Article Title
Overview
Exercise has demonstrated efficacy in mitigating migraine symptoms and reducing disability, yet the most effective exercise modalities for migraine treatment remain unclear.
A systematic review was conducted using databases including PubMed, PEDro, Web of Science, and Google Scholar. This review included clinical trials evaluating the impact of various exercise modalities on the frequency, intensity, duration, and disability associated with migraines. Eight network meta-analyses were performed using both frequentist and Bayesian models to assess the direct and indirect evidence for different exercise interventions. The treatment effects were measured using standardized mean differences (SMD) with 95% confidence intervals (CI) and credible intervals (CrI), calculated based on Hedge’s g and p scores to rank the effectiveness of the exercise modalities.
Introduction
The prevalence and complexity of migraines, along with the substantial costs associated with work absenteeism and healthcare, have led to the development of numerous pharmacological treatments. These include non-steroidal anti-inflammatory drugs, triptans, antiseizure medications, antidepressants, monoclonal antibodies targeting the calcitonin gene-related peptide pathway, ditans, gepants, and onabotulinumtoxinA. While these treatments have proven effective in reducing migraine frequency and halting episodes, their efficacy can vary among individuals, and both short- and long-term side effects can adversely affect the health and quality of life of migraine sufferers.
Non-pharmacological interventions, such as exercise, have been explored as alternatives to reduce the dependency on medication. Exercise has been shown to lower the frequency, intensity, and duration of migraine symptoms, decrease disability, and enhance quality of life. Additionally, exercise can potentially reduce medication use and mitigate some adverse effects associated with pharmacological treatments. Beyond physical benefits, exercise positively impacts psychosocial factors, including anxiety, depression, and social relationships, which in turn can improve migraine symptoms and related disabilities.
Various exercise modalities have been researched, with yoga and aerobic exercise being the most studied and recommended. However, there is no clear consensus on which exercise interventions are most effective for reducing migraine frequency, intensity, duration, and disability based on quantitative data. A previous network meta-analysis (NMA) by Woldeamanuel and Oliveira suggested that strength training might be more effective than aerobic exercise in decreasing migraine frequency. However, their study did not include all exercise modalities, had ambiguous categorizations and outcome measures, and did not examine effects on pain intensity, migraine duration, and disability.
Also read Blood Donation And Relief From Migraine Pain
Utilizing the exercise modality categorization from the clinical practice guideline for exercise prescription in migraine by La Touche et al., a new NMA was conducted. This analysis aimed to evaluate and compare the efficacy of various exercise modalities on migraine frequency, intensity, duration, and disability in patients with migraine, as opposed to standard pharmacological treatments alone. Furthermore, the study compared different exercise modalities with each other to identify the most effective interventions.
Methods
This systematic review and network meta-analysis (NMA) followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Network Meta-Analysis Extension Statement and was registered in the International Prospective Register of Systematic Reviews (PROSPERO: CRD42023399242). Two independent reviewers conducted searches across four databases: MEDLINE (PubMed), Physiotherapy Evidence Database (PEDro), Web of Science, and Google Scholar. The search strategy, which included medical subject headings and non-medical subject headings terms such as “Migraine Disorders” and “Exercise,” was implemented without language or time restrictions. Discrepancies between reviewers were resolved through consensus with a third reviewer. The search was initially performed in April 2023 and updated in May 2023.
Inclusion Criteria
- Population: Adults aged 18 years or older diagnosed with episodic or chronic migraine by a physician.
- Interventions: Exercise modalities such as moderate-intensity continuous aerobic exercise, high-intensity aerobic exercise, yoga, resistance training, relaxation exercise, aerobic exercise with lifestyle recommendations, aerobic and relaxation exercise, and Tai Chi, as defined by La Touche et al.
- Comparators: Other exercise modalities, usual care, pharmacological treatments, or educational interventions.
- Outcome Measures: Migraine frequency (days with migraine per month or similar definitions), intensity (evaluated using visual analog scales or numeric pain rating scales), duration (measured in hours or minutes per migraine attack), and disability (assessed through validated questionnaires such as the Headache Impact Test 6 or the Migraine Disability Assessment Test).
- Study Design: Clinical trials.
Exclusion Criteria
- Studies involving participants younger than 18 years.
- Studies not involving a physician diagnosis of episodic or chronic migraine.
- Studies that did not evaluate the specified exercise modalities.
- Studies lacking appropriate comparators.
- Studies that did not measure the defined outcome variables at the end of the intervention protocol.
- Non-clinical trial designs.
