Magnesium supplementation for migraine prophylaxis

Objectives

This is a protocol for a Cochrane Review (intervention). The objectives are as follows:

To evaluate the benefits and harms of magnesium supplementation for preventing episodic or chronic migraine in children and adults compared to usual pharmacological treatment or placebo.

Secondary objective

To evaluate the benefits and harms of magnesium supplementation in different groups, specifically in terms of equity‐related characteristics such as economic status, age, or sex.

Background

Description of the condition

There are two main types of headaches: primary and secondary [1]. Secondary headaches are triggered by an existing medical problem, such as issues with blood vessels, tumors, infections, or changes in pressure or volume within the skull. Migraine is categorized as a primary headache, which includes headache disorders not caused by another illness [2]. Migraine is a neurological condition characterized by recurrent headaches (4 to 72 hours) that cause moderate to severe pulsating pain, often unilateral and worsened by physical activity [3], fasting and changes in eating habits, exposure to bright light and loud sounds, high caffeine intake, tobacco use [4], sleep disturbances [2], etc. Associated symptoms may include nausea, vomiting, photophobia, and phonophobia [5].

Migraine is generally divided into two groups based on the number of headache days per month: episodic and chronic. Episodic migraine (EM) is characterized by fewer than 15 headache days per month, and this is further subdivided into low‐frequency episodic migraine (LFEM) (1 to 9 headache days per month) and high‐frequency episodic migraine (HFEM) (10 to 14 headache days per month); chronic migraine (CM) is characterized by at least 15 headache days [6, 7]. The International Classification of Headache Disorders (third edition) (ICHD‐3) outlines diagnostic criteria that include migraine with and without aura [8]. Aura is defined as reversible focal neurologic symptoms that develop gradually over a period of 5 to 60 minutes, usually followed by a headache within 60 minutes.

To measure migraine severity, the Migraine Impact Global Severity Evaluation scale (MIGSEV) [9, 10] and the Global Assessment of Migraine Severity (GAMS) rely on patient self‐assessment [11]. These tools categorize migraine attacks into varying degrees of intensity, which are linked to their frequency, duration, and how well they respond to treatment, as well as their influence on daily life [9], and provide a single measure with three severity levels: mild, moderate, and severe [10]. Severe attacks are distinguished by their frequent occurrence, extended duration, resistance to treatment, and substantial negative impact on an individual's quality of life. Migraine directly impacts the quality of life both during attacks and interictally (i.e. between episodes) [12].

In 2019, the global prevalence of migraine was approximately 14,107.3 per 100,000 individuals, representing a 1.7% increase since 1990 [13], with a female‐to‐male ratio of 3:1 and incidence ranging from 8% to 18% [14]. The hormonal fluctuations experienced during pregnancy and breastfeeding may influence both the frequency and severity of migraine attacks in women. Migraine in pregnant women shows an upward trend; research showed that the age‐standardized prevalence of migraine during pregnancy rose from 11.4% in the year 2000 to 17.2% in 2018 [15]. Hormonal shifts during breastfeeding might also affect migraine occurrence, although the evidence is not definitive [16].

Migraine incidence is highest in early adulthood (20s to 30s), impacting the economically active population and causing approximately 20% of work absences [14]. It disproportionately affects females, especially during reproductive years, due to hormonal and genetic factors, necessitating sex‐based treatment customization [17]. Prevalence also varies globally by place of residence, being highest in regions with a high sociodemographic index (SDI) [18, 19].

In terms of years lived with disability, migraine ranks second among all health disorders [20]. The highest prevalence occurs between the ages of 35 and 39 years, with approximately 75% of individuals who experience migraines reporting the condition's onset before reaching the age of 35 years [21]. High prevalence of migraine and considerable impact from migraine are evident in the under‐18s population, as global data analysis indicates a 7.81% migraine prevalence in children and adolescents [22] and a 22.79% increase in headache disorder prevalence, including migraine, in this demographic between 1990 and 2021 [23]. A meta‐analysis estimates a higher migraine prevalence of 11% in children and adolescents, further specifying 8% for migraine without aura and 3% for migraine with aura [24]. Similarly, another systematic review found a 7.7% migraine prevalence in this age group, with girls experiencing migraine more often than boys [25].

