Behavioral interventions for migraine prevention: A systematic review and meta‐analysis

Jonathan R Treadwell; Amy Y Tsou; Benjamin Rouse; Ilya Ivlev; Julie Fricke; Dawn C Buse; Scott W Powers; Mia Minen; Christina L Szperka; Nikhil K Mull

doi:10.1111/head.14914

. 2025 Feb 19;65(4):668–694. doi: 10.1111/head.14914

Behavioral interventions for migraine prevention: A systematic review and meta‐analysis

Jonathan R Treadwell ^1,^✉, Amy Y Tsou ¹, Benjamin Rouse ¹, Ilya Ivlev ^1,², Julie Fricke ³, Dawn C Buse ⁴, Scott W Powers ^5,^6,⁷, Mia Minen ^8,⁹, Christina L Szperka ¹⁰, Nikhil K Mull ^3,¹¹

PMCID: PMC11951403 PMID: 39968795

Abstract

Objectives/Background

This study was undertaken to synthesize evidence on the benefits and harms of behavioral interventions for migraine prevention in children and adults. The efficacy and safety of behavioral interventions for migraine prevention have not been tested in recent systematic reviews.

Methods

An expert panel including clinical psychologists, neurologists, primary care physicians, researchers, funders, individuals with migraine, and their caregivers informed the scope and methods. We searched MEDLINE, Embase, PsycINFO, PubMed, the Cochrane Database of Systematic Reviews, clinicaltrials.gov, and gray literature for English‐language randomized trials (January 1, 1975 to August 24, 2023) of behavioral interventions for preventing migraine attacks. Primary outcomes were migraine/headache frequency, migraine disability, and migraine‐related quality of life. One reviewer extracted data and rated the risk of bias, and a second verified data for completeness and accuracy. Data were synthesized with meta‐analysis when deemed appropriate, and we rated the strength of evidence (SOE) using established methods.

Results

For adults, we included 50 trials (77 publications, N = 6024 adults). Most interventions were multicomponent (e.g., cognitive behavioral therapy [CBT], biofeedback, relaxation training, mindfulness‐based therapies, and/or education). Most trials were at high risk of bias, primarily due to possible measurement bias and incomplete data. For adults, we found that any of three components (CBT, relaxation training, mindfulness‐based therapies) may reduce migraine/headache attack frequency (SOE: low). Education alone that targets behavior may improve migraine‐related disability (SOE: low). For three other interventions (biofeedback, acceptance and commitment therapy, and hypnotherapy), evidence was insufficient to permit conclusions. We also found that mindfulness‐based therapies may reduce migraine disability more than education, and relaxation + education may improve migraine‐related quality of life more than propranolol (SOE: low). For children/adolescents, we included 13 trials (16 publications, N = 1444 children), but the evidence was only sufficient to conclude that CBT + biofeedback + relaxation training may reduce migraine attack frequency and disability more than education alone (SOE: low).

Conclusion

Results suggest that for adults, CBT, relaxation training, and mindfulness‐based therapies may each reduce the frequency of migraine/headache attacks, and education alone may reduce disability. For children/adolescents, CBT + biofeedback + relaxation training may reduce migraine attack frequency and disability more than education alone. Evidence consisted primarily of underpowered trials of multicomponent interventions compared with various types of control groups. Limitations include semantic inconsistencies in the literature since 1975, differential usage of treatment components, expectation effects for subjectively reported outcomes, incomplete data, and unclear dosing effects. Future research should enroll children and adolescents, standardize intervention components when possible to improve reproducibility, consider smart study designs and personalized therapies based on individual characteristics, use comparison groups that control for expectation, which is a known challenge in behavioral trials, enroll and retain larger samples, study emerging digital and telehealth modes of care delivery, improve the completeness of data collection, and establish or update clinical trial conduct and reporting guidelines that are appropriate for the conduct of studies of behavioral therapies.

Keywords: cognitive–behavioral therapy, headache, migraine, mindfulness, relaxation

Plain Language Summary

Behavioral interventions for migraine prevention are an important alternative to medications for both children and adults. To optimize decision‐making by patients and providers, we conducted a comprehensive systematic review and meta‐analysis of 63 randomized trials published since 1978. Due to methodological differences between studies, incomplete data, and unclear dosing effects, the quality of the reviewed evidence was insufficient to permit conclusions; however, some behavioral treatments appear to offer benefits in migraine reduction, and additional research on these interventions in adult and pediatric populations is needed.

Abbreviations

ACT: acceptance and commitment therapy
AHRQ: Agency for Healthcare Research and Quality
CBT: cognitive behavioral therapy
CI: confidence interval
HIT‐6: Headache Impact Test‐6
MBCT: mindfulness‐based cognitive therapy
MBSR: mindfulness‐based stress reduction
MID: minimal important difference
MIDAS: Migraine Disability Assessment
MSQ: Migraine‐Specific Quality of Life Questionnaire v2.1
PCORI: Patient‐Centered Outcomes Research Institute
QoL: quality of life
SMD: standardized mean difference
SOE: strength of evidence
TEP: Technical Expert Panel

INTRODUCTION

One in six Americans has migraine, a chronic disease that globally stands as the second leading cause of disability. Migraine often impacts individuals during crucial periods of their lives, including years of education, career progression, and child‐rearing. ¹ Preventative migraine interventions aim to decrease the frequency, severity, and negative life impact of migraine attacks. Behavioral interventions may be used for migraine prevention as standalone therapies or adjunctive treatments to pharmacologic or other nonpharmacologic interventions. Behavioral therapies may be particularly attractive for individuals seeking to avoid pharmacologic therapies due to side effects or personal preference.

The latest adult clinical practice guidelines from 2012 overlook behavioral therapies that have been more recently applied to migraine, such as mindfulness‐based therapies (mindfulness‐based cognitive therapy [MBCT] and mindfulness‐based stress reduction [MBSR]), ² and the pediatric headache preventive guidelines were limited to studies that included pharmacologic agents. ³ Consensus statements from the American Headache Society and American Academy of Family Physicians on nonpharmacologic migraine prevention were not based on earlier systematic evidence reviews and did not address the specific treatment needs of children and adolescents. ⁴ , ⁵ , ⁶ , ⁷ Recent literature reviews indicate that existing systematic reviews on behavioral therapies may need to be updated. ⁸ The most recent reviews, published in 2018 and 2019, examined biofeedback, CBT, and progressive muscle relaxation (a form of relaxation training) for adult migraine prevention, but did not evaluate all preventive behavioral therapy options. ⁹ , ¹⁰ , ¹¹ Two recent reviews focused on pain prevention in children and adolescents broadly, without specific emphasis on migraine. ¹² , ¹³

To address these evidence gaps and support the potential development of an evidence‐based clinical practice guideline, we performed a systematic review. We defined behavioral interventions as nonpharmacologic strategies intended to modify behavior and/or ways of thinking for adults, adolescents, and children with migraine. This systematic review addressed the (1) effectiveness, (2) comparative effectiveness, (3) contribution of specific behavioral components, (4) effectiveness of non‐migraine‐focused interventions, (5) associations with patient factors, and (6) delivery via telehealth. This article only summarizes the first three areas listed above; additional findings are available in the full report (https://effectivehealthcare.ahrq.gov/products/behavioral‐interventions‐migraine‐prevention/research).

