-interventions) to improve inter-rater reliability.^{58 59} The overall risk of bias was determined by the highest rating across domains.

Unlike the original Cochrane tool, RoB 2 does not automatically assign high risk of bias to unblinded trials, but rather the tool requires reviewers to judge whether lack of masking could cause bias.⁶⁰ We rated unblinded trials as high risk of bias because knowledge of treatment assignment could prompt participants or clinicians to modify care beyond the assigned intervention—for example, by initiating, withholding, or intensifying nonprotocol co-interventions or rescue therapies like physical activity or social engagement—creating differential deviations attributable to the trial context. We also predicted that expectancy effects, shaped by patients' and clinicians' preconceptions about the effectiveness of an intervention, might influence reported outcomes.

In response to growing concerns about untrustworthy trial publications,^{61 62} we applied a preliminary version of the INSPECT-SR (Investigating Problematic Clinical Trials in Systematic Reviews) checklist to eligible trials that reviewers flagged during data extraction as potentially untrustworthy, or that reported results inconsistent with other trials (eg, unusually large effects).⁶³ We used the checklist to identify possible signs of data fabrication, data falsification (manipulation of data or results), or errors in trial conduct or data analysis that could seriously undermine the validity of a trial's conclusions. The checklist includes 21 items in four domains: postpublication notices; study conduct, governance, and transparency; text and figures; and study results. Each domain guides users to judge trustworthiness on a scale of no concerns, some concerns, or serious concerns, culminating in an overall study level judgment. We flagged estimates when the data relied mainly on trials with signs of untrustworthiness. In flagging estimates, we considered the direction in which the suspected untrustworthiness would distort the result, flagging an estimate only when it could plausibly alter the interpretation (eg, untrustworthiness expected to exaggerate benefit would not cause us to flag an estimate that already showed little or no effect).

Data synthesis and analysis

To describe trials and participants, we used descriptive characteristics: means, medians, and ranges for continuous variables, and counts and proportions for dichotomous and categorical variables. To summarise the comparative benefits and harms of drug treatments, we performed frequentist random effects network and pairwise meta-analyses with the restricted maximum likelihood heterogeneity estimator, an overall conservative approach to meta-analysis.⁶⁴ We selected a frequentist approach for feasibility given the size of the network and number of outcomes, and because empirical comparisons suggest that frequentist and bayesian network meta-analyses often give similar point estimates and inferences; differences more commonly reflect choice of priors and the interpretation of uncertainty.^65–69

Data on sleep quality were too sparse across measures and comparisons for a network meta-analysis and therefore we report estimates from pairwise meta-analyses for this outcome. We excluded from the primary analysis withdrawal and run-in enrichment designs, where only patients who tolerated or responded to the active treatment or a treatment within the same class during a run-in period were randomised, because of the tendency of these designs to overestimate benefits and tolerability, but the findings are described in the online supplementary material.^{70 71}

To enable a network meta-analysis, we classified drugs into nodes considering their drug class (eg, selective serotonin reuptake inhibitor antidepressants) and mechanism of action (online supplement 2). When the mechanism of action was unknown or uncertain, we classified drugs as separate nodes, even if the drugs belonged to the same therapeutic class (eg, different types of anticonvulsants). We disregarded drugs or co-interventions that were equally offered or accessible to all arms within a trial. We grouped placebo and usual care into the same node; only six trials compared drugs against usual care. Different doses of the same drug were also grouped within the same node. To test the consistency of effects within each node, we performed pairwise subgroup analyses based on dose or specific drug types. We had planned to separate nodes if credible evidence suggested that different drugs or doses within the same node produced divergent effects, but we found no such evidence. To facilitate meta-analysis, we converted non-parametric outcome data, such as medians and interquartile ranges, into means and standard deviations with established methods when parametric measures were not reported.⁷² In sensitivity analyses, we excluded trials where we converted effect estimates to assess the impact of these conversions.

We rescaled all measures of pain intensity to an 11 point numeric rating scale (0-10), with higher scores indicating greater pain intensity, with linear transformation.⁷³ When the fibromyalgia impact questionnaire was reported on a modified 0-80 scale, excluding two employment items, we converted the scores to the standard 0100 scale.⁵⁰ We applied the same approach to single item sleep quality measures, such as visual analogue or numeric rating scales, converting them to an 11 point scale (0-10), with higher scores reflecting better sleep quality. We did not convert effects across multi-item measures of sleep quality because of potential differences in the construct covered by each instrument. We also avoided standardised mean differences, which can be distorted by differences in variability across trial populations.⁷³

For each trial, we combined arms evaluating different doses of the same drug by summing events and participants for dichotomous outcomes and calculating weighted means and pooled standard deviations for continuous outcomes. We used mean differences to summarise continuous outcomes and relative risks for dichotomous outcomes, except for serious adverse events, which we summarised with risk differences due to the frequent occurrence of zero events. For relative risk, we excluded trials where no events were seen across all arms and applied continuity corrections (+0.5 increments to event counts) when zero events occurred in only one arm. For serious adverse events, to calculate variance for risk differences, we applied continuity corrections for trials in which zero events occurred in one or more arms.

Critical assumptions of a network meta-analysis include transitivity, also called joint randomisability, and coherence, which we assessed with established methods.^{74 75} Specifically, we examined the strength, credibility, and distribution of possible effect modifiers across comparisons, including sex, diagnostic criteria, baseline anxiety or depression, and the instruments used to measure outcomes. Pairwise subgroup analyses provided no evidence of effect modification by these variables. To test for incoherence, we applied node splitting models to detect local inconsistency and design-by-treatment interaction models to assess global inconsistency.^{76 77}

We used the I² statistic to summarise the magnitude of heterogeneity in meta-analyses and interpreted an I² value of 0-40% as not important, 30-60% as moderate, 50-90% as substantial, and 75-100% as considerable.^{78 79} The I² value, however, is prone to misinterpretation because even small degrees of unimportant inconsistency may translate to high I² values if estimates from studies are highly precise.^{80 81} Thus we also considered the absolute magnitude of differences in effect estimates across studies.

For comparisons comprising ≥10 studies, we looked for evidence of publication bias with Egger's test and qualitatively reviewed funnel plots for signs of asymmetry.^{82 83} To enhance the interpretability of the results, we transformed relative risks to absolute effects (number of events per 1000 patients), by using the median risk of the outcome reported across the control groups of trials contributing to the pooled effect as the baseline risk. We performed all analyses with the meta and netmeta packages in R (Vienna, Austria; version 4.1.2).^{84 85}

Certainty of evidence

We used ICEMAN (Instrument to Assess the Credibility of Effect Modification Analyses) to assess the credibility of significant subgroup effects and avoid spurious inferences.⁸⁶ ICEMAN judges subgroup credibility based on several criteria, including the number of trials informing each subgroup, significance of the interaction test, and whether comparisons occurred within rather than between trials. We judged subgroup effects as having at most low credibility when only one trial contributed to one or more subgroups, the interaction P value was >0.05, and comparisons were between rather than within trials.

Reviewers, working independently and in duplicate, assessed the certainty of the evidence with the GRADE approach, resolving disagreements through discussion or with adjudication by a third reviewer.^{74 87 88} The GRADE approach rates evidence as high, moderate, low, or very low certainty, considering risk of bias, inconsistency (heterogeneity), indirectness (differences between study and review questions), publication bias, and imprecision (random error). For network meta-analyses, GRADE also considers intransitivity (differences in effect modifiers across treatment comparisons) and incoherence (differences between direct and indirect estimates). To ensure feasibility, we assessed the certainty of evidence only for comparisons with placebo or usual care. In cases of incoherence, we prioritised the estimate (either direct or indirect) with the higher certainty of evidence.

To interpret effect estimates, we compared effect estimates with the minimum important difference, the smallest difference that patients might perceive as important.⁸⁹ When the point estimate met or exceeded the minimum important difference, we judged the effect important and rated our certainty that an important effect exists. When the point estimate was below the minimum important difference for benefit or harm, we rated our certainty that the effect was trivial or absent. We downgraded one level if the confidence interval crossed one minimum important difference or if the total sample size was 50-100% of the optimal information size; two levels if the interval crossed two minimum important differences or if the sample was <50% of the optimal information size; and three levels if the interval was implausibly wide or the sample too small to be confident in prognostic balance between treatment arms.⁸⁹

We sourced minimum important differences either from the literature or, when not available, based on consensus of the review authors. If we could not source sufficiently credible minimum important differences for instruments specific to fibromyalgia, we considered similar conditions associated with chronic pain. If we identified several candidate minimum important differences, we prioritised those based on their credibility, and where minimum important differences were of comparable credibility, we used the median value.⁹⁰ Based on the literature and insights from review authors, we considered a difference of 10% as minimally important for patient reported global improvement, 1.5 points on the 11 point numeric rating scale for pain intensity,⁹¹ eight points on the fibromyalgia impact questionnaire,⁹² four points on the SF-36 physical component score,^93–97 five points on the SF-36 mental component score,^{93 95–97} 10% for all cause discontinuations, and 5% for serious adverse events. Minimum important differences for sleep varied based on the measure used (eg, 1.5 points for 11 point sleep quality numeric rating scale or visual analogue scale, five points for the Medical Outcomes Study Sleep Problems Index).⁹⁸

We relied on network estimates unless we identified incoherence, where we used the direct or indirect estimate, whichever provided higher certainty evidence. High or moderate certainty evidence indicates confidence that the estimated effect lies on the same side of the minimum important difference as the true effect; low or very low certainty reflects that the true effect may possibly fall on a different side of the minimum important difference. Online supplement 3 presents the GRADE approach used to assess the certainty of evidence.

