Effectiveness of Exercise on Fatigue and Sleep Quality in Fibromyalgia: A Systematic Review and Meta-analysis of Randomized Trials

Objectives: To determine the effects of exercise on fatigue and sleep quality in ﬁbromyalgia (primary aim) and to identify which type of exercise is the most effective in achieving these outcomes (secondary aim). Data Sources: PubMed and Web of Science were searched from inception until October 18, 2018. Study Selection: Eligible studies contained information on population (ﬁbromyalgia), intervention (exercise), and outcomes (fatigue or sleep). Randomized controlled trials (RCT) testing the effectiveness of exercise compared with usual care and randomized trials (RT) comparing the effectiveness of 2 different exercise interventions were included for the primary and secondary

More than 80% of people with fibromyalgia experience severe fatigue 1 or poor sleep quality, 2 both of which are identified by people with fibromyalgia and health care providers as priority targets for treatment. Increased fatigue and poor sleep quality are therefore acknowledged as core symptoms in the updated fibromyalgia diagnostic criteria. 3 Despite the importance of fatigue and sleep quality, most of the research to date has traditionally focused on pain-related outcomes. For instance, the European League Against Rheumatism highlights that exercise is the only therapy supported by "strong" evidence for the management of fibromyalgia. 4 However, the recommendations were based on previous reviews that provided evidence of the benefits of exercise for pain but were unclear for other symptoms [5][6][7][8][9] In their earliest works, Busch et al performed comprehensive reviews including all types of exercise (eg, aerobic, resistance, and flexibility training). 5,9 These reviews concluded that the effects of exercise on fatigue or sleep were unknown owing to the paucity of research at that time. A number of subsequent systematic reviews focused on specific types of exercise have been published (ie, flexibility, 10 aerobic, 6 resistance, 8 and vibration 7 training), which have explored the effects of exercise on fatigue and sleep quality, among other outcomes. Although the contribution of these reviews to the evidence base is acknowledged, the decision to narrow the scope of each review resulted in the inclusion of a restricted number of studies. For instance, for fatigue, only 4 and 2 studies were included in the reviews by Bidonde et al 6 and by Busch et al, 8 respectively. Consequently, it is difficult to make robust conclusions about the effects of exercise interventions on fatigue and sleep in fibromyalgia. Compared with previous reviews, a recent systematic review has focused on mixed exercise training (ie, 2 or more types of exercise combined). 11 This review included a larger number of studies (ie, 11 studies conducted with a total sample of 493 adults with fibromyalgia) and concluded that the effect of mixed exercise resulted in improvements in fatigue, but omitted study of sleep quality. 11 To date, no review has summarized all relevant literature on the effectiveness of exercise interventions (of any type) on fatigue and sleep quality in fibromyalgia. In doing so, the current review will include a large sample size and accurately estimate, for the first time, the effects of physical exercise on these 2 outcomes.
The aims of this systematic review were to determine the effectiveness of exercise for reducing fatigue and improving sleep quality in people with fibromyalgia (primary aim) and to identify which type of exercise interventions might be the most effective in achieving these outcomes (secondary aim).

Methods
A multidisciplinary international task force was established to conduct this review. The PRISMA guidelines were used to guide this systematic review and meta-analysis. 12 The protocol of the present review was specified in advance and registered in the PROSPERO database (registration number: CRD42018118005). . For Web of Science, the search terms were TIZ(fibromyalgia) AND TIZ("exercise" OR "training" OR "yoga" OR "tai chi" OR "tai ji" OR qigong OR hydrotherapy OR "physical activity" OR "body awareness" OR danc)).

Study selection
Two independent researchers (F.E.-L. and C.M.-C.) performed the search, screened the titles and abstracts of all retrieved articles, and examined the final eligibility of the full-text articles. When a study did not report data on fatigue or sleep quality but used questionnaires including these outcomes, the authors were contacted for further information. No restrictions were applied for language. This review followed the PICOS framework.

Population
Adults with fibromyalgia who were diagnosed using 1 of the recognized American College of Rheumatology criteria (1990, 2010, 2011, or 2016) were included in this study.