METHODS

We followed methods outlined in the Agency for Healthcare Research and Quality (AHRQ) Methods Guide for Effectiveness and Comparative Effectiveness Reviews. ¹⁴ We recruited six key informants to refine the topic and provide input on the scope. A seven‐member Technical Expert Panel (TEP) provided input on the protocol. The key informants and TEP included clinical psychologists, adult and pediatric neurologists, primary care physicians, researchers, funders, pediatric and adult patients with migraine, and their caregiver representatives. The protocol was posted online for public comment (www.effectivehealthcare.ahrq.gov) from July 19, 2022, to August 9, 2022, and was registered on PROSPERO (CRD42023397752). The draft systematic review was reviewed by AHRQ, Patient‐Centered Outcomes Research Institute (PCORI), four members of the TEP, and five additional solicited peer reviewers. A revised draft was posted for public comment on the AHRQ Effective Healthcare website (September 5, 2023 to October 20, 2023) that received comments from two people and the American Psychological Association.

Literature search

A research librarian searched MEDLINE and Embase (via Embase.com), PsycINFO (via Ovid), PubMed (in process citations, to capture items not yet indexed in MEDLINE), and the Cochrane Database of Systematic Reviews for randomized controlled trials, systematic reviews, and meta‐analyses published from 1975 to August 24, 2023. A second librarian independently peer‐reviewed the search strategy using the PRESS Checklist. We hand‐searched reference lists of relevant systematic reviews to identify additional trials. File S1 contains the search strategies.

Trial selection

Our eligibility criteria appear in Table 1. The review team was comprised of experienced systematic reviewers and employed dual screening of both abstracts and full‐text articles (in DistillerSR), with discrepancies resolved by discussion between those disagreeing (we did not compute interrater reliability). Trials had to report data on at least one of three primary outcomes (attack frequency, migraine disability, or migraine‐specific quality of life [QoL]). This article describes only these three outcomes; see the full report for other outcomes (https://effectivehealthcare.ahrq.gov/products/behavioral‐interventions‐migraine‐prevention/research). If a trial reported multiple metrics of migraine/headache attack frequency (e.g., both migraine days per month and migraine attacks per month), we selected an outcome using the following order of priority: (1) migraine days per month, (2) headache days per month, (3) migraine attacks per month, (4) headache attacks per month, and (5) percentage of patients experiencing a 50% or more reduction in one of the metrics above. Due to resource constraints, we did not attempt to contact study authors for missing data or methodology information.

TABLE 1.

Trial eligibility criteria.

Aspect	Inclusion	Exclusion
Patients	Children (aged 6–11 years), adolescents (aged 12–17 years), and adults (18 years or older) with migraine headache (episodic or chronic)	Trials conducted exclusively:
	We did not require trials to include only individuals with an International Classification of Headache Disorders diagnosis of migraine headache	Among individuals in institutions (e.g., psychiatric inpatients, long‐term care facilities, incarcerated populations)
	≥80% of trial participants had migraine headache, or the trial reports a subgroup analysis composed of at least 80% patients with migraine	Parents, for trials with interventions targeting children and adolescents
	For trials with participants with other headache types (e.g., medication‐overuse headache, tension‐type headache, cluster headache) in addition to migraine, we included the trial if at least 80% of participants had migraine	Individuals with psychotic disorders
Interventions	Migraine‐focused behavioral interventions used for prevention, administered either alone or with pharmacotherapy, delivered in‐person, via telehealth, or with e‐ or mHealth	Trials focused solely on:
	1. CBT	Physical therapy
	CBT	Exercise
	Cognitive therapy	Catharsis therapy (e.g., written emotional disclosure)
	MBCT ^a	Occupational therapy
	Behavioral therapy	Creative arts therapy (art therapy, music therapy, dance therapy)
	SMT	Massage
	Coping skills training
	LCT
	Parent/caregiver operant training (parent or caregiver reinforces coping behaviors)
	Problem‐solving training
	2. Biofeedback
	Thermal/temperature biofeedback (hand warming/thermal biofeedback; often feedback of skin temperature from finger)
	Electromyographic biofeedback (feedback of electrical activity from muscles of scalp, neck, or upper body)
	Heart rate variability biofeedback
	Electrocardiographic biofeedback
	Pulse
	Blood volume pulse
	Respiratory
	Electroencephalography/neurofeedback
	3. Relaxation training
	Diaphragmatic breathing
	Progressive muscle relaxation (alternatively tensing/relaxing selected muscles)
	Autogenic feedback (use of calm, self‐soothing statements to promote a state of deep relaxation)
	Autogenic training
	Guided imagery/guided visual imagery (children/adolescents)
	4. Mindfulness‐based stress reduction
	Meditation (use of silently repeated word or sound to promote mental calm and relaxation)
	MBCT ^a
	Transcendental meditation
	Guided imagery/guided visual imagery (adults)
	5. Acceptance and commitment therapy
	6. Education
	Education (skills, lifestyle, exercise, nutrition, hydration, stress management, sleep hygiene)
	Neuroscience education therapy
	Healthy lifestyle counseling
	Sleep counseling
	Trigger avoidance
	Weight management (informational)
	Diary/tracking
	7. Hypnotherapy
	8. Trauma‐informed therapy
	EMDR
	Trauma‐focused therapy
	9. DBT
	10. Motivational interviewing and stages of change
	11. Professionally led support groups/peer support
	12. Combination therapies
	Non‐headache‐focused behavioral interventions
	CBT for insomnia or depression/anxiety
	Sleep hygiene counseling
	Parent/caregiver operant training (parent or caregiver reinforces adaptive sleep behaviors)
	• Healthy lifestyle counseling
Comparisons	Effectiveness	Comparators not listed as included
	No intervention (e.g., waitlist, usual care)
	Minimal intervention (e.g., educational materials without skills training)
	Most active: attention control, sham, or placebo
	Comparative effectiveness
	A different eligible behavioral intervention, or a pharmacological intervention in one of the following class:
	Alpha agonists
	Angiotensin‐converting enzyme inhibitors/angiotensin receptor blockers
	Anti‐seizure medications
	Antihistamines (for children and adolescents only)
	Beta‐blockers
	onabotulinumtoxinA
	Calcitonin gene‐related peptide antagonists
	Calcium channel blockers
	Other antidepressants
	Serotonin norepinephrine reuptake inhibitors
	Tricyclic antidepressants
	Component contribution
	The trial design permitted the isolation of the contribution of a single behavioral component (e.g., A + B vs. A measures the contribution of B)
Outcomes	Trial must have reported one or more of three primary outcomes	Some other outcomes were included in the full report, but are not discussed in this article
	Migraine/headache attack frequency
	Migraine/headache count: migraine days per month, migraine attacks per month, headache days per month, or headaches per month
	Responder rate: 50% or more reduction in one of the above quantities
	Functional status/disability
	MIDAS, PedMIDAS, HIT‐6, HANA, MIBS, FIS, FDI (parent form), FDI (child and adolescent), IMPAC, PDI
	QoL
	Migraine‐specific: MSQ
Trial design criteria	Randomized controlled trials reporting outcomes for ≥10 participants per treatment arm	Excluded crossover trials not reporting period 1 data separately
	Period 1 data from crossover RCTs	Excluded reviews, letters, guidelines, position statements, and commentaries
	Published in English language	Excluded single arm or nonrandomized controlled trials
	Published 1975 or after	Unpublished trials/not published as a full‐length article (e.g., conference abstract)
	Subgroup analyses addressing patient factors must have reported outcomes on at least 10 patients per subgroup	SRs were used only to identify potential RCTs for inclusion
Setting	Any noninpatient setting	Hospitalized patients
Setting	Trials conducted in countries rated as “very high” on the 2022 Human Development Index (as defined by the United Nations Development Program). This was to focus our efforts on treatment settings relatively similar to US settings	Trials conducted in other countries
Timing	Trials must have reported 1 of our primary outcomes at 4 weeks or longer after treatment initiation	Earlier timepoints

