Cyanide Dose from Fruit Kernels in Beer¶

Version 1 · Disclaimer

Fruit kernels carry cyanogenic glycosides, mostly amygdalin, which release hydrogen cyanide when hydrolysed. Adding them to a secondary is an old and effective way to get marzipan, stone-fruit and noyau character into a beer, and the dose that does something interesting to the flavour is not far from the dose that matters toxicologically. This page works the arithmetic backwards: from a kernel charge to the cyanide in one bottle, compared against the acute reference dose, the flavourings limit and the documented harm range.

The other half of the question — how much marzipan a charge buys and which processing route gives the most flavour per unit of cyanide — is on Marzipan flavour against cyanide cost.

The numbers here are literature typicals, not measurements. Confidence is tagged where a claim is load-bearing: High / Med / Low. Low means verify before acting on it.

The calculator¶

It also runs on its own at /assets/apps/kernel-cyanide-bench/index.html if the frame above is awkward on a phone.

The model¶

with \(m\) the kernel mass in g, \(C\) the kernel's HCN yield in mg/g, \(f\) the fraction extracted and retained, \(V\) the batch volume in L, \(V_b\) the bottle volume in L and \(bw\) the drinker's mass in kg. Solving for the charge that puts one bottle at one acute reference dose \(R\) (mg CN/kg bw):

Two constants do the work. Amygdalin releases 59.0 mg HCN per gram on complete hydrolysis, which is simple stoichiometry: \(\ce{HCN}\) at 27.03 over amygdalin at 457.4 is 5.91% (High). The 0.963 converts HCN to cyanide ion (26.02 / 27.03), because reference doses are expressed as CN while occurrence data and legal limits are expressed as HCN. Skipping it makes the answer 4% conservative, which is immaterial next to everything else.

What each kernel carries¶

Hydrogen cyanide released on complete hydrolysis, per gram of kernel (the seed inside the stone, not the woody stone itself).

Bitter apricot is the best-covered species: EFSA's range of 0.5 to 3.8 mg/g, BfR's ceiling of 4 mg/g, and ion-chromatography and HPLC studies all converge on 1.7 to 4 mg/g. Peach and plum rest on one good ion-chromatography study. The rest are thinner.

Three traps in the source literature¶

Whole pit against kernel. Picrate studies on freeze-dried whole pits report 68 to 228 µg/g; kernel-only assays on the same fruit run 20 to 40 times higher. Mixing the two understates a dose by more than an order of magnitude.

Impossible units. Secondary sources circulate plum kernel figures of "764 mg HCN/g", which would be 76% cyanide by mass. These are mg/kg misprinted. Discard anything above about 5 mg/g.

Sloe has no published data at all. No quantitative amygdalin or HCN figure for Prunus spinosa kernels appears in EFSA 2016 or 2019, BfR, the New Zealand MPI review or the primary literature (High confidence in the absence, having searched for it). The values above are read across from the plum and greengage group, which is defensible taxonomically but is not a measurement. Treat sloe as the weakest row in the table, and note that this is the species most likely to end up in a hedgerow beer.

The serving-size trap¶

The obvious benchmark for a flavoured alcoholic drink is Regulation (EC) 1334/2008 Annex III Part B, which caps hydrocyanic acid at 35 mg/kg in alcoholic beverages. That limit is a poor guide for beer, and reading it as a safety margin is the main way this calculation goes wrong.

It was written around spirits and liqueurs, where a serving is 25 to 40 mL. At beer serving sizes the same concentration delivers a different dose entirely:

Why full release is the right assumption for beer¶

In a spirit, kernel β-glucosidase is denatured by ethanol, so most of the cyanogenic content stays as intact amygdalin. Controlled human data show isolated amygdalin without an enzyme source produces roughly a quarter of the peak blood cyanide of whole kernels at equal cyanide equivalence (High). That protection does not transfer to beer.

A 4 to 7% ABV aqueous secondary at pH 3.4 to 4.5 does not denature the kernel's own enzyme, and Brettanomyces is a strong β-glucosidase producer: the same activity that liberates varietal thiols and monoterpenes from their glycosides will hydrolyse amygdalin (Med-High). In a mixed-fermentation beer the enzyme is not a question mark, it is one of the things that was pitched on purpose. Assume hydrolysis goes to completion over a long secondary, and set the extraction field to 100% for any decision that matters.

Benzaldehyde is the indicator¶

Amygdalin hydrolysis yields two glucose, one benzaldehyde and one HCN, in equal molar amounts:

The marzipan and stone-fruit aroma is the co-product of the cyanide release, so aroma intensity tracks hydrolysis. A beer that smells strongly of almond has converted its amygdalin. One that carries kernels and smells of nothing has not converted it yet; that is a delay, not a safety margin, and bottle conditioning continues it.