Reporting of results

We report this review according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) extension for network meta-analyses.⁹⁹ We described results with GRADE plain language summaries: high certainty evidence with declarative statements, moderate certainty as likely or probably, low certainty as may, and very low certainty as very uncertain.¹⁰⁰ We interpreted effect sizes against minimum important difference thresholds. When point estimates met or exceeded the minimum important difference, we reported that the intervention improved (or worsened) the outcome. When point estimates were below the minimum important difference, we reported no patient-important effect.

Mean differences capture average treatment effects but can obscure meaningful variation in individual responses. To clarify the interpretation of pain as a continuous outcome, we also reported the proportion of patients who achieved a reduction in pain intensity of at least 30%. We presented these responder data alongside the continuous pain results to support inferences about the proportion of patients likely to have experienced important benefit. For clarity, the main manuscript presents results for drugs studied in more than one eligible trial and the online supplementary material shows findings for all drugs, including those studied in only one trial. Because trials measured sleep outcomes with diverse instruments, we presented these results in a separate summary of findings table.

Deviations from the protocol

We previously published a protocol for this review,¹⁰¹ which proposed a network meta-analysis of all fibromyalgia treatments. Because the evidence base proved large, we divided the review and report here the pharmacological network meta-analysis; we will synthesise non-pharmacological interventions separately. The protocol also proposed searches of PapersFirst and ProceedingsFirst. We excluded conference abstracts because the literature was too large and we did not search those sources.

For crossover trials, the protocol stated that we would pool crossover and parallel trials with paired (within person) methods. Instead, we restricted crossover analyses to the first period before crossover. Many trials tested antidepressants, which can produce carryover effects after discontinuation through longer lasting psychological and behavioural changes, such as learnt coping or altered stress reactivity. To support interpretation of pain, we also extracted the proportion of patients achieving a reduction in pain intensity of at least 30% and synthesised these data with a network meta-analysis.

The protocol specified earlier Cochrane guidance for risk of bias (version 1.0). We assessed risk of bias with the most recent Cochrane endorsed tool for randomised trials (RoB 2). The protocol proposed using standardised mean differences when different instruments measured the same construct. We avoided standardised mean differences because differences in variability across trial populations can distort these data.⁷³ Finally, the protocol planned a bayesian random effects network meta-analysis. We selected a frequentist approach given current evidence suggesting minimal differences from bayesian methods and to avoid the computational burden of fitting large networks under a bayesian framework.⁶⁵

Patient and public involvement

Patients and/or the public were not involved in the design, conduct, reporting, or dissemination plans of this research. We undertook this review to inform a forthcoming parallel BMJ Rapid Recommendation, through which patient partners will interpret the findings and advise on the balance of benefits and harms of the interventions. We will disseminate the results to patient partners and relevant patient and public communities through the guideline process.

Study characteristics

Table 1 lists the characteristics of included trials and online supplement 6 has additional details. The trials tested 88 unique drugs or drug combinations across 67 drug classes (nodes). The most frequently studied were SNRIs, gabapentinoids, tricyclic antidepressants, and selective serotonin reuptake inhibitors (SSRIs). Six trials compared drugs with usual care, 123 with placebo, 15 with other active agents, and seven with different doses of the same drug. Usual care typically involved follow-up with a general practitioner and, in some trials, physicians could prescribe additional drugs at their discretion.

Eighteen trials (11.9%) were unpublished; results were available as part of trial registrations. Half of the trials were published between 2005 and 2020, eight used crossover designs, and nearly half were conducted in the US. More than half of the trials reported industry funding and about half were registered. Among trials that reported the proportion of eligible patients who consented to participation, participation was high (median 87.7%). Trials included very few men (median 5.4%).

Most trials applied the American College of Rheumatology criteria, predominantly the 1990 version (114; 75.5%).¹⁰² Seventeen trials (11.3%) used alternative diagnostic criteria, which were similar to criteria from the American College of Rheumatology, requiring widespread pain, fatigue, and sleep disturbance (online supplement 7). Twenty seven trials excluded patients with depression, 12 excluded those with anxiety, and 10 excluded patients with both depression and anxiety; conversely, most patients had baseline depression in seven trials and anxiety in two. Trials did not report on return to work or education. Seven trials (4.6%) used enrichment designs; only patients who tolerated or responded to treatment during a run-in period were randomised. We excluded these trials from our primary analyses.

Concerns about trustworthiness were raised for 15 trials; 12 on the INSPECT-SR checklist and three in domains not captured by INSPECT-SR (online supplement 8). Reviewers rated six trials as having some concerns and six as having serious concerns on the INSPECT-SR checklist. Concerns about study results were most common (11 trials), including implausible outcome data or treatment effects (seven trials), followed by concerns about study conduct, governance, and transparency, such as missing registration or implausible recruitment (nine trials). Postpublication notices identified a history of retractions among authors of two trials.

Summary of findings

Figure 1 summarises the results of the network meta-analyses, focusing on interventions tested in two or more trials, along with the GRADE ratings and interpretations. Online supplement 9 has the full summary of the findings.

Additional results and sensitivity analyses

Online supplement 48 summarises the findings from withdrawal and run-in enrichment trials. These trials, where only participants who had tolerated or responded to active treatment during a preliminary phase were randomised, generally produced results consistent with the primary analyses. In enriched populations, drugs were typically associated with less discontinuation of treatment.

We performed sensitivity analyses for our key outcomes, patient reported improvement or recovery, pain intensity, and discontinuation of the intervention. We excluded trials that raised concerns about untrustworthy, trials where we converted non-parametric statistics (eg, medians and interquartile ranges) to means and standard deviations to facilitate meta-analysis, trials that did not use the American College of Rheumatology criteria to identify patients with fibromyalgia, trials that did not use the American College of Rheumatology 1990 criteria, and trials where ≥20% of patients had baseline anxiety or depression. Differences in effect estimates between the primary and sensitivity analyses were almost always below the minimum important difference, and in the few instances where effect estimates were not, the direction and clinical interpretation of the effect remained the same (online supplements 49-61).

Main findings

This systematic review and network meta-analysis synthesised evidence from 151 randomised trials of >30 000 patients with fibromyalgia. We found that drug treatments had limited and uncertain benefits. No drug showed moderate or high certainty evidence of important improvement in outcomes. Where moderate or high certainty evidence existed, the evidence indicated little or no effect.

Low certainty evidence suggested that sodium oxybate, selective NRIs, SNRIs, and gabapentinoids may increase patient reported improvement, sodium oxybate and SSRIs may improve fibromyalgia symptoms, and gabapentinoids, sodium oxybate, and tricyclic antidepressants may improve sleep quality. Although mean reductions in pain did not meet the minimum important difference for sodium oxybate, selective NRIs, SNRIs, or cyclobenzaprine, some patients may achieve a reduction in pain intensity of at least 30% compared with placebo or usual care. None of these drugs improved mental health. Except for low potency opioids, which increased the risk of treatment discontinuation, trial drugs seemed to be acceptable and tolerable.

Interpretation of findings

Although we found evidence of modest benefits for some drugs, the evidence was of low certainty. Low certainty evidence may be substantially different from the true effect and is susceptible to change as new evidence emerges.¹²⁴ Moreover, benefits were largely seen for patient reported improvement, typically measured with the Patient Global Impression of Change scale, and did not extend to other core outcomes, such as pain intensity or physical health.^{48 125 126} Several factors may explain this discrepancy. The Patient Global Impression of Change scale requires patients to compare their current health with a previous time point, making the scale susceptible to recall bias and expectancy effects.⁴⁸ Although these biases might be expected to operate equally across trial arms, the side effects of a drug could disclose treatment allocation, introducing differential measurement error.

Nonetheless, the Patient Global Impression of Change scale still offers advantages. Mean differences can understate benefit when treatment responses vary; large gains in a subset are averaged with deterioration or minimal change in others. The scale captures the proportion of patients who report net benefit. The scale also enables patients to weigh the aspects of health that they consider most important and mirrors routine practice where patients' overall impressions guide treatment decisions. Because the scale captures broad perceptions of change, integrating pain, mood, sleep, coping, daily function, and other domains, patients may report improvement even when core outcomes, such as pain or physical health, show little or no change. This distinction may be particularly important for fibromyalgia drugs, such as SNRIs and selective NRIs, which can affect mood, arousal, and overall wellbeing without directly improving pain or physical health. Similarly, broad constructs, such as physical and mental health, may require longer treatment, structured rehabilitation, or other concurrent interventions to change substantially. The Patient Global Impression of Change scale may therefore detect global benefits that are not fully captured by more narrowly defined outcomes or by outcomes less responsive to drug treatment alone.

