Efficacy of Repetitive Transcranial Magnetic Stimulation in Fibromyalgia: A Systematic Review and Meta-Analysis of Randomized Controlled Trials

This article aimed to investigate the efficacy of repetitive transcranial magnetic stimulation (rTMS) in fibromyalgia. The PubMed, Medline, Cochrane Library, and Web of Science databases were searched for articles published through 14 August 2021. We enrolled only randomized controlled trials. The Cochrane Collaboration risk of bias tool was used for quality assessment. Outcomes were analyzed as standardized mean differences (SMDs) with 95% CIs. The beta coefficient and p value were adopted for meta-regression. We included 18 studies comprising 643 participants. A significant reduction in disease influence, as measured by the Fibromyalgia Impact Questionnaire, was observed (SMD, −0.700, 95% CI, −1.173 to −0.228), and the reduction was larger in older patients (β = −0.1327, p = 0.008). The effect persisted at least two weeks after the final treatment session (SMD, −0.784, 95% CI, −1.136 to −0.432). Reductions in pain, depression, and anxiety were discovered, which persisted for at least two weeks after the last intervention. The effects on pain and depression remained significant up to one and a half months after the final session. No serious adverse events were reported by the included articles. In conclusion, our systematic review and meta-analysis revealed that rTMS is safe and effective for managing multiple domains of fibromyalgia-related symptoms and older patients may have a stronger treatment effect. Larger randomized controlled trials with sufficient male populations are warranted to confirm our findings, detect rare adverse events, and determine the optimal stimulation parameters.


Introduction
Fibromyalgia syndrome usually presents as widespread pain accompanied by fatigue and psychiatric symptoms [1]. Although pathophysiology of fibromyalgia remains unclear, it is considered to be associated with central nervous system dysfunction causing central sensitization to pain [2]. The prevalence of fibromyalgia in the general population ranges from 0.2% to 6.6% and is more frequent in women [3]. Although fibromyalgia is characterized by widespread pain, it is often accompanied by many other symptoms such as mood disorders, decreased quality of life, impaired work performance, stiffness, fatigue, and physical functioning [3,4]. To capture the total spectrum of the symptoms, the Fibromyalgia Impact Questionnaire (FIQ) was published in 1991, which has been widely used in the assessment of treatment efficacy for fibromyalgia [3,4]. A later Revised Fibromyalgia Impact Questionnaire (FIQR) was published in 2009, which is easier to score but still correlates well with the original FIQ [5]. Due to the wide spectrum of symptoms, multidisciplinary approaches are necessary to achieve optimal management results [6]. This includes both pharmacological and non-pharmacological methods [7,8]. Among them, repetitive transcranial magnetic stimulation (rTMS) acts as a potential choice, with growing numbers of trials performed recently [9].
Randomized controlled trials (RCTs) have reported that rTMS can alleviate fibromyalgiarelated symptoms with few adverse events [10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27]. Although the mechanism of action of rTMS in fibromyalgia is not fully understood yet, it is believed to modulate the brain areas associated with affective-emotional components of pain, as well as activating the endogenous opioid analgesic system through mediating the motor cortex [28]. No standard protocols have been established so far. Applying both a low-frequency 1 Hz rTMS to the right hemisphere and a high-frequency 10 Hz rTMS to the left hemisphere was found to be effective, and most trials adopted one of the two methods. However, the sample sizes of these experiments were small, and discrepancies existed between the studies. Four articles of meta-analysis [29][30][31] have been published that investigated the efficacy of rTMS in patients with fibromyalgia and detected the sources of between-study heterogeneity [29][30][31][32]. The most recent one, conducted by Sun et al., concluded that rTMS improves pain intensity and FIQ score in patients with fibromyalgia, and the low-frequency rTMS in the dorsolateral prefrontal cortex (DLPFC) region seems to bring an optimal effect regarding the intensity of pain [32]. However, Sun et al. did not measure the modulator effect of different diagnostic criteria for fibromyalgia. Compared with the 1990 version, the 2010, 2011, and 2016 ACR criteria consider additional severity measurements of fibromyalgia-related symptoms and do not include tender point examination [33]. Whether this modification affects the results of rTMS treatment efficacy in patients with fibromyalgia has not been investigated. Besides, considering the wide spectrum of symptoms of fibromyalgia, the modulators for the effectiveness of rTMS measured by FIQ are worth further investigation. Additionally, in patients with major depressive disorder, the effect size of rTMS was determined to be related to age, sex, episode severity, and total rTMS pulses [34][35][36][37]. However, in patients with fibromyalgia, a previous meta-analysis did not recognize a dose-effect response in pain reduction [32]. Whether a dose-effect response measured with FIQ exists is worthy of further research because the FIQ assesses a wider spectrum of symptoms [38]. Furthermore, the influence of age on rTMS efficacy in patients with fibromyalgia has not been surveyed. Finally, although Sun et al. conducted a thorough and comprehensive review of this topic, they did not survey it longitudinally [32]. The duration of the treatment effect of rTMS and its moderators remain unclear.
We conducted a meta-analysis to determine the effect of rTMS for patients with fibromyalgia with longitudinally summarized outcomes. Because fibromyalgia causes a wide spectrum of discomfort and rTMS shows efficacy in multiple categories of symptoms, we chose FIQ/FIQR as the primary outcome. We anticipated filling the knowledge gap of moderator effects of the selected diagnostic criteria, patient demographics, disease severity, and rTMS parameters in the effect size and duration of effectiveness measured by FIQ/FIQR.