Description of the intervention and how it might work

While the underlying mechanisms of migraine remain largely unknown, prophylactic treatment is typically considered for individuals with severe migraine or significantly impaired quality of life. This is defined as the use of pharmacologic or nonpharmacologic interventions with the primary goal of reducing the frequency, severity, or duration of migraine attacks, and minimizing migraine‐related disability, rather than treating individual acute episodes [5]. Preventive therapy is indicated for patients with frequent, severe, or disabling migraines, or those who overuse acute medications or have contraindications to acute therapies [8]. It may involve antihypertensive agents (beta‐blockers such as propranolol, metoprolol and angiotensin receptor blockers including candesartan, telmisartan); calcitonin gene–related peptide (CGRP) pathway‐targeted therapies such as gepants (atogepant, rimegepant) or anti‐CGRP antibodies (erenumab, fremanezumab, galcanezumab, and eptinezumab); antiepileptic drugs (valproic acid/sodium valproate, topiramate, gabapentin, pregabaline, etc); or antidepressants, particularly tricyclic antidepressants like amitriptyline. These medications are associated with a range of adverse effects, including paresthesia, weight loss, fatigue, somnolence, and cognitive and psychiatric effects such as depression and difficulty with memory and concentration [26].

Magnesium supplementation is recognized as a potentially effective alternative prophylactic treatment [27]. The Canadian Headache Society (CHS) guideline for migraine prophylaxis strongly endorses its use for migraine prevention as both a standalone intervention for mild migraines and a non‐pharmacological co‐intervention for treatment‐refractory or more severe migraines [28]. Magnesium, a crucial intracellular ion, is involved in numerous physiological processes, including energy production, protein synthesis, stabilization of mitochondrial membranes, and muscle function [29]. Its role in neuronal excitability is particularly relevant to migraine, as it potentially causes a phenomenon known as cortical spreading depression [30].

Common types of magnesium supplements include magnesium oxide, magnesium citrate, magnesium gluconate, magnesium aspartate, magnesium threonate, and magnesium bisglycinate, among others [31, 32]. Diarrhea and gastrointestinal discomfort are the most common adverse reactions observed with magnesium supplementation, and their occurrence and intensity are dose‐related, providing an indication of the maximum tolerable intake level [33]. Studies exploring the effects of a 12‐week regimen of 600 mg of magnesium daily have found that migraines occurred less often than in participants who received only placebo [27, 34, 35]. In children, a study that tested magnesium supplementation with magnesium oxide, at a daily dosage of 9 mg/kg three times a day for a minimum of four weeks, found equivocal evidence [36].

Why it is important to do this review

Migraine patients often explore complementary therapies to improve their quality of life, with many incorporating clinical supplementation. This is primarily due to issues with current pharmacological preventive treatment for migraine [37], which often has limited efficacy and has adverse effect profiles that many people poorly tolerate; for instance, response rates are relatively low, ranging from a 40% to 50% reduction in migraine attacks, frequency, and severity, and may be accompanied by substantial side effects, like paresthesia (tingling or numbness), fatigue, weight changes, and cognitive effects [38]. In contrast, several nutraceuticals and supplements, such as magnesium, show promising efficacy in migraine prophylaxis and generally possess a more favorable safety profile [39]. Perceptions of the side effects of preventive medications can influence people's willingness to use them. A study revealed that less than 60% of participants were willing to take preventive medications even if they offered a 50% reduction in headache frequency, primarily due to concerns about side effects [40]. This suggests that the perceived safety and tolerability of supplements may be a significant factor in patients' decisions to opt for these alternatives. Therefore, this review of magnesium supplementation as a preventive treatment may have significant implications for patients, clinicians, policymakers, and healthcare systems.

Objectives

To evaluate the benefits and harms of magnesium supplementation for preventing episodic or chronic migraine in children and adults compared to usual pharmacological treatment or placebo.

Secondary objective

To evaluate the benefits and harms of magnesium supplementation in different groups, specifically in terms of equity‐related characteristics such as economic status, age, or sex.

Methods

We will follow the methodological expectations for Cochrane intervention reviews (MECIR) when conducting the review [41], and PRISMA 2020 when reporting our process and findings [42].

Criteria for considering studies for this review

Types of studies

We will include randomized controlled trials (RCTs) and cluster‐RCTs. Following the guidelines in Chapter 23 of the Cochrane Handbook for Systematic Reviews of Interventions [43], we will exclude cross‐over trials, because they may have a potential carry‐over effect after the wash‐out period. Based on the algorithm outlined in Chapter 24 of the Cochrane Handbook for Systematic Reviews of Interventions, we will exclude quasi‐RCTs (i.e. controlled trials that used inappropriate strategies to allocate interventions) and non‐randomized clinical trials, due to the availability of RCTs that address our research questions [44].

