Psychological therapies for the prevention of migraine in adults
Migraine is a common neurological problem associated with the highest burden amongst neurological conditions in terms of years lived with disability. Medications can be used as prophylaxis or rescue medicines, but are costly and not always effective. A range of psychological interventions have been developed to manage migraine.
The objective was to evaluate the efficacy and adverse events of psychological therapies for the prevention of migraine in adults.
We searched CENTRAL, MEDLINE, Embase, PsycINFO and CINAHL from their inception until July 2018, and trials registries in the UK, USA, Australia and New Zealand for randomised controlled trials of any psychological intervention for adults with migraine.
We included randomised controlled trials (RCTs) of a psychological therapy for people with chronic or episodic migraine, with or without aura. Interventions could be compared to another active treatment (psychological or medical), an attention‐placebo (e.g. supportive counselling) or other placebo, routine care, or waiting‐list control. We excluded studies where fewer than 15 participants completed each arm.
Data collection and analysis
We extracted study characteristics and outcome data at post‐treatment and the longest available follow‐up. We analysed intervention versus control comparisons for the primary outcome of migraine frequency. We measured migraine frequency using days with migraines or number of migraine attacks measured in the four weeks after treatment. In addition, we analysed the following secondary outcomes: responder rate (the proportion of participants with a 50% reduction in migraine frequency between the four weeks prior to and the four weeks after treatment); migraine intensity; migraine duration; migraine medication usage; mood; quality of life; migraine‐related disability; and proportion of participants reporting adverse events during the treatment. We included these variables, where available, at follow‐up, the timing of which varied between the studies. We used the GRADE approach to judge the quality of the evidence.
We found 21 RCTs including 2482 participants with migraine, and we extracted meta‐analytic data from 14 of these studies. The majority of studies recruited participants through advertisements, included participants with migraine according to the International Classification of Headache Disorders (ICHD) criteria and those with and without aura. Most intervention arms were a form of behavioural or cognitive‐behavioural therapy. The majority of comparator arms were no treatment, routine care or waiting list. Interventions varied from one 20‐minute session to 14 hours of intervention. No study had unequivocally low risk of bias; all had at least one domain at high risk of bias, and 20 had two to five domains at high risk. Reporting of randomisation procedures and allocation concealment were at high or unclear risk of bias. We downgraded the quality of evidence for outcomes to very low, due to very serious limitations in study quality and imprecision. Reporting in trials was poor; we found no preregistrations stipulating the outcomes, or demonstrating equivalent expectations between groups. Few studies reported our outcomes of interest, most only reported outcomes post treatment; follow‐up data were sparse.
We found no evidence of an effect of psychological interventions for migraine frequency in number of migraines or days with migraine (standardised mean difference (SMD) −0.02, 95% confidence interval (CI) −0.17 to 0.13; 4 studies, 681 participants; very low‐quality evidence).
The responder rate (proportion of participants with migraine frequency reduction of more than 50%) was greater for those who received a psychological intervention compared to control: 101/186 participants (54%) with psychological therapy; 37/152 participants (24%) with control (risk ratio (RR) 2.21, 95% CI 1.63 to 2.98; 4 studies, 338 participants; very low‐quality evidence). We found no effect of psychological therapies on migraine intensity (SMD −0.13, 95% CI −0.28 to 0.02; 4 studies, 685 participants). There were no data for migraine duration (hours of migraine per day). There was no effect on migraine medication usage (SMD −0.06, 95% CI −0.35 to 0.24; 2 studies, 483 participants), mood (mean difference (MD) 0.08, 95% CI −0.33 to 0.49; 4 studies, 432 participants), quality of life (SMD −0.02, 95% CI −0.30 to 0.26; 4 studies, 565 participants), or migraine‐related disability (SMD −0.67, 95% CI −1.34 to 0.00; 6 studies, 952 participants). The proportion of participants reporting adverse events did not differ between those receiving psychological treatment (9/107; 8%) and control (30/101; 30%) (RR 0.16, 95% CI 0.00 to 7.85; 2 studies, 208 participants). Only two studies reported adverse events and so we were unable to draw any conclusions.
