Every entry cites its own source. Each card carries a source line naming the study, dataset or label the figure came from, with the PubMed identifier linked where one exists, so any single number can be checked without having to trust the rest of the index. Where a food has never been analysed the card says so outright rather than showing a figure borrowed from something similar.
UK sources are preferred throughout. All carbohydrate composition comes from the UK's own reference dataset (CoFID) or from UK retail labels, which are declared under assimilated Regulation (EU) No 1169/2011 and are therefore already on the available-carbohydrate basis this index uses. Foreign data is used only where no UK source exists — at present that is espresso alone, taken from USDA FoodData Central because CoFID holds no espresso entry, and the card says so. The histamine side is unavoidably international for the numbers: no UK food-composition dataset has ever measured biogenic amines — not CoFID, not any MAFF or FSA surveillance survey — and no UK retail survey of histamine in British cheese, cured meat or fish exists. The per-food tolerance ratings come from the Swiss SIGHI list because nothing equivalent is published here. But the protocol is UK-sourced, and the dose thresholds and legal limits below are UK law.
What UK law and UK bodies actually say. Histamine is legally limited in Britain in fish and nothing else. Under assimilated Commission Regulation (EC) No 2073/2005, criterion 1.26 sets fishery products from high-histidine species at m = 100 and M = 200 mg/kg across nine samples; criterion 1.27 allows 200/400 mg/kg for brine-matured products, and 1.27a caps fermented fish sauce at 400 mg/kg. Those values were inserted by Regulation (EU) No 1019/2013 and Great Britain has not diverged from them since. There is no UK limit for cheese, cured meat, fermented vegetables or wine — where a cheese incident arises the FSA works from the fish action level as a starting point. For dose rather than concentration, the anchor the FSA itself applies is the EFSA acute reference dose of 50 mg of histamine per healthy person per meal (EFSA BIOHAZ, EFSA J 2011;9(10):2393), with a no-effect level around 50 mg and a lowest-observed-effect level of 75 mg; scaled by body weight that falls to roughly 10 mg per meal for a toddler and 19 mg for a young child. The interpretive bands used here are British in origin — Bartholomew and colleagues at the Central Public Health Laboratory measured 240 fish samples across a decade of UK incidents and set them at under 50 mg/kg safe for most people, 50–200 possibly toxic, 200–1000 probably toxic and above 1000 toxic (PMID 3428380).
How long to restrict, and why that matters more than the food list. Two UK sources give a protocol where none gives a food table. The British Dietetic Association advises eliminating high-histamine foods for two to four weeks and then reintroducing them gradually, under dietitian supervision, and states plainly that serum DAO cannot be used as a definitive diagnostic test and that there is no gold-standard food list — published lists vary widely and agree only on fermented foods and tomatoes. The British Society for Allergy and Clinical Immunology advises avoidance for no longer than three to four weeks. Neither NICE nor BSACI has issued a guideline on histamine intolerance at all, and the only formal guideline anywhere remains a German one from 2021 (PMID 34651098). This matters: an audit of 169 diet websites and apps published in the BDA’s own journal found that 95% recommended elimination but only 56% mentioned reintroduction (PMID 41611642), and a 2025 audit of 50 patient records found restriction lasting up to 78 weeks with 24% of patients losing 2–10 kg unintentionally (PMID 41017170). This index is a lookup tool, not a diet. Prolonged restriction carries real risks of nutrient deficiency and disordered eating, and reintroduction is part of the method, not an optional extra.
Serum DAO tests, genetic tests and DAO supplements. Worth knowing before paying for any of them. In a random Swedish population sample of 1,051 people, 44% fell below the commercial serum DAO cut-off while almost none reported histamine reactions, and DAO level showed no association with symptoms (PMID 37447214). In a single-blind placebo-controlled challenge, histamine intolerance was excluded in 85% of those who believed they had it, and 63% reacted to placebo (PMID 37648152). Direct-to-consumer AOC1 gene tests are sold in the UK, but 66% of 200 healthy newborns carried at least one so-called deficiency variant (PMID 40004469), so the finding is unremarkable rather than diagnostic. Both randomised trials of DAO supplements published since 2023 were funded by the manufacturer and both failed their primary endpoint, with placebo performing numerically better in each (PMID 41437955, PMID 37892588) — note that some review articles restate those results with figures that do not appear in the trials themselves. A properly powered 400-person trial is running in Barcelona and had not reported as of July 2026 (PMID 39796463).
