Sixteen thousand two hundred fifty-one records were identified from the systematic literature search from the inception date of the databases until September 2017. Duplicates were removed, and the title and abstracts of the remaining 9923 articles were scrutinised. A further 9722 articles were excluded. The full text of 201 remaining articles was retrieved for detailed screening and data extraction. After full-text screening, 65 articles did not meet the inclusion criteria and were excluded, leaving a total of 136 articles that met the inclusion criteria. In the update searches undertaken in June 2020, a further 31 articles met the eligibility criteria. Data from 167 RCTs (n = 11,012 participants) investigating 22 nonpharmacological interventions (Table 1) were included in this MA (Figure 1). Of these, there were 15 A + B vs B study designs.[16,41,55,71,96,98,101,104,138,147,154,157,164,192,205]
Screening flow. *All studies included in the systematic review were included in the meta-analysis. RCT, randomised controlled trial.
Of the 11,012 participants included in these trials, 88% were women, and the median age, BMI, and FIQ score were 49 years, 27.6 kg/m2, and 59.2, respectively. Ten percent of the articles reported data on race group or ethnicity. The risk-of-bias assessment is shown in Figure 2. Owing to the exclusion of nonrandomised studies, there were no studies with high RoB on random sequence generation, with 64% at low risk, whereas 36% had unclear RoB. It was unclear if the allocation was adequately concealed in 74% of the studies. There was high RoB on blinding of participant and personnel and also for the outcome assessment in 80% and 70% of studies, respectively.
Funnel plots for publication bias on FIQ, pain, fatigue, sleep, and depression outcomes are presented in Supplement Table 5 (available at https://links.lww.com/PAIN/B513). Overall, asymmetry was apparent on visual inspection suggesting publication bias for each of the outcomes. This was confirmed by the Egger test (P < 0.05 for each outcome).
Of the 35 trials[6,12,14,15,18,26,35,47,63,67,71–73,77,89,91,94,98,110,116,117,126,136,137,139,148,157,161–163,166,167,183–185] (n = 2013 participants) evaluating any type of exercise, 28 (n = 1487) reported data on FIQ. The ES of any exercise intervention on FIQ was moderate (Figure 3) in comparison with usual care; however, there was considerable heterogeneity. As a group, exercise interventions resulted in large improvement in pain, fatigue, and sleep and moderate improvement in depression (Table 2). Substantial heterogeneity (50%-90%) was present for all outcomes. There was evidence of publication bias for FIQ, pain, and fatigue outcomes (Supplement Table 6a, available at https://links.lww.com/PAIN/B513).
Effect size (95% CI)—intervention vs control arm (usual care or placebo or sham): FIQ. CBT, cognitive behavioural therapy; CI, confidence interval; FIQ, Fibromyalgia Impact Questionnaire; MDT, multidisciplinary treatment; tDCS, transcranial direct current stimulation; WBV, whole-body vibration. *Data come from placebo or sham-controlled trials. Negative value favours treatment.
Exercise interventions could be classified into 5 types. Of these, mixed exercise[12,47,67,72,117,137,148,162,163,184,185] was the most often studied (11 trials, n = 487 participants). Mind–body,[6,15,26,35,91,98,110,116,166,183] aerobic,[18,63,73,89,126,136,139,161,167] strengthening,[14,77,94,157] and flexibility exercises[14,71] were examined in 10 (n = 632 participants), 9 (n = 568 participants), 4 (n = 221 participants), and 2 studies (n = 98), respectively (Table 1).
Aerobic, mixed, and strengthening exercises showed improvement in FIQ with considerable heterogeneity (Figure 3). However, flexibility (−0.24; 95% CI −0.98 to 0.51) and mind–body exercises (−0.23; 95% CI −0.87 to 0.41) did not improve FIQ (Figure 3). All exercise types except for flexibility exercise were effective at relieving pain and depression. Only, mind–body and strengthening exercises were effective at improving fatigue, whereas aerobic, flexibility, and strengthening exercises were effective at improving sleep (Table 2).
There was considerable heterogeneity in studies assessing aerobic exercise and substantial or unimportant heterogeneity in other exercise types (Supplement Table 7a, available at https://links.lww.com/PAIN/B513). Because there were fewer than 10 studies, statistical testing for publication bias was not assessed for other exercise types.