Materials and Methods
This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines [39]. We did not register or publish a prior protocol for this review.

Eligibility Criteria
We enrolled RCTs that reported rTMS treatment effects in patients with fibromyalgia. No limitations were imposed regarding the fibromyalgia diagnostic criteria or rTMS protocol. All retrieved articles were required to include 2 or more treatment arms, one of which must be rTMS and another of which must be a sham or any treatment other than NBS. The publication language was restricted to English.

Search Strategy
We searched the PubMed, Web of Science, Cochrane Central Register of Controlled Trials, and Medline databases. The keywords used were "repetitive transcranial magnetic stimulation" AND "fibromyalgia syndrome." The search period was from database inception to the present, with the final search conducted on 14 August 2021 (see File S1 for complete search strategy).

Study Selection and Data Extraction
Two reviewers (YCS and YHG) examined titles and abstracts to identify eligible articles. The reference lists of retrieved works were subsequently searched for related papers. When a consensus was not reached between the 2 reviewers, the senior author (YCL) made the final decision. The following data were extracted using a predetermined form: author, publication year, participant characteristics, rTMS details, comparator arm regimens, clinical outcomes, and adverse events. For articles with two or more intervention arms, we divided the control arm equally to form multiple comparisons. We employed the quantile estimation approach proposed by McGrath et al. [40] when medians and interquartile ranges were reported instead of means and standard deviations. National Institutes of Health image software (imagej.nih.gov, accessed on 12 October 2021) was used for outcomes reported as charts [41]. We set the pretest-posttest correlation coefficients to 0.5 if they were unavailable. We contacted the authors as necessary to resolve any uncertainties.

Quality Assessment
We applied the Cochrane Collaboration tool for assessing the risk of bias for quality assessment [42]. The quality of the eligible articles was evaluated by 2 reviewers (YCS and YHG) independently. Reviewer disagreements were resolved through discussion under the supervision of the senior author (YCL). The results were summarized by the Review Manager software version 5.3 (Cochrane, London, UK) and are presented in a graph and summary table.