Methodology. Values are pooled from peer-reviewed measurements, mostly by reverse-phase HPLC with OPA or dansyl-Cl derivatisation and fluorescence detection. AOAC method 977.13 is the official US reference for fish. Typical limit of detection ~0.01 mg/100 g; "below detection" is reported as <0.1.
Baseline reference for fresh produce. Sánchez-Pérez S et al. Biogenic Amines in Plant-Origin Foods. Foods 2018;7(12):205. PMID 30558197. Tsiasioti A, Tzanavaras PD. Foods 2022;11(20):3234. PMID 36291415 (methodological anchor; LOD 0.1 mg/kg).
Aged & fermented cheese. Linares DM et al. Factors influencing biogenic amines in dairy products. Crit Rev Food Sci Nutr 2012. PMID 22783233. Reinholds I et al. (blue cheeses) Foods 2020;9:93. PMID 31963130. Leuschner RG et al. (Gouda) Int J Food Microbiol 1998;44:15. PMID 9849780. Fernández M et al. (Brie) J Dairy Sci 2007;90:3763. PMID 16960050.
Fish & shellfish. Lehane L, Olley J. Histamine fish poisoning revisited. Int J Food Microbiol 2000;58:1. PMID 10898459. Visciano P et al. Biogenic Amines in Raw & Processed Seafood. Front Microbiol 2012;3:188. PMID 22675321. Visciano P et al. (overview) Foods 2020;9:1795. PMID 33287193. Morello S et al. (5-year N Italy survey) Toxins 2024;16:456. PMID 39591211. Tunç et al. (canned fish) Foods 2025;14:2314. PMID 40647066.
Cured & fermented meats. Michalski M et al. (Polish retail) J Vet Res 2021;65:497. PMID 35112002. Latorre-Moratalla ML et al. (Spanish dry-fermented) Food Chem Toxicol 2017;99:78. PMID 27818315. Suzzi G, Gardini F. (review) Int J Food Microbiol 2003;88:41. PMID 14527784.
Fermented soy. Byun & Mah (miso) J Food Sci 2012;77:T216. PMID 23140362. Li J et al. (Chinese soy sauces) Molecules 2019;24:1522. PMID 30999715. Toro-Funes N et al. (Spanish soy products) Food Chem 2015;173:1119. PMID 25466133.
Cocoa & chocolate. Granvogl M et al. J Agric Food Chem 2006;54:1730. PMID 16506826 (stable-isotope-dilution LC-MS/MS on cocoa beans). Delgado-Ospina J et al. Foods 2020;9:520. PMID 32326283 (effect of fermentation and roasting). Restuccia D et al. Food Addit Contam Part A 2015;32:1156. PMID 25833003. Paparella A et al. Foods 2025;14:255. PMID 39856922 (review tabulating the per-product figures used here). Sánchez-Pérez S et al. Front Nutr 2022;9:897028. PMID 35694170 (putrescine, cadaverine and tyramine competitively inhibit DAO in vitro). Free-histamine figures for chocolate are genuinely contradictory: the most sensitive methods find none at all, while the studies reporting the higher numbers also report more histamine in white chocolate than in dark — which cannot be right, as white chocolate contains no cocoa solids. Tyramine and phenylethylamine, by contrast, are measured consistently across every study, and are what the restriction actually rests on. Cocoa butter, drinking chocolate and carob have never been analysed for amines.
Juices. Visciano P, Schirone M. Foods 2022;11:353. PMID 35159503 — tabulates Preti R et al., Eur Food Res Technol 2016 (commercial juices and nectars; not PubMed-indexed). Vieira SM, Theodoro KH, Glória MBA. Food Chem 2007;100:895 (orange juice; not PubMed-indexed). Sánchez-Pérez S et al. Nutrients 2021;13:1395. PMID 33919293. Kalač P et al. Food Chem 1999;67:275 and 2000;69:309 (sauerkraut, and the finding that sauerkraut juice matches the sauerkraut itself; not PubMed-indexed). Jastrzębska A et al. Food Control 2023;154:109980 (fermented vegetable juices; not PubMed-indexed). Histamine was not detected in a single commercial fruit juice analysed. Orange juice is still the standout — but on putrescine, at 3.5–6.1 mg/100 ml, roughly 20 to 40 times apple juice. Only fermented vegetable juices carry a real histamine load.