The selection process involved two independent reviewers who analyzed titles, abstracts, and keywords, with full-text reviews conducted when necessary. Studies that met the inclusion criteria underwent further full-text analysis. Any discrepancies were resolved by consensus with a third reviewer. Data extraction focused on relevant information about population characteristics, interventions, comparators, outcome measures, and main results, with efforts made to contact original researchers for additional or confirmatory data when needed.
The methodological quality of the included studies was assessed using the PEDro scale, which comprises 11 items evaluating aspects such as eligibility criteria, random allocation, allocation concealment, baseline similarities, blinding of participants, therapists, and assessors, dropout rates, adherence to allocated interventions or intention-to-treat analysis, statistical comparisons, and reporting of point measures and variability. Scores were categorized as “poor” (0-3 points), “fair” (4-5 points), “good” (6-8 points), or “excellent” (9-10 points).
The risk of bias was assessed with the Risk of Bias 2.0 tool, which examines five domains: randomization process, deviations from the intended intervention, missing outcome data, outcome measurement, and selection of the reported result. Each domain was evaluated through a series of questions with possible answers of “yes,” “probably yes,” “probably no,” “no,” or “no information,” leading to an overall judgment of “low risk,” “some concerns,” or “high risk.”
The concordance between independent reviewers in assessing methodological quality and risk of bias was measured using the kappa coefficient (κ), with values interpreted as low (<0.5), moderate (0.5–0.7), or high (>0.7). Any disagreements during this process were resolved by a third reviewer.
Qualitative Analysis
Two independent reviewers utilized the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) tool, adapted specifically for network meta-analyses (NMAs), to categorize the evidence quality of various interventions across different outcome measures into four levels: high quality, moderate quality, low quality, and very low quality. This assessment was conducted based on five domains: study design, imprecision, indirectness, inconsistency, and publication bias.
Evidence Quality Levels
- High Quality: Very confident that the true effect is close to the estimate.
- Moderate Quality: Moderately confident in the effect estimate.
- Low Quality: Limited confidence in the effect estimate.
- Very Low Quality: Minimal confidence in the effect estimate.
Assessment Domains
- Study Design
– Downgraded by one point if there were some concerns.
– Downgraded by two points if there was a high risk of bias.
- Imprecision
– Downgraded by one point if the confidence interval (CI) was wide.
– Downgraded by two points if the CI crossed the non-significant threshold.
- Indirectness
– Evaluated when the transitivity assumption was potentially at risk due to variations in study designs.
- Inconsistency
– Downgraded by one point for heterogeneity within a design or inconsistency between designs.
– Downgraded by two points if both heterogeneity and inconsistency were present, or if heterogeneity could not be calculated.
- Publication Bias
– Downgraded if the funnel plot and Egger’s regression test suggested asymmetry and a significant risk of publication bias.
Statistical Analysis
Two network meta-analyses (NMAs) were conducted for each outcome measure using frequentist and Bayesian methods in RStudio (version 2023.06.0-431) with R software (version 4.3.1). The frequentist model was developed with the “netmeta” package, while the Bayesian model used the “gemtc” package, following guidance from Harrer et al. Exercise interventions were categorized based on La Touche et al., and their effects on migraine frequency, intensity, duration, and disability were analyzed, with network graphs illustrating direct comparisons.
The standardized mean difference (SMD) for each treatment comparison in individual studies and pooled effect sizes were calculated using Hedge’s g. Effect sizes were categorized as trivial (0.0–0.2), small (0.2–0.6), moderate (0.6–1.2), large (1.2–2.0), very large (2.0–4.0), and extremely large (>4.0). Precision was represented with 95% confidence intervals (CIs) and credible intervals (CrIs).
Frequentist model results were summarized in Forest plots for each outcome, showing differences between exercise modalities and their efficacy compared to pharmacological treatment alone. The transitivity assumption was assessed through consistency, using Cochrane’s Q statistic test, inconsistency index (I2), and tau square (T2) value. A full design-by-treatment interaction random-effects model showed significant reduction in Cochrane’s Q statistic for between-design consistency with the random-effects model.
Bayesian model results were also summarized in Forest plots and compared with frequentist Forest plots to identify discrepancies. An effect estimate table for all possible treatment comparisons was created with their respective CIs and CrIs. Bayesian network model fit was evaluated with residual deviance, with significant deviance indicating poor fit. Leverage versus residual deviance plots assessed model fitness, identifying studies contributing to poor fit, which were then subjected to sensitivity analysis. A network meta-regression was performed to determine if variations in patients’ mean ages influenced migraine disability NMA results.
