Are Psychological Interventions Efficacious for Adults With Migraine?

A Systematic Review and Meta-Analysis

Joanne Dudeney PhD; Louise Sharpe PhD; Sarah McDonald PhD; Rachel E. Menzies PhD; Brian McGuire PhD


Headache. 2022;62(4):405-419. 

In This Article


We aimed to investigate the stability of the findings from the recent Cochrane Review on psychological treatments for adults with migraine. Our results showed psychological interventions produced post-treatment improvements on the primary outcome of migraine frequency. Adults who underwent a psychological intervention were also more likely to have a 50% or more reduction in migraine frequency than if allocated to a control group. Similarly, we found broader benefits of psychological interventions through improvements in pain intensity and medication use at post-treatment, and pain-related disability at post-treatment and follow-up. We found no beneficial effect of psychological intervention on improving mood or QoL post-treatment for adults with migraine, though both were trending toward significance. All effects seen were small to moderate, and similar in size to the effects of psychological interventions in other chronic pain conditions in adults,[35] and in migraine or headache in youth.[36–38]

The current results differ from the previous Cochrane Review, which found no beneficial effect of psychological intervention on measured outcomes, except treatment response rate.[21] Our results support the possibility that the stringent study selection criteria of the Cochrane Review may, at least in part, have contributed to the previous null findings, through lack of power. We were able to include additional migraine studies by including data in forms other than mean and standard deviation scores on relevant clinical outcomes. Although it is possible that studies that do not report key mean and standard deviation data may be poorer quality, we are unaware of research supporting this and the risk of bias between these studies did not appear to differ. Further, their inclusion allowed us to estimate outcomes across a greater number of studies. Thus, we were afforded greater power when meta-analyzing each outcome. For example, we tripled the number of studies included for the primary outcome of migraine frequency. As a result, where the Cochrane Review found no effect of psychological intervention on this outcome, our results showed a beneficial effect.

Results of our sensitivity analyses of studies with migraine only and migraine or TTH showed similar beneficial effects of psychological intervention for reducing headache frequency, pain intensity, medication use, and disability. The stability of effects suggests that people with different headache conditions, which include differing symptomatology[25] and burden,[2,39] may respond to psychological intervention in a similar way. Interestingly, contrasted to our findings in a migraine-only population, the sensitivity analysis showed effects for improvements in mood. This could indicate that people with different headache-related conditions do not respond equivalently on mood, with people with TTH accounting for the significant effects for this outcome in the mixed condition analysis. However, more reasonable, given the trend toward significance in the studies with migraine only, is that the significant effect is a result of increased power. This is supported by the GRADE ratings for the migraine-only analysis on mood, which indicate further research would likely have an important impact on the estimate of the effect. However, if people with different headache conditions differ in how they respond to psychological interventions, this has both research and clinical implications. Researchers recruit patients with a range of headache-related conditions into psychological treatment trials,[15,40,41] and neurology clinics often see heterogeneous populations. Thus, it is essential that in future trials authors report the data for each population separately to allow investigation of the condition-specific effects before confirming whether psychological intervention for headache conditions is transdiagnostic for all outcomes.

In line with the Cochrane Review most risk of bias category judgements were unclear and the quality of the evidence for our effects (via GRADE ratings) was very low for all outcomes. Thus, we need to be cautious in the confidence we have in the magnitude of our effects. However, while violations that result in an outcome being graded as low (e.g., unclear--high risk of bias, heterogeneity, small number of studies with low sample sizes) are not desirable, they should not purely be seen as a reflection of poor study methodology. Several studies included in this review predate the recommended guidelines regarding reporting of clinical outcomes, trial registration, protocol publication, and matching treatment expectancy. Further, risk of bias criteria is consistent across all clinical trials (e.g., pharmacological and non-pharmacological). Compared to pharmacological, psychological trials tend to be plagued by smaller sample sizes, which routinely result in a higher risk of bias. As an example, the mean number of participants in meta-analyses of opioids[42] and cannabinoids[43] for chronic pain was shown to be greater than 130, compared to fewer than 60 for reviews of psychological interventions.[44] Indeed, the mean number of participants in the current review was 38. Thus, it is understandable that when using the generic Cochrane risk of bias tool some categories have an uncontrollable greater risk of bias than others.