Open in a new tab

Abbreviations: CBT, cognitive behavioral therapy; DBT, dialectical behavioral therapy; EMDR, eye movement desensitization and reprocessing; FDI, Functional Disability Inventory; FIS, Fatigue Impact Scale; HANA, Headache Needs Assessment; HIT‐6, Headache Impact Test‐6; LCT, learning to cope with triggers; MBCT, mindfulness‐based cognitive therapy; MIBS, Migraine Interictal Burden Scale; MIDAS, Migraine Disability Assessment; MSQ, Migraine‐Specific Quality of Life Questionnaire v2.1; PDI, Pain Disability Inventory; PedMIDAS, Pediatric Migraine‐Specific Disability Assessment; QoL, quality of life; RCT, randomized controlled trial; SMT, stress management training; SR, systematic review.

^{^a}

MBCT was categorized as a combination of cognitive therapy and MBSR, so it appears under two categories.

Intervention categorization

Preventative behavioral interventions can have one component or multiple components. To categorize interventions, first, we considered trial descriptions of interventions, irrespective of how trial authors named their intervention(s). Second, any intervention mentioned in the trial was considered part of the intervention, regardless of duration, delivery intensity, and setting/modality (i.e., self‐guided or in‐clinic). Thus, a trial providing written relaxation technique materials to participants and a trial conducting in‐person weekly relaxation sessions were both considered as relaxation training. Two analysts and four migraine experts reviewed all categorizations to ensure reliability.

We classified education‐only comparison groups in trials as attention controls if: (1) authors indicated their designation as such, (2) the allocated time and interaction with trial staff were comparable to the intervention arm, and (3) the educational component lacked additional behavioral interventions (e.g., relaxation training, cognitive behavioral therapy [CBT]). Interventions not meeting these criteria but still functioning as a minimal intervention control were labeled as minimal intervention. We classified education as a behavioral intervention when aimed to modify behavior or ways of thinking (as per our definition of behavioral interventions).

For analysis, many trials employed multicomponent treatment packages, with different trials often using different combinations. To assess effectiveness, we grouped trials based on the presence of a common component, disregarding the variety or dosage of additional elements within the treatment regimen. For example, the CBT section discusses data on the three‐component intervention by Klan et al. ¹⁵ (CBT + progressive muscle relaxation + education) as well as the six‐component intervention by Lemstra et al. ¹⁶ (CBT + relaxation training + education + exercise + massage therapy + physical therapy). This is because both interventions had a CBT component. These same two studies were also analyzed for relaxation training effectiveness, because both included relaxation training.

Data analysis

We conducted random‐effects meta‐analyses using the restricted maximum‐likelihood approach and the inverse‐variance weighting scheme using the meta package (version 6.2‐1) in R version 4.1.3 (R Foundation for Statistical Computing, Vienna, Austria). ¹⁷ Wherever possible, we computed effect sizes and 95% confidence intervals (CIs) using standard methods. We measured heterogeneity using τ ², complemented by I ². Where necessary, we estimated mean and standard deviation from median and interquartile range, or range using the methods described by Wan et al. ¹⁸ Because trials often used different metrics for the same construct, we conducted meta‐analyses of standardized mean differences (SMDs) using the Hedges g approach. If a trial reported multiple metrics of migraine disability or migraine QoL, we chose for meta‐analysis the metric that was more commonly reported among our included trials (although we extracted all such data). To facilitate interpretation, we used typical standard deviations to convert summary SMDs to a more commonly understood metric (e.g., Pediatric Migraine‐Specific Disability Assessment for migraine disability in youth). We did not conduct sensitivity analyses. Data were extracted by a single team member.

We defined time points from the beginning of the intervention. If a trial reported change from baseline data in addition to at follow‐up data, we prioritized the former. Some trials reported multiple follow‐up time points (and we extracted them all), but for meta‐analyses, we chose the one closest to 12 weeks after the start of treatment, as this was the most commonly reported time point. Our meta‐analyses did not account for different studies having different durations of interventions.

