A systematic search was conducted across Wiley Online, BVS, MEDLINE, and OVID from their inception until November 2, 2024. Additional searches included clinical trial registries, government databases and websites, conference proceedings, patient advocacy group websites, systematic reviews/meta-analyses, and medical society websites, though these were ultimately excluded. The technical expert panel assisted in identifying relevant literature. A medical reference librarian designed and executed the search strategy, which was peer-reviewed by a second librarian and validated by coauthors MK, V-J, and I R-L.

Eligible studies (1) included adult patients (≥18 years) with episodic migraine; (2) evaluated preventive pharmacologic and non-pharmacological treatments; (3) involved randomized clinical trials (RCTs) (phase II/phase III) comparisons of the intervention with placebo, usual care, another pharmacologic therapy, or no treatment (4) reported outcome of interest as reduction of monthly migraine days (MMD), monthly headache days (MHD) and acute medications days (AMD), (5) adverse events (AE) and serious adverse events (SAE). We excluded in vitro, phase I clinical trials, nonrandomized, open-labeled trials, studies without original data, and single-group studies. Therapies in development or intravenous administration terminated development or unavailable in the global market were excluded. Additionally, studies on patients diagnosed with tension headaches or other headache disorders and treated with NSAIDs, triptans, or ergot alkaloids therapies were excluded. Case reports, case series, reviews, post-hoc analyses, or multiple reports of the same study were excluded.

The original study definitions were retained despite evolving migraine criteria, provided they aligned with the current International Classification of Headache Disorders, Third Edition (ICHD-3) standards for episodic migraine (10), characterized by headaches occurring on ≤14 days per month in individuals with migraine. Studies were restricted to those published in English or Spanish.

An extraction form was developed to standardize data collection. Two reviewers independently extracted study characteristics. Inter-rater agreement was assessed using Cohen’s kappa (κ = 0.82), indicating substantial agreement. Discrepancies were resolved through consensus discussion. A third reviewer (K.V.) was consulted when necessary. Authors were contacted for clarifications when data were missing or unclear. The extracted data included the generic name of the drug or device, author, year, study design, sample size, intervention details, administration route, dose, frequency, adverse effects, efficacy and safety outcomes, time frame, and availability in Mexico. Treatments were categorized as monotherapy or combination regimens in migraine prevention.

A comprehensive literature search used detailed search terms and Boolean operators to identify relevant studies. The search focused on RCTs that were double-blind AND placebo-controlled interventions. The following pharmacological and non–pharmacological treatment options were included: Erenumab OR CGRP antagonist OR fremanezumab OR galcanezumab OR Eptinezumab OR gepants OR rimegepant OR atogepant OR topiramate OR propranolol OR beta-blocker OR venlafaxine OR valproate OR oxcarbazepine OR candesartan OR amitriptyline OR antiepileptics, OR antidepressants, OR melatonin OR lanepitant OR aspirin OR NSAIDs OR memantine OR neuromodulation OR nerve blockers OR vestibular treatments OR acupuncture. Additionally, the search incorporated studies involving herbal supplements, oils, and combinations using Rayyan© Software, Cambridge, MA, USA.

The primary efficacy outcome included reducing MMD in the active study group compared with the placebo. The secondary efficacy outcomes were MHD and reduction of AMD, which include specific and non-specific substances. When data on reduction in days or standard deviations were not directly reported in the articles, they were estimated based on comparisons between baseline and final values, reported percentage changes, or visual inspection of figures. Standard deviations were calculated from reported standard errors or visually estimated when necessary. The primary safety outcome included the presence and frequency or percentage of adverse effects and SAE; type, and severity of adverse effects using the Common Terminology Criteria for grading from Grade 1 (mild) to Grade 5 (death), and availability in Mexico.

The risk of bias was evaluated using the Cochrane Risk of Bias Tool for Randomized Trials (RoB 2, v2) (11). This assessment covered five key domains: (1) bias in randomization procedures, (2) bias from deviations in intended interventions, (3) bias due to incomplete outcome data, (4) bias in outcome measurement, and (5) bias in selective reporting. Each domain was rated as “low risk,” “some concerns,” or “high risk,” with an overall bias judgment assigned per study. For this analysis, MMD were the primary outcome to determine bias. Two independent reviewers (D.S. and M.A.M.M.) conducted the assessments.