We rated evidence from all studies as very low quality.
Only four studies reported any follow‐up data. Follow‐ups ranged from four months following intervention to 11 months following intervention. There was no evidence of an effect on any outcomes at follow‐up (very low‐quality evidence).
This review identified 21 studies of psychological interventions for the management of migraine. We did not find evidence that psychological interventions affected migraine frequency, a result based on four studies of primarily brief treatments. Those who received psychological interventions were twice as likely to be classified as responders in the short term, but this was based on very low‐quality evidence and there was no evidence of an effect of psychological intervention compared to control at follow‐up. There was no evidence of an effect of psychological interventions on medication usage, mood, migraine‐related disability or quality of life. There was no evidence of an effect of psychological interventions on migraine frequency in the short‐term or long‐term. In terms of adverse events, we were unable to draw conclusions as there was insufficient evidence. High and unclear risk of bias in study design and reporting, small numbers of participants, performance and detection bias meant that we rated all evidence as very low quality. Therefore, we conclude that there is an absence of high‐quality evidence to determine whether psychological interventions are effective in managing migraine in adults and we are uncertain whether there is any difference between psychological therapies and controls.
Louise Sharpe, Joanne Dudeney, Amanda C de C Williams, Michael Nicholas, Ingrid McPhee, Andrew Baillie, Miriam Welgampola, Brian McGuire
Plain language summary
Psychological therapies for the management of migraine in adults
There was an absence of good‐quality evidence that psychological therapy was effective or harmful in managing frequent migraine immediately following treatment or in the longer term.
Migraine is a condition of the nervous system that is common and associated with lower quality of life and disability. Although medications can help manage migraine, they do not work for all individuals and some individuals experience negative side‐effects (adverse events). Numerous psychological therapies have been evaluated for the management of migraine in adults. Psychological therapies deliver skills such as education, relaxation, or coping strategies to help adults change their behaviour or thoughts about migraine, to try to reduce their migraine‐related symptoms.
We evaluated psychological interventions for adults with chronic or episodic migraine with and without aura (a warning sign that precedes and predicts a migraine). We compared individuals who received psychological therapy for migraine with a 'control' group. Control groups included usual treatment ('standard care'), or waiting to receive treatment, or receiving another type of intervention such as education. We extracted data on the frequency of migraines (i.e. number of days with migraines, or number of migraines, in the month following treatment) as our primary outcome. We also extracted data on the number of responders (people with a 50% reduction in migraine frequency), migraine intensity, migraine duration (number of hours of migraine per day), migraine medication usage, mood, quality of life, and migraine‐related disability. We recorded instances of harm (adverse events) associated with treatment.
We searched databases in July 2018 and found 21 studies with 2482 participants. Most studies investigated one of three interventions, namely a form of psychological therapy called cognitive‐behaviour therapy (CBT), which teaches skills to change thoughts and behaviours. Skills include coping strategies, or biofeedback or relaxation, which teaches people to reduce their tension either by concentrating on relaxing exercises or through a machine that gives feedback about muscle tension or body temperature. The remaining psychological treatments were examined in single studies; they included writing about emotions and eye movement desensitisation, and reprocessing, which uses eye movements to help people accept their pain and other negative experiences. We were interested in outcomes following treatment and at the longest available follow‐up.
We found no evidence that psychological treatments resulted in less migraine frequency in the four weeks following treatment. However, we could only include four studies in this analysis that were not high quality. Four studies reported the proportion of people whose migraines reduced in frequency by 50% or more, and in those studies, people who received psychological treatment were twice as likely to respond to treatment (i.e. 50% reduction in migraine frequency) as those in the control group.
There was no evidence that psychological treatments affected migraine intensity, medication use for migraine, mood or quality of life. Only two studies assessed adverse events, and so we were unable to draw conclusions.
We found very few follow‐up data, and no evidence to support or refute any long‐term effects of psychological treatment.
Quality of evidence
We rated the quality of the evidence using four levels: very low, low, moderate, or high. High‐quality evidence means that we are very confident in the results. Very low‐quality evidence means that we are very uncertain about the results. We judged the quality of evidence as very low.
