Ochratoxin A

Chemistry and toxicology#

Structure. In EFSA’s description, the ochratoxins consist of L-phenylalanine linked by an amide bond to a substituted dihydroisocoumaric acid.³ OTA carries a chlorine substituent, which distinguishes it from ochratoxin B. Its formula is C₂₀H₁₈ClNO₆ and its molar mass 403.8 g/mol.³ The isocoumarin part gives OTA a strong fluorescence, the basis of its detection.³ The FAO guidelines describe OTA as a heat-stable mould metabolite.²

Heat products. Heating converts part of the OTA into 2′R-ochratoxin A (2′R-OTA), first reported as 14R-OTA. It is a diastereomer of OTA.³ In model experiments between 120 and 270 °C, isomerisation began at 120 °C. At 210 °C and above, 25% 2′R-OTA formed in less than one minute.¹⁶ EFSA states that roasted coffee appears to be the major source of 2′R-OTA.³ Its toxicity was not assessed in the sources opened for this entry.

Bound OTA. EFSA cites evidence that OTA binds covalently to coffee polysaccharides, through esterification of its carboxyl group, as a further heat reaction.³ It considers it likely that some of these esters are split during brewing and release OTA into the brew, because higher amounts of OTA have been found in brewed coffee than in the roasted beans.³

EFSA’s 2020 assessment. EFSA concluded that OTA is genotoxic in vitro and in vivo, but that the mechanisms are unclear; direct and indirect genotoxic and non-genotoxic modes of action might each contribute to tumour formation.³ As a consequence, it declared the tolerable weekly intake (TWI) of 120 ng/kg body weight it had set in 2006 no longer valid.³ It used a margin of exposure (MOE) approach instead, based on two reference points:³

• Non-neoplastic effects: a benchmark dose lower confidence limit (BMDL₁₀) of 4.73 µg/kg body weight per day, from kidney lesions in pigs. An MOE of 200 or more indicates low health concern.
• Neoplastic effects: a BMDL₁₀ of 14.5 µg/kg body weight per day, from kidney tumours in rats. MOEs were below 10,000 for almost all exposure scenarios.

EFSA concluded that the tumour MOEs would indicate a possible health concern if genotoxicity is direct, and that the uncertainty is high.³ It named preserved meat, cheese, and grains and grain-based products as the most important contributors to chronic exposure; coffee is not named among them.³ The 2009 Codex code quotes the values current at the time: International Agency for Research on Cancer (IARC) Group 2B, possibly carcinogenic to humans, and a JECFA provisional tolerable weekly intake of 100 ng/kg body weight.¹

The producing fungi in detail#

The Codex code lists A. ochraceus and related species (A. westerdijkiae, A. steynii), A. niger and related species, and A. carbonarius as the OTA producers in coffee.¹ A 2013 review places them in Aspergillus section Circumdati and section Nigri.⁹ A 2024 review names A. westerdijkiae and A. steynii as the strongest producers in section Circumdati, and A. carbonarius in section Nigri.⁸

Toxin production varies sharply between species. In a 2003 Brazilian study, A. niger was the most common of the three, but only 3% of its isolates produced OTA. Of the A. ochraceus isolates tested, 75% did, and 77% of A. carbonarius isolates.¹⁰ Finding black aspergilli on coffee is therefore not the same as finding an OTA risk.

Penicillium verrucosum and P. nordicum make OTA mainly in temperate and colder zones.⁹ Both the Codex code and the FAO guidelines limit the coffee producers to Aspergillus.^1,2 This matches the moisture and water activity and microorganisms entries.

Contamination along the chain#

Before harvest. The Codex code names two possible infection routes on the plant: through the flowers without visible signs, or by the coffee berry borer (Hypothenemus hampei) carrying spores into the fruit.¹ It states that it is not certain whether OTA forms on the plant.¹ The FAO guidelines see good evidence that it can.² In the 2003 Brazilian study, infection rates of cherries taken from trees were very low; mean OTA rose from about 0.1 µg/kg at the tree stage to 3.3 µg/kg in storage.¹⁰ More mature, drier fruit and by-products such as husk and pulp can carry more spores of OTA producers.²

Ground and over-ripe fruit. Fruit on the ground is more likely to be contaminated as contact time lengthens.^1,2 If fallen fruit must be collected, the Codex code asks for separate storage and rapid processing and drying.¹ Unripe and over-ripe cherries should be removed before processing.¹

Delays and fermentation. Holding harvested fruit in bags or heaps for up to a week causes high temperatures and quick fermentation, and raises the OTA risk.¹ For washed coffee, the Codex code asks for fermentation as short as possible, 12 to 36 hours, and clean water.¹ The harvesting and sorting and fermentation entries give the details.