The absence of compelling benefits on pain may reflect a narrow focus on intensity rather than the broader experience of pain. Some drugs may not reduce intensity directly but could influence how pain is perceived, its emotional salience, or patients’ capacity to cope. If such effects were meaningful, however, these effects would also appear as parallel improvements in physical health, mental health, or overall fibromyalgia symptoms, none of which were seen in this review. Furthermore, for some drugs, although mean reductions did not achieve the minimally important difference, 1-2 in 10 patients reported a reduction in pain intensity of at least 30% compared with placebo or usual care. These estimates suggest that, even where average effects fall short of clinical importance, a meaningful subset of patients may derive important pain relief.

Although sodium oxybate seemed to be one of the most effective drugs in our review, serious safety concerns restrict its clinical use. Sodium oxybate has FDA boxed warnings for central nervous system depression and for abuse and misuse; the boxed text notes the potential for serious harms, including seizures, respiratory depression, decreased consciousness, coma, and death.¹²⁷ Because of these risks, sodium oxybate is classified as a heavily controlled substance. In the US, sodium oxybate is only available under a risk evaluation and mitigation strategy, in Canada, special authorisation and risk protocols are required, and in the UK, the drug can only be prescribed by narcolepsy specialists. Although trial data suggest that sodium oxybate is tolerable, follow-up in trials lasted only 8-14 weeks, and randomised trials were not optimised to detect rare or long term adverse events.^128–132

Gabapentinoids also warrant attention to safety. In the US, pregabalin is a schedule V controlled substance.¹³³ In 2019, the FDA added warnings about respiratory depression with gabapentinoids, particularly when used with opioids or other central nervous system depressants, or in patients with underlying respiratory disease.^{134 135} Other concerns include suicidality (a class warning for anticonvulsants), dizziness, somnolence, weight gain, and withdrawal symptoms with abrupt discontinuation.^{136 137} Although our review found that gabapentinoids did not increase treatment discontinuation (high certainty) or serious adverse events (moderate certainty), the trials were brief and not designed to detect misuse or longer term harms.

Several limitations restrict the strength of inferences drawn from the available evidence. Most trials reported considerable attrition and failed to clarify whether patients who discontinued treatment continued to be assessed for outcomes. Investigators frequently used last observation carried forward to impute missing data, a method that risks masking clinical deterioration and underestimating harms.

The generalisability of these findings to contemporary clinical practice may be limited. Nearly all trials enrolled patients based on the 1990 American College of Rheumatology criteria, which require widespread pain and pain at specific tender points.¹⁰² Contemporary diagnostic criteria, however, no longer include tender points and emphasise non-musculoskeletal symptoms.^{138 139} Observational data also suggest that clinicians frequently bypass formal diagnostic criteria entirely.^{138 140 141} Consequently, historical trial populations likely represent a narrower subset of patients than those who received a diagnosis in contemporary clinical practice. Antidepressants such as SNRIs may show greater effectiveness in this broader population characterised by more diverse, non-musculoskeletal symptoms.

The brevity of follow-up was another limitation. Only nine of 151 trials reported outcomes after six months of follow-up (median duration of follow-up 12 weeks). Evidence suggests that many patients discontinue drug treatments within a year because of poor tolerability or diminishing benefit.51 52 142 143 As a result, our review likely overestimated both efficacy and tolerability compared with findings in clinical practice.

Most of the trials in this review included patients who had not previously used drugs for fibromyalgia. Hence, our review does not provide direct evidence on the effects of drugs for patients who did not initially respond to drug treatments. Also, no evidence exists on the order in which drug treatments may be trialled, with limited evidence on the effectiveness of combinations of these drugs.

Our review relied on patient reported outcomes. Although concerns about the reliability of these measures may exist, symptoms of fibromyalgia, such as pain, are subjectively experienced, and no objective laboratory measures have been established to predict benefit in terms of how patients feel or function. To consider objective measures of function, we included return to full time work and education as one of our outcomes of interest, but trials did not report on this outcome. Given the short duration of most trials, meaningful effects on work or education would have been unlikely.

We analysed outcomes at the time point closest to the end of treatment. In some cases, available data may have fallen outside the expected therapeutic window. For example, a trial comparing five day courses of intravenous lidocaine and ketamine reported outcomes at four weeks after treatment, a time point likely beyond the period of benefit.¹⁴³ More broadly, given the chronic nature of fibromyalgia, drugs with only short lived effects are unlikely to offer sustained clinical value when discontinued. Also, we identified 15 trials with concerns about trustworthiness, affecting the comparisons of eight drugs with placebo or usual care. These concerns may not be immediately apparent to evidence users, potentially misleading patients and healthcare providers, and adversely affecting care.

Comparison with other studies

The absence of head-to-head comparisons for drug treatments for fibromyalgia is a well recognised barrier to optimising care.¹⁴⁴ Our systematic review and network meta-analysis attempted to address this research gap by synthesising both direct and indirect evidence. Our findings align with previous reviews showing that drug treatments provide limited clinical benefit.²⁸ Consistent with previous reviews, we found that gabapentinoids and SNRIs showed signals of efficacy, but their benefits were modest and of low certainty.^{25 27} Although tricyclic antidepressants seemed to improve sleep quality, unlike earlier reviews we found no evidence that these drugs reduced pain or improved quality of life.^{24 25} These discrepancies may be attributed to different operationalisations of outcomes (mean pain v proportion achieving meaningful pain reduction), inadequate consideration of the importance of effect sizes, and whether reviews assessed the certainty of evidence.^{24 25}

Our review also synthesised evidence on the selective NRIs, ampreloxetine and esreboxetine, which are investigational drugs and, to our knowledge, have not been reviewed in previous reports. The evidence supporting this class, although limited in certainty and modest in magnitude, parallels that for approved agents, such as SNRIs and gabapentinoids. Our finding that sodium oxybate improved fibromyalgia outcomes aligns with the position of the European Alliance of Associations for Rheumatology, which acknowledges efficacy but advises against use because of its safety profile.¹⁷

Strengths and limitations of this study

The strengths of our systematic review include a rigorous and comprehensive search for eligible studies, screening and extraction of data in duplicate, focus on patient important outcomes, and the use of GRADE methods to evaluate the certainty of evidence. Despite our rigorous search of the literature, we may have missed eligible studies. We attempted to mitigate this issue by reviewing the references of similar systematic reviews and soliciting experts about eligible trials not identified by our search. We are also aware of the results of one trial becoming available since the last date of our search and several ongoing trials.^145–148 These trials may warrant future updates to this review.

To facilitate our network meta-analysis, we grouped drugs of the same class and mechanism of action into the same node. The mechanism of action for many drugs is uncertain, however, and even drugs with similar mechanisms can differ in their affinity for receptors across the body. Invariably, some investigators may consider select combinations of drugs in the same node suboptimal. When the mechanism of action was unknown or uncertain, we classified drugs as separate nodes, even if the drugs belonged to the same therapeutic class (eg, different types of anticonvulsants). We also conducted subgroup analyses to assess whether drugs grouped within the same node produced different effects.

Several planned subgroup analyses were not feasible due to limitations of the evidence. Most trials used the 1990 American College of Rheumatology criteria, with too few using the 2010 or 2016 versions to allow comparison. Trials also enrolled predominantly women: only four trials included >30% men, and none reported outcomes stratified by sex. No trials reported disability benefits status of participants. Future trials that adopt contemporary diagnostic criteria, enrol adequate numbers of men, report results grouped by sex, and record receipt of disability benefits may clarify whether these factors modify treatment effects.

Also, we applied a prespecified hierarchy when trials reported more than one measure of pain intensity to minimise the potential for selective inclusion of pain measures, to avoid multiplicity, and to maximise comparability across studies by prioritising the most commonly reported and best validated global pain intensity measure used in fibromyalgia trials. Combining different measures of pain, however, could introduce clinical heterogeneity and complicate the interpretation of pooled estimates. In our dataset, nearly all of the trials reported pain intensity as average daily or weekly pain severity. Only four trials used alternative constructs (current pain intensity, spontaneous pain intensity, morning pain, and pain sensations during activity and rest), suggesting this limitation is unlikely to influence our findings.

Our review prioritised outcomes that core outcome sets and patient input identified as most important. Other outcomes, such as cognitive dysfunction, may be important in a subset of patients but fell outside the scope of this review. Similarly, our focus on pain intensity may not have captured the broader experience of pain, such as patients’ capacity to cope.

Sparse data for measures and treatment comparisons precluded network meta-analysis for sleep quality. This limitation restricted indirect comparisons between drugs and reduced the precision that indirect evidence could provide. Consensus on a core sleep measure that captures the dimensions most important to fibromyalgia patients would enable more uniform reporting across trials and, in turn, more informative synthesis.

The networks for commonly studied interventions were largely star shaped, with active treatments connected mainly through placebo or usual care, and few closed loops. Because statistical tests for incoherence require closed loops, we could rarely assess incoherence statistically. We mainly presented and interpreted comparisons with placebo or usual care, which almost always were based on direct evidence and are unaffected by the absence of closed loops. Comparisons among active drugs rely on indirect evidence and the transitivity assumption, and may warrant more cautious interpretation.