Types of participants

We will include any person aged two years or older who has an established diagnosis of episodic or chronic migraine, with or without aura, according to the International Classification of Headache Disorders (ICHD) criteria [45]; specifically, we will allow ICHD‐2, ICHD3 beta and ICHD 3 (see Table 1). We expect to find trials involving pregnant and breastfeeding women because the prevalence of migraine is higher among women during their reproductive years [46].

1. International Classification of Headache Disorders, 3rd Edition (ICHD‐3).

Migraine without aura	Migraine with aura	Chronic migraine
A. At least 5 attacks fulfilling criteria B to D B. Headache attacks lasting 4 to 72 hours (when untreated or unsuccessfully treated) C. Headache has at least 3 of the following 4 characteristics: 1. unilateral location 2. pulsating quality 3. moderate or severe pain intensity 4. aggravation by or causing avoidance of routine physical activity (e.g. walking or climbing stairs) D. During headache at least one of the following: 1. nausea and/or vomiting 2. photophobia and phonophobia E. Not better accounted for by another ICHD‐3 diagnosis	A. At least 2 attacks fulfilling criteria B and C B. One or more of the following fully reversible aura symptoms: 1. visual 2. sensory 3. speech and/or language 4. motor 5. brainstem 6. retinal C. At least 3 of the following 6 characteristics: 1. at least one aura symptom spreads gradually over ≥ 5 minutes 2. two or more aura symptoms occur in succession 3. each individual aura symptom lasts 5 to 60 minutes 4. at least one aura symptom is unilateral 5. at least one aura symptom is positive 6. the aura is accompanied, or followed within 60 minutes, by headache D. Not better accounted for by another ICHD‐3 diagnosis	A. Headache (migraine‐like or tension‐type‐like) on ≥ 15 days/month for > 3 months, and fulfilling criteria B and C B. Occurring in a patient who has had at least five attacks fulfilling criteria B to D for 1.1 Migraine without aura and/or criteria B and C for 1.2 Migraine with aura C. On ≥ 8 days/month for > 3 months, fulfilling any of the following: 1. criteria C and D for 1.1 Migraine without aura 2. criteria B and C for 1.2 Migraine with aura 3. believed by the patient to be migraine at onset and relieved by a triptan or ergot derivative D. Not better accounted for by another ICHD‐3 diagnosis

We will exclude other primary headaches (such as trigeminal autonomic cephalalgias and tension‐type headaches) and secondary headaches, due to their distinct underlying pathologies and differential treatment responses.

We will include studies enrolling participants with migraine, regardless of the frequency of attacks, the intensity of attacks, or previous use of prophylactic treatment.

If we come across studies that include only a portion of participants pertinent to this review, we will incorporate these studies if separate data are available for the relevant subset, or if over 80% of participants satisfy the inclusion criteria.

Types of interventions

We will include studies implementing a minimum four‐week magnesium supplementation of any kind (magnesium oxide, citrate, dicitrate, sulphate, chloride, lactate, malate, taurate, glycinate, bisglycinate, orotate, and threonate), administered orally in a dose range of 200 to 600 mg daily according to the general recommendations of the CHS, the Swiss Headache Society, or other guidelines [47, 48] or any other dose used by the trial authors. We will consider co‐interventions, provided they are not part of the randomization process and are provided in all groups in a study.

We anticipate that most trials will have permitted the use of medication for acute migraine attacks experienced during the trial period. We therefore will record descriptions of trial rules concerning the use of acute medication in a 'Characteristics of included studies' table whenever such information is provided. If a trial includes multiple arms, we will include any arm that meets the inclusion criteria for this review.

We plan to investigate the following comparisons of interventions and comparators.

Interventions

• Magnesium supplementation

Comparators

• Placebo/no intervention

• Pharmacological treatment grouped by drug class or mechanism of action:

  • antihypertensive agents;

  • antiepileptic (antiseizure) drugs;

  • antidepressants;

  • calcitonin gene–related peptide (CGRP) pathway‐targeted therapies; or

  • onabotulinumtoxinA.

Outcome measures

We will not consider measurement of the outcomes assessed in this review as a criterion for eligibility.

Critical outcomes

• Acute migraine attacks/migraine frequency: number of headache days in the past three months and after one month of treatment, measured by attack report form (diary) [49] or as defined by the study authors. We will analyze the data as continuous data, with the primary measure being the mean change in score from baseline. We will use an effect estimate of 0.5 standard deviations to establish a minimally important clinical difference (MCID) [50].

• Headache duration: defined as the length of time a migraine attack lasts, assessed by the attack report form (diary) [51]. We will analyze the data as continuous data, with the primary measure being the mean change in score from baseline.