Where the model is conservative, and where it is not¶

HCN boils at 25.6 °C and beer pH sits far below its pK_a of 9.2, so it is present as the free, volatile acid. Active secondary strips volatiles efficiently and an airlocked vessel will lose some fraction of what is released. Whole, uncracked kernels with intact seed coats also extract slowly. None of this is quantified for beer (Low), which is why the calculator defaults to 100% release. Once the beer is capped that route closes: anything generated in the bottle stays in the bottle.

Against that, the conservative column is an upper literature value, not an upper bound. Bitterness is the signal, since amygdalin is what makes a kernel bitter, and a batch of hedgerow stones is not a uniform population. The model also treats one bottle as one sitting; two bottles in an evening is two reference doses.

Reference doses and legal limits¶

The agencies disagree about the acute reference dose by a factor of eighteen: EFSA 20 µg/kg, UK COT 5, BfR 75, FSANZ 80, JECFA 90. That spread is wider than the between-study variation in kernel content for any species in the table above, so the choice of reference dose matters more than the choice of kernel figure. EFSA's 20 µg/kg is the value that applies in the UK and is what the calculator defaults to; the selector shows what the others say about the same beer.

Worth noticing: the spirit-drinks limit expressed per litre of pure alcohol is savage when applied to a low-alcohol product. A 5% ABV beer meeting 7 g/hL of 100% vol would be capped at 3.5 mg HCN/L. That category does not legally cover beer, but it is the closest thing to a limit written with a kernel-derived cyanide source specifically in mind, and it lands within a factor of two of what the ARfD implies for a 330 mL serving.

Worked example¶

100 g of bitter almond kernels into 5 L, bottled at 330 mL, conservative content (3.0 mg HCN/g), full release: 300 mg HCN in the batch, 60 mg/L in the beer, which is already above the flavourings limit, and about 19 mg CN in one bottle. For an 80 kg adult that is 0.238 mg CN/kg bw: roughly twelve times the EFSA ARfD, 2.3 times EFSA's human no-effect dose, and 43% of the lowest fatal oral dose on record. At the central 1.2 mg/g it is still about five times the ARfD.

Backwards, one bottle equalling one ARfD in a 5 L batch corresponds to 8 g of bitter almond kernels at the conservative figure, 21 g at the central one — roughly 7 to 19 kernels. Single-digit to low-double-digit grams, not 100.

Getting the real number¶

Total cyanide in a finished beer is a routine analysis: acid or enzymatic hydrolysis followed by distillation and either ion chromatography or a picrate colourimetric method. Ask for total hydrocyanic acid including that bound in cyanogenic glycosides, not free HCN. A free-HCN assay on a young beer whose amygdalin has not yet hydrolysed reads near zero and says nothing about what the bottle will be in six months. A food-testing lab will run it on 100 mL. For a batch that the calculator puts anywhere near the reference dose, that is the answer this page can only bracket.

Sources¶

1. EFSA CONTAM Panel. Acute health risks related to the presence of cyanogenic glycosides in raw apricot kernels. EFSA Journal 2016;14(4):4424. doi:10.2903/j.efsa.2016.4424
2. EFSA CONTAM Panel. Evaluation of the health risks related to the presence of cyanogenic glycosides in foods other than raw apricot kernels. EFSA Journal 2019;17(4):5662. doi:10.2903/j.efsa.2019.5662
3. Abraham K, Buhrke T, Lampen A. Bioavailability of cyanide after consumption of a single meal of foods containing high levels of cyanogenic glycosides. Archives of Toxicology 2016;90:559–574. doi:10.1007/s00204-015-1479-8
4. Cho H-J et al. Determination of cyanogenic compounds in edible plants by ion chromatography. Toxicological Research 2013;29(2):143–147.
5. Bolarinwa IF, Orfila C, Morgan MRA. Amygdalin content of seeds, kernels and food products commercially available in the UK. Food Chemistry 2014;152:133–139.
6. Yıldırım FA, Aşkın MA. Variability of amygdalin content in seeds of sweet and bitter apricot cultivars. African Journal of Biotechnology 2010;9(39):6522–6524.
7. Rodríguez Madrera R, Suárez Valles B. Development and validation of an analytical method for cyanogenic compounds in Rosaceae and Sambucus. Molecules 2021;26(24):7563.
8. Grgurič M, Kraner Šumenjak T, Kristl J. Cyanide contents in pits of cherries, gages and plums using a modified sensitive picrate method. Agricultura 2022;19(1):43–50.
9. New Zealand MPI / ESR. Evaluation of food safety risks associated with foods containing cyanogenic glycosides, 2022.
10. BfR Stellungnahme 009/2015. Zwei bittere Aprikosenkerne pro Tag sind für Erwachsene das Limit.
11. ATSDR. Toxicological Profile for Cyanide.
12. Regulations (EC) 1334/2008 Annex III Part B; (EU) 2019/787; (EU) 2023/915; assimilated Reg. (EC) 1881/2006 as amended by (EU) 2017/1237 for Great Britain.