Knowledge Library · Evidence vault

Caffeine and Health Evidence Vault.

Source-validated peer-reviewed evidence on caffeine, coffee, and the engineered products that deliver them. The popular conflation of coffee and caffeine does most of the marketing work; the regulatory and academic literatures keep them separate. Every line of evidence below is tagged for what it is actually about — coffee (a complex beverage carrying caffeine plus chlorogenic acids, diterpenes, polyphenols, magnesium) or caffeine (the isolated molecule).

Stale-date reminder: re-check November 2026 — the upper-bound stroke evidence is contested at the high end (Pang 2021 vs Smyth 2024 INTERSTROKE).

⭐ The regulatory ceiling

EFSA 2015 Scientific Opinion: 400 mg per day, 200 mg in pregnancy.

Primary source. European Food Safety Authority Panel on Dietetic Products, Nutrition and Allergies (NDA Panel), Scientific Opinion on the safety of caffeine, EFSA Journal 2015. DOI 10.2903/j.efsa.2015.4102.

Tagged: caffeine evidence. Caffeine the molecule, from all dietary sources. Not coffee specifically.

What the opinion says.

Why this matters. Canonical European safety reference. NHS, ACOG, RCOG, and the UK Food Standards Agency align with the EFSA pregnancy figure. This is the all-sources daily-intake ceiling, against which every shelf vehicle below is read.

Citation. EFSA NDA Panel. Scientific Opinion on the safety of caffeine. EFSA J. 2015. DOI 10.2903/j.efsa.2015.4102.

BMJ Umbrella 2017 — coffee evidence

201 meta-analyses on coffee and health: protective at moderate intake.

Primary source. Poole R, et al. (2017). Coffee consumption and health: umbrella review of meta-analyses of multiple health outcomes. BMJ, 359:j5024.

Tagged: coffee evidence. Coffee the beverage, not caffeine in isolation. The distinction matters: protective associations may be driven by chlorogenic acids and other non-caffeine bioactives. Decaffeinated coffee shares many of the protective liver and metabolic associations — suggesting the effect is largely not the caffeine.

OutcomeMagnitude (3–4 cups/day vs none)
All-cause mortalityRR 0.83 (17% lower)
Cardiovascular mortalityRR 0.81 (19% lower)
Cardiovascular diseaseRR 0.85 (15% lower)
Incident cancer (high vs low)RR 0.82 (18% lower)

What's protective. Statistically significant inverse associations across type 2 diabetes, renal stones, gout, gallstones, Parkinson disease, Alzheimer disease, depression, leukaemia (overall), colorectal cancer, liver cancer, chronic liver disease, cirrhosis, and endometrial cancer.

What's adverse. Lung cancer (residual confounding by smoking suspected), urinary tract cancer, pregnancy loss, low birth weight, preterm birth, acute leukaemia in childhood, fracture risk in women.

Authors' conclusion. Moderate coffee consumption appears safe or beneficial for the general adult population, with the explicit exceptions of pregnancy and women at high fracture risk.

Citation. Poole R et al. BMJ 2017;359:j5024.

The contested upper bound

Stroke at high coffee intake: prospective cohort vs case-control.

Two large recent papers reach different conclusions about coffee intake above 4 cups per day. Both findings hold within their study designs. The contradiction is real and unresolved in the published literature.

Pang 2021 — protection plateaus, no harm.

Primary source. Pang Y, et al. (2021). J Stroke Cerebrovasc Dis, PMID 33188952. Pooled 21 studies, 30 cohorts, n > 2.4 million. Coffee, not caffeine isolate.

Smyth 2024 INTERSTROKE — harm at > 4 cups per day.

Primary source. Smyth A, et al. (2024). Tea, coffee and risk of stroke: an INTERSTROKE international case-control study. Int J Stroke, 19(9):1053–1063. DOI 10.1177/17474930241264685. n ≈ 27,000 across 27 countries; ~13,500 first-stroke cases. Coffee, not caffeine isolate.

Critical reading discipline. The high-intake stroke signal in INTERSTROKE 2024 is in the general study population. Hypertension was a covariate adjusted for in the multivariable model — not a subgroup driving the finding. Any copy stating "this only applies to people with high blood pressure" is wrong canon and creates regulatory exposure. The honest reading is: prospective cohorts plateau; case-control finds harm; the upper bound is contested at the high end across study designs.

UK Biobank 2024 — both

Caffeine 200–300 mg/day, coffee 3 cups: lowest cardiometabolic multimorbidity risk.

Primary source. Lu Z, et al. (2024). Habitual coffee, tea, and caffeine consumption and cardiometabolic multimorbidity. J Clin Endocrinol Metab. DOI 10.1210/clinem/dgae552.