Data analysis was performed with SPSS software (v.31; IBM Corp., Armonk, NY, USA). Heterogeneity across studies was evaluated using the Chi-square test and quantified via the I² statistic. A fixed-effect model was applied for analyses with low heterogeneity (I² < 50%), while a random-effects model was used when I² ≥ 50%. No subgroup or sensitivity analyses were pre-specified due to the limited number of homogeneous studies available. Funnel plots were generated to assess publication bias for the primary efficacy comparison (allopathic pharmacological treatments vs. placebo for MMD). Egger’s test was applied and showed no evidence of publication bias (p = 0.27). Continuous outcomes were expressed as mean differences (MD) with 95% confidence intervals (CIs). Statistical significance was defined as p < 0.05. The meta-analyses conducted in this review were limited to efficacy outcomes, specifically MMD, due to heterogeneity in the reporting and classification of AE and SAE across studies.

Our search until November 30, 2024, identified 605 scientific papers through database and trial registry screening; after removing duplicates or illegible by automation tools, 202 records remained. Figure 1 shows the flowchart of the clinical studies screened and excluded, and finally, 39 RCTs were included and analyzed. All studies were published between 1987 and 2024. All studies were randomized double-blinded clinical trials classified in pharmacological treatments and non-pharmacological interventions.

The included studies predominantly employed RCT designs, with the majority (27) utilizing double-blind, placebo-controlled methodologies. Additionally, 12 studies implemented double-blind RCTs with active comparators, directly comparing the efficacy and safety of different pharmacological treatments.

The included studies demonstrated considerable variability in sample sizes ranging from 28 to 1,001 participants, with a distinct trend toward larger trials (>300 participants) evaluating newer therapeutic classes such as anti-CGRP monoclonal antibodies (e.g., erenumab, fremanezumab) and gepants (e.g., rimegepant, atogepant). In contrast, smaller-scale trials were more frequently observed for established drug classes, including anti-seizure medications (e.g., topiramate, valproate) and beta-blockers (e.g., propranolol), as well as non-pharmacologic approaches such as acupuncture.

All interventions were administered in outpatient or clinical trial environments, with routes including oral, subcutaneous (SC), intravenous (IV), transcutaneous, and topical.

Table 1 summarizes the study design, sample size, interventions, clinical outcomes, adverse events profile, and availability in Mexico. This section presents a structured narrative synthesis of the efficacy and safety findings of each pharmacologic class used for episodic migraine prevention.

Table 2 summarizes a structured narrative synthesis of non-pharmacological. This section provides a structured narrative synthesis of non-pharmacological interventions for episodic migraine prevention.

Non-pharmacological interventions included neuromodulation techniques such as occipital nerve stimulation, caloric vestibular stimulation, and acupuncture. These modalities were assessed primarily over a 12-week period, except for occipital nerve stimulation, which reported outcomes at 1 month.

Occipital nerve stimulation showed reductions in MHD ranging from −2.0 to −5.5 days at 1 month, depending on the stimulation frequency. MMD also improved, although the data were not consistently reported across all frequency subgroups. AE were reported in up to 25% of patients and included local pain, hematoma at the stimulation site, nausea, and dizziness. SAE were not reported in these trials.

Caloric vestibular stimulation was evaluated over 12 weeks and demonstrated a reduction in both MMD and MHD of −3.9 days. The same intervention showed a reduction in AMD of −3.9 days. The most frequent AE included nausea, dizziness, ear discomfort, and tinnitus. No SAE were reported.

Acupuncture, evaluated over a 12-week period, showed a reduction of −2.5 days in MMD compared to a −1.0-day reduction in the sham acupuncture control group. MHD and AMD were not reported. AE occurred in 25% of patients and included local pain and hematoma at the puncture sites. No SAE were reported.

Although these interventions yielded promising effects in reducing migraine frequency and associated medication use, the wide variability in protocols, outcome definitions, and follow-up times hindered direct comparison and aggregation of results. Nonetheless, the generally favorable safety profile across studies supports the potential role of these non-pharmacological strategies as adjunctive treatments in individualized preventive regimens.

Unfortunately, the meta-analysis of the efficacy of the non-pharmacological treatments was not feasible for the heterogeneity of the outcome measurements used in each study.

A structured GRADE assessment was conducted to determine the certainty of evidence for the main pharmacological comparisons included in this review. This approach complements the narrative synthesis and meta-analysis by addressing potential limitations in risk of bias, inconsistency, indirectness, imprecision, and publication bias.

