Skip to content

Preventive treatment

TL;DR — Prevention is warranted when attack frequency, disability, acute-treatment failure/contraindication or overuse risk makes attack-only care inadequate; the threshold is individualized rather than a single monthly-day cutoff. A 74-trial network meta-analysis (32,990 participants) found high-certainty ≥50%-response evidence for CGRP monoclonal antibodies, preventive gepants and topiramate, moderate-certainty evidence for beta-blockers, valproate and amitriptyline, and more discontinuation harms with several repurposed oral drugs (Lampl 2023, PMID 37208596). Pivotal CGRP trials typically yield active–placebo differences around 1–2 monthly migraine days, with large placebo responses and meaningful responder minorities (Goadsby 2017, PMID 29171821; Dodick 2018, PMID 29800211). Direct trials found lower adverse-event discontinuation and higher ≥50% response with erenumab or atogepant than topiramate, although both were sponsor-funded and studied defined populations (Reuter 2022, PMID 34743579; Reuter 2026, PMID 42492556). No validated biomarker selects the best first preventive, so comorbidity, reproductive safety, contraindications, cost, route and patient priorities drive choice.

What prevention is trying to change

Goal Suitable measure Common misreading
Fewer attacks Monthly migraine days (MMD) Mean reduction hides responders/nonresponders
Less total headache Monthly headache days Not identical to MMD, especially in chronic migraine
Lower disability MIDAS/HIT-6 or functional diary Score change may lag frequency change
More reliable acute control Acute-medication days and response Fewer doses can reflect avoidance rather than benefit
Better tolerability/adherence Discontinuation and persistence Trial completion overstates real-world persistence
Disease modification Sustained benefit after withdrawal Rarely tested; suppression is not modification

A preventive trial should be long enough for titration and an adequate observation window, then judged on frequency, severity, function, acute use and harms. A ≥50% MMD reduction is a common episodic endpoint; ≥30% can be meaningful in chronic migraine, but response should be prespecified rather than retrofitted.

Comparative evidence and its limits

The 2015 network meta-analysis found several repurposed drugs better than placebo but few robust differences between active agents and substantial trial heterogeneity (Jackson 2015, PMID 26172390). A 2019 clinical update framed gepants, ditans, CGRP antibodies and neuromodulation as additions to—not replacements for—outcome-based individual selection (Goadsby 2019, PMID 31341711). The 2023 update found CGRP antibodies/gepants and topiramate supported by high-certainty responder evidence; CGRP-targeted therapies had the most favourable combined efficacy/tolerability profile (Lampl 2023, PMID 37208596).

Network comparisons are not randomized head-to-head comparisons. Older oral-drug trials were smaller, used variable criteria and reporting, while newer registration programs are larger and more standardized. Better apparent tolerability can be real and also partly reflect ascertainment and trial-era differences.

Class Evidence position Frequent limiting issues
Topiramate High-certainty efficacy Paresthesia, cognitive symptoms, weight loss, teratogenicity
Beta-blocker Moderate-certainty efficacy Fatigue, bradycardia, hypotension, asthma/exercise limits
Candesartan Two small crossover RCTs Hypotension; pregnancy contraindication; smaller database
Valproate Moderate-certainty efficacy Weight/metabolic effects and major reproductive risk
Amitriptyline Moderate-certainty efficacy Sedation, anticholinergic effects, weight gain
CGRP antibody High-certainty efficacy, low trial discontinuation Cost/access, injection/infusion, long-term special-population gaps
Preventive gepant High-certainty class evidence Cost, adherence, interactions, shorter long-term history

Conventional oral preventives

Topiramate

Topiramate is among the best-supported conventional preventives, but titration and adverse effects determine usable dose. Cognitive slowing, word-finding difficulty, paresthesia, appetite/weight change and mood effects commonly drive discontinuation; kidney-stone and ocular risks are less common but important. Pregnancy-prevention considerations are central.

The evidence asymmetry is instructive: topiramate increases ≥50% response with high certainty, yet moderate-certainty evidence shows more adverse-event discontinuation than placebo (Lampl 2023, PMID 37208596). A drug can be efficacious and still lose effectiveness at population level through non-persistence.

Beta-blockers and candesartan

Propranolol/metoprolol have long use and moderate-certainty evidence, particularly useful when another indication aligns. Fatigue, reduced exercise tolerance, low blood pressure/pulse and asthma can make the same choice poor.