We assessed the certainty of the evidence with the GRADE approach. Although GRADE provides a structured and transparent framework for evaluating certainty, its use is subjective, and reasonable reviewers may reach different conclusions.¹⁴⁹ We evaluated imprecision against minimum important differences based on either published sources or, when unavailable, on consensus among review authors. What constitutes an important effect, however, may vary across patients; some evidence suggests that individuals with chronic conditions may value even small improvements.¹⁵⁰

We applied a preliminary version of the INSPECT-SR checklist to some trials to identify possible signs of data fabrication, falsification (manipulation of data or results), or analytic error that could compromise their conclusions.⁶³ We may have missed some of these concerns or misidentified trials. Methods to detect these problems may have poor sensitivity and specificity. We did not interpret the identification of a trial as untrustworthy as an accusation of research misconduct. Trials may be flagged for reasons that include genuine error, overly optimistic interpretation (intentional or unintentional), differences in research environments or scientific cultures from conventional western norms, or non-standard methodological practices.

We interpreted effect estimates in relation to the minimum important difference. For continuous outcomes, mean differences summarise average effects and may obscure clinically important improvements when treatment responses are heterogeneous; a small mean difference can still coexist with important improvement in a subset of participants. This approach presents a general interpretive challenge for continuous outcomes.³⁵ Finally, we disregarded drugs that were equally offered or accessible to all arms within a trial. Also, some drug combinations may produce synergistic or antagonistic effects, and future component network meta-analyses may inform these interactions.

Study implications

This systematic review highlights a gap in the evidence between current prescribing practices and guidance. Guidance continues to favour SNRIs, tricyclic antidepressants, and gabapentinoids, but the best available data showed only marginal or uncertain benefit.^{17 151} Drugs suspected of being most effective for fibromyalgia, including duloxetine, milnacipran, and pregabalin, did not seem to markedly improve physical health or improve fibromyalgia symptoms. Although mean reductions in pain did not seem to be clinically important for sodium oxybate, selective NRIs, SNRIs, or cyclobenzaprine, some patients may achieve a reduction in pain intensity of at least 30% compared with placebo or usual care. Opioids, although sometimes prescribed, did not show compelling evidence of benefit and are associated with hyperalgesia, tolerance, opioid use disorder, and other long term side effects not captured in trials.

In August 2025, the FDA approved a sublingual formulation of cyclobenzaprine for fibromyalgia. Two pre-approval trials (AFFIRM and RALLY) failed to meet their primary endpoints and were subsequently excluded from the FDA submission.^{152 153} We did not find evidence of important differences between the sublingual and oral formulations of cyclobenzaprine; oral cyclobenzaprine already has moderate bioavailability and a long half life. Although the sublingual formulation is half the dose of the oral formulation, presumably to reduce sedating effects, using half the tablet would allow patients to administer the same dose.

Evidence for some emerging or increasingly used treatments, including medical cannabis and low dose naltrexone, is limited.^{154 155} One small trial comparing cannabis (Bediol; 6.3% tetrahydrocannabinol and 8% cannabidiol), oxycodone, and their combination reported no differences.¹⁰⁶ Moderate certainty evidence suggests that naltrexone has little or no effect on pain intensity, fibromyalgia symptoms, or sleep quality.

When drugs are considered, physicians might reasonably trial antidepressants in patients with fibromyalgia and coexisting insomnia, anxiety, or depression, but efficacy is likely to be modest, and long term adherence may be poor because of limited benefit and adverse effects.^{51 52 142} Furthermore, clinical visits may stigmatise people with fibromyalgia. Clinicians sometimes attribute patients’ symptoms to psychological causes and dismiss the symptoms on that basis.¹⁵⁶ Antidepressants may produce nocebo effects in patients who perceive these drugs negatively.¹⁵⁷ Decisions about whether and when to start drug treatment, and how drug treatments compare with non-pharmacological care, require head-to-head trials with non-pharmacological treatments, and were outside the scope of this review. Moreover, current guidelines for managing fibromyalgia have not been updated since 2012 in Canada¹⁶ and since 2017 in Europe.¹⁷ Current guidance in the US is restricted to the diagnosis of fibromyalgia and does not address its management.¹⁸ Our results call for an updated guideline that re-evaluates new evidence.

Other priorities for future research may include direct comparisons between pharmacological and non-pharmacological interventions, investigation of commonly co-prescribed drug combinations, such as SNRIs with gabapentinoids, and assessment of long term effects, ideally for more than 52 weeks. Our finding that half of all fibromyalgia trials were unregistered, and 12% of trial data were available only through trial registries, highlights potential challenges with reporting and publication bias.

Conclusions

This systematic review and network meta-analysis suggests that drug treatments for fibromyalgia offer, at best, modest and uncertain benefits; no drug showed moderate or high certainty evidence of an important improvement in core outcomes. Low certainty evidence suggested that sodium oxybate, selective NRIs, SNRIs, and gabapentinoids may increase patient reported measures of global improvement, sodium oxybate and SSRIs may improve fibromyalgia symptoms, and sodium oxybate, gabapentinoids, and tricyclic antidepressants may improve sleep quality. Effects were typically small, approximating the minimum important difference.

On average, no drug reduced pain intensity by an important margin, but responder analyses suggested that a meaningful proportion of patients treated with sodium oxybate, selective NRIs, SNRIs, and cyclobenzaprine may achieve at least a 30% reduction in pain. Except for low potency opioids, which increased the risk of discontinuing treatment, trial drugs seemed to be tolerable and acceptable. Although sodium oxybate also showed low certainty evidence of benefit, the serious safety profile of the drug warrants caution and limits its use.