Inconsistency within each intervention was evaluated using net splitting (frequentist) and node splitting (Bayesian) methods, comparing direct and indirect effect estimates. Significant differences were indicated by p < 0.05.
Publication bias was assessed with comparison-adjusted funnel plots and Egger’s regression test for funnel plot asymmetry. Visual asymmetry and p < 0.05 suggested publication bias.
The “viscomp” package was used to create a rank heat plot to rate exercise modalities based on p scores, with higher scores indicating greater superiority. These p scores should be interpreted in conjunction with the Forest plot, using pharmacological treatment alone as the comparison group.
Results
A total of 28 studies were included in this network meta-analysis (NMA), with a selection process depicted in a flowchart (Figure 1). The studies involved 1501 migraine patients, comprising 86.3% females and 13.7% males. The majority, 70.7%, had episodic migraine, 7.3% had chronic migraine, and 22% had unspecified migraine types. The mean age ranged from 23 to 51 years, and body mass index (BMI) ranged from 20 to 36 kg/m². Ultimately, 1430 participants were analyzed.
Methodological quality and risk of bias were assessed using the PEDro scale, where eight studies scored well, fourteen scored fair, and six scored poorly, with high reviewer concordance (κ = 0.92). In terms of bias risk, none of the studies had a low risk of bias; four had some concerns, while twenty-four had a high risk of bias, with moderate reviewer concordance (κ = 0.662).
Migraine frequency was evaluated in 25 studies, showing that yoga, high-intensity aerobic exercise, and moderate-intensity continuous aerobic exercise were significantly more effective than pharmacological treatment alone. Yoga was the most effective intervention, followed by high-intensity aerobic exercise, resistance training, and moderate-intensity continuous aerobic exercise. High heterogeneity and inconsistency were observed, but these decreased significantly under a random-effects model. No publication bias was detected.
Migraine intensity was assessed in 23 studies, with yoga and high-intensity aerobic exercise showing significant superiority over pharmacological treatment in the frequentist model. High-intensity aerobic exercise ranked first, followed by yoga, relaxation exercise, and moderate-intensity continuous aerobic exercise. High heterogeneity and inconsistency were present, but model fit was satisfactory, and no publication bias was detected.
Migraine duration was examined in 17 studies, demonstrating that high-intensity and moderate-intensity continuous aerobic exercise were significantly more effective than pharmacological treatment. High-intensity aerobic exercise ranked highest, followed by moderate-intensity continuous aerobic exercise and relaxation exercise. Moderate to high heterogeneity and inconsistency were noted, but these improved under a random-effects model. No significant inconsistency or publication bias was found.
Migraine disability was assessed in 14 studies, with moderate-intensity continuous aerobic exercise being the only intervention significantly better than pharmacological treatment in the frequentist model. High heterogeneity and inconsistency were observed, but these decreased significantly under a random-effects model. Age differences between studies did not significantly influence the outcomes. No publication bias was detected.
The strength of evidence, according to the GRADE evaluation, was very low for the effectiveness of exercise interventions compared to pharmacological treatment across all outcomes, including high-intensity aerobic exercise, yoga, and moderate-intensity continuous aerobic exercise. There was very low-quality evidence supporting the superiority of high-intensity aerobic exercise and yoga over migraine education in reducing migraine intensity. Similarly, very low-quality evidence suggested no significant superiority of high-intensity aerobic exercise over yoga and moderate-intensity aerobic exercise in decreasing migraine frequency, intensity, and duration.
Conclusion
The findings from both frequentist and Bayesian NMAs indicated very low-quality evidence supporting the efficacy of certain exercise interventions for migraine management. Specifically, yoga, high-intensity aerobic exercise, and moderate-intensity continuous aerobic exercise were effective in reducing migraine frequency. Yoga alone was effective in significantly decreasing migraine intensity. Both high- and moderate-intensity aerobic exercises were shown to significantly reduce migraine duration.
The frequentist model uniquely demonstrated that high- and moderate-intensity aerobic exercises significantly reduced migraine intensity, and moderate-intensity aerobic exercise significantly reduced migraine disability. Yoga emerged as the top-ranked intervention for reducing migraine frequency. High-intensity aerobic exercise was ranked highest for reducing both migraine intensity and duration, while moderate-intensity aerobic exercise was most effective for reducing migraine disability. However, the ranking of high-intensity aerobic exercise should be cautiously interpreted due to the limited number of studies and their high risk of bias.
Oncology Related Tools
Other
Latest Research