We considered a two‐tailed p‐value < 0.050 to represent statistical significance. To assess the clinical importance of findings, we consulted our TEP and the published literature to determine a minimal important difference (MID) for each primary outcome. For migraine/headache attack frequency, the MID was set at 1 migraine day/month. For disability, the MID was 3 points on the Migraine Disability Assessment (MIDAS; which ranges 0–90). We identified no empirical literature on the MID for MIDAS, and we chose 3 points because it represents approximately 3% of the scale range, similar to our MID for attack frequency. For migraine‐specific QoL, the MID was 19 points on the 0–100 Migraine‐Specific Quality of Life v2.1 (based on a trial by Cole et al.). ¹⁹

Evidence rating

We evaluated risk of bias in individual trials as well as the strength of evidence (SOE) for each body of evidence related to a specific comparison and outcome. To assess risk of bias, we employed the revised Cochrane risk of bias tool for randomized trials. ²⁰ File S2 contains our study‐by‐study risk of bias ratings.

For assessing SOE, we adhered to the 2013 AHRQ Methods Guide, ²¹ which considers multiple inputs (trial design, risk of bias, consistency of results across trials, directness of the evidence, effect estimate precision, reporting bias, strong dose–response association, large magnitude of effect, and whether controlling for all plausible confounders would increase the effect). The SOE rating was high, moderate, low, or insufficient. This rating is made separately for each outcome of each comparison. A rating of insufficient is given if the evidence does not permit a conclusion for that outcome (e.g., the CI was wide enough to include both favors A and favors B). Two analysts rated the SOE independently, with discrepancies resolved by consensus.

RESULTS

Evidence base

For the full report, searches identified 1791 potentially relevant references (Figure 1). Among these, 127 full‐text articles were excluded, predominantly due to reasons such as not meeting population criteria, not evaluating a comparison of interest, or not assessing a key outcome of interest. For a list of all articles excluded at full text, see File S3.

Trial flow diagram. This figure shows the counts and various stages of our article screening process. Searches identified 1791 potentially relevant references, of which 106 were duplicates. An additional 61 potentially relevant references were identified from the reference lists of relevant systematic reviews. We excluded 1526 citations at the abstract level and ordered the remaining 220 for full‐text consideration. Of these, we excluded 127 studies, with the most common reasons for exclusion being “Does not meet population criteria,” “Does not evaluate a comparison of interest,” and “Does not evaluate a key outcome of interest.” As a result, we included 63 studies in 93 publications.

For adults, we included 50 randomized trials (published from 1978 to 2023) assessing behavioral interventions (see Table 2). ¹⁵ , ¹⁶ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² , ⁵³ , ⁵⁴ , ⁵⁵ , ⁵⁶ , ⁵⁸ , ⁵⁹ , ⁶⁰ , ⁶¹ , ⁶² , ⁶⁴ , ⁶⁵ , ⁶⁶ , ⁶⁷ , ⁶⁹ , ⁷⁰ , ⁷¹ , ⁷² Collectively, these trials enrolled 6024 adults; 36 trials addressed effectiveness, 16 comparative effectiveness, and three addressed the contribution of a specific behavioral component. Some trials addressed multiple areas. Twenty‐four trials were conducted in the United States, five in the Netherlands, four in Germany, three in Canada, three in Italy, two in Sweden, two in the United Kingdom, and one each in six other countries (one trial did not report the country in which patients were enrolled). The median number of patients at baseline was 75 (interquartile range = 37–111), the weighted average patient age was 42 years, and 84% of participants were women. Fourteen US trials reported race or ethnicity, with a median 73% White, 14% Black, 4% Asian/Pacific Islander, 8% Hispanic, and 1% Native American or Alaska Native. The baseline number of days per month that patients experienced migraine or headache attacks per month (reported by 25 trials) ranged from 4 to 25 (median = 10). Few studies reported on comorbid anxiety and depression; 47% of patients had comorbid anxiety (four trials), and 46% had comorbid depression (seven trials). In two thirds of trials (33 trials), behavioral interventions employed multicomponent interventions, whereas 34% (17 trials) assessed single behavioral interventions. The trial duration (time from treatment initiation to last follow‐up) ranged from 1 to 37 months (median = 6 months).

TABLE 2.

Included trials of adults on effectiveness, comparative effectiveness, or component contributions.