• Adverse events (AEs): defined as the number of participants with any unfavorable symptoms that occurred during the study. We will include all AEs, such as diarrhea and gastrointestinal discomfort. We will summarize the types of AEs reported narratively, in order to avoid unit‐of‐analysis errors, as study participants may experience multiple adverse events. There is no reported threshold for dichotomous outcomes; therefore, we will consider the MCID for these outcomes as a relative risk reduction of at least 25% [52].

Important outcomes

• Migraine‐specific pain: measured by tools like the Headache Impact Test (HIT‐6), pain intensity scale: 11‐ point NRS (anchors ‘no pain’ and ‘pain as bad as you can imagine’) [51], four‐point scale (where 0 = no headache; 1 = mild headache; 2 = moderate headache; 3 = severe headache); visual analog scales (VAS), or any other self‐rated measure chosen by the authors. If a study reports several scale measures for this outcome, we will prioritize visual analog scales. We will analyze the data as continuous data, with the primary measure being the mean change in score from baseline.

• Quality of life: defined as the impact of health conditions and their treatments on an individual's life, measured by any self‐rated validated scale, such as Migraine‐Specific Quality of Life Questionnaire (MSQ) [53], Migraine Disability Assessment (MIDAS) [54], Headache Impact Test‐6 (HIT‐6) scores [55], SF‐36 [56], or any other validated questionnaire. If a study reports several scale measures for this outcome, we will prioritize MSQ. The MCID for the EQ‐5D will range from 0.08 to 0.12. We will analyze the data as continuous data, with the primary measure being the mean change in score from baseline.

• Migraine severity: measured by the MIDAS scale, which relies on patient self‐assessment (grade I for “little or no disability” (0 to 5); grade II for “mild disability” (6 to 10); grade III for “moderate disability” (11 to 20); and grade IV for “severe disability” (≥ 21) and also includes a migraine severity global question, with responses ranging between 0 (no pain at all) and 10 (very severe pain)) [57], or Migraine Severity (MIGSEV) [58] and Global Assessment of Migraine Severity (GAMS) scale [11], providing a single measure with three severity levels: mild, moderate, and severe. The MCID of the MIDAS varies depending on the baseline severity of the migraine disability. According to the study by Ruscheweyh and colleagues, for participants with a baseline MIDAS score greater than 20 (indicative of severe disability, MIDAS grade IV), the MCID is estimated to be a reduction of approximately 30% in the MIDAS score. For patients with a baseline MIDAS score between 6 and 20 (MIDAS grades II and III), the MCID is estimated to be a reduction of 4 points [59]. The MCIDs for the MIGSEV and GAMS scales are not explicitly detailed in the medical literature. We will use an effect estimate of 0.5 standard deviations to establish an MCID [50]. If a study reports several scale measures for this outcome, we will prioritize the MIDAS. We will also analyze this outcome as a dichotomous outcome, assessing the absolute number of participants who achieved "little or no disability" or "mild disability" (grade I and II migraine severity) according to MIDAS scale, or mild severity according to MIGSEV and GAMS scales.

• Acute headache medication days: defined as the number of days per month on which acute headache or migraine‐specific medication is used. Assessed by the attack report form (diary) [60]. We will analyze the data as continuous data, with the primary measure being the mean change in score from baseline.

Following the guidelines of the Cochrane Handbook for Systematic Reviews of Interventions [61], we will prioritize the most appropriate scale when a study provides several options for a given outcome. Our selections will be the PSQI (sleep quality), ESS (daytime functioning), and EQ‐5D (quality of life).

Timing of outcome measurement

We will categorize outcomes into two sets of time points:

• short‐term follow‐up (defined as 0 to 12 weeks); or

• long‐term follow‐up (defined as longer than 12 weeks).

When multiple results are reported for a given outcome, we will include the longest follow‐up in each category.

Search methods for identification of studies

Electronic searches

We will search the following databases from their inception. We will not impose restrictions based on language, location, or publication status.

• Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library (from inception to date of search)

• MEDLINE (Ovid MEDLINE ALL; 1946 to date of search)

• Embase (Ovid; 1974 to date of search)

We will search the following trial registries.

• WHO International Clinical Trials Registry Platform (ICTRP; www.who.int/trialsearch)

• ClinicalTrials.gov (www.clinicaltrials.gov)

Details of the search strategies can be found in Supplementary material 1.

Searching other resources

We will search for gray literature using the following resources.

• Open Grey (opengrey.eu/)

• ProQuest; dissertation and theses

In addition, we will examine the reference lists of relevant trials to identify further published, unpublished, ongoing, or planned trials. We will contact experts about unpublished or ongoing clinical trials and study authors for additional details as needed. We will also review all studies before publication for modifications, comments, or retractions.