Statistical Analysis
The primary outcome was FIQ/FIQR score. The secondary outcomes were fibromyalgiarelated pain intensity, Brief Pain Inventory (BPI) interference subscale score, McGill Pain Questionnaire (MPQ) score, number of tender points, Beck Depression Inventory (BDI) score, Hamilton Depression Rating Scale (HDRS) score, Hospital Anxiety, and Depression Scale anxiety subscale (HADS-A) score, and fatigue severity scale (FSS) score. The data were extracted for the following time points: at baseline and 2 weeks to 1 month and 1.5 to 3 months after the final rTMS treatment. A meta-analysis was conducted if the outcomes were appropriately reported for 3 or more comparisons in similar populations. We used a random-effects model for effect size pooling; the results are presented as standardized mean differences (SMDs) with a 95% CIs. Between-study heterogeneity was assessed using I 2 , and considerable heterogeneity was defined as an I 2 of >50% [43]. Subgroup analyses for all outcomes were conducted for the stimulation site, fibromyalgia diagnostic criteria, and frequency of stimulation to identify any moderator effects. A significant difference between effect sizes was indicated by nonoverlapping 95% CIs. Furthermore, to explore the reasons for between-study heterogeneity, we performed post hoc analyses for outcomes with I 2 values >50%; such analyses comprised random-effects meta-regression exploring the correlations between the effect sizes and the studies' distinct characteristics. Publication year, age, fibromyalgia disease duration, rTMS frequency, rTMS intensity, pulses per session, total pulses, number of treatment weeks, number of sessions per week, baseline pain intensity, baseline BDI score, and baseline FIQ/FIQR score were treated as quantitative variables. Sex, the fibromyalgia diagnostic criteria, stimulated hemisphere, and targeted brain area were treated as categorical variables. The meta-regression results were considered statistically significant when p < 0.05. Funnel plots and Egger tests were used to detect publication bias, and a two-tailed p < 0.1 was regarded as statistically significant [44]. We conducted a sensitivity analysis for the primary outcome by removing one trial at a time and analyzing the remaining trials to estimate each study's contribution to the overall effect size. All analyses were performed in Comprehensive Meta-Analysis software version 3 (Biostat, Englewood, NJ, USA).

Certainty of Evidence
We used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology to assess the certainty of the evidence of the primary outcome. Because our study included only RCTs, the results begin as high certainty, and the final rating depends on the overall risk of bias, imprecision, inconsistency, indirectness, and publication bias [45].

Study Selection and Description
The initial search returned 455 articles. Eighteen RCTs [10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27] with 643 participants entered qualitative synthesis ( Figure 1). Two studies [11,18] included two intervention arms; therefore, the control groups for this research were divided for separate comparisons, creating 20 comparisons in total. 17 RCTs entered quantitative analysis for all the outcomes, and one was not included since none of the outcomes reported in the study were in our interest of quantitative synthesis. The included trials' characteristics are presented in Table 1.  The number of participants ranged from 15 to 86, and the mean age ranged from 40.4 to 53.9 years. Three studies [12,13,17] recruited fibromyalgia patients diagnosed with the 2016 ACR criteria. One article enrolled participants with fibromyalgia diagnosed with the 2011 ACR criteria [18]. Four [10,14,23,24] studies included individuals with fibromyalgia diagnosed with the 2010 ACR criteria. Nine papers [11,15,[19][20][21][22][25][26][27] enrolled patients with fibromyalgia diagnosed with the 1990 ACR criteria. One paper [16] did not mention the diagnostic criteria. The number of treatment sessions ranged from 8 to 20. The duration from the first to the last treatment session ranged from 2 to 21 weeks. The total number of pulses ranged from 12,000 to 60,000, and the pulses per session ranged from 1200 to 4000. The targeted brain area was M1 in 10 interventions and DLPFC in nine interventions; one trial [22] did not specify the targeted brain area. Additional data are presented in Table 2.

Subgroup Analysis
The effect sizes associated with each stimulation site is listed in Table 3. rTMS over the M1 area was effective in reducing FIQ/FIQR score, pain intensity, BPI interference subscale score, MPQ score, and BDI score; rTMS over the DLPFC reduced FIQ/FIQR score, pain intensity, MPQ score, number of tender points, and HDRS score. However, no significant difference was detected between subgroups for any outcome.  The results of the subgroups analysis by distinct diagnostic criteria are presented in Table 4. Among patients with diagnoses based on the 1990 ACR criteria, rTMS reduced FIQ/FIQR score, pain intensity, BPI interference subscale score, MPQ score, and number of tender points. For patients with diagnoses based on the 2010, 2011, or 2016 ACR criteria, rTMS reduced pain intensity, BPI interference subscale score, MPQ score, BDI score, HDRS score, and HADS-A score. Nonetheless, none of the differences between subgroups reached statistical significance.  The outcomes of the subgroup analysis by high frequency (HF) and low frequency (LF) stimulation of rTMS are shown in Table 5. For the LF group, rTMS improved FIQ/FIQR score, pain intensity, MPQ score, and HDRS score. In the HF group, FIQ/FIQR score, intensity of pain, BPI interference, MPQ score, number of tender points, BDI score, and HADS-A score improved after treatment. Nevertheless, no significant difference appeared between subgroups for any outcome.