Intervention (exposure)
Intervention was based on exercise. Mixed interventions that consisted of exercise combined with other interventions (ie, cointerventions) were considered, as long as exercise comprised at least 50% of the intervention.

Comparison
Studies should have either an intervention group with exercise and a non-intervention control group (eg, treatment as usual) or 2 exercise groups. Therefore, the primary subset of studies included randomized controlled trials (RCT) and the second subset of studies included randomized trials (RT).

Outcome
Outcomes measured were fatigue and sleep quality. When a study included more than 1 assessment per outcome, all the figures were extracted but only the most common assessment among the included studies was meta-analyzed.

Study design
RCTs and RTs were included for the primary and secondary aims, respectively.

Data extraction and quality assessment
Two independent researchers (C.M.-C. and D.R.) extracted the key information from each eligible study. When the information to be extracted was unavailable, authors were contacted. Disagreements were solved in a consensus meeting between the independent reviewers with a third reviewer (C.M.H.).
The Grading of Recommendations Assessment, Development and Evaluation framework was used to assess the quality of the evidence across studies for fatigue and sleep quality separately. Risk of bias of individual studies was assessed using the Cochrane Risk of Bias tool. Studies with a score of at least 5 points were

List of abbreviations:
CI confidence interval RCT randomized controlled trial RT randomized trial SMD standardized mean difference considered as having high risk of bias. Inconsistency across studies was considered serious when heterogeneity was high (I 2 50%). Indirectness was considered serious when interventions included both exercise and additional components (ie, cointerventions). Imprecision was considered serious when the 95% confidence interval (CI) was wide and crossed the line of no effect, and as such the interpretation of the data would be different if the true effect were at 1 end of the CI or the other. Finally, publication bias was assessed via funnel plots.
Two researchers independently assessed risk of bias (I.C.A.-G. and M.R.-A.) and the inconsistency, indirectness, imprecision, and publication bias (F.E.-L. and J.G.M.) of each eligible study. Disagreements on these assessments were solved in a consensus meeting between the independent reviewers with a third reviewer (C.M.H.).

Data synthesis and analysis
For the primary aim, quantitative synthesis of RCTs (ie, metaanalyses) were performed using Review Manager version 5.3. a Statistical significance was set at a P value less than .05. Standardized mean differences (SMD) between the exercise and control groups were computed for both outcomes separately. When a control group was used as a comparator twice in the same study, we halved the sample size of the control group. Weighted mean differences were calculated using a random effects model. Heterogeneity was measured using the I 2 statistic (the percentage of total variability attributed to between-study heterogeneity). When heterogeneity was high (I 2 50%), further explorations based on subgroups analyses were computed. SMD effect sizes were calculated using Hedges' adjusted g (similar to Cohen's d). Effect sizes of 0.2, 0.4, and 0.8 were considered small, moderate, and large, respectively.
For the secondary aim, a narrative synthesis structured around each outcome was conducted. When at least 3 of the included studies presented similar comparisons, we performed metaanalyses using the same methods that have been described for the primary aim.

Study selection and characteristics
Thirty-seven unique studies were included in this review.  Of them, 4 studies included 3 arms (ie, control group and 2 exercise intervention groups, each with a different exercise training such as aerobic in 1 group and flexibility in another group) and therefore they were included for addressing both aims of the present review. [40][41][42][43] Thus, a total 20 RCTs 13-20,25,36,40-49 and 21 RTs 21-24,26-35,37-43 were included in the review. Figure 1 shows a PRISMA diagram.
A moderate risk of bias was present in most of the included RCTs and RTs (see supplemental figs S1 and S2 for overall summaries and supplemental figs S3 and S4 [available online only at http://www.archives-pmr.org/] for specific information on each individual included work per study design). No study reported having conflicts of interest. Figure 2 presents a meta-analysis conducted in 1003 people with fibromyalgia (61% randomly allocated into exercise interventions). In comparison with usual care, exercise interventions were effective for reducing fatigue in fibromyalgia (pooled SMD, e0.47; 95% CI, e0.67 to e0.27). This finding was robust across 2 sensitivity analyses, as shown in supplemental figures S5 and S6 (available online only at http://www.archives-pmr.org/): (1) when a study with high risk of bias 40 was not included in the metaanalysis (pooled SMD, e0.49; 95% CI, e0.71 to e0.27) and (2) when fixed effects models were computed (pooled SMD, e0.40; 95% CI, e0.53 to e0.26). Supplemental figure S7 (available online only at http://www.archives-pmr.org/) demonstrates the funnel plot, which did not indicate publication bias.

