Burgon 2023 — Inorganic contaminants and ochratoxin A in Brazilian cocoa beans and “bean-to-bar” chocolates
This ITAL master’s thesis (Campinas, Brazil) validated and applied ICP OES, ICP-MS and HPLC-FLD methods to quantify seven trace elements (As, Cd, Co, Cu, Hg, Pb, Se) and ochratoxin A in cocoa beans and artisanal “bean-to-bar” chocolates produced from Brazilian cacao between 2020 and 2022. Across 23 cocoa-bean samples and 65 bean-to-bar chocolates, total arsenic and total mercury were below detection in every sample; cadmium in Amazonas cocoa beans averaged 0.68 mg/kg, exceeding the Brazilian maximum tolerated limit of 0.30 mg/kg for cocoa paste, and one Amazonas-origin 70% dark chocolate reached 0.74 mg/kg Cd, exceeding the 0.40 mg/kg limit for chocolates >40% cocoa solids in both BRASIL IN 160/2022 and EU 2014. Cadmium exposure modelled at 15.75 g/day of dark chocolate reached 94% of the JECFA PTMI in children (15 kg body weight). Cobalt at the maximum dark-chocolate level (1.88 mg/kg) modelled to 123% of the EFSA upper level in children. Ochratoxin A was detected in 30% of cocoa beans and roughly 19% of chocolates, with all values below Brazilian regulatory limits.
Key numbers
Bean-to-bar chocolates (n=65) — mean and range by cocoa-solids classification (mg/kg, as-sold basis): (Table 2)
| Cocoa class | n | Cd mean (range) | Co mean (range) | Cu mean (range) | Pb mean (range) | Se mean (range) |
|---|---|---|---|---|---|---|
| White (0% cocoa) | 2 | <0.001 | 0.05 (<0.002–0.012) | 0.15 (0.11–0.20) | 0.020 (<0.005–0.039) | <0.019 |
| Milk (40–50%) | 12 | 0.025 (0.009–0.094) | 0.23 (0.06–0.94) | 3.90 (1.63–9.24) | 0.012 (<0.005–0.037) | 0.018 (<0.019–0.077) |
| Semi-sweet (50–70%) | 17 | 0.041 (0.004–0.10) | 0.37 (0.12–0.63) | 8.22 (4.13–12.20) | 0.012 (<0.005–0.06) | 0.087 (<0.019–0.25) |
| Dark (70–100%) | 34 | 0.078 (0.008–0.74) | 0.47 (<0.002–1.88) | 10.10 (2.96–19.3) | 0.013 (<0.005–0.04) | 0.088 (<LOD–0.28) |
Chocolate LODs (ICP-MS): Cd 0.001, Co 0.002, Cu 0.007, Pb 0.005, Se 0.019, Hg 0.007, As 0.015 mg/kg. As and Hg were not detected in any sample. (Table 1)
Bean-to-bar chocolates by region of bean origin (n=65) — mean and range, mg/kg: (Table 3)
| Region of origin | n | Cd | Co | Cu | Pb | Se |
|---|---|---|---|---|---|---|
| Amazonas | 7 | 0.145 (0.03–0.74) | 0.170 (0.03–0.54) | 7.34 (0.198–12.8) | 0.014 (<0.005–0.03) | 0.075 (<0.019–0.28) |
| Amapá | 2 | 0.054 (0.007–0.1) | 0.15 (0.054–0.25) | 7.17 (5.67–8.66) | 0.019 (0.02–0.02) | 0.044 (<0.019–0.19) |
| Bahia | 35 | 0.046 (0.01–0.18) | 0.39 (0.03–0.92) | 8.50 (1.63–15.7) | 0.012 (<0.005–0.05) | 0.08 (<0.019–0.25) |
| Espírito Santo | 8 | 0.05 (0.01–0.09) | 0.49 (<0.002–0.94) | 6.3 (0.01–11.40) | 0.022 (<0.005–0.06) | 0.055 (<0.019–0.23) |