Certainty of Evidence
Overall evidence was assessed using GRADE. The certainty of the evidence of the improvement of FIQ/FIQR scores after rTMS treatment revealed a low quality of evidence. The level was downgraded due to large CI and significant between-study heterogeneity. As for the outcome two weeks to one month after the last session, the certainty of the evidence was moderate. The details are presented in Table 6. The upper and lower limit of 95% CI ranged from large to small effect size. e This was calculated by pooling the sham group of the 11 comparisons included in the primary outcome, comparing the FIQ/FIQR score before and after treatment. f This was calculated by pooling the rTMS group of the 11 comparisons included in the primary outcome, comparing the FIQ/FIQR score before and after treatment. g This was calculated by pooling the rTMS group of the 5 studies included in the primary outcome and provided baseline data, comparing the FIQ/FIQR score before and 2 weeks to 1 month after treatment. h This was calculated by pooling the rTMS group of the 5 studies included in the primary outcome and provided baseline data, comparing the FIQ/FIQR score before and 2 weeks to 1 month after treatment.

Discussion
We systematically reviewed 18 RCTs investigating the effect of rTMS on fibromyalgiarelated symptoms. Compared with sham treatment, patients receiving rTMS had lower FIQ scores as well as less pain, depression, and anxiety. These effects persisted for at least two weeks after the final treatment session, and the improvement of pain and depression remained significant at up to one and a half months after the final session. Moreover, the efficacy was stronger in older patients. However, no reduction was detected in fatigue, and the correlations between FIQ score and diagnostic criteria, disease severity, and rTMS parameters were not significant.
Several systematic reviews with quantitative synthesis have yielded discrepant results. Knijnik et al. [31] performed a meta-analysis of five studies. They concluded that rTMS improved quality of life but did not reduce depression or pain. Saltychev et al. [29] conducted a meta-analysis of seven trials. They reported that the decrease of pain after rTMS did not reach clinical significance. Hou et al. [30] performed a meta-analysis of 16 studies treating fibromyalgia with NBS. Among them, 11 treated fibromyalgia with rTMS, and the pooled effect size revealed significant reductions in pain, depression, fatigue, and the number of tender points as well as general improvements in health and function. Finally, the most recent meta-analysis including 14 RCTs revealed improvements in pain intensity and FIQ score [32]. By enrolling up-to-date RCTs counting 17 in total, our meta-analysis further revealed treatment effects not only in FIQ score and pain but also in depression and anxiety. The higher statistical power in our review may explain the discrepancies. As for the effect sizes of previous meta-analyses, pain reduction was most surveyed, which had small to medium effect sizes [30][31][32]46]. This corresponds to our study, which also revealed a medium effect size [46].
The improvement in FIQ/FIQR score as well as the secondary outcomes in our study implies that an overall improvement of the total spectrum of problems related to fibromyalgia might exist, which includes fibromyalgia-related symptoms, overall impact, and functional impairment [38]. Considering the high positive correlation between FIQR and suicide risk revealed in previous articles [47,48], as well as higher health economic costs in patients with higher FIQ scores [49], this improvement in the FIQ score has an important impact at both the individual and public health levels.
We found improvements in pain intensity, pain quality, and physical functioning measured by VAS/NPRS, MPQ, and BPI interference score [50]. Although the pathophysiology of fibromyalgia is unknown, central sensitization that affects the pain modulatory system is believed to play an important role. Research has demonstrated that rTMS may attenuate symptoms of fibromyalgia by moderating the cortical excitability of brain structures associated with pain modulation [9,51]. Moreover, the M1 and DLPFC areas are crucial in top-down pain control and opioid release [23], possibly explaining the analgesic effect observed in our review. Some studies have found a relationship between the total number of tender points and the severity of central sensitization [52][53][54][55]. Hence, the decreased numbers of tender points after treatment revealed in our study may also imply an improvement in central sensitization.
Relieved emotional functioning measured by BDI, HDRS, and HADS-A were noticed by meta-analyses [50]. Similar results regarding fibromyalgia-related anxiety and depression have been reported by studies targeting the M1 [12] or DLFPC [23]. Because both anxiety and depression are related to pain [56], the symptom reduction may stem from the analgesic effects of rTMS. Furthermore, research has revealed that the DLPFC is related to the anterior insula and amygdala [57], which are associated with anxiety [58] and depressive [59] symptoms. The aforementioned evidence potentially explains the effects observed in our meta-analysis.