Trial	Baseline N	Episodic, chronic, or mixed	Behavioral treatment(s)	Control group(s)	Effectiveness	Comparative effectiveness	Components contribution
Aguirrezabal (2019) ²²	116	NR	Education	TAU or no intervention	✓
Blanchard (1978) ²³	37	NR	Relaxation training (PMR) Thermal biofeedback + relaxation training (autogenic training)	TAU or no intervention	✓	✓
Bromberg (2012) ²⁴	189	100% chronic	CBT + biofeedback + relaxation training + education	TAU or no intervention	✓
Brown (1984) ²⁵	39	NR	MBSR via guided imagery, relaxing statements MBSR via guided imagery, scene details	Attention control	✓	✓
Cousins (2015) ²⁶	73	Mixed (% NR)	CBT + relaxation training (PMR + deep breathing)	TAU or no intervention	✓
Cuneo (2023) ²⁷	36	100% chronic	Biofeedback	TAU or no intervention	✓
D'Souza (2008) ²⁸	90	NR	Relaxation training (autogenic training + deep breathing)	Attention control Written emotional disclosure (not an included treatment)	✓
Day (2014) ²⁹	36	NR	MBCT	TAU or no intervention	✓
de Tommaso (2017) ³⁰	33	100% episodic	Nociceptive blink reflex biofeedback Nociceptive blink reflex biofeedback + topiramate	Topiramate		✓
Dindo (2020) ³¹	103	100% episodic	ACT therapy + education Relaxation training + education + deep breathing	None		✓
Dittrich (2008) ³²	30	NR	Relaxation training (PMR) + exercise	TAU or no intervention	✓
Fritsche (2010) ³³	40	100% episodic	CBT + relaxation training (PMR) + education PMR + education	None			✓
Flynn (2019) ³⁴	150	NR	Hypnotherapy	TAU or no intervention	✓
Grazzi (2021) ³⁵	35	100% episodic	ACT	TAU or no intervention	✓
Hedborg (2011) ³⁶	76	100% episodic	Relaxation training + healthy lifestyle counseling + sleep counseling + stress management Relaxation training + healthy lifestyle counseling + sleep counseling + stress management + massage therapy	Minimal control	✓
Holroyd (1988) ³⁷	37	100% chronic	Thermal biofeedback + relaxation training	Attention control	✓
Holroyd (2010) ³⁸	232	NR	CBT + relaxation training (PMR) + propranolol CBT + relaxation training (PMR)	Placebo Propranolol	✓	✓
Janssen (1986) ³⁹	NR	NR	Relaxation training (PMR) Relaxation training (autogenic training)	None		✓
Kewman (1980) ⁴⁰	23	100% episodic	Thermal biofeedback	Attention control	✓
Klan (2022) ¹⁵	106	96% episodic, 4% chronic	CBT + relaxation training (PMR) + education Relaxation training	TAU or no intervention	✓	✓
Kleiboer (2014) ⁴¹	368	Mixed (% NR)	CBT + relaxation training + education	TAU or no intervention	✓
Kohlenberg (1981) ⁴²	117	100% episodic	CBT + thermal biofeedback + education + relaxation training + MBSR (meditation)	Attention control	✓
Kropp (1997) ⁴³	38	Mixed (% NR)	CBT + relaxation training Biofeedback	None		✓
Lemstra (2002) ¹⁶	80	100% chronic	CBT + relaxation training + education + exercise + massage therapy + physical therapy	TAU or no intervention	✓
Matchar (2008) ⁴⁴	614	100% chronic	Relaxation training + education	TAU or no intervention	✓
Mathew (1981) ⁴⁵	715	NR	Thermal and EMG biofeedback Thermal and EMG biofeedback + propranolol Thermal and EMG biofeedback + amitriptyline Biofeedback + amitriptyline + propranolol	TAU or no intervention Propranolol Amitriptyline Propranolol + amitriptyline	✓	✓
Mérelle (2008) ⁴⁶	127	100% episodic	Relaxation training (autogenic training) + education	TAU or no intervention	✓
Minen (2020) ⁴⁷	62	NR	Relaxation training (PMR)	TAU or no intervention	✓
Minen (2020) ⁴⁸	139	NR	Relaxation training (PMR)	TAU or no intervention	✓
Minen (2021) ⁴⁹	52	NR	Heart rate variability biofeedback	TAU or no intervention	✓
Odawara (2015) ⁵⁰	27	100% episodic	Thermal biofeedback + EMG biofeedback + PMR	TAU or no intervention	✓
Pickering (2012) ⁵¹	42	100% episodic	Relaxation training	TAU or no intervention	✓
Rausa (2016) ⁵²	47	100% chronic	EMG biofeedback	Attention control	✓
Reich (1989) ⁵³	392	100% chronic	Relaxation training Thermal biofeedback Combination of any two interventions from the other groups	Microcurrent electrical therapy (not an included treatment)		✓
Richardson (1989) ⁵⁴	47	Mixed (% NR)	CBT + relaxation training (PMR; in clinic) CBT + relaxation training (PMR; self‐administered)	TAU or no intervention	✓
Rothrock (2006) ⁵⁵	100	19% episodic, 81% chronic	Education (general migraine information)	Minimal control	✓
Sargent (1986) ⁵⁶ , ⁵⁷	136	NR	Thermal biofeedback + relaxation training EMG biofeedback + relaxation training Relaxation	TAU or no intervention	✓	✓	✓
Seminowicz (2020) ⁵⁸	98	100% episodic	MBSR Stress management training + education	None		✓
Seng (2019) ⁵⁹	60	48% episodic, 52% chronic	MBCT + education	TAU or no intervention	✓
Simshäuser (2022) ⁶⁰	54	100% Episodic	MBCT	TAU or no intervention	✓
Sorbi et al. (1984) ⁶¹	21	100% episodic	CBT + PMR + thermal biofeedback CBT + PMR	None			✓
Sorbi (1986) ⁶² , ⁶³	29	NR	CBT Relaxation training (autogenic training)	None		✓
Underwood (2023) ⁶⁴	727	45.5% episodic, 54.5% chronic	Education (healthy living, general migraine information)	Minimal control	✓
Varkey (2011) ⁶⁵	91	99% episodic, 1% chronic	Relaxation training	Topiramate		✓
Vasiliou (2021) ⁶⁶	94	Mixed (% NR)	ACT	TAU or no intervention	✓
Wachholtz (2008) ⁶⁷ , ⁶⁸	83	NR	Spiritual meditation Internally focused secular meditation Externally focused secular meditation Relaxation training (PMR)	None		✓
Wells (2021) ⁶⁹	89	Mixed (% NR)	MBSR Education (general migraine information, stress)	None		✓
Wittchen (1983) ⁷⁰	20	NR	CBT + relaxation training + education	TAU or no intervention	✓

Open in a new tab

Note: A checkmark (✓) indicates that the trial addressed that aspect of the treatment. See the Methods section for definitions of the different types of comparison groups.

Abbreviations: ACT, acceptance and commitment therapy; CBT, cognitive behavioral therapy; EMG, electromyographic; MBCT, mindfulness‐based cognitive therapy; MBSR, mindfulness‐based stress reduction; NR, not reported; PMR, progressive muscle relaxation; TAU, treatment as usual.

For the 50 adult trials, we rated the overall risk of bias to be low for three trials (6%), some concerns for 10 trials (20%), high for 35 trials (70%), and a mix of some concerns and high (depending on the type of treatment comparison) for two trials (4%). Our concerns mostly involved incomplete data and possible measurement bias. Specifically, incomplete data often occurred due to the number of patients not providing data at follow‐up. Measurement bias was mostly due to differences in attention time from trial staff and differential expectations of benefit in control groups (e.g., waitlist) versus treatment groups. The purpose of an attention control group, which six trials used, is to try to account for these concerns. Specifically, if the control group is seen for approximately the same amount of time by trial staff as the actively treated group, and the control group's treatment could reasonably be believed to improve migraine‐related outcomes, then differential expectations are an unlikely reason for why outcomes may have differed. We recognize that blinding is rarely possible in behavioral trials, but some trials were able to control for differential expectations in other ways (e.g., sham treatment). See the Appendix for the full report on domain‐specific risk of bias ratings.

We included 13 trials of children/adolescents (published from 1984 to 2015) that had enrolled a total of 1414 children/adolescents (Table 3). ⁷³ , ⁷⁴ , ⁷⁵ , ⁷⁶ , ⁷⁷ , ⁷⁸ , ⁷⁹ , ⁸⁰ , ⁸¹ , ⁸² , ⁸³ , ⁸⁴ , ⁸⁵ Eight trials were conducted in the United States, three in Germany, one in Canada, and one in Sweden. Twelve trials employed multiple behavioral components together, and one trial employed only a single behavioral component. The median number at baseline was 36 (interquartile range = 30–42), the weighted average age was 14.5 years, and 61% of participants were girls. Eight US trials reported race or ethnicity, with a median of 90% White and 10% Black. The migraine or headache days per month (reported by four trials) ranged from 8 to 23 (median = 10), and the baseline number of migraine episodes or headaches per month (reported by six trials) ranged from three to 48 (median = 14). Only one trial ⁸⁵ reported comorbid anxiety (present in 25% [6/24]) and depression (13% [3/24]). The trial duration (time between the start of treatment and the last follow‐up) ranged from 1 to 16 months (median = 7 months).