| Floresta Amazônica | 2 | 0.08 (0.04–0.12) | 0.77 (0.46–1.07) | 12.63 (7.36–17.9) | <0.005 (<0.005–0.01) | 0.009 (<0.019–0.02) |
| Mata Atlântica | 4 | 0.036 (0.01–0.06) | 0.29 (0.15–0.43) | 7.6 (3.6–8.42) | 0.0038 (<0.005–0.008) | 0.024 (<0.019–0.06) |
| Pará | 5 | 0.025 (<0.001–0.1) | 0.60 (0.01–1.88) | 10.14 (0.11–19.3) | 0.011 (<0.005–0.04) | 0.10 (<0.019–0.18) |
| Rondônia | 2 | 0.024 (0.016–0.031) | 0.15 (<0.002–0.15) | 4.46 (2.96–5.96) | 0.015 (0.01–0.02) | 0.11 (0.08–0.14) |
| All-region mean | 65 | 0.057 (<LOD–0.74) | 0.38 (<LOD–1.88) | 7.84 (0.01–19.3) | 0.012 (<LOD–0.06) | 0.058 (<LOD–0.28) |
Cocoa beans (n=23) by region — mean and range, mg/kg: (Table 4)
| Region | n | Cd | Cu | Pb |
|---|---|---|---|---|
| Amazonas | 8 | 0.68 (0.52–0.89) | 18.39 (15.2–24.8) | <0.06 (<0.06–0.09) |
| Bahia | 11 | 0.056 (<0.17–0.16) | 19.76 (10.34–28.69) | 0.04 (<0.001–0.313) |
| Pará | 2 | <0.17 | 22.38 (16.62–28.12) | <0.06 |
| Espírito Santo | 2 | 0.22 (<0.17–0.44) | 16.6 (10.2–23.0) | <0.06 |
| All-region mean | 23 | 0.24 (<0.17–0.89) | 19.28 (10.2–28.69) | 0.010 (<0.06–0.313) |
Cocoa-bean LODs (ICP-OES): As 0.171, Cd 0.021, Cu 0.043, Pb 0.057 mg/kg. As was not detected in any bean sample. (Table 1)
Regulatory comparison. Brazil IN 160/2022 maximum tolerated limits: cocoa paste/beans 0.30 mg/kg Cd (and 0.50 mg/kg As, 0.50 mg/kg Pb, 40 mg/kg Cu); chocolate with >40% cocoa 0.40 mg/kg Cd. EU 2014 LMT for cocoa powder 0.60 mg/kg Cd. Amazonas cocoa beans averaged 0.68 mg/kg Cd (8/8 samples above the 0.30 mg/kg Brazil LMT for cocoa paste, with one sample at 0.89 mg/kg also exceeding the 0.60 mg/kg EU limit). One 70%-cocoa dark chocolate from an Amazonas bean source reached 0.74 mg/kg Cd, exceeding the 0.40 mg/kg LMT for chocolates >40% cocoa solids.
Exposure modelling (15.75 g/day chocolate, 60 kg adult / 15 kg child):
| Analyte | Max-finding concentration | Adult intake | Child intake | Adult % of reference | Child % of reference | Reference |
|---|---|---|---|---|---|---|
| Cd | 0.74 mg/kg (dark, AM origin) | 5.91 µg/kg pc/day | 23.6 µg/kg pc/day | 24% PTMI | 94% PTMI | JECFA PTMI 25 µg/kg pc |
| Pb | 0.06 mg/kg (semi-sweet) | 0.012 µg/kg pc/day | 0.048 µg/kg pc/day | 0.1% | 0.4% | EFSA BMDL₀.₁ 12 µg/kg pc |
| Cu | 19.3 mg/kg (dark) | 0.005 mg/kg pc/day | 0.02 mg/kg pc/day | 1% PTMDI | 4% PTMDI | WHO 1982 PTMDI 0.5 mg/kg pc |
| Co | 1.88 mg/kg (dark, PA origin) | 0.5 mg/kg pc/day | 1.97 mg/kg pc/day | 31% UL | 123% UL | EFSA 2012 UL 0.0016 mg/kg pc — child exceedance |
Considering the higher per-capita chocolate consumption in countries such as Germany (11.1 kg/year, ~30.4 g/day), the modelled child intake of Cd would reach up to 180% of the PTMI and Co up to 238% of the UL — described in the thesis as a potential health risk for that consumer group.