Drying. The Codex code divides drying into three stages.¹ In the first, moisture falls slightly over 1–3 days for cherry and 1 day or less for parchment; aw stays above 0.95. The second stage has the fastest moisture loss and favourable conditions for OTA producers. In the third, the coffee is dry enough that they no longer grow.¹ Rewetting can occur at any stage and favours rapid fungal growth.¹

Storage and transport. The Codex code states that coffee keeps drying below 60% relative humidity and starts to absorb water above 80%.¹ Moisture can enter from damp floors and walls, leaks, wind-driven rain, dead air and mixing dry with wet coffee.¹ In containers, temperature swings can condense the remaining water and rewet the coffee locally.¹ The green coffee storage entry covers warehouse practice.

Defects and husk. In lower-grade coffee, fruits with black and sour defects contained the highest OTA levels.¹ Defects are not evenly distributed across the bean classes separated from bulk coffee, and defective beans and husk sometimes contain more OTA than sound beans.¹ According to a 2002 review chapter, husk is a significant OTA source, and cleaning and grading green coffee are effective ways to reduce OTA.¹⁸

Water activity thresholds#

The thresholds below come from the Codex code and the FAO guidelines; they are the same as in the moisture and water activity entry.^1,2

The FAO guidelines relate these thresholds to moisture. In a data set of more than 2,000 samples, aw 0.76 corresponded on average to about 18% moisture in robusta cherry and about 16% in arabica parchment.² At a 99% confidence level, the moisture figure fell to about 13% for both.² The guidelines conclude that the recommended maxima of 12% (wb) for dry parchment and 13% (wb) for dry cherry include a substantial safety margin.² The Codex code sets 12.5% for both.¹

Codex measures by stage#

The Codex code of practice CAC/RCP 69-2009 sets out good practice to prevent and reduce OTA.¹ The main measures:

The Codex code cites the FAO guidelines of 2006, written so that producing countries could build their own national programmes.¹ The guidelines name two tools: managing water availability from the start of drying, and favouring competing microorganisms and restrictive growth conditions before that point.²

Fate in roasting#

Roasting reduces measurable OTA, but studies disagree on how much. The values below are reported reductions; they are not comparable one to one, because samples, contamination levels and roasting equipment differ.

Kinetics. A 2010 Brazilian study inoculated green arabica with A. westerdijkiae to 247 ng/g OTA. It roasted the beans in a spouted-bed roaster at 180–240 °C for 5, 8 or 12 min.¹¹ Reductions ranged from 8.2% (180 °C, 5 min) to 98.9% (240 °C, 12 min). Loss followed first-order kinetics and the Arrhenius equation, with an apparent activation energy of 49.2 kJ/mol.¹¹ Roast colour ranged from very light (Agtron 95) to very dark (Agtron 25).¹¹

Other studies. A 2001 study roasted for 2.5–10 min, from light medium to dark. It found a 69% reduction over all conditions, with no significant differences between times or colours.¹² It noted that nine studies on OTA reduction in roasting were then available, seven of which reported 69–96%.¹² A 2003 Italian study found reductions above 90% in naturally contaminated samples roasted for espresso.¹³ A 2008 Italian study reported 65–100%, the same range the Codex code gives.^1,14 EFSA summarises losses of 70% to more than 90% at temperatures above 180 °C.³ A 1994 Swiss study of spoiled beans found no significant reduction and judged a loss of less than 30% likely.¹⁵ A 2024 review gives 12–90%, depending on roasting temperature and initial contamination.⁸

Why results differ. The sources point to several factors: roasting temperature and time, the roaster type and initial contamination.^8,11,13 Part of the OTA can be converted to 2′R-OTA or bound to coffee polysaccharides rather than broken down.^3,16 In 51 coffee samples from France, Germany and Guatemala, OTA was found in 96% and 2′R-OTA in 35%. The OTA to 2′R-OTA ratio ranged from 2.5:1 to 10:1, with a mean of 5.5:1.¹⁶ The highest values, 28.4 µg/kg OTA and 3.9 µg/kg 2′R-OTA, were in Guatemalan coffee.¹⁶

Decaffeination, soluble coffee and brewing#

Decaffeination. No source opened for this entry reported OTA changes during decaffeination. The decaffeination entry covers the processes.