Despite careful attention to methodology, this review has some limitations that should be considered when interpreting our results and planning future psychological intervention trials for adults with migraine. First, we were unable to analyze outcomes for adverse events post-treatment, and most outcomes at follow-up. The absence of measures of adverse events has been noted in the literature,[45] and it is imperative that studies measure adverse events or symptoms (e.g., increased migraine frequency or migraine-related disability) and report rates of, and if possible reasons for, dropout regardless of whether they are deemed to be related to the trial.[46] The absence of follow-up data limited our analyses and thus leaves uncertainty about the long-term efficacy of psychological intervention, which is problematic given the chronicity of migraine. While small, short-term improvements are encouraging, they are insufficient for us to recommend psychological treatment as a routine part of care. Research into other chronic pain conditions shows sustained effects of psychological intervention at follow-up, indicating the potential for similar patterns in adults with migraine.[35] However, researchers in this space must attempt to assess outcomes at follow-up, when funding permits. Second, there were limitations in both the participant characteristic and outcome data we were able to include. Consistent with other reviews,[17,23,24] we included studies that did not require participants have a ICHD or ICD diagnosis, nor report diagnostic group (e.g., chronic vs. episodic, presence of aura). However, guidelines on behavioral treatment for headache generally,[26] and chronic migraine specifically,[25] note the importance of including diagnostic information to increase validity and reliability of research findings. Of note, most studies included participants using diagnostic criteria, and of those that did not, more than half pre-dated the publication of these recommendations. Third, while we increased the number of studies compared to previous reviews, there were still a small number of studies included in the analyses. Of 39 studies, our largest outcome analyses, pain intensity and disability in the mixed headache group analysis, included only 17 studies. In 2005, recommendations were published regarding which core outcomes should be considered in the design of clinical trials for treatments of chronic pain.[47] These include pain; physical functioning; emotional functioning; participant rating of global improvement, symptoms, and adverse events; and participant disposition. While we observed increased adherence to these recommendations since 2005, especially in the most recently published studies, no study included in the current review assessed all outcomes. To increase standardization of reporting and the ability to pool data to assess treatment efficacy and effectiveness, researchers should be continually encouraged to collect and publish demographic and outcome data in line with recommendations.

Finally, given the small number of studies and heterogenous sample and study characteristics, we were limited in our ability to conduct moderator analyses. These analyses can be useful to discern active or beneficial components of treatments. While not an initial aim of the current paper, in line with the recent review into non-headache chronic pain,[35] we conducted post hoc moderator analyses on treatment modality. We were able to assess the benefit of behavioral treatments (e.g., PMR, autogenic training, coping skills) on four outcomes, where k ≥ 4: migraine frequency (k = 6, d = 0.28, p = 0.078), pain intensity (k = 6, d = 0.30, p = 0.031), disability (k = 5, d = 0.22, p = 0.098), and QoL (k = 6, d = 0.26, p = 0.122). Analyses on other modalities (e.g., CBT, biofeedback) were unable to be conducted, as k < 4 for all outcomes. From the above results, we could conclude that behavioral treatment has no beneficial effect on reducing migraine frequency, disability, or QoL. However, the heterogeneity of other study factors (e.g., dose, clinician, type of control) or sample factors (e.g., diagnosis, aura, comorbidities) means we cannot confidently confirm that any significant or nonsignificant effects were due to treatment modality alone. However, identifying active or beneficial components of treatments warrants further consideration in research trials in the future.

These limitations notwithstanding, our results have shown that psychological interventions are likely efficacious in reducing migraine frequency, pain intensity, and medication use in adults immediately following treatment, as well as reducing disability, both following treatment and over time. These effects appear stable for adults with either migraine or TTH and support the idea that psychological interventions may be transdiagnostic for headache conditions. This review has updated and extended previous reviews and comprehensively represents the current state of the evidence. However, further trials are needed to broaden our understanding of whether psychological interventions can improve mood and QoL for adults with migraine, extend our confidence in the current findings, and begin to determine which interventions are most efficacious.