Tagged: both. Two parallel analyses on the UK Biobank cohort — reading them separately is the disciplined move.

Both metrics converge on the same picture: moderate intake sits at the bottom of the U-curve.

Citation. Lu Z et al. J Clin Endocrinol Metab 2024. DOI 10.1210/clinem/dgae552.

The structural critique

Caffeine the molecule vs caffeine the engineered delivery vehicle.

Most of the protective associations in the literature come from cohorts whose dominant caffeine source is filter coffee or tea, consumed without large added-sugar loads. The category-level findings do not transfer cleanly to:

The structural pattern: caffeine is dose-dependent and well-studied; the engineered products that deliver it are formulated to overshoot the dose. The manufacturer is creator of the gap — sourcing the ingredients, engineering the formulation, marketing at adolescents, funding the research that biases the science (see Industry Funding Bias in Nutrition Research).

Shelf vehicles read against the EFSA 400 mg/day ceiling.

Typical mid-range values from manufacturer disclosures and the USDA FoodData Central database. Individual products vary.

VehicleCaffeine% of daily ceilingSugar load
Energy drink, 250 ml standard can80 mg20%27 g
Energy drink, 500 ml large can160 mg40%54 g
Energy drink, high-strength (568 ml Monster, 473 ml Reign)160–300 mg40–75%varies by SKU
Ready-to-drink coffee, 250 ml80–130 mg20–33%8–18 g
Takeaway latte, large (16 oz)150–200 mg38–50%0 g (unsweetened) / 25–45 g (syrup-flavoured)
Filter coffee, home-brewed mug (240 ml)80–120 mg20–30%0 g
Espresso shot (30 ml)60–80 mg15–20%0 g
Cola, 330 ml can32–42 mg8–11%35 g
Dark chocolate, 50 g serving (70%)30–40 mg8–10%varies

Two adolescents drinking one large 500 ml energy drink each on a school morning are at 40% of the adult EFSA daily caffeine ceiling and over the WHO adult daily free-sugars ceiling, before lunch. The molecule-versus-vehicle distinction is not abstract; it is the actual food environment.

High-risk groups

Six populations for whom the population-average reading does not apply.

Pregnancy.
EFSA ceiling 200 mg per day from all sources. The BMJ umbrella review reports residual associations with low birth weight, preterm birth, and pregnancy loss even at intakes below the ceiling, leading some authors to argue for a more conservative threshold. NHS, RCOG, ACOG all align on 200 mg per day. Caffeine clearance halves in the third trimester; perceived intake and effective intake diverge.
Children and adolescents.
EFSA basis: 3 mg per kg body weight per day. For a 30 kg primary-school child this is 90 mg per day; for a 50 kg early-adolescent it is 150 mg per day. One large energy drink can carry more than this. Caffeine in adolescents is associated with poorer sleep, higher anxiety, and (in mixed-formulation energy drinks) cardiovascular events. The engineered-vehicle pattern is most acute here.
Hypertension.
Caffeine acutely raises blood pressure in non-habitual users; tolerance develops with regular use. Moderate coffee intake in well-controlled hypertension is acceptable per the BMJ umbrella review; AMA clinical guidance advises avoiding caffeine before BP measurement and certain cardiac tests. INTERSTROKE 2024 does not single out hypertensive subgroups; the high-intake stroke signal is in the general population.
Arrhythmia (atrial fibrillation, supraventricular tachycardia).
Population-level evidence does not support a strong association between moderate coffee intake and AF incidence; observational studies suggest moderate coffee may be neutral or mildly protective. However, individual symptomatic patients commonly identify caffeine as a personal trigger. Clinical guidance is individualised.
Anxiety and panic disorder.
Caffeine antagonises adenosine and increases noradrenaline; both are direct mechanisms for panic-symptom provocation. The DSM-5 recognises caffeine-induced anxiety disorder as a discrete diagnostic entity. Patients with panic disorder are commonly more sensitive than the general population. The threshold for symptomatic provocation can be well below the EFSA general ceiling.
Insomnia and shift workers.
EFSA flags single doses of 100 mg as enough to extend sleep latency in some adults. Caffeine half-life is 4–6 hours in healthy adults, extending to 7–9 hours in pregnancy, with oral contraceptive use, and with reduced CYP1A2 activity. A 200 mg afternoon coffee has measurable sleep effects for many adults at bedtime. Shift workers using caffeine for alertness should track timing relative to off-shift sleep.
Conflicts and uncertainties

Three live contradictions in the literature.

Naming them strengthens the evidence base; hiding them weakens it.