Nine studies[5,31,64,65,78,99,142,176,190] (n = 1004) assessed education vs usual care (Table 1). Three trials (n = 251) that reported data on fatigue showed an improvement (−0.31; 95% CI −0.55 to −0.06). However, education alone was no better than usual care for other outcomes with moderate heterogeneity (30%-60%) (Supplement Table 8, available at https://links.lww.com/PAIN/B513).
Twenty-nine trials with 2447 participants evaluated psychological treatments[8,10,28,29,34,37,52,58,68,84,90,105,112,122,123,133,134,144,149,151,152,168,170,174,178,189,193,198,199] including CBT[52,58,90,105,122,123,189,199] and mindfulness[8,10,37,134,149,151,168,170] (Table 1). Psychological treatments were more efficacious than usual care for FIQ, pain, sleep, and depression (Figure 3, Table 2). However, they showed no improvement in fatigue.
There was substantial heterogeneity for FIQ, fatigue, and sleep outcomes and moderate heterogeneity for pain and sleep outcomes. The Egger test did not detect publication bias for any outcome (Supplement Table 9, available at https://links.lww.com/PAIN/B513).
Eight studies (n = 566) assessed CBT, and 8 studies (n = 630) assessed mindfulness (Table 1). Both improved FIQ. Although CBT significantly improved pain, mindfulness was superior relative to usual care for fatigue and depression but not for pain. There was no improvement in depression and sleep outcomes for these interventions (Table 2).
Ten trials with 1023 participants[7,9,40,43,95,106,108,120,153,165] showed significant improvements in FIQ, pain, sleep, and depression with MDT compared with usual care (Table 2; Supplementary Table 10, available at https://links.lww.com/PAIN/B513). There was no improvement in fatigue with MDT. There was evidence of publication bias for pain outcome (P = 0.043). Owing to <10 studies, publication bias could not be assessed for other outcomes.
This intervention type covered all aquatic interventions that involve adopting a static position in water with different minerals or in sea water. It was examined in 13 trials (n = 685).[13,17,33,51,55,56,59,60,96,101,145,195,205] These showed efficacy for FIQ, pain, and depression relative to usual care with considerable heterogeneity (Supplement Table 11a, available at https://links.lww.com/PAIN/B513). Evidence from 3 studies (n = 147) that reported fatigue outcomes demonstrated no effect (−0.23; 95% CI −0.56 to 0.09). However, it showed significant improvements on fatigue when it was applied as an adjunctive treatment to exercise in A + B vs B design (−0.94; 95% CI −1.47 to −0.41)[55,96,101,205] (Supplement Table 11b, available at https://links.lww.com/PAIN/B513).
Publication bias could not be assessed because there were less than 10 trials for each outcome. I2 test identified considerable heterogeneity for FIQ, pain, and depression (Supplement Table 11a, available at https://links.lww.com/PAIN/B513).
Eight acupuncture trials[36,50,80,92,118,177,180,188] (n = 518) were analysed. Six of these (n = 340) had sham acupuncture as a comparator, whereas the remaining 2 (n = 178) had usual care comparison arms.
Acupuncture was more effective than sham acupuncture (−0.88; 95% CI −1.75 to −0.02) and usual care (−0.59; 95% CI −0.99 to −0.19) for FIQ. It showed greater pain reduction compared with sham acupuncture (−0.98; 95% CI −1.56 to −0.40) and usual care (−0.64; 95% CI −0.96 to −0.31) as well as greater improvement in fatigue compared with sham acupuncture (−0.50; 95% CI −0.90 to −0.10) and usual care (−0.41; 95% CI −0.80 to −0.01). There was no improvement in sleep. Although acupuncture was more effective than sham for depression (−0.57; 95% CI −1.03 to −0.10), there was no effect relative to usual care (−0.23; 95% CI −0.63 to 0.16). Considerable heterogeneity of placebo-controlled trials was observed in FIQ, pain, and sleep (Supplement Table 12, available at https://links.lww.com/PAIN/B513).