In a 57-person crossover trial, 12-week headache days averaged 13.6 with candesartan 16 mg versus 18.5 with placebo; migraine days were 9.0 versus 12.6, and 40.4% achieved ≥50% reduction in migraine days relative to placebo (Tronvik 2003, PMID 12503978). A later 72-person triple-blind crossover found migraine days/4 weeks of 2.95 with candesartan, 2.91 with propranolol and 3.53 with placebo; responder rates were 43%, 40% and 23% (Stovner 2014, PMID 24335848). Small crossover samples and carryover assumptions limit precision.

Valproate and antidepressants

Anticonvulsant reviews support valproate efficacy, but reproductive toxicity and metabolic harms sharply constrain its role (Chronicle 2004, PMID 15266476). The efficacy question cannot be separated from avoidable fetal exposure.

Amitriptyline may align with insomnia or another pain syndrome, while sedation, anticholinergic load and weight gain limit use. Antidepressant meta-analysis found benefit signals but heterogeneous, often older trials (Xu 2017, PMID 28557171). Choice should not imply that migraine is a depressive disorder.

CGRP monoclonal antibodies

Agent Target/route Anchor outcome Trial boundary
Erenumab Receptor; monthly SC STRIVE: MMD −3.2/−3.7 vs −1.8 placebo; ≥50% response 43.3%/50.0% vs 26.6% Episodic; 6 months (Goadsby 2017, PMID 29171821)
Fremanezumab Ligand; monthly/quarterly SC MMD difference vs placebo −1.5 (95% CI −2.01 to −0.93) monthly; −1.3 (−1.79 to −0.72) quarterly Excluded failure of two preventive classes (Dodick 2018, PMID 29800211)
Galcanezumab Ligand; monthly SC EVOLVE-1 reduction −4.7/−4.6 vs −2.8 placebo; no meaningful efficacy difference between doses Episodic; 6 months (Stauffer 2018, PMID 29813147)
Eptinezumab Ligand; quarterly IV PROMISE-2 chronic MMD −7.7/−8.2 vs −5.6 placebo Infusion and large placebo change (Lipton 2020, PMID 32209650)

The antibody class was built specifically for migraine; early trial synthesis and galcanezumab-focused review describe mechanism-specific development and placebo-controlled efficacy (Israel 2018, PMID 29623520; Gklinos 2021, PMID 33803190). Injection-site reactions are common class events; constipation and blood-pressure concerns are particularly tracked with receptor blockade. Rare or delayed harms require postmarketing denominators larger and more diverse than RCTs.

DELIVER enrolled adults with two to four previous preventive failures and showed eptinezumab benefit in a more treatment-experienced population, partly addressing registration-trial selection (Ashina 2022, PMID 35716692). Real-world reviews generally find effectiveness consistent with trials but inherit uncontrolled selection, reimbursement rules and missing follow-up (Mavridis 2021, PMID 34358126).

Preventive gepants

Rimegepant 75 mg every other day reduced MMD by 4.3 days versus 3.5 with placebo in weeks 9–12; least-squares mean difference was −0.8 days (95% CI −1.46 to −0.20), with adverse events in 36% of both groups (Croop 2021, PMID 33338437). The result is statistically positive and clinically modest on the mean; individual responder distributions and dual acute/preventive utility matter.

Atogepant has daily regimens and evidence in episodic and chronic migraine; class network meta-analysis across 19 phase-3 studies and 14,584 participants supports efficacy of antibodies and gepants but cannot establish a universal within-person ranking (Haghdoost 2023, PMID 36855951). A chronic-migraine-focused gepant review treats encouraging RCT results as an update rather than proof of equivalence across phenotypes or agents (Cho 2023, PMID 37656319). Gepants offer oral, short-half-life prevention and avoid injection, while daily/every-other-day adherence and CYP/transporter interactions remain practical constraints (Moreno-Ajona 2020, PMID 32251023).

Head-to-head evidence

HER-MES was the first pivotal active comparison. In 777 participants, discontinuation for adverse events was 10.6% with erenumab and 38.9% with topiramate (OR 0.19, 95% CI 0.13–0.27); ≥50% response was 55.4% versus 31.2% (OR 2.76, 2.06–3.71) over the last three months (Reuter 2022, PMID 34743579).

