Effectiveness of Therapeutic Exercise in Fibromyalgia Syndrome: A Systematic Review and Meta-Analysis of Randomized Clinical Trials

Objective The aim of this study was to summarize evidence on the effectiveness of therapeutic exercise in Fibromyalgia Syndrome. Design Studies retrieved from the Cochrane Plus, PEDro, and Pubmed databases were systematically reviewed. Randomized controlled trials and meta-analyses involving adults with fibromyalgia were included. The primary outcomes considered in this systematic review were pain, global well-being, symptoms of depression, and health-related quality of life. Results Effects were summarized using standardized mean differences with 95% confidence intervals using a random effects model. This study provides strong evidence that physical exercise reduces pain (−1.11 [95% CI] −1.52; −0.71; overall effect p < 0.001), global well-being (−0.67 [95% CI] −0.89, −0.45; p < 0.001), and symptoms of depression (−0.40 [95% CI] −0.55, −0.24; p < 0.001) and that it improves both components of health-related quality of life (physical: 0.77 [95% CI] 0.47; 1.08; p < 0.001; mental: 0.49 [95% CI] 0.27; 0.71; p < 0.001). Conclusions This study concludes that aerobic and muscle strengthening exercises are the most effective way of reducing pain and improving global well-being in people with fibromyalgia and that stretching and aerobic exercises increase health-related quality of life. In addition, combined exercise produces the biggest beneficial effect on symptoms of depression.


Introduction
Fibromyalgia Syndrome (FMS) is a rheumatic disease of unknown etiology [1] which is characterized by widespread pain and associated with multiple other symptoms including fatigue, anxiety, and depression [2]. The global mean prevalence of FMS in the general population is 2.7% with a femaleto-male ratio of 3 : 1 [3] and the diagnosis is most often made in the middle age [4].
There is evidence from randomized controlled trials (RCTs) that some treatments, for example, pharmacotherapy, patient education, behavioral therapy, and physiotherapy, are effective in reducing symptoms [5]. Physiotherapy techniques used with this patient group include massage therapy, 2 BioMed Research International

Methods
This review was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement [13] and the recommendations of the Cochrane Collaboration [14,15].

Data Sources and Searches.
A systematic review of publications retrieved from the Cochrane Plus, PEDro, and Pubmed databases was performed. A manual search of the journals FisioterapiaandCuestiones de Fisioterapia was also carried out. The search strategy is detailed in Additional File (see Supplementary Material available online at https://doi.org/10.1155/2017/2356346). Only fully published material in Spanish or English was reviewed. The keywords used in database searches were "fibromyalgia", "physical activity", "exercise", and "exercise therapy". The search strategy was adapted as necessary for each database. This comprehensive search was performed from April 2016 to May 2017.

Study Selection.
The search was conducted by two authors (DS, SN) who screened the titles and abstracts of potentially eligible studies. DS and SN also independently examined the full text of articles which passed the initial screening in order to determine whether they met the selection criteria. Cases where there was a discrepancy between the two reviewers were reevaluated and a consensus decision was achieved by discussion.

Type of Study.
RCTs comparing types of therapeutic exercise or comparing therapeutic exercise with a control group receiving another intervention or standard care were included.

Participants.
Studies with participants older than 18 years, diagnosed with FMS in the absence of significant comorbidity, were included.

Type of Intervention.
Studies using aerobic, strengthening, or stretching exercises or a combination of these were considered. Studies of exercise interventions based on activities such as yoga or tai-chi were excluded.

Comparisons.
All included studies compared the effect of at least one type of exercise with a control treatment, either another form of physical activity or standard care.

Outcomes Measures.
All included studies assessed at least one key domain of FMS symptoms (pain; symptoms of depression; global well-being; health-related quality of life (HRQOL)).

Data Extraction.
Two authors (DS, TG) extracted the data independently using standard extraction forms. Data collected included participants, sample sizes, duration of studies, interventions, outcomes, results, and methods to measure outcomes. Discrepancies were rechecked and consensus was achieved by discussion.
Data extracted after treatment were considered an experimental group and values presented by the patients before treatment as a control group. When two different treatments were compared in the same study they were treated as independent studies for the purposes of the meta-analysis, because the aim of this study was to compare the effects of various therapies.
On the other hand, for each variable two subgroups were differentiated depending on whether the analysis by intention-to-treat or per protocol was performed in the study. When standard deviations (SDs) were not reported in the publication, they were calculated based on what was published from -values, confidence intervals, or standard errors or used the mean of the SDs from other studies using the same outcome scale.