• 1.Siracusa R, Paola RD, Cuzzocrea S, et al. Fibromyalgia: pathogenesis, mechanisms, diagnosis and treatment options update. Int J Mol Sci 2021;22:3891. 10.3390/ijms22083891 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 2.Pomares FB, Funck T, Feier NA, et al. Histological underpinnings of grey matter changes in fibromyalgia investigated using multimodal brain imaging. J Neurosci 2017;37:1090–101. 10.1523/JNEUROSCI.2619-16.2016 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 3.Bourke JH, Wodehouse T, Clark LV, et al. Central sensitisation in chronic fatigue syndrome and fibromyalgia; a case control study. J Psychosom Res 2021;150:110624. 10.1016/j.jpsychores.2021.110624 [DOI] [PubMed] [Google Scholar]
• 4.Marques AP, Santo A de S do E, Berssaneti AA, et al. Prevalence of fibromyalgia: literature review update. Rev Bras Reumatol Engl Ed 2017;57:356–63. 10.1016/j.rbre.2017.01.005 [DOI] [PubMed] [Google Scholar]
• 5.Branco JC, Bannwarth B, Failde I, et al. Prevalence of fibromyalgia: a survey in five European countries. Semin Arthritis Rheum 2010;39:448–53. 10.1016/j.semarthrit.2008.12.003 [DOI] [PubMed] [Google Scholar]
• 6.Jones GT, Atzeni F, Beasley M, et al. The prevalence of fibromyalgia in the general population: a comparison of the American College of Rheumatology 1990, 2010, and modified 2010 classification criteria. Arthritis Rheumatol 2015;67:568–75. 10.1002/art.38905 [DOI] [PubMed] [Google Scholar]
• 7.Heidari F, Afshari M, Moosazadeh M. Prevalence of fibromyalgia in general population and patients, a systematic review and meta-analysis. Rheumatol Int 2017;37:1527–39. 10.1007/s00296-017-3725-2 [DOI] [PubMed] [Google Scholar]
• 8.Johnson CM, Makai GEH. Fibromyalgia and irritable bowel syndrome in female pelvic pain. Semin Reprod Med 2018;36:136–42. 10.1055/s-0038-1676090 [DOI] [PubMed] [Google Scholar]
• 9.Aaron LA, Buchwald D. A review of the evidence for overlap among unexplained clinical conditions. Ann Intern Med 2001;134:868–81. 10.7326/0003-4819-134-9_part_2-200105011-00011 [DOI] [PubMed] [Google Scholar]
• 10.Bileviciute-Ljungar I, Norrefalk J-R, Borg K. Pain burden in post-COVID-19 syndrome following mild COVID-19 infection. JCM 2022;11:771. 10.3390/jcm11030771 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 11.Robinson RL, Kroenke K, Mease P, et al. Burden of illness and treatment patterns for patients with fibromyalgia. Pain Med 2012;13:1366–76. 10.1111/j.1526-4637.2012.01475.x [DOI] [PubMed] [Google Scholar]
• 12.Fitzcharles M-A, Da Costa D, Pöyhiä R. A study of standard care in fibromyalgia syndrome: a favorable outcome. J Rheumatol 2003;30:154–9. [PubMed] [Google Scholar]
• 13.Wolfe F, Anderson J, Harkness D, et al. Work and disability status of persons with fibromyalgia. J Rheumatol 1997;24:1171–8. [PubMed] [Google Scholar]
• 14.Martinez JE, Ferraz MB, Sato EI, et al. Fibromyalgia versus rheumatoid arthritis: a longitudinal comparison of the quality of life. J Rheumatol 1995;22:270–4. [PubMed] [Google Scholar]
• 15.Walitt B, Fitzcharles MA, Hassett AL, et al. The longitudinal outcome of fibromyalgia: a study of 1555 patients. J Rheumatol 2011;38:2238–46. 10.3899/jrheum.110026 [DOI] [PubMed] [Google Scholar]
• 16.Fitzcharles M-A, Ste-Marie PA, Goldenberg DL, et al. 2012 Canadian Guidelines for the diagnosis and management of fibromyalgia syndrome: executive summary. Pain Res Manag 2013;18:119–26. 10.1155/2013/918216 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 17.Macfarlane GJ, Kronisch C, Dean LE, et al. EULAR revised recommendations for the management of fibromyalgia. Ann Rheum Dis 2017;76:318–28. 10.1136/annrheumdis-2016-209724 [DOI] [PubMed] [Google Scholar]
• 18.Arnold LM, Bennett RM, Crofford LJ, et al. AAPT diagnostic criteria for fibromyalgia. J Pain 2019;20:611–28. 10.1016/j.jpain.2018.10.008 [DOI] [PubMed] [Google Scholar]
• 19.Forte ML, Butler M, Andrade KE, et al. Table 1, fda-approved drugs for the treatment of fibromyalgia. In: Treatments for fibromyalgia in adult subgroups. Rockville (MD): Agency for Healthcare Research and Quality (US), 2015. [Google Scholar]
• 20.Giorgi V, Sarzi-Puttini P, Pellegrino G, et al. Pharmacological treatment of fibromyalgia syndrome: a practice-based review. Curr Pain Headache Rep 2024;28:1349–63. 10.1007/s11916-024-01277-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 21.Moore RA, Derry S, Aldington D. Amitriptyline for neuropathic pain in adults. Cochrane Database Syst Rev 2019. 10.1002/14651858.CD007076.pub3 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 22.Tofferi JK, Jackson JL, O’Malley PG. Treatment of fibromyalgia with cyclobenzaprine: a meta-analysis. Arthritis Rheum 2004;51:9–13. 10.1002/art.20076 [DOI] [PubMed] [Google Scholar]
• 23.Agarwal A, Emary PC, Gallo L, et al. Physicians’ knowledge, attitudes, and practices regarding fibromyalgia: a systematic review and meta-analysis of cross-sectional studies. Medicine (Baltimore) 2024;103:e39109. 10.1097/MD.0000000000039109 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 24.Farag HM, Yunusa I, Goswami H, et al. Comparison of amitriptyline and US Food and Drug Administration-approved treatments for fibromyalgia: a systematic review and network meta-analysis. JAMA Netw Open 2022;5:e2212939. 10.1001/jamanetworkopen.2022.12939 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 25.Alberti FF, Becker MW, Blatt CR, et al. Comparative efficacy of amitriptyline, duloxetine and pregabalin for treating fibromyalgia in adults: an overview with network meta-analysis. Clin Rheumatol 2022;41:1965–78. 10.1007/s10067-022-06129-8 [DOI] [PubMed] [Google Scholar]
• 26.Mascarenhas RO, Souza MB, Oliveira MX, et al. Association of therapies with reduced pain and improved quality of life in patients with fibromyalgia: a systematic review and meta-analysis. JAMA Intern Med 2021;181:104–12. 10.1001/jamainternmed.2020.5651 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 27.Migliorini F, Maffulli N, Eschweiler J, et al. Duloxetine for fibromyalgia syndrome: a systematic review and meta-analysis. J Orthop Surg Res 2023;18:504. 10.1186/s13018-023-03995-z [DOI] [PMC free article] [PubMed] [Google Scholar]
• 28.Moore A, Bidonde J, Fisher E, et al. Effectiveness of pharmacological therapies for fibromyalgia syndrome in adults: an overview of Cochrane Reviews. Rheumatology (Oxford) 2025;64:2385–94. 10.1093/rheumatology/keae707 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 29.Migliorini F, Maffulli N, Eschweiler J, et al. Pharmacological management of fibromyalgia: a Bayesian network meta-analysis. Expert Rev Clin Pharmacol 2022;15:205–14. 10.1080/17512433.2022.2044792 [DOI] [PubMed] [Google Scholar]
• 30.Shing CH, Wang F, Lau LNL, et al. Skeletal muscle relaxant for the treatment of fibromyalgia: a systematic review and meta-analysis of randomized controlled trials. Reg Anesth Pain Med 2025;50:937–47. 10.1136/rapm-2024-105776 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 31.da Rocha AP, Mizzaci CC, Nunes Pinto ACP, et al. Tramadol for management of fibromyalgia pain and symptoms: Systematic review. Int J Clin Pract 2020;74:e13455. 10.1111/ijcp.13455 [DOI] [PubMed] [Google Scholar]
• 32.Hudson J, Imamura M, Robertson C, et al. Effects of pharmacologic and nonpharmacologic interventions for the management of sleep problems in people with fibromyalgia: systematic review and network meta-analysis of randomized controlled trials. Arthritis Care Res 2025;77:1095–105. 10.1002/acr.25505 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 33.Turk DC, Dworkin RH, Allen RR, et al. Core outcome domains for chronic pain clinical trials: IMMPACT recommendations. Pain 2003;106:337–45. 10.1016/j.pain.2003.08.001 [DOI] [PubMed] [Google Scholar]
• 34.Turk DC, Dworkin RH, Revicki D, et al. Identifying important outcome domains for chronic pain clinical trials: an IMMPACT survey of people with pain. Pain 2008;137:276–85. 10.1016/j.pain.2007.09.002 [DOI] [PubMed] [Google Scholar]
• 35.Busse JW, Bartlett SJ, Dougados M, et al. Optimal strategies for reporting pain in clinical trials and systematic reviews: recommendations from an OMERACT 12 workshop. J Rheumatol 2015;42:1962–70. 10.3899/jrheum.141440 [DOI] [PubMed] [Google Scholar]
• 36.Murad MH, Wang Z, Chu H, et al. When continuous outcomes are measured using different scales: guide for meta-analysis and interpretation. BMJ 2019;364:k4817. 10.1136/bmj.k4817 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 37.Guyatt GH, Oxman AD, Vist GE, et al. GRADE: an emerging consensus on rating quality of evidence and strength of recommendations. BMJ 2008;336:924–6. 10.1136/bmj.39489.470347.AD [DOI] [PMC free article] [PubMed] [Google Scholar]
• 38.Mills EJ, Thorlund K, Ioannidis JPA. Demystifying trial networks and network meta-analysis. BMJ 2013;346:bmj.f2914. 10.1136/bmj.f2914 [DOI] [PubMed] [Google Scholar]
• 39.Siemieniuk RA, Agoritsas T, Macdonald H, et al. Introduction to BMJ rapid recommendations. BMJ 2016;354:i5191. 10.1136/bmj.i5191 [DOI] [PubMed] [Google Scholar]
• 40.World Health Organization . World health organization model list of essential medicines – 23rd list. In: Executive summary of the report of the 24th WHO Expert Committee on the Selection and Use of Essential Medicines. Geneva, 2023. [Google Scholar]