Synthesis of the data
Owing to the high heterogeneity (ie, I 2 Z51%) observed across RCTs testing the effects of exercise on fatigue, we explored several post hoc analyses. Most of them were not significant, as the effects on fatigue were similar between levels of adherence: (1) studies in which participants had to attend to at least 80% of the training sessions (ie, adherence) to be included in the analyses and those studies with a lower or no adherence criterion, (2) sex of participants: studies in which only women participated versus those in which both sexes were included, (3) type of intervention: only exercise versus cointerventions, (4) type of exercise: meditative exercise programs (ie, tai chi, yoga, qigong) versus others (ie, aerobic, muscular resistance, flexibility), (5) sample size: those with at least 20 participants in each group versus others, (6) type of setting in which exercise was performed: land-based versus water-based, and (7) training intensity: low-to-moderate versus moderate-to-high. Supplemental figures S8 to S14 (available online only at http://www.archives-pmr.org/) show these nonsignificant findings. Interestingly, those studies in which fatigue was the primary outcome ( fig 3) and used a shorter (<24wk) nonaerobic exercise intervention resulted in greater effect on fatigue (greater effect sizes) than comparative studies (supplemental figs S15 and S16; available online only at http://www.archives-pmr. org/). Figure 4 depicts a meta-analysis conducted in 731 people with fibromyalgia (59% randomly allocated into exercise interventions). In comparison with usual care, exercise interventions had a small effect on enhancing sleep quality in fibromyalgia (pooled SMD, e0.17; 95% CI, e0.32 to e0.01). This finding was robust across 2 sensitivity analyses, as shown in supplemental figures S17 and S18 (available online only at http://www. archives-pmr.org/): (1) when a study with high risk of bias 45 was not included in the meta-analysis (pooled SMD, e0.19; 95% CI, e0.35 to e0.02), and (2) when fixed effects model were computed (pooled SMD, e0.17; 95% CI, e0.32 to e0.02). Owing to the small heterogeneity (ie, I 2 Z5%), post hoc analyses were not needed. Supplemental figure S19 (available online only at http:// www.archives-pmr.org/) depicts the funnel plot, which did not indicate publication bias. Table 1 shows that when comparing exercise versus usual care, there was low to moderate quality evidence for the beneficial effects of exercise on fatigue, whereas the evidence was moderate for benefits on sleep quality.
In the 21 RTs included in the present review, a wide range of exercise interventions were implemented and compared in a total of 1254 individuals with fibromyalgia who were randomly allocated into different interventions. Thus, it was difficult to perform robust comparisons. However, we were able to quantify 1 comparison for sleep quality and 3 for fatigue. First, when comparing different types of exercise, meditative exercise programs were more effective for improving sleep quality but not for lowering fatigue