Data collection and analysis

Selection of studies

We will use Covidence software for selecting studies [62]. Two review authors (JPR and EQ) will independently assess the titles and abstracts of identified references. A third review author (LIG) will mediate any disagreements. We will retrieve the complete texts of all potentially relevant reports. The same two review authors (JPR and EQ) will then independently screen these full‐text articles against the inclusion criteria. If necessary, a third review author (LIG) will resolve any conflicts in this process. If disagreements arise that cannot be resolved, we will classify the study as pending and will contact the original authors for more information. We will create a PRISMA flow diagram to illustrate the steps involved in study selection [42]. Additionally, we will provide a table outlining the characteristics of all articles excluded following full‐text review, including the justifications for their exclusion [42].

Data extraction and management

For studies meeting our predetermined inclusion criteria, two independent review authors (JPR and EQ) will extract essential data on participants, interventions, and comparators using a pre‐tested data extraction form. We will solve any discrepancies between the two review authors through discussion, and if necessary, a third review author (LIG) will provide input to reach a consensus.

One review author (JPR) will be responsible for entering the collected data into Review Manager [63]. Subsequently, two other review authors (LIG and EQ) will verify and transfer the study data for analysis.

We will extract the following information from the study reports.

• Methods

  • Study design: including the start and end dates (if unavailable, this will be noted), study settings and country applying the World Bank Country Classification by income level (https://www.worldbank.org), language of publication, and study identifier

• Participants

  • Inclusion and exclusion criteria used for participant selection

  • Details of the participants at baseline, such as mean age, age range, sex/gender, place of residence, diagnostic criteria for migraine, severity of the condition, attack frequency, the presence or absence of aura symptoms, number of painkillers taken per month, associated symptoms such as nausea, vomiting, photophobia, and phonophobia, and the full inclusion and exclusion criteria

  • The number of participants in each treatment group who experienced the specified outcome events

  • Additional participant characteristics related to equity and defined by PROGESS‐Plus [64, 65], such as race/ethnicity/culture/language, occupation, religion, education, socioeconomic status, and social capital

• Interventions and comparisons, according to the Template for Intervention Description and Replication (TIDieR) checklist [66]

  • The name of the intervention (types of magnesium)

  • Doses of magnesium supplementation

  • The physical or informational materials used, as well as the procedures, activities, or processes involved

  • Information on who delivered the intervention, including their expertise, background, and any specific training provided

  • A description of the modes of delivery

  • Information on the frequency and duration of the intervention delivery

  • Whether any personalization or adaptations were planned

  • Any modifications made to the intervention during the study

  • Measures of adherence to or fidelity of the intervention

• Outcomes: clear definitions of the relevant outcomes, the methods and timing of their measurement, and any pertinent subgroups for the review

• Study funding sources

• Declarations of interest from the primary investigators

We will present this extracted information in an 'overview of included studies and synthesis' table. We will also contact the authors of the included studies to inquire if they are willing to answer any questions regarding their research, and these communications will be documented. If necessary, we will seek any relevant missing study information directly from the study authors.

Risk of bias assessment in included studies

Using the Cochrane RoB 2 tool at the final follow‐up time point, two review authors (JPR and EQ) will independently assess the potential for bias in all outcomes included in the summary of findings tables (see Certainty of the evidence assessment) [67, 68]. We will assess the risk of bias based on the intention‐to‐treat principle (i.e. the effect of assignment to the intervention) [69]. We will solve any discrepancies in assessment through consultation with a third review author (LIG). Should insufficient details be present in the published reports, trial protocols, clinical study reports, or other resources, we will reach out to the original study investigators to obtain further information regarding the risk of bias elements.

We will evaluate the risk of bias across all RoB 2 categories, classifying each as 'high risk of bias,' 'some concerns,' or 'low risk of bias.' This classification will be based on the answers to the signaling questions and the algorithms embedded within the RoB 2 tool. These categories encompass bias arising from the randomization process, variations in intended treatments, incomplete outcome data, the measurement of outcomes, and the selection of reported findings. The tool's algorithm will determine the overall risk of bias for each outcome. We will provide evidence to justify our ratings in the risk of bias table, and if our judgment deviates from the algorithm's recommendation, we will explicitly state the reasons for this divergence. We will use the RoB 2 Excel tool to manage our risk of bias assessments [68]. All data will be made publicly accessible as supplementary material in a designated repository (available at https://osf.io/).