Interpretation constraints:

  • double-dummy design preserved blinding but created route burden in both groups;
  • topiramate titration/tolerability is intrinsically different from fixed antibody dosing;
  • the primary endpoint was discontinuation, not MMD change;
  • sponsor and Germany-specific access context affect generalizability;
  • completion-on-treatment post-hoc efficacy creates selection and should not replace ITT results (Ehrlich 2022, PMID 36380284).

Still, the tolerability gap is large enough to challenge mandatory multi-oral failure policies.

TEMPLE subsequently randomized 545 participants to atogepant 60 mg daily or tolerated-dose topiramate. Adverse-event discontinuation was 12% versus 30% (RR 0.4, 95% CI 0.3–0.6), ≥50% response was 64% versus 39% (RR 1.6, 1.4–2.0), and the MMD contrast over months 4–6 was −1.8 days (95% CI −2.5 to −1.0) in favour of atogepant (Reuter 2026, PMID 42492556). The 24-week double-blind result supplies direct oral-versus-oral evidence; 96% White enrollment and sponsor context constrain transportability.

Starting, monitoring and stopping

Phase Minimum record
Baseline ≥4 weeks of headache/migraine days, acute days, disability and current exposures
Titration Dose, adherence, emerging harms and reasons for missed doses
Efficacy window MMD/MHD change, ≥50%/≥30% response, function and acute use
Continuation Durability, burden between attacks, patient-valued benefit
De-escalation Relapse after interval extension/withdrawal and retreatment response

Stopping for inefficacy should occur only after an adequate dose/window unless harms require earlier cessation. Conversely, statistically detectable improvement without meaningful functional benefit may not justify ongoing cost or risk.

The American Headache Society’s 2024 position statement moved CGRP-targeted therapies to first-line-option status without requiring failure of nonspecific preventives, based on efficacy, tolerability and migraine-specific development (Charles 2024, PMID 38466028). This is a policy position, not proof that every patient should start with the most expensive option.

Combination and sequencing

Evidence for combining oral preventives, CGRP therapy and onabotulinumtoxinA remains observational rather than randomized factorial evidence. A 2026 exploratory meta-analysis pooled six extractable observational studies and reported a 7.9-day MHD reduction (95% CI 5.7–10.2) and 51% ≥50% response, but lacked controlled comparative designs, showed moderate-to-high heterogeneity and imputed dispersion measures in several studies (Sarvari Soltani 2026, PMID 41721358). Mechanistic non-redundancy makes combination plausible, particularly in chronic migraine with residual burden, but additive cost and rare-harm detection matter.

After one CGRP antibody fails, switching ligand↔receptor, to a gepant or to onabotulinumtoxinA is common. A prospective non-randomized 53-person cohort found improvement after onabotulinumtoxinA following CGRP-antibody failure, but selection and regression to the mean prevent a comparative sequencing inference (Silvestro 2026, PMID 42646770). Nonresponse may reflect target biology, dose/exposure, adherence, regression to mean or inadequate time; no validated rule distinguishes these. Cross-trial differences should not be used as indirect proof that a particular antibody is superior (Mitsikostas 2017, PMID 28240610).

Special safety boundaries

Preventive selection changes with pregnancy potential, hypertension, constipation, kidney stones, mood disorder, asthma, body-weight goals and interacting medicines. Long half-lives make antibody exposure persist after stopping. Conventional drugs have longer observational experience but not necessarily better safety.

A network safety analysis found CGRP antibodies/gepants generally similar to placebo for discontinuation and serious events in phase-3 trials, but such trials are not powered for rare events or pregnancy (Messina 2023, PMID 36786548). “Well tolerated in trials” should remain bounded by who was enrolled and how long they were followed.