Histamine liberators — how strong is the evidence? Weaker than the diet literature implies, and worth knowing. Sánchez-Pérez et al. (PMID 33919293) found that 68% of the foods excluded from published low-histamine diets contain no significant histamine, and state that no human clinical study supports the belief that foods release the body's own histamine. Comas-Basté et al. (PMID 32824107) name papaya, kiwi, strawberry, pineapple and plum as putative releasers while noting the mechanism "has not yet been elucidated". A better-supported explanation for several of these foods — citrus above all — is that their putrescine and cadaverine content competes for the same DAO enzyme that clears dietary histamine. Since 2023 the concept has finally been tested: a randomised controlled trial of 140 children with chronic spontaneous urticaria compared drugs alone against drugs plus a pseudoallergen-free, low-histamine diet and found no significant difference (p = 0.07), although its own abstract concludes in favour of the diet (PMID 40880715). Set against that, a 2024 systematic review in the BSACI’s journal concluded the strongest evidence for a low-histamine diet is in chronic urticaria specifically rather than the multi-system syndrome (PMID 39236849). Where an entry here is flagged a liberator on clinical tradition rather than measurement, its card says so.
Clinical / regulatory anchors. EFSA BIOHAZ Panel. Risk based control of biogenic amine formation in fermented foods. EFSA J 2011;9(10):2393. Maintz L, Novak N. Histamine and histamine intolerance. Am J Clin Nutr 2007;85:1185. PMID 17490952. Hrubisko M et al. (review) Nutrients 2021;13:2228. PMID 34209583.
SIGHI overlay. Each entry shows the rating from the SIGHI Food Compatibility List (Schweizerische Interessengemeinschaft Histamin-Intoleranz, Swiss Interest Group Histamine Intolerance), the standard clinical reference for histamine intolerance. Scale: 0 well tolerated, 1 moderately compatible, 2 incompatible, 3 very poorly tolerated. Annotations: H high histamine content, L liberator (releases mast-cell mediators), A contains other biogenic amines, B DAO blocker, ! highly perishable (histamine forms rapidly), ? insufficient or contradictory information — meaning unassessed, not cleared. Source: SIGHI Food Compatibility List v2024-08-29, © Heinz Lamprecht / SIGHI, histaminintoleranz.ch. Two things to understand about these ratings. First, SIGHI states in its own preamble that the histamine column "does NOT show the histamine content of the ingredients, but the perceived COMPATIBILITY" — it is an experiential list compiled from patient reports, not a measurement, which is precisely why measured values are shown alongside it here. Second, a rating is only given where SIGHI actually lists that item; where it does not, the badge reads SIGHI ? and the card names the nearest listed food and its rating. Ratings are not inherited, because SIGHI's own are not: a whole orange rates 3 but orange juice rates 2, while a whole lemon rates 2 and lemon juice rates 3. Where measured values and SIGHI tolerance ratings diverge, both are shown — patients should discuss with their clinician.
Carbohydrate — source and method. All carbohydrate figures are computed from McCance & Widdowson's The Composition of Foods Integrated Dataset (CoFID), 2021 edition, published by the Office for Health Improvement and Disparities. Contains public sector information licensed under the Open Government Licence v3.0. Each entry carries its CoFID food code so any figure can be checked against the source. CoFID expresses carbohydrate as monosaccharide equivalents, which overstates the weight of starch by about 9% because hydrolysing starch adds water; figures here are converted back to actual weight using CoFID's own published divisors (starch ÷ 1.10, sugars ÷ 1.05). Absorbed = starch + sugars by weight. Total = absorbed + oligosaccharides + AOAC dietary fibre. Oligosaccharides — the raffinose and stachyose of beans, the fructans of onions and wheat — are counted in the total but not the absorbed figure, because humans lack the enzymes to break them down. CoFID's starch values already exclude resistant starch. Where a food has no published AOAC fibre value, fibre is estimated from its non-starch polysaccharide figure using the ×1.29 relationship implied by SACN's own conversion of the 18 g/day NSP reference to 30 g/day AOAC fibre; those entries are marked "fibre est." on the card.