TABLE 3.

Included trials of children/adolescents on effectiveness, comparative effectiveness, or component contributions.

Trial	Baseline N	Episodic, chronic, or mixed	Behavioral treatment(s)	Comparison group(s)	Effectiveness	Comparative effectiveness	Component contribution
Albers (2015) ⁷³	900	100% episodic	CBT + relaxation training + education	TAU or no intervention	✓
Allen (1998) ⁷⁴	27	100% episodic	Thermal biofeedback + education Thermal biofeedback	None			✓
Connelly (2006) ⁷⁵	37	76% episodic, 24% chronic	CBT + relaxation training + education	TAU or no intervention	✓
Cottrell (2007) ⁷⁶	30	100% episodic	CBT + thermal biofeedback + relaxation training (PMR) + education + activity pacing	Attention control	✓
Fichtel (2001) ⁷⁷	37	100% episodic	Relaxation training (PMR)	TAU or no intervention	✓
Gerber (2010) ⁷⁸	34	100% episodic	Thermal biofeedback + EMG biofeedback + relaxation training + education CBT+ PMR + education	None		✓
Labbé (1984) ⁷⁹	28	100% episodic	Thermal biofeedback + relaxation training (autogenic training)	TAU or no intervention	✓
Labbé (1995) ⁸⁰	30	100% episodic	Thermal biofeedback + relaxation training (autogenic training) Relaxation training (autogenic training)	TAU or no intervention	✓		✓
Powers (2013) ⁸¹	135	100% chronic	CBT + thermal biofeedback + EMG biofeedback + relaxation training + amytriptyline Education + amytriptyline	None		✓
Rapoff (2014) ⁸²	35	Mixed (% NR)	CBT + relaxation training + pain management education	Minimal control	✓
Richter (1986) ⁸³	42	100% episodic	CBT Relaxation training (PMR + deep breathing)	Attention control	✓	✓
Sartory (1998) ⁸⁴	43	100% episodic	CBT+ PMR CBT + blood volume pulse biofeedback	Metoprolol		✓
Scharff (2002) ⁸⁵	36	100% episodic	CBT + thermal biofeedback + relaxation training	Attention control TAU or no intervention	✓

Open in a new tab

Note: A checkmark (✓) indicates that the trial addressed that aspect of the treatment. See the Methods section for definitions of the different types of comparison groups.

Abbreviations: CBT, cognitive behavioral therapy; EMG, electromyographic; NR, not reported; PMR, progressive muscle relaxation; TAU, treatment as usual.

For children/adolescents, we rated the overall risk of bias to be low for one trial (8%), some concerns for two trials (15%), high for eight trials (62%), and a mix of some concerns and high (depending on the type of treatment comparison) for two trials (15%). The reasons for our risk of bias ratings were similar in the pediatric literature as compared to the adult literature.

Adults: Effectiveness

Cognitive behavioral therapy

We included 12 trials assessing the effectiveness of CBT for adults. Notably, control arms differed across studies; 10 trials compared CBT to treatment as usual care/waitlist, one trial used an attention control, and one trial used a placebo control. Variability in control groups also occurred with other treatments (e.g., biofeedback, relaxation training; see Table 2).

For migraine/headache attack frequency, our meta‐analysis of 10 trials (Figure 2A, combining data from 839 patients) found a summary SMD of −0.33 (95% CI = −0.55 to −0.11, with a τ ² of 0.0668), favoring CBT, translating to a difference of −1.1 migraine days per month (95% CI = −1.8 to −0.4) with CBT compared with no treatment or treatment as usual. Neither Bromberg et al. ²⁴ nor Lemstra et al. ¹⁶ reported data on this outcome.

Meta‐analyses of cognitive behavioral therapy in adults. (A) Migraine or headache frequency. (B) Migraine disability. (C) Migraine‐specific quality of life. B, biofeedback; CBT, cognitive behavioral therapy; Chr, only patients with chronic migraine; CI, confidence interval; E, education; Epi, only patients with episodic migraine; HA, headaches (unreported timeframe); HA d/w, headache days per week; HA d/28 d, headache days per 28‐day period; HA/w, headaches per week; HDI, Headache Disability Inventory; HIT‐6, HIT‐6, Headache Impact Test‐6; MBSR, mindfulness‐based stress reduction; Mi d/w, migraine days per week; Mi d/30 d, migraine days per 30‐day period; Mi/w, migraine attacks per week; MIDAS, Migraine Disability Assessment; Mix, both episodic and chronic patients; MSQoL, Migraine‐Specific Quality of Life; NA, not applicable; NR, not reported; O, other (neither behavioral nor pharmacologic); P, pharmacologic; PDI, Pain Disability Inventory; R, relaxation training; SMD, standardized mean difference; T, tailored treatment; TAU, treatment as usual.

Seven trials used one of the key instruments for measuring migraine disability, and our meta‐analysis (Figure 2B, combining data from 712 patients) found a summary SMD of −0.32 (95% CI = −0.66 to 0.03). This result corresponds to 7 points on the MIDAS (95% CI = −17 to 3). This result is insufficient to permit conclusions, as the CI included both no effect (MIDAS difference of 0) and a clinically important effect (MIDAS difference of −17). The likely reason for the wide CI is differing effect sizes rather than small study sizes, as evidenced by τ ² = 0.1643. Wittchen ⁷⁰ (not included in the meta‐analysis due to unreported dispersion) found similar reductions in disability in both the CBT and control groups, with no statistical test reported. Overall, this outcome is insufficient to permit conclusions.

Only two of 11 trials reported migraine‐specific QoL (meta‐analysis in Figure 2C, combining data from 464 patients). Whereas Holroyd et al. ³⁸ found a statistically significant advantage for the groups receiving CBT, Kleiboer et al. ⁴¹ found no statistically significant difference between groups. The inconsistency produced a wide confidence interval in the meta‐analysis, precluding a conclusion.