• 41.Dickersin K, Min YI. Publication bias: the problem that won’t go away. Ann N Y Acad Sci 1993;703:135–46. 10.1111/j.1749-6632.1993.tb26343.x [DOI] [PubMed] [Google Scholar]
• 42.Higgins JPT, Thomas J, Chandler J, et al. Cochrane handbook for systematic reviews of interventions version 6.5 (updated august 2024) [Cochrane]. 2024. Available: www.training.cochrane.org/handbook
• 43.Forbes C, Greenwood H, Carter M, et al. Automation of duplicate record detection for systematic reviews: Deduplicator. Syst Rev 2024;13. 10.1186/s13643-024-02619-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 44. The Retraction Watch Database. New York: The Center for Scientific Integrity, 2018. Available: http://retractiondatabase.org [Google Scholar]
• 45.Mease PJ, Clauw DJ, Christensen R, et al. Toward development of a fibromyalgia responder index and disease activity score: OMERACT module update. J Rheumatol 2011;38:1487–95. 10.3899/jrheum.110277 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 46.Choy EH, Arnold LM, Clauw DJ, et al. Content and criterion validity of the preliminary core dataset for clinical trials in fibromyalgia syndrome. J Rheumatol 2009;36:2330–4. 10.3899/jrheum.090368 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 47.Mease P, Arnold LM, Choy EH, et al. Fibromyalgia syndrome module at OMERACT 9: domain construct. J Rheumatol 2009;36:2318–29. 10.3899/jrheum.090367 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 48.Rampakakis E, Ste-Marie PA, Sampalis JS, et al. Real-life assessment of the validity of patient global impression of change in fibromyalgia. RMD Open 2015;1:e000146. 10.1136/rmdopen-2015-000146 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 49.Ware JE, Sherbourne CD. The MOS 36-item short-form health survey (SF-36). Med Care 1992;30:473–83. 10.1097/00005650-199206000-00002 [DOI] [PubMed] [Google Scholar]
• 50.Burckhardt CS, Clark SR, Bennett RM. The fibromyalgia impact questionnaire: development and validation. J Rheumatol 1991;18:728–33. [PubMed] [Google Scholar]
• 51.Ben-Ami Shor D, Weitzman D, Dahan S, et al. Adherence and persistence with drug therapy among fibromyalgia patients: data from a large health maintenance organization. J Rheumatol 2017;44:1499–506. 10.3899/jrheum.170098 [DOI] [PubMed] [Google Scholar]
• 52.Nantz E, Liu-Seifert H, Skljarevski V. Predictors of premature discontinuation of treatment in multiple disease states. Patient Prefer Adherence 2009;3:31–43. 10.2147/ppa.s4633 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 53.Little RJ, Cohen ML, Dickersin K, et al. The design and conduct of clinical trials to limit missing data. Stat Med 2012;31:3433–43. 10.1002/sim.5519 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 54.Molnar FJ, Man-Son-Hing M, Hutton B, et al. Have last-observation-carried-forward analyses caused us to favour more toxic dementia therapies over less toxic alternatives? A systematic review. Open Med 2009;3:e31–50. [PMC free article] [PubMed] [Google Scholar]
• 55.Mavridis D, White IR. Dealing with missing outcome data in meta-analysis. Res Synth Methods 2020;11:2–13. 10.1002/jrsm.1349 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 56.Mathur MB, Shpitser I, VanderWeele TJ. Resurrecting complete-case analysis: a defense. Am J Epidemiol 2026;195:1765–78. 10.1093/aje/kwaf284 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 57.Sterne JAC, Savović J, Page MJ, et al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ 2019;366:l4898. 10.1136/bmj.l4898 [DOI] [PubMed] [Google Scholar]
• 58.Minozzi S, Dwan K, Borrelli F, et al. Reliability of the revised Cochrane risk-of-bias tool for randomised trials (RoB2) improved with the use of implementation instruction. J Clin Epidemiol 2022;141:99–105. 10.1016/j.jclinepi.2021.09.021 [DOI] [PubMed] [Google Scholar]
• 59.Minozzi S, Cinquini M, Gianola S, et al. The revised Cochrane risk of bias tool for randomized trials (RoB 2) showed low interrater reliability and challenges in its application. J Clin Epidemiol 2020;126:37–44. 10.1016/j.jclinepi.2020.06.015 [DOI] [PubMed] [Google Scholar]
• 60.Moustgaard H, Clayton GL, Jones HE, et al. Impact of blinding on estimated treatment effects in randomised clinical trials: meta-epidemiological study. BMJ 2020;368:l6802. 10.1136/bmj.l6802 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 61.Carlisle JB. False individual patient data and zombie randomised controlled trials submitted to Anaesthesia . Anaesthesia 2021;76:472–9. 10.1111/anae.15263 [DOI] [PubMed] [Google Scholar]
• 62.Ioannidis JPA. Hundreds of thousands of zombie randomised trials circulate among us. Anaesthesia 2021;76:444–7. 10.1111/anae.15297 [DOI] [PubMed] [Google Scholar]
• 63.Wilkinson J, Heal C, Antoniou GA, et al. Protocol for the development of a tool (INSPECT-SR) to identify problematic randomised controlled trials in systematic reviews of health interventions. BMJ Open 2024;14:e084164. 10.1136/bmjopen-2024-084164 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 64.Langan D, Higgins JPT, Jackson D, et al. A comparison of heterogeneity variance estimators in simulated random‐effects meta‐analyses. Res Synth Methods 2019;10:83–98. 10.1002/jrsm.1316 [DOI] [PubMed] [Google Scholar]
• 65.Sadeghirad B, Foroutan F, Zoratti MJ, et al. Theory and practice of Bayesian and frequentist frameworks for network meta-analysis. BMJ Evid Based Med 2023;28:204–9. 10.1136/bmjebm-2022-111928 [DOI] [PubMed] [Google Scholar]
• 66.Seide SE, Jensen K, Kieser M. A comparison of Bayesian and frequentist methods in random-effects network meta-analysis of binary data. Res Synth Methods 2020;11:363–78. 10.1002/jrsm.1397 [DOI] [PubMed] [Google Scholar]
• 67.Kiefer C, Sturtz S, Bender R. A simulation study to compare different estimation approaches for network meta-analysis and corresponding methods to evaluate the consistency assumption. BMC Med Res Methodol 2020;20:36. 10.1186/s12874-020-0917-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 68.Hong H, Carlin BP, Shamliyan TA, et al. Comparing Bayesian and frequentist approaches for multiple outcome mixed treatment comparisons. Med Decis Making 2013;33:702–14. 10.1177/0272989X13481110 [DOI] [PubMed] [Google Scholar]
• 69.Carlin BP, Hong H, Shamliyan TA, et al. AHRQ methods for effective health care. In: Case study comparing Bayesian and frequentist approaches for multiple treatment comparisons. Rockville (MD): Agency for Healthcare Research and Quality (US), 2013. [PubMed] [Google Scholar]
• 70.Furlan AD, Chaparro LE, Irvin E, et al. A comparison between enriched and nonenriched enrollment randomized withdrawal trials of opioids for chronic noncancer pain. Pain Res Manag 2011;16:337–51. 10.1155/2011/465281 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 71.Temple R. Enrichment of clinical study populations. Clin Pharmacol Ther 2010;88:774–8. 10.1038/clpt.2010.233 [DOI] [PubMed] [Google Scholar]
• 72.Wan X, Wang W, Liu J, et al. Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range. BMC Med Res Methodol 2014;14:135. 10.1186/1471-2288-14-135 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 73.Thorlund K, Walter SD, Johnston BC, et al. Pooling health‐related quality of life outcomes in meta‐analysis—a tutorial and review of methods for enhancing interpretability. Res Synth Methods 2011;2:188–203. 10.1002/jrsm.46 [DOI] [PubMed] [Google Scholar]
• 74.Brignardello-Petersen R, Mustafa RA, Siemieniuk RAC, et al. GRADE approach to rate the certainty from a network meta-analysis: addressing incoherence. J Clin Epidemiol 2019;108:77–85. 10.1016/j.jclinepi.2018.11.025 [DOI] [PubMed] [Google Scholar]
• 75.Brignardello-Petersen R, Tomlinson G, Florez I, et al. Grading of recommendations assessment, development, and evaluation concept article 5: addressing intransitivity in a network meta-analysis. J Clin Epidemiol 2023;160:151–9. 10.1016/j.jclinepi.2023.06.010 [DOI] [PubMed] [Google Scholar]
• 76.van Valkenhoef G, Dias S, Ades AE, et al. Automated generation of node‐splitting models for assessment of inconsistency in network meta‐analysis. Res Synth Methods 2016;7:80–93. 10.1002/jrsm.1167 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 77.Jackson D, Barrett JK, Rice S, et al. A design‐by‐treatment interaction model for network meta‐analysis with random inconsistency effects. Stat Med 2014;33:3639–54. 10.1002/sim.6188 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 78.Higgins JP, Thomas J, Chandler J, et al. Cochrane handbook for systematic reviews of interventions.2nd edn. Chichester (UK): John Wiley & Sons, 2019. [Google Scholar]
• 79.Higgins JPT, Thompson SG. Quantifying heterogeneity in a meta‐analysis. Stat Med 2002;21:1539–58. 10.1002/sim.1186 [DOI] [PubMed] [Google Scholar]
• 80.Mittlböck M, Heinzl H. A simulation study comparing properties of heterogeneity measures in meta‐analyses. Stat Med 2006;25:4321–33. 10.1002/sim.2692 [DOI] [PubMed] [Google Scholar]