For cluster‐RCTs, we will employ the RoB 2 tool alongside an additional domain specific to this trial design. This domain (1b – "bias arising from the timing of identification and recruitment of participants") is available at www.riskofbias.info, and includes its associated signaling questions. We will adhere to the guidelines outlined in the Cochrane Handbook for Systematic Reviews of Interventions (section 23.1.2 and Table 23.1.a) [43].

Measures of treatment effect

We will present binary data, such as adverse events, as a relative risk (RR) with its 95% confidence intervals (CIs). For continuous outcomes, like acute migraine attacks, we will estimate the treatment effect by calculating the mean difference (MD) with 95% CIs. If we combine data from studies that used different tools to assess the same outcome, we intend to compute standardized mean differences (SMDs) with 95% CIs. We will enter data presented on a scale with a uniform direction of effect and multiply the SMD by a standard deviation that is representative of the combined studies (e.g. the standard deviation from a widely recognized scale used in several studies included in the analysis that informed the result). We will multiply the mean values from studies using the opposite direction by –1 so that higher scores consistently indicate better (or worse) outcomes across studies. This standardization will allow for appropriate pooling of results in meta‐analysis. We will conduct meta‐analyses only when appropriate, that is, when the treatments, participants, and the fundamental clinical questions are sufficiently similar to allow for meaningful pooling of data.

Unit of analysis issues

The participant with episodic or chronic migraine, with or without aura, will serve as the unit of analysis. The level of randomization will be a key consideration, and we will account for repeated outcome observations. In instances where a single study yields multiple eligible comparisons for a given meta‐analysis, we will either aggregate groups to form a singular pairwise comparison or implement a reduction in sample size by partitioning any shared group into two or more groups. While the latter approach provides some means of adjusting the precision of the comparison, it does not fully address the correlation arising from the inclusion of identical participant sets in multiple comparisons [61]. If we include studies with multiple arms, we will only use the information from the arms relevant to this review.

For cluster‐randomized controlled trials, the cluster, and not the individual participant, will be treated as the unit of analysis to mitigate unit‐of‐analysis errors, as specified in Section 23.1.1 of the Cochrane Handbook for Systematic Reviews of Interventions [43]. Should the effect measure for the cluster not be determined through appropriate methodologies in the included studies, we will inflate the standard error of the effect estimate (obtained from an analysis that ignores clustering) by multiplying it by the square root of the design effect. We will calculate this design effect using an intracluster (or intraclass) correlation coefficient (ICC) of 0.02, consistent with the recommendations outlined in sections 23.1.4 and 23.1.5 of the Cochrane Handbook for Systematic Reviews of Interventions [43].

Dealing with missing data

We will prioritize obtaining any missing data by reaching out to the authors of the selected studies, if this is possible. Our assessment of the risk of bias will include a careful review of key numerical data, such as details of the screening process, the number of participants randomized, and the characteristics of intention‐to‐treat, as‐treated, and per‐protocol groups. In this process, we will analyze attrition rates (e.g. dropouts, participants lost to follow‐up, and withdrawals) and critically evaluate concerns surrounding missing data and the methods used for imputation (such as last observation carried forward). Our primary analysis will be based solely on the data we have. However, in situations where missing data cannot be obtained and are considered to introduce significant bias, we will conduct a sensitivity analysis to explore how including such studies affects the overall findings.

Reporting bias assessment

To reduce reporting bias, we will perform an extensive literature search without any restrictions on publication date or language. We will also use study protocols and trial registrations to identify potential selective reporting in the studies. If an outcome is reported in at least 10 studies, we will use funnel plots to evaluate small‐study effects. Funnel plot asymmetry can be for a variety of reasons, including true variations in effect size by study size, poor study design leading to bias in smaller studies, and selective non‐reporting [70]. Therefore, we will interpret these plots cautiously [71]. Following the recommendations in Chapter 13 of the Cochrane Handbook for Systematic Reviews of Interventions [61], we will not conduct statistical tests for funnel plot asymmetry due to their typically low power.

Synthesis methods

We will conduct all data synthesis using Review Manager [63]. We intend to perform a meta‐analysis only when we determine that the participants, interventions, comparisons, and outcomes are similar enough to yield a clinically significant result. Our main analysis will include all studies, irrespective of their assessed risk of bias. We will primarily synthesize data using the random‐effects model [72], given the anticipated clinical diversity. We will employ the Restricted Maximum Likelihood (REML) estimator to calculate the variance between trials. To determine the confidence interval for the meta‐analysis effect estimate, we will apply the Hartung‐Knapp‐Sidik‐Jonkman method when we have at least three studies and the heterogeneity estimate is greater than zero. In all other cases—for example, with pooled analyses of only two studies or a heterogeneity estimate of zero—we will use the Wald‐type method [73]. When interpreting random‐effects meta‐analyses, we will consider the entire distribution of effects and present confidence intervals. Our statistical analyses will adhere to the guidelines outlined in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions [73]. If a meta‐analysis is not feasible, we will summarize the findings using narrative synthesis, following the Synthesis Without Meta‐analysis guidelines, rather than a pooled statistical approach. We will follow Chapter 12 of the Cochrane Handbook for Systematic Reviews of Interventions [74]. We will perform subgroup analyses using the methodology described by Deeks and colleagues, as recommended in Section 10.11.3 of the Cochrane Handbook for Systematic Reviews of Interventions [61].