Discussion
This systematic review aimed to determine the effectiveness of exercise on fatigue and sleep quality in those with fibromyalgia and to identify which type of exercise interventions might be the most effective in achieving these outcomes. In the current review, we found that, compared with usual care, exercise has moderate effects for lowering fatigue and small effects for improving sleep quality. We also observed that, compared with other types of exercise, meditative exercise programs were more effective for improving sleep quality but not for lowering fatigue. In interpreting the findings of this review, several factors must be noted. First, most of the studies were based on aerobic exercise. Thus, the effect sizes of the present metaanalyses may more accurately reflect the effectiveness of aerobic training on fatigue and sleep quality than the effects of other types of exercise. Indeed, we observed that those exercise interventions that did not include aerobic exercise appeared to be more effective at reducing fatigue. Second, the effects of exercise on fatigue were highly variable across studies (I 2 Z51%) and remarkably higher when fatigue was the main outcome. Third, there was a lack of high-quality studies in the field and, consequently, the quality of evidence provided in the present review is low to moderate for the effectiveness of exercise in reducing fatigue (the evidence is in favor of exercise but the effect size is unclear and likely to be moderate) and moderate for small effects of exercise (of any type) on enhancing sleep quality.   11 Using similar statistical methods to previous metaanalyses but in a larger sample size (nZ1003), our pooled estimation showed that exercise produces significant and probably meaningful (moderate effect size) reduction in fatigue in fibromyalgia. Thus, the comprehensive approach followed in the present meta-analysis allowed us to robustly determine, for the first time, the overall effects of exercise on fatigue in large sample of individuals with fibromyalgia.

Effectiveness of exercise for improving sleep quality in fibromyalgia
Previous meta-analyses were unable to determine the effectiveness of exercise on sleep quality in fibromyalgia owing to the paucity of research. Indeed, most failed to find RCTs on this topic. Given the extent of sleep dysfunction in individuals with fibromyalgia, it is important to determine the effectiveness of exercise for improving sleep quality in this population. The most comprehensive review to date included only 2 studies examining sleep and reported moderate effects of exercise for enhancing sleep quality (nZ104). 51 The number of included studies in the present work was considerably higher (13 RCTs, nZ806), allowing us to better estimate the effectiveness of exercise for improving sleep quality in fibromyalgia. The effectiveness of exercise (of any type) in enhancing sleep quality in fibromyalgia was limited (small effect). However, meditative exercise programs (ie, tai chi, yoga, qigong) may offer a promising approach. Although there are potential mechanisms that can provide a rationale to support the effectiveness of meditative exercise on improving sleep quality, our finding is based on an imprecise estimation (SMD, e0.80; 95% CI, e1.57 to e0.02) from a relatively small sample size (141 participants in meditative exercise vs 177 participants in other types of exercise). Thus, further high-quality experimental research is required to confirm or refute our findings.

Exercise mechanisms for fatigue and sleep quality in fibromyalgia
Aberrations in the central nervous system are well-known in fibromyalgia. [52][53][54] For example, compared with nonfibromyalgia controls, abnormal levels of metabolites (eg, reductions in the ratio of N-acetylaspartate to creatine) have been observed in the hippocampus of people with fibromyalgia, 52 as well as structural abnormalities (eg, lower volume) 53 and functional changes (eg, increased activation). 54 Another system that might be altered in fibromyalgia is the hypothalamic-pituitary-adrenal axis, as well as a sympathetic hyperactivity mediated by a dysfunction in the autonomic nervous system. 55 These alterations may be related to increased levels of fatigue. 52 Interestingly, exercise may revert these aberrations by regulating the levels of metabolites and promoting angiogenesis, neurogenesis, and connectivity of the hippocampus. 56,57 The present meta-analyses demonstrated that exercise had a small beneficial effect on sleep quality in fibromyalgia. In this disease, hyperactivity of the sympathetic nervous system is welldocumented and, thus, stress levels are considerably high. 58,59 Physiological responses to exercise often include a decrease in this sympathetic tone and a shift toward parasympathetic activity, which may be related to muscular and nervous relaxation, leading to reductions in stress levels and, finally, to improvement in sleep quality. [60][61][62] In this respect, our review showed that meditative exercise programs were more effective in improving sleep quality than other types of exercise. Although meditative exercise is safe in fibromyalgia, little is known about their mechanisms of action. It is likely that this type of exercise can enhance the parasympathetic activity and reduce sympathetic tone by decreasing activation of the hypothalamic-pituitary-adrenal axis. Moreover, meditative exercise may facilitate enhanced rapid eye movement sleep by increasing central nervous system inhibitory c-aminobutyric acid and serotonin levels. 63