Ochratoxin A (HPLC-FLD; LOD 0.04 µg/kg, LOQ 0.1 µg/kg for both matrices):
OTA was detected in 13 of 43 cocoa-bean samples (30%) and in 12 of 62 bean-to-bar chocolates (~19%). (Tables 7–10)
| Matrix | Region/class | n | Positive | Mean (range, µg/kg) |
|---|---|---|---|---|
| Cocoa beans | Amazonas | 8 | 4 | 0.50 (<0.04–0.64) |
| Cocoa beans | Bahia | 23 | 8 | 0.76 (<0.04–1.17) |
| Cocoa beans | Espírito Santo | 2 | 0 | <0.04 |
| Cocoa beans | Pará | 10 | 0 | <0.04 |
| Bean-to-bar chocolate | Amazonas | 9 | 2 | 0.50 (<0.04–0.16) |
| Bean-to-bar chocolate | Amapá | 1 | 0 | <0.04 |
| Bean-to-bar chocolate | Bahia | 34 | 6 | 0.01 (<0.04–1.11) |
| Bean-to-bar chocolate | Espírito Santo | 12 | 3 | 0.11 (<0.04–0.48) |
| Bean-to-bar chocolate | Pará | 4 | 1 | 0.06 (<0.04–0.24) |
| Bean-to-bar chocolate | Rondônia | 2 | 0 | <0.04 |
Maximum OTA in any single sample was 1.19 µg/kg (Ibirapitanga-BA cocoa bean) and 1.11 µg/kg in a Bahia semi-sweet chocolate. All values were below the Brazilian IN 160/2022 maximum tolerated limits (10 µg/kg cocoa beans; 5 µg/kg chocolate). Modelled OTA intake did not reach the CAC/JECFA PTWI of 100 ng/kg pc/week.
A bean-processing time-course on an Arataca-BA farm (6 fermentation days followed by drying) detected OTA only in one drying-stage sample at 1.17 µg/kg, identifying drying as the critical step for ochratoxigenic fungal development as water activity drops.
Methods (brief)
Sampling. Cocoa beans collected 2020–2022 from Bahia (Arataca, Ibirapitanga, Itabuna, Itajuípe, Porto Seguro), Pará (Novo Repartimento, Mocajuba), Amazonas (Borba, Manicoré, Nova Olinda) and Espírito Santo (Linhares); transferred to identified plastic bags and stored refrigerated until analysis. 65 bean-to-bar chocolate samples were purchased from Brazilian e-commerce retailers; 6 conventional chocolate bars were purchased in Campinas-SP local retail for comparison. Chocolates were stratified by cocoa-solids percentage as white (0%), milk (40–50%), semi-sweet (50–70%), dark (70–100%). For ochratoxin A, additional pre-processing samples (6-day fermentation cochos plus drying-stage beans) were collected from two farms in Arataca-BA (n=14) and Mocajuba-PA (n=8).
Sample preparation. Beans were ground in a mill (Ika, Staufen, Germany); chocolates were grated. 0.5 g aliquots were digested in closed-vessel microwave (Milestone Start D, Sorisole, Italy) with 6 mL HNO₃ (Berghof sub-boiling, 65%) plus 2 mL 30% H₂O₂ (Merck) at 1000 W with a 70°C → 170°C ramp programme, then diluted to 20 mL with deionised water.
Instrumentation. Cocoa beans were analysed by ICP-OES (Agilent 5100 VDV, 1200 W RF, axial view, seaspray nebulizer; analytical lines As 193.696 nm, Cd 214.439 nm, Cu 324.754 nm, Pb 220.353 nm). Chocolates were analysed by ICP-MS (Thermo iCAP RQ, 1550 W RF, KED with 5 mL/min He; isotopes ⁷⁵As, ⁵⁹Co, ⁶³Cu, ⁸⁰Se, ¹¹¹Cd, ²⁰²Hg, ²⁰⁸Pb; internal standard mix of ⁷²Ge, ¹⁰³Rh, ⁴⁵Sc, ²⁰⁹Bi, ¹⁹⁵Pt at 50 µg/L). Multielement standards (Specsol/Quimlab, Brazil) at 100 mg/L.