Investigation of heterogeneity and subgroup analysis

We will visually assess the variability of point estimates and the overlap of their CIs. We will use the I² statistic to evaluate the degree of heterogeneity present among the trials in each analysis [75]. If considerable unexplained heterogeneity is detected, we will document it and explore potential reasons using prespecified subgroup analyses. Our interpretation of the I² statistic will follow the approximate scale provided in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions [73]:

• 0% to 40%: potentially not important;

• 30% to 60%: potentially moderate heterogeneity;

• 50% to 90%: potentially substantial heterogeneity;

• 75% to 100%: considerable heterogeneity.

The characteristics listed below are expected to introduce clinical heterogeneity, and we will therefore conduct subgroup analyses for each, including an exploration of interactions.

• Intervention based on types of magnesium: magnesium oxide, citrate, dicitrate, sulphate, chloride, lactate, malate, taurate, glycinate, bisglycinate, orotate, and threonate or any type of magnesium chosen by the authors.

• Baseline migraine severity: according to the MIDAS scale (migraine‐related disability with grade I for “little or no disability” (0 to 5); grade II for “mild disability” (6 to 10); grade III for “moderate disability” (11 to 20); and grade IV for “severe disability” (≥ 21), or using each trial's cutoff for migraine baseline severity classification.

• Sex/gender: men versus women, considering trials conducted with male or female participants exclusively, or with over 80% of participants being female or male

• Socioeconomic status based on country income level: low‐, middle‐, and high‐income countries, based on the World Bank classification (available at the World Bank website: https://www.worldbank.org/)

• Age: using the age group classification outlined in the Medical Subject Headings (MeSH) 0 to 12 years old (pediatric group), 13 to 18 years old (adolescent group), 19 to 64 years old (adult group) and ≥ 65 years old (aged group) (https://www.ncbi.nlm.nih.gov/mesh/68009273)

We will use the formal test for subgroup interactions in Review Manager [63], being aware of its constraints due to its observational nature and lower power to detect differences when each category has fewer than 10 studies [61].

Equity‐related assessment

We will explore health inequity through three characteristics defined by PROGRESS‐Plus [64, 65]: age, sex, and place of residence. See Supplementary material 2.

Age: migraine typically emerges during early adulthood, with the highest incidence seen in the second and third decades of life. This population is economically active, which is why it is of utmost importance to study the impact of this disease on different age groups. Individuals experiencing migraine headaches are significantly more likely to miss workdays due to their condition. Collectively, migraine is responsible for approximately 20% of all work absences attributed to illness [14]. We will address this health issue by assessing the effects of the intervention in different age ranges, using the age group classification outlined in the MeSH (see Investigation of heterogeneity and subgroup analysis).

Sex: on the other hand, there is a higher proportion of females affected with CM than with EM [6]. Research indicates that women experience migraines more frequently and severely than men, due to a combination of hormonal, genetic, and possibly epigenetic factors, along with variations in how migraines manifest clinically and the presence of other health conditions. Migraine prevalence is higher in women during their reproductive years, and hormonal shifts associated with menstruation, pregnancy, and menopause can substantially affect migraine attack frequency and intensity. Treatment responses also differ between males and females [17]. These findings highlight the need to account for sex‐based differences when managing migraine, customizing treatment plans to achieve the best results for both men and women. By evaluating the impact of magnesium supplementation on migraine in males and females, this review will provide insights that can help address a significant health disparity. Consequently, we will conduct a subgroup analysis to assess whether there is evidence of a differential effect in males and females.

Place of residence: the prevalence and incidence of chronic migraine vary significantly across countries with different income levels, according to medical literature. Global studies, such as those from the Global Burden of Disease (GBD), have provided detailed data on these variations. Regions with a high SDI exhibited the highest prevalence rates in both 1990 and 2019 [18, 19]. The burden of chronic migraine also poses a significant challenge in middle‐ and low‐income regions, particularly in Asia [76, 77], underscoring the need for public health interventions tailored to the socioeconomic and geographic characteristics of each region. Therefore, we will analyze this issue by including subgroups of countries by income level, that is, low‐, middle‐, and high‐income countries, based on the World Bank classification (see Investigation of heterogeneity and subgroup analysis).