Biofeedback

We included 13 trials on effectiveness that administered biofeedback to adults. For migraine/headache attack frequency, six of the 13 reported data for meta‐analysis (Figure 3A, combining data from 212 patients). The summary SMD was insufficient to permit conclusions (−0.37, 95% CI = −0.87 to 0.12, τ ² = 0.255); as the CI overlapped with a null effect, it was not narrow enough to indicate the absence of a difference. Of the other seven trials, three reported this outcome. Sorbi and Tellegen ⁶¹ found no statistically significant difference between groups in the percentage reduction in frequency. Neither Sargent et al. ⁵⁶ , ⁵⁷ nor Blanchard et al. ²³ reported a statistical test between biofeedback and no treatment.

For migraine disability, only two of the 13 trials reported effect size calculable information on included instruments measuring this outcome, and our meta‐analysis was insufficient to permit conclusions (Figure 3B, combining data from 160 patients). Of the other trials, three reported this outcome. Odawara et al. ⁵⁰ did not report the disability scale they used but did report that the biofeedback group improved statistically significantly more than the control group. The other two trials, ³⁰ , ⁴⁹ reported that there was no statistically significant difference between groups.

Only Minen et al. ⁴⁹ reported data on the Migraine‐Specific Quality of Life v2.1, with no statistically significant difference between groups.

Relaxation training

We included 23 trials assessing relaxation training for migraine prevention in adults. For migraine/headache attack frequency, our meta‐analysis of 13 trials (Figure 4A, combined n = 1091) favored relaxation training, with an SMD of −0.31 (95% CI = −0.50 to −0.11), translating to 1 fewer migraine day/month (95% CI = 0.4–1.6 fewer days/month). Four additional trials reported this outcome but had insufficient data for effect size calculation and did not report statistical tests. Three ³² , ⁵⁶ , ⁵⁷ had point estimates in the direction favoring relaxation training, and the other ⁴⁸ was in the direction favoring the control group.

Our meta‐analysis of migraine disability included 10 of 23 trials (Figure 4B, combining data from 1409 patients). The summary SMD was −0.19 (95% CI = −0.41 to 0.03, which translates to a MIDAS result [range 0–90] of −5 points [95% CI = −11 to 1]), which is insufficient to permit conclusions. The wide CI was likely due to differing effect sizes (τ ² = 0.079). Three additional trials reported this outcome but were not included in meta‐analyses. One found a statistically significant advantage for the relaxation training group, ⁵¹ another reported no statistically significant between‐group difference, ³⁷ and the third did not report a statistical test, ⁷⁰ but we conducted it and found no statistical significance.

Three of the 23 trials reported migraine‐specific QoL data, and our meta‐analysis (Figure 4C, combining data from 572 patients) was insufficient to permit conclusions (SMD −0.56, 95% CI = −1.38 to 0.26). The CI overlapped with a null effect and was not narrow enough to indicate the absence of a difference.

Mindfulness‐based treatments

Five trials assessed mindfulness‐based treatments in adults. Brown ²⁵ employed single‐component treatment with MBSR. Seng et al. ⁵⁹ reported combination treatment with CBT, mindfulness, and education (CBT and mindfulness together are called MBCT); Simshäuser et al. ⁶⁰ and Day et al. ²⁹ used MBCT; and Kohlenberg and Cahn ⁴² reported a multicomponent treatment that included CBT, relaxation training, biofeedback, and meditation.

All five reported migraine/headache attack frequency and enough information for effect size calculation (Figure 5A, combining data from 227 patients). The result of this meta‐analysis was statistically in favor of MBSR‐based treatments, with an SMD of −0.30 (95% CI = −0.54 to −0.05), corresponding to 1 fewer migraine day/month (95% CI = 0.2–1.8 fewer days/month). There was little between‐trial heterogeneity (τ ² = 0.006). For migraine disability, we meta‐analyzed the two trials that reported data in this category (Figure 5B, combining data from 103 patients). A random‐effects meta‐analysis was insufficient to permit conclusions (SMD = −0.48, 95% CI = −1.02 to 0.06, indicating too wide to indicate a difference or a lack of a clinically important difference). None of the five trials reported migraine‐specific QoL.

Meta‐analyses of mindfulness‐based therapy in adults. (A) Migraine or headache frequency. (B) Migraine disability. B, biofeedback; CBT, cognitive behavioral therapy; CI, confidence interval; E, education; Epi, only patients with episodic migraine; HA d/w, headache days per week; HA/d, headaches per day; HA/w, headaches per week; HDI, Headache Disability Inventory; MBSR, mindfulness‐based stress reduction; Mi/w, migraine attacks per week; Mix, both episodic and chronic patients; NA, not applicable; NR, not reported; PDI, Pain Disability Inventory; R, relaxation training; SMD, standardized mean difference; TAU, treatment as usual.

Education alone

Three trials specifically examined education's impact in adults. ²² , ⁵⁴ , ⁵⁵ , ⁶⁴ , ⁷¹ This section differs from those above on CBT, biofeedback, and relaxation training, because those sections considered any use of a single component regardless of other additional components, whereas this section examines only education when used in isolation as a way to change behavior.

For migraine/headache attack frequency, Underwood et al. ⁶⁴ found a statistically significant effect against education (the effect corresponded to a difference of 0.8 migraine days/month, 95% CI = 0.2–1.4). In contrast, Rothrock et al. ⁵⁵ reported only mean reductions and no dispersion, but at follow‐up, the education group had improved by six headaches per month, whereas there was zero change in the control group. Thus, the results conflict with Underwood et al. ⁶⁴ Aguirrezabal et al. ²² did not report data on this outcome.

For migraine disability, all three trials found a statistically significant advantage of education. Rothrock et al. ⁵⁵ reported that MIDAS scores improved statistically significantly more in the education group than in the control group (effect size not calculable). Aguirrezabal et al. ²² found that the percentage of patients who experienced at least a 50% improvement in MIDAS score was statistically significantly higher in the education group (70% vs. 35%). Underwood et al. ⁶⁴ reported a statistically significant advantage of education (corresponding to a Headache Impact Test‐6 [HIT‐6] difference of 1.2 points, 95% CI = 0.1–2.3).

None of the trials reported migraine‐specific QoL.

Acceptance and commitment therapy

Two trials assessed acceptance and commitment therapy (ACT) compared to usual care for migraine prevention in adults. ³⁵ , ⁶⁶ Only one small trial (n = 35), Grazzi et al., ³⁵ reported on headache frequency, with the authors reporting that ACT reduced migraine/headache attack frequency (−2.3 migraine days per month, 95% CI = −4.5 to −0.03).