We discovered a positive correlation between age and FIQ score reduction in metaregression. No studies have found this relationship between rTMS efficacy for fibromyalgia and age. However, a study compared pain sensitivity and structural changes to the brain in fibromyalgia between patients aged above and below 50 years [60]. Distinct patterns of change in the thickness of gray matter were detected, as well as increased pain sensitivity in only the older group. Moreover, insular gray matter significantly decreased with age across all patients with fibromyalgia. The authors concluded that the brain structures and functions involved in pain modulation might shift from being adaptive in younger individuals to being maladaptive in older patients with fibromyalgia. Because the insular cortex is critical for pain modulation [61] and anterior insula change is believed to be the mechanism underlying the efficacy of rTMS, we anticipate greater improvements among older patients with fibromyalgia.
We did not identify a correlation between effect size and the total number of rTMS pulses. This result may imply a ceiling of rTMS efficacy for patients with fibromyalgia. However, this result may be solely due to the insufficient power of our small sample sizes. Future studies are warranted to fill this knowledge gap.
The adverse events resulting from rTMS treatment are generally tolerable. Dizziness, nausea, headache, neck pain, stimulation site discomfort, and several neurobehavioral adverse events were reported in the enrolled studies. However, no serious adverse events were reported. Although rTMS has been reported to carry the risk of inducing seizures, no seizures were observed in our review.
The strength of our study exists in several aspects. First, we are the first to summarize fibromyalgia-related symptoms not only widely but also longitudinally. Second, this is the first meta-analysis to assess the moderators for FIQ/FIQR score, which revealed a relationship between effect size and age. This inspires future studies to assess the possible difference in pathophysiology of fibromyalgia between younger and older patients, and such correlation may also encourage succeeding randomized controlled trials to compare the effect sizes of rTMS in fibromyalgia patients of different ages. Third, we are the first to assess the moderator effect longitudinally in order to find out the possible factors that determine the duration of effectiveness. Fourth, we included up-to-date RCTs, further reducing the possibilities of false negatives compared with the previous meta-analyses [29][30][31][32].
This review and meta-analysis has several limitations. First, all of the included studies had small numbers of participants, and the patient demographics, study designs, and stimulation parameters were heterogeneous. Although, according to our meta-analyses, most potential moderators were unrelated to the treatment effects, the low statistical power meant that the possibility of false negatives was high. Second, all of the enrolled studies had a female majority. Although no correlation between treatment effect and sex was detected in meta-regression, this may result from the underrepresentation of men with fibromyalgia. Third, several distinct sets of diagnostic criteria were adopted. Only three studies used the latest 2016 ACR criteria, and the low statistical power may have caused a false negative in the detection of the correlation of criteria adopted with effect size. Moreover, the scant use of the 2016 criteria, the latest ACR criteria for fibromyalgia, might impede the generalizability of our results. Fourth, most of the included investigations allowed concurrent medication during the study period. Therefore, rTMS acted more as an add-on therapy to medication treatment. The possibility of an interaction between pharmacological and rTMS treatment effects requires further research. Fifth, we did not include participant ratings of improvement, concomitant pain treatments, deposition of participants, and adverse events as outcomes of meta-analysis because of insufficient data and high between-study heterogeneity. However, these outcomes are important in the evaluation of treatments for chronic pain [50], and future trials may estimate these outcomes to fill the gap. Sixth, FIQ/FIQR were not the primary outcome in most of the RCTs included in our review, which might bias our results. Finally, although no serious adverse events were reported, the relatively low number of participants in most of the studies may limit our ability to conclude that such events are rare [62]. Larger RCTs using the 2016 ACR criteria with a sufficient male population are warranted to confirm our findings and to delineate the optimal dose, treatment frequency, and stimulation target for rTMS.

Conclusions
This meta-analysis revealed that rTMS is safe and effective for treating multiple domains of fibromyalgia-related symptoms, and older patients may have a stronger effect. Future studies are required to detect rare adverse events and determine the optimal stimulation parameters.

Conflicts of Interest:
The authors declare no conflict of interest.