To explain the primary results (acute migraine attacks, headache duration, and adverse events) in the context of health inequity, we will construct a separate summary of findings table and discuss the applicability of the results in the 'Equity‐related implications for practice' and 'Equity‐related implications for research' sections of the conclusions in terms of potential differences and context‐dependent aspects that may affect both the outcomes and their practical implementation.

Sensitivity analysis

For the critical outcomes (acute migraine attacks, headache duration, and adverse events), quality of life, and migraine severity, we intend to conduct several sensitivity analyses to investigate how the following factors influence the effect sizes, as outlined in Section 10.14 of the Cochrane Handbook for Systematic Reviews of Interventions [61].

• Study design: by excluding studies that used cluster‐randomization

• Methodological quality: by removing studies judged to have an overall high risk of bias

• Publication status: by excluding data that has not been formally published

Certainty of the evidence assessment

We will assess the overall certainty of the evidence based on the GRADE approach. This method considers factors related to internal validity (overall risk of bias, inconsistency, imprecision, and publication bias) and external validity (directness of findings). Two review authors (JPR and EQ) will independently assess the certainty of the evidence for each comparison and outcome. Any disagreements in these assessments will be resolved through discussion, and if necessary, we will consult a third review author (LIG).

We will summarize the evidence in summary of findings tables, which will include crucial information regarding the best estimate of the effect's magnitude, both as relative terms and absolute differences, for each relevant comparison of alternative management strategies. It will also detail the number of participants and studies contributing to each important outcome, along with an overall confidence rating for the effect estimates of each outcome. We will justify all decisions to lower the certainty of the evidence using informative footnotes and the GRADE guidelines [78].

The summary of findings tables will be generated using the methods outlined in the Cochrane Handbook for Systematic Reviews of Interventions, Review Manager, and GRADEpro GDT software [61, 63, 79, 80]. We will create summary of findings tables for the following comparisons.

• Magnesium compared to placebo/no intervention

• Magnesium compared to antihypertensive drugs

• Magnesium compared to antiepileptic (antiseizure) drugs

• Magnesium compared to antidepressants

• Magnesium compared to calcitonin gene–related peptide (CGRP) pathway‐targeted therapies

• Magnesium compared to onabotulinumtoxinA

We will include the following outcomes, prioritizing data at the longest follow‐up time point.

• Acute migraine attacks/migraine frequency (long term)

• Migraine severity, both as a continuous outcome and a dichotomous outcome (long term)

• Quality of life (long term)

• Adverse events (long term)

• Migraine‐specific pain (long term)

Consumer involvement

We will not involve consumers in this review, although we used core outcome sets when we selected the review's outcomes. These core outcome sets have been developed with active involvement from consumers [51, 60].

Supporting Information

Supplementary materials are available with the online version of this article: 10.1002/14651858.CD016307.

Supplementary materials are published alongside the article and contain additional data and information that support or enhance the article. Supplementary materials may not be subject to the same editorial scrutiny as the content of the article and Cochrane has not copyedited, typeset or proofread these materials. The material in these sections has been supplied by the author(s) for publication under a Licence for Publication and the author(s) are solely responsible for the material. Cochrane accordingly gives no representations or warranties of any kind in relation to, and accepts no liability for any reliance on or use of, such material.

Supplementary material 1 Search strategies

Supplementary material 2 Summary of the characteristics of participants we would expect to see in the evidence and the actual participant characteristics extracted from the included studies

New

Additional information

Contributions of authors

All review authors read and approved the final draft of the protocol.

JPR: conceptualization, methodology, writing original draft

EQ: conceptualization, methodology, review and edit of draft

LBV: conceptualization, methodology, review and edit of draft, supervision

CMEL: methodology, data curation (search strategies), writing original draft, review and edit of draft

LIG: conceptualization, methodology, data curation, investigation, writing original draft, review and edit of draft, supervision, project administration

Declarations of interest

JPR: none known

EQ: none known

LBV: none known

CMEL: none known

LIG: none known

Sources of support

Internal sources

• Universidad Hospital Italiano de Buenos Aires, Argentina

  Provides salaries for the research team

External sources

• No sources of support provided

Registration and protocol

Cochrane approved the proposal for this review in April 2025.

Data, code and other materials

Data sharing is not applicable to this article as it is a protocol, so no datasets were generated or analysed.