For migraine‐related disability, Grazzi et al. ³⁵ found no statistically significant difference for either MIDAS or HIT‐6, whereas Vasiliou et al. ⁶⁶ reported a statistically significant advantage of ACT (SMD corresponding to a MIDAS difference of 18 in the direction of the ACT group, 95% CI = 4–31). Our meta‐analysis indicated insufficient evidence to permit conclusions due to a wide CI (SMD corresponding to a MIDAS difference of −3, 95% CI = −33 to 27). Neither trial reported data on migraine‐related QoL.

Hypnotherapy

Flynn ³⁴ compared hypnotherapy (the only treatment component) with usual care. For migraine/headache attack frequency, there was no statistically significant difference (SMD corresponding to a difference of 0, 95% CI = −2 to 2). For migraine‐related disability, the trial reported a statistically significant advantage of hypnotherapy (SMD corresponding to a MIDAS difference of 42, 95% CI = 23–61). The trial did not report migraine‐specific QoL.

Adults: Comparative effectiveness

We examined 17 different comparisons in adults, but only four had sufficient evidence for conclusions (listed in four sections below). The other comparisons (listed in Table 4) were each made by a single trial, and we rated the evidence as insufficient, typically due to statistically nonsignificant effects in single small trials at high risk of bias. See the full report for details. In the sections below, we only describe outcomes that permitted conclusions (i.e., an SOE rating of low, moderate, or high).

TABLE 4.

Other comparisons with insufficient evidence.

Population	Comparison	Trial
Adults	CBT vs. relaxation training	Sorbi et al. (1986) ⁶² , ⁶³
Adults	Relaxation training vs. biofeedback	Reich (1989) ⁵³
Adults	Relaxation training vs. MBSR meditation	Wachholtz et al. (2008) ⁶⁷ , ⁶⁸
Adults	MBSR vs. education	Wells et al. (2021) ⁶⁹
Adults	Relaxation training vs. ACT	Dindo et al. (2020) ³¹
Adults	EMG biofeedback vs. thermal biofeedback	Sargent et al. (1986) ⁵⁶ , ⁵⁷
Adults	Relaxation training using PMR vs. relaxation using autogenic training	Janssen and Neutgens (1986) ³⁹
Adults	MBSR using guided imagery of scene details vs. relaxation using guided imagery and relaxing statements	Brown (1984) ²⁵
Adults	MBSR meditation spiritual vs. internally focused secular vs. externally focused secular	Wachholtz et al. (2008) ⁶⁷ , ⁶⁸
Adults	Biofeedback vs. topiramate	de Tommaso and Delussi (2017) ³⁰
Adults	Relaxation training vs. topiramate	Varkey et al. (2011) ⁶⁵
Adults	CBT + relaxation training + education vs. relaxation	Klan et al. (2022) ¹⁵
Adults	MBSR vs. stress management training + education	Seminowicz et al. (2020) ⁵⁸
Adults	Adding CBT to relaxation training and education	Fritsche et al. (2010) ³³
Adults	Adding biofeedback to CBT and relaxation training	Sorbi and Tellegen (1984) ⁶¹
Adults	Adding biofeedback to relaxation training	Sargent et al. (1986) ⁵⁶ , ⁵⁷
Pediatric	CBT vs. biofeedback	Gerber et al. (2010) ⁷⁸
Pediatric	CBT vs. relaxation training	Richter et al. (1986) ⁸³
Pediatric	Biofeedback vs. relaxation training vs. metoprolol	Sartory et al. (1998) ⁸⁴
Pediatric	Adding biofeedback to relaxation training	Labbé et al. (1995) ⁸⁰
Pediatric	Adding education to biofeedback	Allen & Shriver (1998) ⁷⁴

Open in a new tab

Note: See the full report for details on these comparisons.

Abbreviations: ACT, acceptance and commitment therapy; CBT, cognitive behavioral therapy; EMG, electromyographic; MBSR, mindfulness‐based stress reduction; PMR, progressive muscle relaxation.

MBSR versus education

Wells et al. ⁶⁹ found that MBSR was associated with a greater improvement in migraine‐related disability, as measured by the MIDAS, compared with education (SMD = 0.70, 95% CI = 0.27–1.13). This difference corresponds to an 18‐point shift on the 0–90 MIDAS scale (95% CI = 7.18–30.06). MBSR also resulted in a greater improvement in HIT‐6 scores at 8‐week follow‐up (SMD = 0.86, 95% CI = 0.42–1.29). This equates to a HIT‐6 difference of 5.8 (95% CI = 2.81–8.64). Because there was only one trial of this outcome/comparison, the evidence was insufficient to permit a conclusion.

CBT + relaxation training versus propranolol

Holroyd et al. ³⁸ compared a combination of CBT and relaxation training with propranolol, which had a starting dosage of 60 mg/day (titrated as tolerated or switched to nadolol). After 10 months of follow‐up, the results indicated a statistically significantly greater reduction in migraine attack frequency with the use of propranolol than with use of the combination of CBT and relaxation training. This equated to an absolute reduction of −1.40 (95% CI = −2.63 to −0.16) migraine days per month. We also note that the trial compared a combination of CBT, relaxation training, and propranolol with propranolol alone. The combination of the behavioral and pharmacologic interventions was superior to the pharmacologic intervention alone, with an SMD of −2.63 (95% CI = −3.11 to −2.14) for attack frequency.

For migraine‐specific QoL, the trial found that the combined behavioral intervention was associated with a greater improvement than with propranolol. The improvement with the behavioral intervention at the 10‐month follow‐up was both clinically important and statistically significant, with an SMD of 0.89 (95% CI = 0.49–1.28). This SMD corresponds to a difference of 12 points on the Migraine‐Specific Quality of Life Questionnaire v2.1 (MSQ; 95% CI = 7–18). The combined behavioral and pharmacologic group had an even greater improvement than pharmacologic treatment alone, with an SMD of 2.78 (95% CI = 2.28–3.28).

MBSR versus stress management training + education

For migraine attack frequency, Seminowicz ⁵⁸ found a statistically significant difference favoring MBSR over stress management training + education (SMD = −0.64, 95% CI = −1.06 to −0.23, corresponding to a difference in migraine days/month of −2.1, 95% CI = −3.5 to −0.8).

CBT + relaxation training versus biofeedback

For migraine attack frequency, Kropp et al. ⁴³ found a statistically significant difference favoring biofeedback (SMD = −0.69, 95% CI = −1.34 to −0.03), corresponding to a difference in migraine days/month of −2.2 (95% CI = −4.4 to −0.1).

Adults: Component contributions

Both comparisons in this category (the effect of adding CBT, and the effect of adding biofeedback) were insufficient to permit conclusions. See Table 4 and the full report for details.