1. Stroke at high intake: prospective cohort vs case-control.

Pang 2021 (prospective cohort meta-analysis, n > 2.4M) shows protection peaking at 3–4 cups per day with a plateau beyond. Smyth 2024 INTERSTROKE (case-control, n ≈ 27,000) shows higher stroke odds at > 4 cups per day in the general population. Both findings hold within their study designs. The contradiction is between study designs, not between subgroups. Resolution would require an individual-participant-data meta-analysis spanning both designs. Until that exists, the upper bound of safe coffee intake is genuinely contested.

2. Caffeine in pregnancy below 200 mg per day.

EFSA and the major obstetric bodies set 200 mg per day as the pregnancy ceiling. Several authors of the BMJ umbrella review have argued the underlying meta-analyses show residual signal for pregnancy loss and low birth weight even at intakes below this, suggesting a more conservative threshold. The regulatory threshold and the conservative-author threshold are not the same. Both readings are defensible; the gap between them is genuinely a judgement call.

3. Long-term effects of energy drinks in adolescents.

Cross-sectional surveys (BMJ Open 2017 systematic review and successors) document associations with poor sleep, headache, irritability, and mental-health symptoms. Longitudinal cohort evidence is the gap. The PHE 2018 evidence base for the proposed under-16 sales restriction was acknowledged as observational. The UK voluntary-restriction landscape sits on this evidence base. The honest reading is that the precautionary case is strong; the longitudinal-causal case is incomplete.

Sources — full citation list

Copy-paste-ready primary sources.

  1. EFSA NDA Panel. Scientific Opinion on the safety of caffeine. EFSA J. 2015. DOI 10.2903/j.efsa.2015.4102.
  2. Poole R, et al. Coffee consumption and health: umbrella review of meta-analyses. BMJ 2017;359:j5024.
  3. Pang Y, et al. Coffee consumption and risk of stroke: a meta-analysis. J Stroke Cerebrovasc Dis 2021. PMID 33188952.
  4. Smyth A, et al. Tea, coffee and risk of stroke: an INTERSTROKE international case-control study. Int J Stroke 2024;19(9):1053–1063. DOI 10.1177/17474930241264685.
  5. Lu Z, et al. Habitual coffee, tea, and caffeine consumption and cardiometabolic multimorbidity. J Clin Endocrinol Metab 2024. DOI 10.1210/clinem/dgae552.
  6. Simonin C, Duru C, et al. Association between caffeine intake and age at onset in Huntington's disease. Neurobiol Dis 2013;58:179–182.
  7. Adrissi J, et al. Caffeine consumption in Huntington disease: a single-centre survey. Tremor Other Hyperkinet Mov 2024.
  8. Visram S, et al. Consumption of energy drinks by children and young people: a rapid review. BMJ Open 2017.
  9. Public Health England (2018). Energy drinks consultation evidence base.
  10. Caffeine Consumption Patterns review. PMC12269801, 2024.

Reading note: the AMA "What doctors wish patients knew about the impact of caffeine" piece is a representative consumer-facing claim surface that the FSMA template is built to deconstruct — included for context, not as a primary source. It compresses associational findings into clinical-style guidance and presents the EFSA 400 mg/day figure as a clinical ceiling rather than a population-safety threshold.

🗺️ The regulation map

UK 2026: caffeine across food, drink, supplements, and advertising.

Caffeine sits across multiple UK regulatory frames. The map below covers the surfaces a typical UK shopper encounters.