• 81.Rücker G, Schwarzer G, Carpenter JR, et al. Undue reliance on I2 in assessing heterogeneity may mislead. BMC Med Res Methodol 2008;8:79. 10.1186/1471-2288-8-79 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 82.Mueller KF, Meerpohl JJ, Briel M, et al. Methods for detecting, quantifying, and adjusting for dissemination bias in meta-analysis are described. J Clin Epidemiol 2016;80:25–33. 10.1016/j.jclinepi.2016.04.015 [DOI] [PubMed] [Google Scholar]
• 83.Sterne JAC, Sutton AJ, Ioannidis JPA, et al. Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. BMJ 2011;343:bmj.d4002. 10.1136/bmj.d4002 [DOI] [PubMed] [Google Scholar]
• 84.Balduzzi S, Rücker G, Schwarzer G. How to perform a meta-analysis with R: a practical tutorial. Evid Based Mental Health 2019;22:153–60. 10.1136/ebmental-2019-300117 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 85.Balduzzi S, Rücker G, Nikolakopoulou A, et al. netmeta: an R package for network meta-analysis using frequentist methods. J Stat Softw 2023;106:1–40. 10.18637/jss.v106.i02 [DOI] [Google Scholar]
• 86.Schandelmaier S, Briel M, Varadhan R, et al. Development of the Instrument to assess the Credibility of Effect Modification Analyses (ICEMAN) in randomized controlled trials and meta-analyses. CMAJ 2020;192:E901–6. 10.1503/cmaj.200077 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 87.Brignardello-Petersen R, Bonner A, Alexander PE, et al. Advances in the GRADE approach to rate the certainty in estimates from a network meta-analysis. J Clin Epidemiol 2018;93:36–44. 10.1016/j.jclinepi.2017.10.005 [DOI] [PubMed] [Google Scholar]
• 88.Puhan MA, Schünemann HJ, Murad MH, et al. A GRADE Working Group approach for rating the quality of treatment effect estimates from network meta-analysis. BMJ 2014;349:g5630. 10.1136/bmj.g5630 [DOI] [PubMed] [Google Scholar]
• 89.Zeng L, Brignardello-Petersen R, Hultcrantz M, et al. GRADE Guidance 34: update on rating imprecision using a minimally contextualized approach. J Clin Epidemiol 2022;150:216–24. 10.1016/j.jclinepi.2022.07.014 [DOI] [PubMed] [Google Scholar]
• 90.Devji T, Carrasco-Labra A, Qasim A, et al. Evaluating the credibility of anchor based estimates of minimal important differences for patient reported outcomes: instrument development and reliability study. BMJ 2020;369:m1714. 10.1136/bmj.m1714 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 91.Carrasco-Labra A, Devji T, Qasim A, et al. Minimal important difference estimates for patient-reported outcomes: a systematic survey. J Clin Epidemiol 2021;133:61–71. 10.1016/j.jclinepi.2020.11.024 [DOI] [PubMed] [Google Scholar]
• 92.Bennett RM, Bushmakin AG, Cappelleri JC, et al. Minimal clinically important difference in the Fibromyalgia Impact Questionnaire. J Rheumatol 2009;36:1304–11. 10.3899/jrheum.081090 [DOI] [PubMed] [Google Scholar]
• 93.Badhiwala JH, Witiw CD, Nassiri F, et al. Minimum clinically important difference in SF-36 scores for use in degenerative cervical myelopathy. Spine 2018;43:E1260–6. 10.1097/BRS.0000000000002684 [DOI] [PubMed] [Google Scholar]
• 94.Fu V, Weatherall M, McNaughton H. Estimating the minimal clinically important difference for the Physical Component Summary of the Short Form 36 for patients with stroke. J Int Med Res 2021;49. 10.1177/03000605211067902 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 95.Park KB, Shin J-S, Lee J, et al. Minimum clinically important difference and substantial clinical benefit in pain, functional, and quality of life scales in failed back surgery syndrome patients. Spine (Phila Pa 1976) 2017;42:E474–81. 10.1097/BRS.0000000000001950 [DOI] [PubMed] [Google Scholar]
• 96.Díaz-Arribas MJ, Fernández-Serrano M, Royuela A, et al. Minimal clinically important difference in quality of life for patients with low back pain. Spine (Phila Pa 1976) 2017;42:1908–16. 10.1097/BRS.0000000000002298 [DOI] [PubMed] [Google Scholar]
• 97.Grönkvist R, Vixner L, Äng B, et al. Measurement error, minimal detectable change, and minimal clinically important difference of the Short Form-36 Health Survey, Hospital Anxiety and Depression Scale, and Pain Numeric Rating Scale in patients with chronic pain. J Pain 2024;25:104559. 10.1016/j.jpain.2024.104559 [DOI] [PubMed] [Google Scholar]
• 98.Rejas J, Ribera MV, Ruiz M, et al. Psychometric properties of the MOS (Medical Outcomes Study) Sleep Scale in patients with neuropathic pain. Eur J Pain 2007;11:329–40. 10.1016/j.ejpain.2006.05.002 [DOI] [PubMed] [Google Scholar]
• 99.Hutton B, Salanti G, Caldwell DM, et al. The PRISMA extension statement for reporting of systematic reviews incorporating network meta-analyses of health care interventions: checklist and explanations. Ann Intern Med 2015;162:777–84. 10.7326/M14-2385 [DOI] [PubMed] [Google Scholar]
• 100.Santesso N, Glenton C, Dahm P, et al. GRADE guidelines 26: informative statements to communicate the findings of systematic reviews of interventions. J Clin Epidemiol 2020;119:126–35. 10.1016/j.jclinepi.2019.10.014 [DOI] [PubMed] [Google Scholar]
• 101.Busse JW, Ebrahim S, Connell G, et al. Systematic review and network meta-analysis of interventions for fibromyalgia: a protocol. Syst Rev 2013;2:18. 10.1186/2046-4053-2-18 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 102.Wolfe F, Smythe HA, Yunus MB, et al. The American College of Rheumatology 1990 criteria for the classification of fibromyalgia. Arthritis Rheum 1990;33:160–72. 10.1002/art.1780330203 [DOI] [PubMed] [Google Scholar]
• 103.Lederman S, Sullivan G, inventors; Tonix Pharmaceuticals Holding Corp, assignee. Cyclobenzaprine treatment for fibromyalgia. Canada patent CA3204202A1. 2021.
• 104.Branco JC, Cherin P, Montagne A, et al. Long-term therapeutic response to milnacipran treatment for fibromyalgia: a European 1-year extension study following a 3-month study. J Rheumatol 2011;38:1403–12. 10.3899/jrheum.101025 [DOI] [PubMed] [Google Scholar]
• 105.European phase III, multicentre, double-blind, randomised, monotherapy, 12-month study of milnacipran for the treatment of the fibromyalgia syndrome - FMS European long-term study. 2006. Available: https://trialsearchwhoint/Trial2aspx?TrialID=EUCTR2006-001230-41-FI
• 106.van Dam CJ, Kramers C, Schellekens A, et al. Cannabis combined with oxycodone for pain relief in fibromyalgia pain: a randomized clinical self-titration trial with focus on adverse events. Front Pain Res (Lausanne) 2024;5:1497111. 10.3389/fpain.2024.1497111 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 107.Goldenberg DL, Clauw DJ, Palmer RH, et al. Durability of therapeutic response to milnacipran treatment for fibromyalgia: results of a randomized, double-blind, monotherapy 6-month extension study. Pain Med 2010;11:180–94. 10.1111/j.1526-4637.2009.00755.x [DOI] [PubMed] [Google Scholar]
• 108.Nasser K, Kivitz AJ, Maricic MJ, et al. Twice daily versus once nightly dosing of pregabalin for fibromyalgia: a double-blind randomized clinical trial of efficacy and safety. Arthritis Care Res 2014;66:293–300. 10.1002/acr.22111 [DOI] [PubMed] [Google Scholar]
• 109.Santandrea S, Montrone F, Sarzi-Puttini P, et al. A double-blind crossover study of two cyclobenzaprine regimens in primary fibromyalgia syndrome. J Int Med Res 1993;21:74–80. 10.1177/030006059302100202 [DOI] [PubMed] [Google Scholar]
• 110.Calandre EP, Rico-Villademoros F, Galán J, et al. Quetiapine extended-release (Seroquel-XR) versus amitriptyline monotherapy for treating patients with fibromyalgia: a 16-week, randomized, flexible-dose, open-label trial. Psychopharmacology (Berl) 2014;231:2525–31. 10.1007/s00213-013-3422-0 [DOI] [PubMed] [Google Scholar]
• 111.Celebi E, Ataoglu S, Bahar Ataoglu B, et al. The investigation effects of pregabalin and duloxetine treatment according to personality characteristics groupe with fibromyalgia patients. Duzce Med J 2016;18:47–53. [Google Scholar]
• 112.Bidari A, Moazen-Zadeh E, Ghavidel-Parsa B, et al. Comparing duloxetine and pregabalin for treatment of pain and depression in women with fibromyalgia: an open-label randomized clinical trial. DARU J Pharm Sci 2019;27:149–58. 10.1007/s40199-019-00257-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 113.Murakami M, Osada K, Mizuno H, et al. A randomized, double-blind, placebo-controlled phase III trial of duloxetine in Japanese fibromyalgia patients. Arthritis Res Ther 2015;17:224. 10.1186/s13075-015-0718-y [DOI] [PMC free article] [PubMed] [Google Scholar]
• 114.A Study of Duloxetine in Fibromyalgia, 2012. Available: https://clinicaltrialsgov/show/NCT01552057
• 115.Matthey A, Cedraschi C, Piguet V, et al. OC027—Dual reuptake inhibitor milnacipran and spinal pain pathways in fibromyalgia patients: a randomized double-blind placebo-controlled trial. Clin Ther 2013;35:e11. 10.1016/j.clinthera.2013.07.026 [DOI] [PubMed] [Google Scholar]
• 116.Clauw DJ, Mease P, Palmer RH, et al. Milnacipran for the treatment of fibromyalgia in adults: a 15-week, multicenter, randomized, double-blind, placebo-controlled, multiple-dose clinical trial. Clin Ther 2008;30:1988–2004. 10.1016/j.clinthera.2008.11.009 [DOI] [PubMed] [Google Scholar]