There is no evidence that psychological treatments affect the frequency of migraine. More responders (i.e. those reporting a 50% reduction) received psychological treatment than control, but this was based on very low‐quality evidence and therefore we are uncertain of this result. In terms of adverse events, we were unable to draw conclusions as there was insufficient evidence. There were very few long‐term data available, and no indication that psychological interventions had any long‐term effects. Overall there was an absence of high‐quality evidence for the effect of psychological treatment on migraines and therefore we are uncertain whether there is any difference between psychological therapies and controls. Funding of high‐quality studies is needed and additional studies may change the conclusions of this review.
Louise Sharpe, Joanne Dudeney, Amanda C de C Williams, Michael Nicholas, Ingrid McPhee, Andrew Baillie, Miriam Welgampola, Brian McGuire
Implications for practice
There is an absence of high‐quality evidence to support or refute the efficacy of psychological treatments in the management of migraine. We did find that compared to mostly inactive controls, participants who received a psychological intervention were more likely to have a 50% or greater reduction in migraine frequency. Our results showed that twice as many participants who received a psychological therapy reported a 50% reduction in migraine frequency compared to control, such that participants receiving psychological interventions were more likely to report a reduction in frequency following intervention (540/1000 in the psychological therapy group compared to 240/1000 in the control group; RR 2.21). It is important to note that these results are based on poor‐quality data, and therefore we cannot be confident in the effects reported here. They are also at odds with the results on our primary outcome of migraine frequency. Nevertheless, it has been noted above, that three out of four trials in the migraine frequency analysis were minimal interventions (i.e. self‐directed, home‐based or internet‐delivered interventions) with minimal input from the therapists involved. In contrast, the majority of studies in the responder analysis were face to face, and related to multimodal behavioural interventions. Therefore the differing results could relate to the type of psychological therapy. Unfortunately, there were too few studies to test this possibility. It is also possible that while psychological interventions do not reduce migraine frequency across all participants, but a subgroup of participants respond well. From the analyses presented here approximately only three participants needed to be treated for one to benefit, which is comparable to commonly prescribed medications (e.g. topiramate (Linde 2013)), and compares favourably to others (e.g. gabapentin or pregabalin (Linde 2013b); other antiepileptics (Linde 2013a)). We urge caution, however, because the quality of the evidence was very low according to GRADE, and as a result, we have little confidence in the estimate of effect. From the two available studies, we were unable to draw conclusions as there was insufficient evidence.
For adults with migraine
We found an absence of high‐quality evidence for whether psychological interventions have a role in the management of migraine and therefore we are uncertain whether there is any difference between psychological therapies and controls. People who received a psychological intervention did not overall have a different migraine frequency from those who received a control condition (like a waiting list) although more people who received a psychological intervention responded to treatment than in the control group (i.e. had a 50% or greater reduction in migraine frequency). According to our analysis of number needed to treat for an additional beneficial outcome, one in three participants responded. Therefore, psychological therapy could be an alternative for some people. However, the evidence base was very low quality and we are not very confident in these estimates.
Given the lack of an effect of psychological therapies on migraine frequency, and the lack of an effect of intervention on long‐term outcomes, clinicians should proceed cautiously.
For policy makers
Given the absence of high‐quality evidence to support the efficacy of psychological treatments in the management of migraine, policy implications revolve largely around the need for high‐quality evidence on which to base future policy decisions.
For funders of the intervention
The absence of high‐quality evidence for psychological interventions for managing migraine means that this review has few implications for funders of the intervention in routine care. However, this review strongly suggests that there is a need to fund high‐quality randomised controlled trials (RCTs) of psychological interventions with appropriate follow‐up to improve the evidence base for psychological therapies for managing migraine.
Implications for research
There is a strong need for high‐quality RCTs of psychological interventions for the management of migraine. Despite a long history of psychological intervention for migraine, the evidence is very low quality and, as a result, estimates of the effect of psychological treatments are uncertain, due to the high risk of bias identified in the majority of trials and due to very serious study limitations and imprecision. The International Headache Society (IHS) position statement on the design of clinical trials for chronic migraine guides researchers to ensure that future research overcomes the problems we identified in this review (Tassorelli 2018). The specific implications for research design, measurement, and transparency of research data are listed below.