Soluble coffee. The EU sets a higher maximum level for soluble coffee (5.0 µg/kg) than for roasted coffee (3.0 µg/kg).⁴ No primary study on OTA through soluble coffee manufacture was opened for this entry.

Brewing. A 2014 study prepared seven kinds of coffee beverage from ground roasted coffee naturally contaminated at 0.92 ng/g. Between 22.3% and 66.1% of the OTA passed into the drink, depending on the process.¹⁷ A 2008 study found a 50–75% reduction with a moka pot, but little reduction in Turkish coffee.¹⁴ EFSA notes that higher OTA levels have been observed in brewed coffee than in the roasted beans, and considers release of bound OTA during brewing a likely cause.³

Sampling and analysis#

Uneven distribution. OTA is concentrated in some beans, defects and husk, not spread evenly through a lot.¹ The Codex code ties any re-use of out-sorted defects to a representative sampling plan and direct OTA analysis.¹ The FAO guidelines ask for a spear sample from each sack, as a lot may be a blend of different sources.²

EU official sampling. Implementing Regulation (EU) 2023/2782 sets the official sampling plan for coffee and coffee products in Annex I, Part G.⁶ An incremental sample weighs about 100 g. Lots of 15 t or more are divided into sublots of 15–30 t; each sublot gives 100 incremental samples, combined into a 10 kg aggregate sample. Smaller lots give 10–100 incremental samples and an aggregate sample of 1–10 kg.⁶ The laboratory sample is mixed thoroughly, with fine grinding if needed.⁶ A lot is rejected only if the result, minus the expanded measurement uncertainty, is above the maximum level.⁶

Standard method. EN 14132 specifies immunoaffinity column clean-up and high-performance liquid chromatography (HPLC) for OTA in barley and roasted coffee.⁷ It is validated for roasted coffee from 0.2 to 5.5 µg/kg.⁷ An immunoaffinity column holds an antibody that binds OTA selectively, so other coffee compounds can be washed away.¹¹ Detection uses OTA’s fluorescence (HPLC-FLD).^3,11,14

A worked example. The 2010 kinetic study extracted 25 g of coffee with methanol and 3% aqueous sodium bicarbonate (50:50). It cleaned the extract on an immunoaffinity column and measured at 333 nm excitation and 477 nm emission.¹¹ Its limits of detection and quantification were 0.1 and 0.3 ng/g; recoveries were 78–86.5%.¹¹

Performance criteria. For OTA in most foods, the EU requires a limit of quantification of at most half the maximum level, preferably at most one fifth.⁶ For roasted coffee this means at most 1.5 µg/kg, preferably 0.6 µg/kg (our calculation from the 3.0 µg/kg level).^4,6 Mean recovery should lie between 70 and 120%; repeatability RSD_r at most 20% and reproducibility RSD_R preferably at most 25%.⁶

Typical levels. The FAO guidelines state that most commercial samples contain no detectable OTA at a detection limit of 0.1–0.5 µg/kg. Most positive samples fall below 5 µg/kg, and anything above 20 µg/kg is considered exceptionally high.²

Legal limits#

EU. Regulation (EU) 2023/915 sets maximum levels of 3.0 µg/kg for roasted coffee beans and ground roasted coffee (entry 1.2.14) and 5.0 µg/kg for soluble coffee (entry 1.2.15).⁴ No entry names green coffee.⁴ The first EU levels for coffee, in 2005, were 5.0 µg/kg for roasted and 10.0 µg/kg for soluble coffee.⁵ That regulation identified coffee, wine and beer as significant contributors to OTA exposure. It left a level for green coffee to be considered after further EFSA assessment.⁵

Codex. The Codex code of practice sets no maximum level; it describes practices.¹ A 2024 review states that neither Codex nor the US Food and Drug Administration has set an OTA limit for coffee.⁸

Switzerland. The Swiss values were not opened for this entry; the contaminants and food safety entry gives them.

Reference values#

µg/kg equals ng/g and parts per billion (by mass).

Open questions#

• Infection on the plant. Whether OTA producers infect and produce OTA before harvest, and how important that route is, is not settled (see Contested).
• Toxicity of 2′R-OTA and bound OTA. The sources opened here describe how they form, not how toxic they are; whether they should count towards limits was not addressed.
• Decaffeination and soluble coffee. No primary study on OTA through these processes was opened for this entry.
• Coffee’s share of exposure. EFSA’s 2020 summary does not name coffee among the main contributors; no coffee-specific exposure figure was found in the sources opened.

Key studies#

General information, not medical advice.