Clinical applications
The recent European League Against Rheumatism recommendations for the management of fibromyalgia highlight exercise as the only therapy with a strong level of evidence. 4 These recommendations were based on the findings provided by systematic review of previous reviews. As we have discussed, although previous systematic reviews showed reliable findings for pain management, they have provided limited evidence on the effectiveness of exercise for reducing fatigue and increasing sleep quality in fibromyalgia. 6,8 The present meta-analyses suggest that the effectiveness of exercise may differ for different outcomes. This means that it cannot be assumed that the benefits of exercise for pain automatically extend to other symptoms of the condition. An interesting finding for health care providers has emerged from our review in that fatigue reductions were higher when the main outcome of the study was fatigue. Therefore, instead of designing a "fix-all" exercise protocol for fibromyalgia, exercise programs should be designed as outcomespecific by considering how fibromyalgia manifests in the person who is going to engage in the program. For example, meditative exercise programs (eg, tai chi or qigong) may be more advisable for people with fibromyalgia who experience difficulty sleeping.
The studies included in the present systematic review investigated a wide range of exercise programs, including different types of exercise, intensities, frequencies, and program durations. Although we explored several post hoc analyses, we were unable to determine the most effective exercise intervention for reducing fatigue. From our approach to subgrouping the effects of different exercise interventions compared with usual care, we observed that the ideal intervention for reducing fatigue in fibromyalgia appears to be specifically designed for such an outcome, lasts less than 24 weeks, and does not involve aerobic exercise. Collectively, the high heterogeneity that emerged from the effects of exercise on fatigue limits the establishment of evidence-based guidelines. Although the American College of Sports Medicine has released specific recommendations to consider when conducting exercise interventions in fibromyalgia, 64 a recent review has reported poor therapeutic validity of studies that accomplish these American College of Sports Medicine exercise recommendations. 65 Standard exercise interventions for the average or most common profile of people with fibromyalgia appears misjudged as people with fibromyalgia are heterogeneous. 66 Thus, personalized exercise programs are warranted. In this context, some people with fibromyalgia may experience fears related to engaging in exercise 67 or a discordance of being more capable to engage in exercise than is self-perceived. 68 Individuals with these characteristics may be more likely to experience exercise as stressful. Therefore, exercise interventions should not only be tailored to how fibromyalgia manifests in each person but also to (more) general characteristics of the person.

Implications for research agenda
Findings of the present study provide evidence indicating that exercise is effective for reducing fatigue in fibromyalgia. However, future research is needed to determine what type of exercise is most beneficial for people with fibromyalgia, which intensity is best, the optimal length of the training, and the most beneficial delivery method. Our findings indicate that exercise appears to promote only small benefits on sleep quality in fibromyalgia. Although relaxation is a potential mechanism by which exercise might improve sleep quality, not all types of exercise promote relaxation. Thus, meditative exercise programs that suppose a lower physical load than other types of exercises could be more effective for enhancing sleep quality in fibromyalgia. Therefore, large, high-quality experimental studies testing the effectiveness In favor of exercise Exercise, fatigue, and sleep in fibromyalgia of very gentle exercise specifically designed for enhancing sleep quality in fibromyalgia are warranted in the future. In addition, further research testing the effectiveness of exercise in objectively measured fatigue or sleep quality is warranted.

Study limitations
The most common limitations among the included studies were as follows: (1) the long-term effects of the interventions were not reported; (2) results were not stratified by sex, and most of the participants were women. Moreover, we did not include conference proceedings and other types of gray literature owing to the often low quality of reporting in conference abstracts.

Conclusions
We provided low-to-moderate quality evidence that exercise is moderately effective for lowering fatigue and demonstrated that there is moderate evidence of small effects of exercise for enhancing sleep quality in fibromyalgia. Although speculative, meditative exercise programs may be a promising approach for improving sleep quality in fibromyalgia. As most of the interventions involved aerobic exercise, research using other types of exercise is warranted. Instead of designing "fix-all" and "one size fits all" protocols, exercise programs should be specifically designed for the outcome that is targeted and tailored to the characteristics of the person who is going to engage in the exercise in order to be as effective as possible.