The efficacy of other nonpharmacological interventions including massage,[39,41] manual therapy,[1,25,38,135,147,187] electrotherapy,[48,49,75,88,103,138,146,158,181,192] laser,[75,146,158,192] transcranial direct current stimulation,[57,62,74,79,97,102,104,127,130,150,172,182,203] biofeedback,[16,191] nutritional supplement,[4,19,53,119,129,131,156,159,197] homeopathy,[21,154] magnetotherapy,[2,46,115,179] music,[3,42,128,143,186,196] weight loss, cupping therapy, material of cloth,[11,100] hyperbaric oxygen therapy, topical oil, cryotherapy, and whole-body vibration[141,164] are presented in Supplement Tables 13 a to q (available at https://links.lww.com/PAIN/B513). Although electrotherapy showed significant improvements on FIQ, fatigue, and sleep, manual therapy was better than usual care for pain. Some of the interventions showed larger effect sizes such as hyperbaric oxygen therapy (one trial with 50 participants) and material of cloth (2 trials with 89 participants) that refers to wearing a wool shirt or t-shirt with bioceramic solution. There were 1 or 2 A + B vs B study designs for some of the other interventions included in the review, and firm inferences on their effect size could not be drawn[16,41,71,157]
Subgroup analyses were undertaken for (1) all nonpharmacological interventions and (2) exercise interventions alone. There was no effect of age, BMI, setting, and funding source on improvement in FIQ when all nonpharmacologic interventions were considered (Supplement Table 14, available at https://links.lww.com/PAIN/B513). Studies with <50 participants (−0.79; 95% CI −0.99 to −0.60) showed greater improvement in FIQ compared with studies with > 50 participants (−0.51; 95% CI −0.66 to −0.35) with similar levels of heterogeneity (87.4% vs 71.1%) when all nonpharmacological interventions were analysed. Similar results were observed within exercise interventions alone, where studies with <50 participants had higher ES than those with >50 participants (−0.99 [−1.41 to −0.57] vs −0.45 [−0.85 to −0.05]) with similar levels of heterogeneity (89.7% vs 72.9%) (Supplement Table 14b, available at https://links.lww.com/PAIN/B513).
A secondary analysis for time-dependent effects was undertaken on FIQ for commonly studied interventions. Outcome assessment time points were grouped into 3 intervals of similar duration (1–8 weeks, 9–15 weeks and ≥16 weeks).
All interventions except for electrotherapy showed a peak effect size in the first 8 weeks. After this, the treatment responses diminished for all interventions except electrotherapy that showed a greater magnitude of effect size (Supplement Table 15, available at https://links.lww.com/PAIN/B513).
We conducted a further subgroup analysis to examine the influence of the time gap between the end of treatment and outcome assessment. All included studies were subgrouped based on the number of weeks between these 2 time points: 0 to 0.5 weeks, 2 to 6 weeks, 10 to 14 weeks, and 18 to 44 weeks. The ESs remained stable to up to 14 weeks after theend of intervention and diminished if the gap was greater than 18 weeks (Supplement Table 16, available at https://links.lww.com/PAIN/B513). Substantial heterogeneity was still present in each of these subgroups.
Sensitivity analysis was conducted for all nonpharmacological interventions, as well as within-exercise interventions for FIQ. Although the pooled ES for all nonpharmacological interventions was −0.63 (−0.75 to −0.50), it was −0.75 (−0.95 to −0.55) for studies that performed an intention-to-treat analysis and −0.62 (−0.86 to −0.39) for studies that properly concealed the allocation (Supplement Table 17, available at https://links.lww.com/PAIN/B513). Sensitivity analysis on exercise trials showed similar findings (Supplement Table 17b, available at https://links.lww.com/PAIN/B513).
Using mean baseline FIQ score of 62 and the estimated minimally clinically important difference (14%; 95% CI 13–15), from Bennett et al., the minimally clinically important difference on a 0 to 100 scale is estimated as 8.4. The ES of each intervention on FIQ was multiplied by the SD from the Bennett et al. study to back translate to 0 to 100 score. All nonpharmacological interventions showed a clinically meaningful improvement except for psychological treatment, CBT, MDT, and weight loss that provided less than clinically relevant change (Supplement Table 18, available at https://links.lww.com/PAIN/B513).
Pain. 2022;163(8):1432-1445. © 2022 Lippincott Williams & Wilkins