What is the efficacy of aerobic exercise versus strength training in the treatment of migraine? A systematic review and network meta-analysis of clinical trials

Abstract

Background: Multiple clinical trials with different exercise protocols have demonstrated efficacy in the management of migraine. However, there is no head-to-head comparison of efficacy between the different exercise interventions.

Methods: A systematic review and network meta-analysis was performed involving all clinical trials which determined the efficacy of exercise interventions in reducing the frequency of monthly migraine. Medical journal search engines included Web of Science, PubMed, and Scopus spanning all previous years up to July 30, 2022. Both aerobic and strength/resistance training protocols were included. The mean difference (MD, 95% confidence interval) in monthly migraine frequency from baseline to end-of-intervention between the active and control arms was used as an outcome measure. Efficacy evidence from direct and indirect comparisons was combined by conducting a random effects model network meta-analysis. The efficacy of the three exercise protocols was compared, i.e., moderate-intensity aerobic exercise, high-intensity aerobic exercise, and strength/resistance training. Studies that compared the efficacy of migraine medications (topiramate, amitriptyline) to exercise were included. Additionally, the risk of bias in all included studies was assessed by using the Cochrane Risk of Bias version 2 (RoB2).

Results: There were 21 published clinical trials that involved a total of 1195 migraine patients with a mean age of 35 years and a female-to-male ratio of 6.7. There were 27 pairwise comparisons and 8 indirect comparisons. The rank of the interventions was as follows: strength training (MD = -3.55 [- 6.15, - 0.95]), high-intensity aerobic exercise (-3.13 [-5.28, -0.97]), moderate-intensity aerobic exercise (-2.18 [-3.25, -1.11]), topiramate (-0.98 [-4.16, 2.20]), placebo, amitriptyline (3.82 [- 1.03, 8.68]). The RoB2 assessment showed that 85% of the included studies demonstrated low risk of bias, while 15% indicated high risk of bias for intention-to-treat analysis. Sources of high risk of bias include randomization process and handling of missing outcome data.

Conclusion: Strength training exercise regimens demonstrated the highest efficacy in reducing migraine burden, followed by high-intensity aerobic exercise.

Keywords: Aerobic exercise; Exercise; Migraine; Network meta-analysis; Strength training; Systematic review.

Fig. 1

PRISMA (Preferred Reporting of Items in Systematic Reviews and Meta-Analysis) flowchart depicting the identification, screening, and inclusion of the studies included

Fig. 2

Network plot of all studies included. The size of the nodes and the thickness of edges depend on the number of people randomized and the number of trials conducted, respectively

Fig. 3

A Risk of Bias summary for included studies which applied intention-to-treat analysis. B Risk of Bias summary for included studies which applied per-protocol analysis

Fig. 4

A Forest plot for comparison of all interventions against the reference (i.e. placebo) using the frequentist random-effect model network meta-analysis. B Individual study results (for all studies) grouped by treatment comparison. Forest plot was based on random-effects model for network meta-analysis. MD = mean difference between active and placebo arms in monthly migraine frequency from baseline to end of intervention

Fig. 5

A Leverage versus residual deviance. The leverage plot helps to assess model fitness. Points that lie outside the line with average leverage of 3 can generally be identified as contributing to the model's poor fit. All included studies (represented by the green dots) were congregated between the average leverage parabola lines of 1.5 and 2.5, indicating that all studies fitted well in the model. B The stem plot represents the posterior residual deviance of each study arm. The number of stems represents the 43 study arms compared in the network meta-analysis. Each stem indicates the residual deviance of each arm in each included study. The shorter the stem, the smaller the residual deviance, the better the model fit for each data point. None of the stems featured residual deviance greater than 2 – reflecting the model’s fitness. C This plot represents each arm’s contribution to the residual deviance for the NMA (x-axis) and the unrelated mean effect (UME) inconsistency models (y-axis) along with the line of equality. The points on the equality line means there is no improvement in model fit when using the inconsistency model, suggesting that there is no evidence of inconsistency. Points above the equality line means they have a smaller residual deviance for the consistency model indicating a better fit in the NMA consistency model and points below the equality line means they have a better fit in the UME inconsistency model. Most of the included study arms were found on the equality line indicating minimal inconsistency and optimum model fitness