SurfaceMechanismUK status 2026Upstream actorInternational parallel
EFSA 2015 ceiling 400 mg / day adults; 200 mg / day pregnancy; 3 mg / kg / day children & adolescents Reference standard. EFSA Scientific Opinion 2015 (DOI 10.2903/j.efsa.2015.4102); UK retained; NHS, RCOG, FSA align. EFSA NDA Panel; NHS / RCOG dietary advice. FDA 400 mg/day adult ceiling; ACOG pregnancy 200 mg/day; Health Canada 400 mg/day.
Caffeine ingredient labelling "Contains caffeine" / "high caffeine content" labels on beverages Mandatory. FIC 1169/2011 retained Article 10: drinks with caffeine >150 mg/L (other than tea, coffee, and their preparations) must carry "high caffeine content. Not recommended for children or pregnant or breastfeeding women" plus the caffeine content in mg per 100ml. Beverage manufacturer. EU FIC; broadly similar in major Western markets.
Energy drinks — under-16 sales Voluntary retailer restrictions on sale of high-caffeine energy drinks to under-16s Voluntary (retailer-led). Most UK major multiples and convenience-store chains restrict sale to under-16s under voluntary policy; government consultation on statutory under-16 ban (2018) did not progress to legislation but voluntary practice has held. Retailers (predominantly); some manufacturers signal under-16 marketing avoidance. Some EU member states have or have considered statutory under-16 restrictions; varies by jurisdiction.
Caffeine in supplements / sports nutrition Caffeine as added ingredient in food supplements Partial. Food supplements regulated under Food Supplements (England) Regulations 2003 (SI 2003/1387) and parallel SIs; caffeine-as-supplement above EFSA general safe-intake levels approaches MHRA medicine-classification territory. Supplement manufacturer. FDA dietary supplement classification distinct from food regulation; EU food-supplement Directive 2002/46/EC.
HFSS placement and advertising Sugar-sweetened caffeinated beverages within SI 2021/1368 and ASA/CAP HFSS rules Restricted (HFSS only). Sugar-sweetened energy drinks and sweetened caffeinated beverages fall within HFSS placement (Oct 2022), volume promotions (Oct 2025), and advertising (Jan 2026, first enforcement April 2026) regimes where they exceed NPM threshold. Beverage manufacturer; retailer arrangement. See Impulse Buying Triggers and Food Marketing to Kids.
Soft Drinks Industry Levy Sugar-sweetened caffeinated beverages within SDIL bands Statutory. SDIL in force April 2018; major energy-drink reformulation documented (decoded in Reformulation Tracking: Scarborough 2020 PLOS Med; Pell 2021 BMJ). Beverage manufacturer. Mexico (2014), South Africa (2018), and equivalents internationally.

How to read the map. The EFSA reference ceiling is widely-adopted but not statutorily binding on consumers; it shapes UK dietary advice. The FIC labelling requirement is statutory and applies universally. The under-16 sales restriction is voluntary retailer practice. The SDIL and HFSS frames bear on sweetened caffeinated beverages specifically. Caffeine-as-supplement at high doses approaches the medicine-classification boundary — MHRA scrutiny applies where the dose-and-claim combination crosses the threshold.

Defamation-safety statement

What this brief does not claim.

This evidence vault contains no allegation of unlawful conduct against any named UK or international manufacturer, retailer, broadcaster, or food business operator. Discussion of caffeine, coffee, and engineered-delivery-vehicle products is general industry-practice description and peer-reviewed evidence summary supported by primary sources (EFSA 2015 Scientific Opinion; Poole 2017 BMJ umbrella review; Pang 2021; Smyth 2024 INTERSTROKE; Lu 2024; Geha 2000 J Nutr; multiple Public Health England, FSA, NHS, RCOG references).

Named-party reference policy. Where energy-drink and caffeinated-product brands are referenced (Monster, Reign, generic energy-drink and caffeinated-product categories), references are limited to public-record product specifications — caffeine content per pack, pack size, sugar load — as disclosed on the products themselves and in manufacturer corporate communications. No factual claim is made about any specific manufacturer's commercial conduct, marketing practice, or contractual arrangement beyond what the named parties have themselves placed in the public record. The structural critique (the molecule-versus-vehicle distinction; engineered products formulated to overshoot the EFSA safe-intake reference) is applied to the industry pattern rather than to any specific named party's conduct.

Related & further reading

Where to go next.

The full Knowledge Library carries five streams. The peer-reviewed evidence base for the structural critique is in Industry Funding Bias in Nutrition Research. The cognitive and brain-health evidence is in UPF Brain & Cognitive Claims. The behaviour-change defensibility argument for decision-point capture is in the Behaviour Change & Decision-Point Capture vault. The frozen-aisle expression of the same structural critique is in Frozen Food in the UK. The FSMA gold-standard evidence-vault companions to this brief are Impulse Buying Triggers (the engineered-shopping-environment piece), Food Marketing to Kids (developmental psychology and HFSS marketing), Brand vs Manufacturer (the structural transparency gap), Reformulation Tracking (SDIL evidence including energy-drink reformulation), Cultural Food Myths (diaspora-community decoder framework), Global Staple Foods, Dietary Patterns, and Carbohydrate Types (the comprehensive carbohydrate decoder underpinning the sugar load on caffeinated products). The label-reading mechanics are decoded in The SCANSMART Method, Ingredient Rules, Nutrition Claims, Decoded, and Front-of-Pack Labels.

Caffeine and Health Evidence Base v1.3 (gold-standard depth) · Compiled 2 May 2026; gold-standard upgrade 11 May 2026 · Stale-date reminder: re-check November 2026 — the upper-bound stroke evidence is contested at the high end (Pang 2021 vs Smyth 2024 INTERSTROKE) · Defamation-safe; named-party references public-record-only.