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


Search Process

From an initial search of 1020, 39 studies were included in the review. Twenty studies were included from the Cochrane Review,[21] with our revised search dates and criteria producing a further 19 eligible studies.

Sample and Study Characteristics

Sample and study characteristics are reported in Table 1 (migraine-only data, k = 31; combined migraine and TTH data, k = 8). Overall, studies were published over a 40-year period (1981–2021), with the majority (k = 26) published from 2008 onward. Studies included 3155 participants post-treatment; n = 1678 in the treatment arms (mean n = 38, standard deviation SD = 30, range 15–195). The age range of participants was 18–71 years. Twenty-eight studies required participants have a migraine/headache diagnosis using International Classification of Headache Disorders (ICHD) criteria or International Classification of Disease (ICD) coding criteria, with the remaining 11 reporting various diagnosis requirements (e.g., neurologist letter, self-report, unreported). Sixteen studies reported including patients with migraine with and without aura, one study excluded participants with aura, and the remaining did not report these data. For the eight studies that included participants with TTH, four included participants with either episodic or chronic TTH, two included participants with chronic TTH, one with episodic TTH, and one study did not report data in which type of TTH could be ascertained. The percentage of participants with TTH ranged from 21% to 41% of the total sample.

The type, mode, and duration of psychological intervention varied. Most interventions were behavioral (e.g., progressive muscle relaxation [PMR], autogenic training, behavioral management, k = 13), followed by CBT (k = 6). The remaining studies employed biofeedback, mindfulness-based stress reduction, hypnosis, EMDR, psychoeducation, or a combination of the above-mentioned interventions (e.g., CBT and biofeedback; biofeedback and relaxation). Studies mainly used non-active control groups (e.g., waitlist, standard care; k = 19), with the remaining employing a control to account for time and/or attention (k = 13), pharmacotherapy (k = 6), or a placebo (k = 1). All but one study included information on treatment duration. The number of sessions ranged from 1 to 22, with 13 studies reporting interventions of 4 sessions or fewer.

Thirty studies reported on face-to-face interventions in a clinic setting, with three studies including additional phone contact. Twelve face-to-face interventions were provided in a group format, 18 provided individual treatment, and two included both group and individual components. Face-to-face interventions were primarily delivered by a qualified psychologist or therapist (k = 17). Seven studies did not report face-to-face delivery. Ten studies reported remotely delivered interventions (one study delivered both a face-to-face and remote treatment arm[34]). Nine of the remotely delivered interventions were individual, and four involved clinician support.

Risk of Bias

As can be seen through both Table 2 and Figure 2, all studies and risk of bias categories had some degree of unclear or high risk of bias. The total risk of bias score ranged from 1 to 7, with higher scores indicating greater risk of bias. T-tests revealed no significant differences in total bias scores between studies in the Cochrane Review[21] (n = 20, M = 4.89, SD = 1.41) and studies included in the current review only (n = 19; M = 4.58, SD = 1.74; t[37] = 0.614, p = 0.543).

Figure 2.

Risk of bias graph, by category [Color figure can be viewed at]

Meta-analytic Findings

Table 3 presents a summary of all post-treatment outcomes for the studies including adults with migraine only, including GRADE ratings. Of note, with the addition of 19 studies, all reported effect sizes differ from the original Cochrane Review.[21] Figure 3 presents the forest plot for the primary outcome (migraine frequency) and forest plots for all secondary outcomes can be seen in the supporting information.

Figure 3.

Forest plot for migraine frequency, post-treatment. CI, confidence interval

Migraine Frequency. For the primary outcome of change in migraine frequency, post-treatment, 16 studies (6, 7, 10, 11, 12, 13, 18, 19, 25, 26, 34, 35, 36, 37, 38, 39; n = 1402) were included in the analysis. The analyses demonstrated that psychological interventions had a small beneficial effect in reducing migraine frequency, post-treatment, compared to control conditions (d = 0.28, 95% confidence interval [CI]: 0.12, 0.44; p = 0.001). This outcome had moderate heterogeneity, I 2 = 52%. The GRADE quality rating was very low, meaning further research is extremely likely to have an important impact on our confidence in the estimate of effect.

Six studies (13, 14, 16, 17, 25, 34; n = 435) assessed the proportion of participants who reduced the frequency of their migraine in the 4 weeks following intervention compared to the 4 weeks prior to the intervention (i.e., treatment responders). Forty-five percent (108/240) of participants who received psychological intervention had a 50% or greater reduction in migraine frequency compared with 24% (53/219) of those who received a control intervention (RR = 1.82, 95% CI: 1.19, 2.80; p = 0.006), indicating that participants who received psychological treatment were more likely to be classified as responders. There was moderate heterogeneity for this outcome, I 2 = 53%. The NNTB based on these results is 4.81 (95% CI: 3.42, 8.11). The GRADE quality rating was very low.