Data
Items. The following items were extracted: author/ year, design of the study, participants, interventions, comparisons, outcomes studied in this meta-analysis, and conclusions.
When researchers reported more than one indicator for an outcome a predefined order of preference for analysis was used. These preferences were predefined according to the specificity of each outcome measure (in descending order): Pain. Visual Analogue Scale (VAS), VAS, from Fibromyalgia Impact Scale (FIQ), and Multidimensional Pain Inventory subscale Global Well-Being. FIQ total score Symptoms of Depression. Beck Depression Inventory (BDI), Hospital Anxiety and Depression Scale (HAD), and VAS from FIQ HRQOL. Total SF-36 questionnaire (SF-36) score.

Risk of Bias within Studies and Methodological Quality.
Two pairs of reviewers (DS, SN and TG, DP) worked independently to assess the methodological quality in accordance with the CONSORT 2010 [16] statement (Consolidated Standards of Reporting Trials), which contains 25 items scored as zero or one. Only studies that scored over 15 on the CONSORT checklist were included. In addition, the Cochrane Collaboration's tool was used to assess the risk of bias. Sequence generation, allocation concealment, blinding, completeness of outcome data, and absence of selective outcome reporting were also assessed. Risk of bias was classified as low, unclear, or high in each domain.

Data Synthesis and Analysis
2.7.1. Summary Measures. The meta-analysis was conducted using the Review Manager Analysis software (RevMan 5.3) from the Cochrane Collaboration. Standardized mean differences (SMDs) were calculated from the means and SMDs for each intervention. The SMD used in RevMan software is the measure of effect size known as Hedge's (adjusted) , which is the difference between the 2 means divided by the pooled SD, with a correction for small sample bias. Hedge's (adjusted) was chosen because most of the studies included in this meta-analysis were small (<40 subjects per group). As it uses quantitative measures and continuous variable, the statistical analysis method used was the inverse variance [15].
The combined results were assessed using a random effects model, which is more conservative than a fixed effects model and incorporates both within-and between-study variance. Cohen's was used to evaluate the magnitude of the effect size, calculated as SMD, using the following criteria: > 0.2 to 0.4 small effect size; > 0.4 to 0.8 medium effect size; > 0.8 large effect size. Overall effects were assessed using the statistic; < 0.05 was the criterion for rejection of the null hypothesis, that is, concluding that a systematic effect had been demonstrated [17]. The results of the metaanalysis were classified using the following modified level of evidence descriptors: strong = consistent results in at least two RCTs of moderate quality; moderate = consistent results in at least two low quality RCTs and/or one moderate quality RCT; limited = results in low quality RCTs; conflicting = inconsistent results in multiple RCTs; without evidence = no RCT evidence available.

Planned Methods of Analysis.
Heterogeneity was assessed using the 2 statistic: 2 < 40% heterogeneity might not be important; 2 = 30-60% may represent moderate heterogeneity; 2 = 50-90% may represent substantial heterogeneity; 2 = 75-100% may represent considerable heterogeneity. The significance of 2 depends on the magnitude and the impact of heterogeneity tests (e.g., Chi-squared test). Cochran's statistic was also calculated. This statistic is associated with the chi-squared statistic of heterogeneity with − 1 degrees of freedom, where is the number of included studies. If is significant, < 0.10, it is likely that at least one of the included studies is different from the others.
In the random effects model tau 2 ( 2 ) is also used to estimate the variance in the distribution of effects across studies. If 2 = 0 the results of random effects meta-analysis would be almost identical to those of a fixed effects analysis, indicating that there is no heterogeneity [15].

Sensitivity Analysis.
In order to examine the influence of individual studies on the overall results, pooled analyses were conducted with each study individually deleted from the model. This enabled us to investigate causes of heterogeneity [15].

Subgroup Analysis.
The effects of the various types of exercise (aerobic, strengthening, stretching, and combined) were also analyzed separately.

Risk of Bias across Studies.
Potential publication bias was assessed by visually inspecting the funnel plot (plots of effect estimates against standard error) produced by the RevMan Analysis software. Publication bias tends to result in asymmetrical funnel plots [15,18]. Data on all variables from intention-to-treat analysis were combined to produce the funnel plot.

Study Selection.
The literature search produced 704 citations, of which 262 were double hits (studies found in at least two data sources). Screening of title and abstracts resulted in exclusion of 393 studies. After reading the full text of the remaining articles, 33 studies were excluded. 16 RCTs were included in the qualitative synthesis, but only 14 were included in the quantitative analyses because the required measures were not available for 2 studies (Figure 1).