Why total and absorbed differ, and what that means. Energy conversion factors under assimilated Regulation (EU) No 1169/2011, Annex XIV: carbohydrate 4 kcal/g, polyols 2.4, fibre 2, erythritol 0. Fibre is not absorbed as glucose but fermented in the colon to short-chain fatty acids, which is why it carries half the energy factor. Definitions of resistant starch and dietary fibre follow SACN, Carbohydrates and Health (2015). Note that absorption is not the same as glycaemic response: fructose is absorbed completely via the GLUT5 transporter yet raises blood glucose very little, while lactose needs an enzyme most adults worldwide stop producing after childhood. Glycaemic index, and where these numbers come from. The compiled GI tables — the University of Sydney database and the published International Tables — all restrict commercial re-use, and while an individual GI value is a fact, and facts carry no copyright, a curated compilation attracts database right. So no table has been copied here. Every GI value is taken from the primary study that actually measured it, cited on the food's own card with its PubMed identifier, and foods for which only a compiled-table value could be found are left blank rather than filled in. Values are on the glucose = 100 scale, and the bands are the ISO 26642 ones: 55 or under low, 56–69 medium, 70 or over high. Treat the band as the useful signal rather than the exact number — differences under about 15 points cannot be reliably told apart within one person (PMID 42177949). Glycaemic load is computed here, not copied, as the food's GI multiplied by this index's own absorbed-carbohydrate figure per 100 g. That is the method the tables' own authors recommend, because published GL figures use standardised portions that will not match a UK serving. Showing it per 100 g keeps every food on the same footing — which is what makes the difference visible: white bread and boiled kidney beans differ roughly tenfold on glycaemic load even though both are ordinary staples. The scientific case is more finely balanced than it looked until recently, and it is only fair to give both sides. Postprandial glucose response varies widely between individuals eating identical meals — 68% variation in the PREDICT 1 cohort of 1,002 adults (PMID 32528151; PMID 26590418), which was widely read as undermining a single published GI per food. A 2026 analysis of 1,116 food tests argues the opposite: that an individual’s response is simply their own glucose tolerance multiplied by the food’s average GI, that prediction error on this basis is smaller than a person’s own test-to-test variability, and that person-specific food effects are largely day-to-day noise — though it also concludes that GI differences below about 15 points cannot be reliably told apart within one person (PMID 42177949). That paper comes from the group that created GI testing, which is worth weighing. Supporting it, ten cohorts of over 100,000 people each link high-GI diets to type 2 diabetes and cardiovascular disease at effect sizes comparable to fibre (PMID 38272606), and a randomised trial of personalised dietary advice failed to beat standard advice on most outcomes (PMID 38714898). One practical warning from UK work at the University of Bath: continuous glucose monitors read about 0.9 mmol/L high and misclassify GI — a smoothie measured 69 by monitor against 53 by finger-prick — so figures derived from CGM data, including several cook-and-cool studies, should be treated with caution (PMID 40021059). Composition, meanwhile, is a property of the food and does not move with the person eating it, which is why this index reports it.
Caveats. Histamine in fermented and aged foods is determined by microbial activity, not the food's intrinsic chemistry, and the variation is far wider than an order of magnitude: a 2025 survey of 125 ripened cheeses on the Polish retail market found histamine above 100 mg/kg in 11% of them, with cadaverine spanning roughly 470-fold and tyramine 115-fold across ostensibly similar products (PMID 40724300). Stilton lacks dedicated quantitative studies (pooled with other blue cheeses, Reinholds 2020). Walnut is listed as "polyamine-rich" rather than histamine-containing — its low-histamine-diet status is based on putrescine/spermidine content, not free histamine. Strawberry contains no measurable histamine but is classed as a clinical liberator. Chocolate is the same case, more starkly: no cocoa product carries SIGHI's high-histamine flag, and the restriction is driven by tyramine and phenylethylamine. Juices marked "No published data" have simply never been analysed — an absent measurement is not a low one, so those entries sort to the end of the histamine orders rather than the safe end.