Chocolate, cheese and caffeine are blamed for migraine far more often than tested

The classic dietary migraine triggers are reported by a minority of patients and have never been established as causes. A meta-analysis published this month pooled seven observational studies and found that chocolate was named as a trigger by 13.9% of patients (95% CI 7.0%–21.0%), caffeine by 12.0% (6.0%–19.0%) and cheese by 8.0% (3.0%–12.0%) [s1]. Every one of those figures comes from people reporting what they think set off their attacks — which is the weakest kind of evidence that exists for a causal claim, and the reviewers say so.

The review searched PubMed, MEDLINE, Ovid and Scopus from database inception to 11 November 2024, included eight studies and carried seven into the quantitative synthesis [s1]. Four studies were rated good quality and four fair on the NHLBI assessment tool [s1]. Heterogeneity was extreme: I² exceeded 95% for every analysis, meaning the studies disagreed with each other far more than chance would explain [s1]. The authors' conclusion is that effects vary considerably between individuals, that routine avoidance of these foods cannot be universally recommended, and that prospective studies are needed to clarify whether the relationships are causal at all [s1].

The problem the field named thirty years ago

A review in Current Pain and Headache Reports titled itself after the logical fallacy in question — post hoc ergo propter hoc, "after this, therefore because of this" [s4]. Its subject is precisely the gap this meta-analysis measures: migraine patients commonly report foods, alcohol, weather and atmospheric changes, light, sound and odours as factors that trigger or aggravate attacks, physicians frequently repeat those attributions as advice, and most clinical studies attempting to test them have produced conflicting results [s4]. The review's stated aim is to analyse what supports or refutes a correlation between environmental stimuli and attacks, and it highlights the methodological difficulty of investigating the link at all [s4].

That difficulty is structural. A migraine attack is not an instantaneous event; it unfolds over hours. Anything a person did in the hours beforehand is available to be blamed, and the things people notice themselves doing — eating a particular food, going out into bright light — are the ones that get recorded. An attribution built that way will be reported at some rate regardless of whether the association exists.

What people actually name

The list is broader than food. In a hospital cohort of 1,487 children and young people aged 5 to 17 attending a paediatric headache clinic, 568 — 38% — reported any trigger at all [s5]. Of those, 71% named one trigger, 21% two and 7% three or more [s5]. Environmental factors dominated, cited by 424 of 568 (75%) [s5]. The specific items most often named were stress (142 of 568, 25%), loud noise (140, 25%), light (135, 24%) and inadequate fluid intake (117, 20%) [s5]. More than half — 309 of 568, or 54.5% — said school was the only source of their triggers [s5].

That last figure is worth sitting with. It can be read as evidence that school environments provoke migraine, or as evidence that people attribute attacks to wherever they spend their time. Self-report cannot distinguish the two.

What holds up when it is deliberately administered

The cleanest test available is to give a substance under controlled conditions and see whether an attack follows. A systematic review of human pharmacological provocation studies in migraine with aura assembled fourteen such studies covering seven agents: CGRP, the potassium channel opener levcromakalim, glyceryl trinitrate, sildenafil, cilostazol, endothelin-1 and histamine [s2].

Even here, provocation is unreliable. Aura induction occurred far less frequently than headache induction across the studies [s2]. CGRP provoked aura in 17 of 53 participants — 32% — across three studies, with latencies ranging from 10 to 360 minutes [s2]. Opening vascular ATP-sensitive potassium channels with levcromakalim elicited aura in 14 of 52 participants (27%) in randomised crossover trials, with onset between 20 and 120 minutes [s2]. The remaining agents produced minimal or no aura response [s2].

The two substances with the strongest claim to being genuine biological triggers, administered directly at experimental doses, produce an attack in roughly a third of the people given them. No dietary item has anything close to that kind of evidence behind it.

Where environmental data is stronger than diaries

Air pollution and weather can be measured without asking anyone. A population-based case-crossover study drew on the Negev Migraine Cohort — 7,032 adult patients with migraine living in Be'er Sheva, Israel, identified from electronic medical records covering 2000 to 2023 — and linked daily pollutant and meteorological readings from fixed monitoring stations to emergency migraine-related encounters [s3].

Short-term exposure to nitrogen dioxide was associated with increased risk of an emergency migraine encounter (odds ratio 1.41, 95% CI 1.13–1.77), as was solar radiation (OR 1.23, 1.07–1.42) [s3]. Cumulative NO₂ exposure was associated with higher triptan use (incidence rate ratio 1.10, 1.00–1.21), as was fine particulate exposure over the preceding quarter (IRR 1.09, 1.00–1.19) [s3]. Weekly climate conditions modified the short-term effects: hot, dry summer weeks amplified the NO₂ association (OR 2.18, 1.06–3.30), and cold, humid winter weeks intensified the PM₂.₅ association (OR 3.78, 1.74–5.82) [s3].

Case-crossover designs compare each patient with themselves on different days, which removes confounding by anything stable about the person. The study's own stated limitation is the outcome: emergency healthcare encounters are a proxy for attacks, and who goes to an emergency department depends on more than whether their head hurts [s3].

What this leaves

The honest summary is uncomfortable for both sides of the trigger argument. Dietary triggers are reported by a real minority of patients, with wide and unstable estimates, and have not been shown to cause attacks [s1] [s4]. Some environmental exposures show a measurable association when measured objectively rather than recalled [s3]. And the substances with genuine experimental provocation credentials are not foods at all [s2].

The reviewers' own recommendation stops well short of an avoidance list: individualised identification supported by headache diaries, rather than blanket dietary restriction [s1]. Which is another way of saying that after decades of the same list being repeated, the field cannot tell a given patient which items on it apply to them.

This article is informational and is not medical advice.

Sources

• Assessment of Chocolate, Cheese, and Caffeine as Migraine Triggers: A Systematic Review and Meta-Analysis of Patient-Reported Trigger Rates — Journal of Nutrition and Metabolism, 2026-08-26
• Molecular triggers of migraine aura: A systematic review of human pharmacological provocation studies — Cephalalgia, 2026-04-01
• Acute Environmental Triggers and Intermediate-Term Modulators of Emergency Migraine-Related Health Care Encounters — Neurology, 2026-04-15
• Migraine and triggers: post hoc ergo propter hoc? — Current Pain and Headache Reports, 2013-10-01
• Patient- and parent-reported triggers of migraine attacks in children and adolescents — Developmental Medicine & Child Neurology, 2026-06-24