Open questions

  • Which baseline clinical or biological marker selects topiramate, antibody, gepant or another oral preventive before serial trials? (Lampl 2023, PMID 37208596)
  • Does first-line CGRP access improve long-term function and total cost compared with step therapy? (Charles 2024, PMID 38466028; Reuter 2022, PMID 34743579)
  • When and how can successful prevention be withdrawn without avoidable relapse? (Schoenen 2020, PMID 32758365)
  • Do combination regimens produce additive benefit in randomized trials, and which components can then be removed? (Mavridis 2021, PMID 34358126)
  • Can early effective prevention alter chronification rather than suppress attacks only during exposure? (Ashina 2021, PMID 33773610)

References

  1. Lampl C, et al. Comparative effectiveness of migraine preventive drugs: systematic review and network meta-analysis. J Headache Pain. 2023. PMID 37208596
  2. Jackson JL, et al. Comparative effectiveness meta-analysis of drugs for migraine prophylaxis. PLoS One. 2015. PMID 26172390
  3. Chronicle E, Mulleners W. Anticonvulsant drugs for migraine prophylaxis. Cochrane Database Syst Rev. 2004. PMID 15266476
  4. Xu XM, et al. Antidepressants for prevention of migraine in adults: meta-analysis. Eur J Neurol. 2017. PMID 28557171
  5. Tronvik E, et al. Candesartan for migraine prophylaxis: randomized controlled trial. JAMA. 2003. PMID 12503978
  6. Stovner LJ, et al. Candesartan versus propranolol for migraine prophylaxis. Cephalalgia. 2014. PMID 24335848
  7. Goadsby PJ, et al. Controlled trial of erenumab for episodic migraine. N Engl J Med. 2017. PMID 29171821
  8. Dodick DW, et al. Fremanezumab for prevention of episodic migraine. JAMA. 2018. PMID 29800211
  9. Stauffer VL, et al. Galcanezumab for prevention of episodic migraine: EVOLVE-1. JAMA Neurol. 2018. PMID 29813147
  10. Lipton RB, et al. Eptinezumab in chronic migraine: PROMISE-2. Neurology. 2020. PMID 32209650
  11. Ashina M, et al. Eptinezumab after two-to-four preventive failures: DELIVER. Lancet Neurol. 2022. PMID 35716692
  12. Croop R, et al. Oral rimegepant for preventive treatment of migraine. Lancet. 2021. PMID 33338437
  13. Reuter U, et al. Erenumab versus topiramate for migraine prevention. Cephalalgia. 2022. PMID 34743579
  14. Ehrlich M, et al. Erenumab versus topiramate: post-hoc efficacy analysis from HER-MES. J Headache Pain. 2022. PMID 36380284
  15. Charles AC, et al. CGRP-targeting therapies are a first-line option: AHS position statement update. Headache. 2024. PMID 38466028
  16. Messina R, et al. Safety and tolerability of CGRP antibodies and gepants: network meta-analysis. Cephalalgia. 2023. PMID 36786548
  17. Haghdoost F, et al. Efficacy of CGRP mAbs and gepants: network meta-analysis. Cephalalgia. 2023. PMID 36855951
  18. Moreno-Ajona D, et al. Gepants: what could be their role in migraine treatment? Curr Opin Neurol. 2020. PMID 32251023
  19. Cho SJ, et al. Update of gepants in chronic migraine. Curr Pain Headache Rep. 2023. PMID 37656319
  20. Israel H, et al. CGRP monoclonal antibodies for preventive treatment. Curr Pain Headache Rep. 2018. PMID 29623520
  21. Mitsikostas DD, et al. CGRP monoclonal antibodies: comparisons across randomized studies. Curr Opin Neurol. 2017. PMID 28240610
  22. Mavridis T, et al. CGRP monoclonal antibodies from clinical studies to real-world evidence. Pharmaceuticals. 2021. PMID 34358126
  23. Schoenen J, et al. Monoclonal antibodies blocking CGRP transmission: added value in prevention. Rev Neurol (Paris). 2020. PMID 32758365
  24. Gklinos P, et al. Role of galcanezumab in migraine prevention. Pharmaceuticals. 2021. PMID 33803190
  25. Goadsby PJ. Primary headache disorders: five new things. Neurol Clin Pract. 2019. PMID 31341711
  26. Ashina M, et al. Migraine: disease characterisation, biomarkers, and precision medicine. Lancet. 2021. PMID 33773610
  27. Reuter U, et al. Atogepant versus topiramate for migraine prevention: TEMPLE. Lancet Neurol. 2026. PMID 42492556
  28. Sarvari Soltani M, et al. Dual-mechanism onabotulinumtoxinA and anti-CGRP antibody prevention: exploratory meta-analysis. Eur J Med Res. 2026. PMID 41721358
  29. Silvestro M, et al. OnabotulinumtoxinA after CGRP-antibody treatment failure: multicentre real-world study. Toxins (Basel). 2026. PMID 42646770