Participants.
The number of groups compared in the studies varied: one study compared four groups (two exercise groups, one self-help course group, and a combination of exercise and self-help course group) [28], two studies compared three groups [19,21,24] (two interventions groups and a control group), two studies compared one type of exercise with a control group [20,31], and four studies [26,27,32,33] compared two different types of exercise without a control group; one of them was identified as an equivalence study [26]. In total 715 participants were studied before and after treatment. Three studies included men in the sample, in total 15 men of 165 patients. Almost all the participants were women ( = 700, 97.90%); there were 15 (2.10%) male participants. The average age of participants was 42.36 years.

Variables.
There was much variability in the outcome measures used in the included studies. Pain intensity was assessed using the VAS in five studies [19,21,22,26,33], and two used the SF-36 pain subscale [23,25], one the Multidimensional Pain Inventory [26], and three the FIQ pain scale [27,28,30].

Risk of Bias within Studies and Methodological Quality.
After critical review of each study included, it was concluded that all the studies included in this exceeded minimum thresholds for methodological and scientific quality.
However, since it is impossible to blind participants to group assignment in exercise intervention protocols, all studies were considered to be at a high risk of bias with respect to blinding of participants and personnel (Table 2) ( Figure 2).
The remaining data on outcome variables were homogeneous and therefore other sensitive analyses were not necessary.

Risk of Bias across Studies.
On visual inspection, the funnel plot of posttreatment outcomes was symmetrical and there was thus no evidence of publication bias. Due to heterogeneity produced by Kayo et al. 's study, data from this study were excluded for this analysis (Figure 4).

Discussion
The use of various exercise interventions in the studies presented above notwithstanding any physical activity is damaging for people with FMS. This is the first meta-analysis to assess the most effective exercise for improving some symptoms or conditions in fibromyalgia. Aerobic exercise for 30 to 60 minutes at an intensity of 50-80% of maximum heart rate 2 or 3 times per week for a period of 4-6 months and muscle strengthening exercises (1 to 3 sets of 8-11 exercises, 8-10 repetitions with a load of 3.1 kg or 45% of 1 repetition maximum (RM)) seem to be most effective in decreasing the pain and severity of FMS. Stretching the major muscle groups and aerobic exercise can improve the physical and mental component of HRQOL, respectively. Combined exercise programs consisting of aerobic exercise, muscle strengthening, and stretching exercises performed for 45-60 minutes 2 or 3 times per week for 3-6 months seem to be the most effective in reducing the symptoms of depression. The findings of this research are consistent with two previous equivalence studies [21,26] which concluded that aerobic and strengthening exercise have similar effects on pain intensity and FMS severity. In addition, like in this study, Kayo et al. [21] and Bircan et al. [33] also found that both aerobic and strengthening exercise were equally effective in improving HRQOL. However, this meta-analysis found that stretching exercise produces a greater improvement in the physical component of HRQOL than the rest of types of exercise that were studied whereas aerobic and combined exercise seem to be better at improving mental quality of life. Kayo et al. [21] noted that after 12 weeks without exercise the group who had performed muscle strengthening exercises had experienced recurrence of symptoms, whereas the beneficial effects of aerobic exercise persisted longer.
The results of other meta-analyses were also considered. In 2010, Häuser et al. [34] compared various types of aerobic exercise and found that exercising in the water and on dry land was similarly effective, this one being mild to moderate intensity, with a frequency of 2-3 times per week for 20-30 minutes at least for 4 weeks. Like García-Martínez et al. [23] Häuser et al. concluded that for the effects on physical condition and depression to persist the patients must maintain the exercise regime and therefore need to be motivated [34].
In another meta-analysis published by Kelley et al. [35] in the same year, seven studies were collected, to investigate the effects of physical activity, consisting of 15-60-minute sessions of aerobic and/or muscle strengthening exercises 2 or 3 times per week for a period of 12-23 weeks. The metaanalysis by Kelley et al. [35], like this study, suggested that physical activity improves the general well-being of women with FMS.
This meta-analysis had several limitations, one of which is the sample size of the included studies; most used relatively few participants. In addition, studies included in this metaanalysis were performed predominantly in women due to the fact that fibromyalgia is a syndrome with a significant female predominance [3]. This could be a limitation of the present study because it is not known whether the results obtained could be extrapolated to the male population suffering from fibromyalgia.
Unlike pharmacological studies, which are easily blinded, behavioral and physical treatment requiring the active participation of patients is virtually impossible to be blinded.
It is also important to take into account the heterogeneity of the studies, primarily due to the inclusion of the studies of Kayo et al. and Jones et al. [21,27]. This heterogeneity should be taken into consideration when drawing conclusions from the analysis of this study.
Another important limitation is that each type of therapeutic exercise was investigated in only a small number of studies. Aerobic exercise is the most commonly studied type of exercise treatment for FMS.

Conclusions
Exercise is beneficial for people with FMS but it is unable to draw any conclusions about what type of exercise is most