Pain Intensity. Twelve studies (3, 6, 10, 18, 19, 20, 22, 30, 34, 35, 37, 38; n = 1105) reported on rates of pain intensity post-treatment. Psychological interventions had a moderate beneficial effect on reducing pain intensity for adults with migraine compared to control (d = 0.31, 95% CI: 0.11, 0.51; p = 0.002). This outcome had considerable heterogeneity, I 2 = 59%. The GRADE quality rating was very low.

Mood. Ten studies (5, 6, 9, 12, 13, 14, 17, 36, 38, 39; n = 733) reported on mood post-treatment. Psychological interventions demonstrated no beneficial effect on improving mood for adults with migraine compared to control conditions (d = 0.31, 95% CI: −0.01, 0.63; p = 0.058). There was substantial heterogeneity for this outcome, I 2 = 77%. The GRADE quality rating for mood post-treatment was very low.

Seven studies (5, 9, 12, 13, 14, 36, 39; n = 606) reported on mood at follow-up. Psychological interventions had no significant effect on improving mood for adults with migraine compared to control conditions at follow-up (d = 0.55, 95% CI: −0.10, 1.19, p = 0.098). There was considerable heterogeneity for this outcome, I 2 = 91%.

Pain-related Disability. Fifteen studies (5, 7, 9, 10, 11, 12, 13, 19, 25, 27, 28, 34, 35, 37, 39; n = 1663) reported on pain-related disability post-treatment. Psychological interventions had a significant small to medium beneficial effect on reducing disability, post-treatment, compared to control conditions (d = 0.33, 95% CI: 0.16, 0.49, p < 0.001). There was moderate heterogeneity for this outcome, I 2 = 59%. The GRADE quality rating was very low.

Four studies (5, 12, 13, 39; n = 383) reported on disability at follow-up. Psychological interventions had a significant medium effect on reducing disability at follow-up, compared to control conditions (d = 0.44, 95% CI: 0.09, 0.79, p = 0.014). There was moderate heterogeneity for this outcome, I 2 = 63%.

Quality-of-life. Ten studies (3, 7, 14, 19, 25, 28, 30, 36, 38, 39; n = 1008) reported on QoL post-treatment. Psychological interventions produced no difference on QoL outcomes, compared to control conditions (d = 0.23; 95% CI: −0.02, 0.47; p = 0.075). There was considerable heterogeneity for this outcome, I 2 = 69%. The GRADE quality rating was very low.

Medication use. Four studies (7, 12, 13, 36; n = 247) reported on medication use post-treatment. Psychological interventions had a significant small effect on reducing medication use post-treatment, compared to control conditions (d = 0.30; 95% CI: 0.05, 0.55; p = 0.021). There was no heterogeneity for this outcome, I 2 = 0%. The GRADE quality rating was very low.

Adverse Events and Follow-up Analyses. Too few studies (k < 4) reported data on adverse events, precluding analysis. Apart from mood and disability (as previously reported), there were too few studies (k < 4) that included follow-up data for calculation of overall effect sizes.

Sensitivity Analysis. To determine the stability of findings across studies that include populations with mixed headache conditions (e.g., migraine and TTH) to studies of migraine-only populations, we conducted sensitivity analyses, adding the eight studies that combined data from adults with migraine or TTH to the migraine-only studies. Note, as with the migraine-only studies, not all studies included data on all outcomes. We were able to analyze six post-treatment outcomes. No studies with migraine and TTH participants reported data on QoL, and few studies (k < 4) reported follow-up data for outcomes. See Table 4 for the results of the sensitivity analysis. GRADE ratings were not conducted on the sensitivity analyses.

Including studies with migraine and TTH participants, we continued to find a beneficial effect of psychological intervention on the following outcomes: migraine/headache frequency, treatment response, pain intensity, and pain-related disability. The magnitude of the effect sizes remained stable across analyses (largest difference between the effects was 0.06 for medication), as did the level of heterogeneity (largest I 2 change was <10%). With the addition of studies with mixed headache conditions, the effect size for mood became significant, showing a moderate beneficial effect of treatment on improving mood (d = 0.32).

Summary and Comparison of Results to Cochrane Review

The prior Cochrane Review[21] included data extracted from 14 studies, compared to the 31 migraine-only studies that contributed data to the main findings. The Cochrane Review showed evidence for a difference in the number of people with migraine who responded to treatment to those who responded in the control groups, with no other significant benefit demonstrated. Our current results showed small but significant improvements for those receiving psychological treatments compared to controls for a range of outcomes, including migraine frequency, pain intensity, disability, and medication use, but not mood or QoL. Despite these benefits, the quality of evidence in both reviews was judged to be very low.