Table 3 summarizes the GRADE evidence profiles for anti-CGRP monoclonal antibodies, gepants, and combination therapies, using monthly migraine days (MMD) as the primary outcome.

The certainty of evidence was rated as moderate for anti-CGRP therapies due to some imprecision across trials, high for gepants based on robust and consistent findings, and low for combination treatments, mainly due to heterogeneity, indirect comparisons, and small sample sizes. These ratings provide a useful framework for interpreting the strength and applicability of the observed clinical effects.

Bias was evaluated following the guidelines of the Cochrane Handbook for Systematic Reviews. The details are illustrated in Figures 6, 7. We did not find any risk of bias in the RCTs using the intention-to-treat modality. However, in the Per-protocol approach, participants were randomly assigned to groups using computer-generated random sequences through an interactive web-response system in all two studies. One trial noted that pharmacists were unblinded; however, their role was limited to drug preparation and inventory management. Another trial did not provide details on allocation concealment and blinding of outcome assessment. All two trials reported patient follow-up losses; each predefined outcome was clearly described. Studies meeting the inclusion criteria were included in this meta-analysis.

The present review also has some limitations. First, the present study was restricted to eligibility criteria, in which merely “Number of studies” were included in the analysis. Some unpublished and missing data from studies also influence aggregate results. Furthermore, some of the studies were completed by the same researchers, which may lead to publication bias. In addition, the double-blind period of these included studies ranged from 3 to 6 months, and the difference might result in heterogeneity. To reduce heterogeneity, only studies with 12-week follow-up were included in the final quantitative synthesis. While this approach improved comparability across trials, it also excluded a significant number of potentially relevant studies and may have limited the scope of the analysis, particularly with regard to long-term efficacy and safety outcomes.

Additionally, pharmacological treatments were grouped into broad therapeutic classes (e.g., antidepressants, anti-seizure medications), despite marked differences in their mechanisms of action and clinical profiles. For example, amitriptyline and venlafaxine, although both classified as antidepressants, have distinct pharmacodynamic properties; similarly, topiramate and valproate differ substantially in their molecular targets and tolerability. This classification may oversimplify treatment effects and obscure clinically meaningful differences between individual agents. Greater granularity, as reflected in the compound-specific data provided in Supplementary material, is likely to be more informative for guiding clinical decision making. Also, the analysis excluded older, non-specific pharmacological therapies commonly used in migraine prevention—such as other beta-blockers and certain calcium channel blockers—owing to the lack of recent or high-quality RCTs meeting inclusion criteria. Some of them are treatments that remain widely prescribed in routine practice in some countries, and their omission from the current synthesis may limit the generalizability of the findings.

In addition, non-pharmacological and nutraceutical interventions were grouped into heterogeneous categories, despite having distinct therapeutic mechanisms and varying levels of supporting evidence. This broad classification complicates interpretation and precludes firm conclusions about the relative efficacy of individual non-drug strategies.

Moreover, this review included only studies published in English or Spanish, which may have introduced language bias and limited the inclusion of potentially relevant trials from other regions. Otherwise, Meta-analyses were conducted using SPSS due to software availability at the institution. While SPSS is appropriate for basic fixed- and random-effects models, it does not offer the advanced options or flexibility of specialized platforms such as RevMan or R-based packages like meta or metafor. This may limit some statistical nuance in modeling or subgroup analysis.

Finally, due to the exclusion of patients older than 65 years, pregnant individuals, and those with significant cardiovascular or cerebrovascular comorbidities, the generalizability of the results is limited. These populations, which are frequently encountered in clinical practice, remain underrepresented in current trials. Moreover, future studies should prioritize the investigation of subgroup-specific responses to preventive treatments, including stratification by migraine frequency (e.g., high-frequency episodic vs. chronic migraine), sex, and age group. Such analyses are essential to advancing a more tailored and equitable approach to migraine management.

Furthermore, studies with longer follow-ups and larger sample sizes should be performed to identify the confirmative safety profile of gepants and monoclonal antibodies and determine the duration of its therapeutic effects.

Taken together, these limitations highlight the need for further high-quality, head-to-head trials of both pharmacological (e.g., gepants vs. monoclonal antibodies) and non-pharmacological treatments, with mechanistic specificity, standardized outcomes, and longer follow-up durations, and evaluations of cost-effects of the treatment to better inform personalized approaches to migraine prevention.

Preventive therapy for episodic migraine should be individualized.