One difficulty in conducting this review was the number of studies in which authors only reported group aggregate data for participants with migraine and tension‐type headache, where the data were combined within the study. There are no available data to assess whether adults with migraine and tension‐type headache respond similarly or differently. Therefore it is important that when interventions are developed for the management of adults with migraine and tension‐type headache, data should be reported for each of the headache types separately. Tassorelli 2018 outlines the way in which individuals with types of headache other than migraine (e.g. tension‐type headache, medication overuse headache) should be treated in clinical trials. For example, they recommend that adults with migraine can be included if they also have tension‐type headache, as long as the person is able to distinguish the two types of headache clearly. Similarly they recommend that, if adults with medication overuse headache are included in studies, they are stratified as part of the randomisation procedure.
The lack of preregistered studies was a limitation. From 1 July 2005, registration of studies of interventions has been mandatory, as part of the CONSORT process (Schulz 2010). We were unable to locate most study registrations and the vast majority of studies did not refer to a trial registry. This is likely because many studies were published before 2005. In addition, few studies nominated a primary outcome and there was little consensus about how to measure outcomes across studies (see measurement issues below). We rated the majority of studies at high risk of bias for performance bias, which is generally a problem in psychological trials where it is difficult to blind participants to conditions. It is possible to include attention placebo conditions and measure participants' treatment expectancy to try and ensure that there is a low risk of performance bias.
Measurement (end points)
There is a need for more conformity in the literature in terms of the outcomes that are measured. If psychological interventions are to be viewed as a primary intervention for the management of migraine, then they should be judged by the same criteria as other interventions (such as medications). There needs to be standardisation about which outcome measures are important. The results of this meta‐analysis indicate that the field needs data on migraine intensity as none currently exists. This has been achieved in the rheumatology area with OMERACT, which is an initiative that organises consensus conferences biannually to agree on important outcomes that should be included in clinical trials (see Tugwell 2007 for a description of the initiative and its history), and through IMMPACT in chronic pain (Turk 2003). In the headache literature, Tassorelli 2018 has attempted to outline appropriate measures and time points for clinical trials for interventions to manage migraine. They make the following recommendations:
- baseline recording periods for adults with migraine should be at least 28 days of prospective recording, preferably electronically, where time stamps can confirm that participants are recording their headache features prospectively rather than retrospectively;
- the headache diary should include information on migraine‐associated symptoms, migraine medication usage, duration (defined as how long the migraine persists), severity/intensity, whether an aura was present, and impact on daily functioning;
- primary outcomes and primary end points (e.g. end of treatment, follow‐up) should be preregistered and prospectively defined. Primary outcomes should be either: change in number of headache days per month (and based on at least 28 days' monitoring), or change in number of headache days with moderate to severe migraine, or the proportion of participants whose migraine frequency reduces by 50% or more. In addition to the inclusion of each of these outcomes, they recommend the following secondary outcomes: migraine intensity, cumulative hours per month where migraine intensity is moderate to severe, conversion from chronic to episodic migraine (for chronic migraine), migraine medication usage, conversion of medication overuse to appropriate use (where relevant), depression, anxiety, functional impairment, and general impression of improvement;
- it is also important to record adverse events.
Increasingly, it is being suggested that data should be made publicly available. Had we been able to obtain outcome data from each trial, this meta‐analysis would have reported more comparisons and included additional data. This is particularly the case with headache diary data, where researchers captured but did not report means and standard deviations for key outcome data. Researchers should avail themselves of public repositories for their data so that they can be easily accessible and contribute to the evidence base.
It would be helpful for manuals or detailed summaries of interventions to be made available so that the content of interventions is clear. This would allow for work to be replicated. With the increase in publication of journals online, manuals or intervention summaries could be published as supplementary material, available through the journal's website.
There were some interventions that have been used in the literature, but no data were available for analysis in this review, so we were unable to draw conclusions about their effects. These included mindfulness and eye movement desensitisation and reprocessing. Future research should investigate novel interventions and compare them to other active treatments with some evidence of efficacy, such as behavioural approaches.