• 117.Martin S, Chandran A, Zografos L, et al. Evaluation of the impact of fibromyalgia on patients’ sleep and the content validity of two sleep scales. Health Qual Life Outcomes 2009;7:64. 10.1186/1477-7525-7-64 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 118.Buysse DJ, Reynolds CF, Monk TH, et al. The Pittsburgh sleep quality index: a new instrument for psychiatric practice and research. Psychiatry Res 1989;28:193–213. 10.1016/0165-1781(89)90047-4 [DOI] [PubMed] [Google Scholar]
• 119.Juhola J, Arokoski JPA, Ervasti J, et al. Internal consistency and factor structure of Jenkins Sleep Scale: cross-sectional cohort study among 80 000 adults. BMJ Open 2021;11:e043276. 10.1136/bmjopen-2020-043276 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 120.Parrott AC, Hindmarch I. The Leeds Sleep Evaluation Questionnaire in psychopharmacological investigations - a review. Psychopharmacology (Berl) 1980;71:173–9. 10.1007/BF00434408 [DOI] [PubMed] [Google Scholar]
• 121.Buysse DJ, Yu L, Moul DE, et al. Development and validation of patient-reported outcome measures for sleep disturbance and sleep-related impairments. Sleep 2010;33:781–92. 10.1093/sleep/33.6.781 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 122.Hunt SM, McEwen J, McKenna SP. Measuring health status: a new tool for clinicians and epidemiologists. J R Coll Gen Pract 1985;35:185–8. [PMC free article] [PubMed] [Google Scholar]
• 123.Safety and tolerability study comparing sodium oxybate given as an oral solution to a single-blinded combination of oral tablets plus oral solution in subjects with fibromyalgia, 2008. Available: https://clinicaltrialsgov/show/NCT00803023
• 124.Djulbegovic B, Koletsi D, Hozo I, et al. High certainty evidence is stable and trustworthy, whereas evidence of moderate or lower certainty may be equally prone to being unstable. J Clin Epidemiol 2024;171:111392. 10.1016/j.jclinepi.2024.111392 [DOI] [PubMed] [Google Scholar]
• 125.Schwartz CE, Powell VE, Rapkin BD. When global rating of change contradicts observed change: Examining appraisal processes underlying paradoxical responses over time. Qual Life Res 2017;26:847–57. 10.1007/s11136-016-1414-3 [DOI] [PubMed] [Google Scholar]
• 126.Correction. RMD Open 2015;1:e000146corr1. 10.1136/rmdopen-2015-000146corr1 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 127. Xyrem (sodium oxybate) [prescribing information]. Palo Alto, CA: Jazz Pharmaceuticals Inc, 2025. [Google Scholar]
• 128.Spaeth M, Bennett RM, Benson BA, et al. Sodium oxybate therapy provides multidimensional improvement in fibromyalgia: results of an international phase 3 trial. Ann Rheum Dis 2012;71:935–42. 10.1136/annrheumdis-2011-200418 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 129.A safety and efficacy study of xyrem® (sodium oxybate) to treat fibromyalgia. 2007. Available: https://clinicaltrialsgov/show/NCT00423813
• 130.Moldofsky H, Inhaber NH, Guinta DR, et al. Effects of sodium oxybate on sleep physiology and sleep/wake-related symptoms in patients with fibromyalgia syndrome: a double-blind, randomized, placebo-controlled study. J Rheumatol 2010;37:2156–66. 10.3899/jrheum.091041 [DOI] [PubMed] [Google Scholar]
• 131.Russell IJ, Perkins AT, Michalek JE, et al. Sodium oxybate relieves pain and improves function in fibromyalgia syndrome: a randomized, double‐blind, placebo‐controlled, multicenter clinical trial. Arthritis Rheum 2009;60:299–309. 10.1002/art.24142 [DOI] [PubMed] [Google Scholar]
• 132.Trial Comparing the Effects of Xyrem (Sodium Oxybate) With Placebo for the Treatment of Fibromyalgia, 2004. Available: https://clinicaltrialsgov/show/NCT00087555
• 133.Evoy KE, Morrison MD, Saklad SR. Abuse and misuse of pregabalin and gabapentin. Drugs 2017;77:403–26. 10.1007/s40265-017-0700-x [DOI] [PubMed] [Google Scholar]
• 134.U.S. Food and Drug Administration . Neurontin, Gralise, Horizant (gabapentin) and Lyrica, Lyrica CR (pregabalin): drug safety communication—serious breathing problems. 2019. Available: https://www.fda.gov/drugs/drug-safety-and-availability/neurontin-gralise-horizant-gabapentin-and-lyrica-lyrica-cr-pregabalin-drug
• 135.Gomes T, Juurlink DN, Antoniou T, et al. Gabapentin, opioids, and the risk of opioid-related death: a population-based nested case-control study. PLoS Med 2017;14:e1002396. 10.1371/journal.pmed.1002396 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 136.Patorno E, Bohn RL, Wahl PM, et al. Anticonvulsant medications and the risk of suicide, attempted suicide, or violent death. JAMA 2010;303:1401–9. 10.1001/jama.2010.410 [DOI] [PubMed] [Google Scholar]
• 137.U.S. Food and Drug Administration . Statistical review and evaluation: antiepileptic drugs and suicidality. 2008. [Google Scholar]
• 138.Wolfe F, Clauw DJ, Fitzcharles M, et al. The American College of Rheumatology preliminary diagnostic criteria for fibromyalgia and measurement of symptom severity. Arthritis Care Res 2010;62:600–10. 10.1002/acr.20140 [DOI] [PubMed] [Google Scholar]
• 139.Wolfe F, Clauw DJ, Fitzcharles M-A, et al. 2016 Revisions to the 2010/2011 fibromyalgia diagnostic criteria. Semin Arthritis Rheum 2016;46:319–29. 10.1016/j.semarthrit.2016.08.012 [DOI] [PubMed] [Google Scholar]
• 140.Wolfe F, Clauw DJ, Fitzcharles M-A, et al. Fibromyalgia criteria and severity scales for clinical and epidemiological studies: a modification of the ACR preliminary diagnostic criteria for fibromyalgia. J Rheumatol 2011;38:1113–22. 10.3899/jrheum.100594 [DOI] [PubMed] [Google Scholar]
• 141.Kumbhare D, Ahmed S, Sander T, et al. A survey of physicians’ knowledge and adherence to the diagnostic criteria for fibromyalgia. Pain Med 2018;19:1254–64. 10.1093/pm/pnx271 [DOI] [PubMed] [Google Scholar]
• 142.Wigers SH, Veierød MB, Mengshoel AM, et al. Healthcare experiences of fibromyalgia patients and their associations with satisfaction and pain relief: a patient survey. Scand J Pain 2024;24. 10.1515/sjpain-2023-0141 [DOI] [PubMed] [Google Scholar]
• 143.Collazo E, Ramirez Urbano I, Gonzalez-Ripoll C. Clinical trial on the efficacy of various therapeutic procedures for the relief of pain in patients with fibromyalgia. Rev Int Acupuntura 2015;9:87–94. 10.1016/j.acu.2015.07.002 [DOI] [Google Scholar]
• 144.Perrot S. Fibromyalgia: do I tackle you with pharmacological treatments? PR9 2025;10:e1222. 10.1097/PR9.0000000000001222 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 145.A Study to Evaluate eFFIcacy and Safety of Sublingual TNX-102 SL Tablet Taken at Bedtime in Patients With fibRoMyalgia (AFFIRM), 2025. Available: https://wwwclinicaltrialsgov/study/NCT02436096
• 146.Metformin for fibromyalgia symptoms (inform trial). 2025. Available: https://clinicaltrialsgov/study/NCT05900466
• 147.Lidocaine Versus Magnesium Sulphate in Management of Myofascial Pain Dysfunction Syndrome, 2024. Available: https://clinicaltrialsgov/study/NCT06595017
• 148.A Clinical Study of ONO-1110 in Patients With Fibromyalgia, 2025. Available: https://clinicaltrialsgov/study/NCT06752590
• 149.Murad MH, Mustafa R, Morgan R, et al. Rating the quality of evidence is by necessity a matter of judgment. J Clin Epidemiol 2016;74:237–8. 10.1016/j.jclinepi.2015.11.018 [DOI] [PubMed] [Google Scholar]
• 150.Busse JW, Vankrunkelsven P, Zeng L, et al. Medical cannabis or cannabinoids for chronic pain: a clinical practice guideline. BMJ 2021;374:n2040. 10.1136/bmj.n2040 [DOI] [PubMed] [Google Scholar]
• 151.Agarwal A, Oparin Y, Glick L, et al. Attitudes toward and management of fibromyalgia: a national survey of Canadian rheumatologists and critical appraisal of guidelines. J Clin Rheumatol 2018;24:243–9. 10.1097/RHU.0000000000000679 [DOI] [PubMed] [Google Scholar]
• 152.Tonix Pharmaceuticals Holding Corp . Tonix Pharmaceuticals announces phase 3 RALLY study results for TNX-102 SL 5.6 mg for the management of fibromyalgia [Internet]. Chatham, NJ: Tonix Pharmaceuticals, 2022. Available: https://ir.tonixpharma.com/news-events/press-releases/detail/1304/tonix-pharmaceuticals-announces-phase-3-rally-study-results [Google Scholar]
• 153.Sullivan G, Gendreau R, Gendreau J, et al. A phase 3 randomized, double-blind, placebo-controlled trial of bedtime sublingual cyclobenzaprine (TNX-102 SL) for the treatment of fibromyalgia (FM): evidence for a broad spectrum of activity on the FM syndrome [abstract]. Arthritis Rheumatol 2019;71. [Google Scholar]
• 154.Singla A, Anstine CV, Huang L, et al. A cross-sectional survey study of cannabis use for fibromyalgia symptom management. Mayo Clin Proc 2024;99:542–50. 10.1016/j.mayocp.2023.12.018 [DOI] [PubMed] [Google Scholar]
• 155.Yang J, Shin K-M, Do A, et al. The safety and efficacy of low-dose naltrexone in patients with fibromyalgia: a systematic review. J Pain Res 2023;16:1017–23. 10.2147/JPR.S395457 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 156.Colombo B, Zanella E, Galazzi A, et al. The experience of stigma in people affected by fibromyalgia: a metasynthesis. J Adv Nurs 2025;81:6317–32. 10.1111/jan.16773 [DOI] [PMC free article] [PubMed] [Google Scholar]
• 157.Colloca L, Barsky AJ. Placebo and nocebo effects. N Engl J Med 2020;382:554–61. 10.1056/NEJMra1907805 [DOI] [PubMed] [Google Scholar]

Supplementary Materials

Data Availability Statement

Data are available in a public, open access repository. All data are available on Open Science Framework (https://osf.io/ezfrc/).