Method validation. Bias evaluated against three certified reference materials: ERM BD512 (Dark Chocolate) for Cd and Cu; NIST SRM 1547 (Peach Leaves) for Pb, Se, Co; NRCC TORT-3 (Lobster Hepatopancreas) for As and Hg in chocolates; NRCC TORT-2 for As, Cd, Cu, Pb in cocoa beans. Recovery 88–98% (cocoa-bean method, ICP-OES) and 84–105% (chocolate method, ICP-MS). Precision (CV, n=7): 4–13% (beans), 2–14% (chocolates). Linearity r ≥ 0.999 across all curves. LODs and LOQs reported per matrix in Table 1.
Ochratoxin A. 10 g sample homogenised in 65°C water bath, extracted with 200 mL methanol:3% sodium bicarbonate (1:1 v/v) in Ultra-Turrax at 10,000 rpm for 10 min; filtered through Whatman paper, then through fibreglass filter; 20 mL of filtrate diluted with 20 mL PBS containing 0.01% Tween 20, passed through an Ochraprep R-Biopharm immunoaffinity column at 2–3 mL/min, washed with 20 mL water, eluted with 4 mL methanol:acetic acid (98:2 v/v), evaporated under N₂ at 40°C, reconstituted in 0.4 mL mobile phase. Analysed by HPLC-FLD (Agilent Infinity 1260; Shimadzu Shimpack C18 250 × 4.6 mm, 5 µm; mobile phase methanol:acetonitrile:water:acetic acid 35:30:34:01 v/v/v/v at 1 mL/min; excitation 333 nm, emission 477 nm). Method adapted from Copetti et al. 2010 and validated for recovery (97% / 95% at 1.7 µg/kg, 94% / 84% at 4.3 µg/kg, 86% / 88% at 10.8 µg/kg for beans / chocolates) and precision (CV 15% beans, 13% chocolates).
Exposure modelling. Maximum-concentration scenario using Brazilian São Paulo per-capita chocolate consumption of 5.75 kg/year (15.75 g/day, 479.16 g/month, 110.25 g/week); adult body weight 60 kg, child body weight 15 kg (MDCI 2018; FAO/WHO 2013). Reference values: JECFA PTMI 25 µg/kg pc for Cd (WHO 2013), EFSA BMDL₀.₁ 12 µg/kg pc for Pb (EFSA 2013), WHO PTMDI 0.5 mg/kg pc for Cu (WHO 1982), EFSA 2012 UL 0.0016 mg/kg pc for Co, EFSA 2023 UL 255 µg/day adults and 130 µg/day children 7–10 years for Se, CAC PTWI 100 ng/kg pc for OTA.
Limitations
This is a defended master’s thesis from ITAL/Campinas; Chapter 3 (ochratoxin A) was prepared for submission to Food Control but had not been peer-reviewed at the time of access. Sample sizes per region of origin are uneven — Bahia dominates (n=35 chocolates, n=11 beans) while Pará, Espírito Santo and Rondônia each carry n≤8. The cocoa-bean ICP-OES LODs are relatively high (Cd 0.171, Pb 0.057 mg/kg), so the bean-level Pb estimate of “0.010 (<0.06–0.313) mg/kg” is dominated by left-censored values; the maximum Pb of 0.313 mg/kg in a Bahia bean is a single-sample value. Total As and total Hg were not detected anywhere, but iAs and MeHg were not speciated, so this dataset cannot rule out trace iAs or MeHg below the tHg/tAs LODs. The exposure model uses single maximum-finding concentrations, not the population mean or 95th percentile, so the modelled intake fractions are upper-bound rather than expected exposures. The 15.75 g/day São Paulo consumption figure is a national average; subgroups of high-consuming children may exceed it. Conventional chocolates were a small n=6 convenience sample for comparison and are not market-representative.
Implications
- Certification: Provides matched bean-and-finished-chocolate Cd, Cu, Pb data across the major Brazilian production regions (Amazonas, Bahia, Pará, Espírito Santo) using ICP-MS/ICP-OES with NIST/ERM/NRCC certified reference material validation. The Amazonas cocoa-bean Cd mean of 0.68 mg/kg with an 0.89 mg/kg maximum sits above both the Brazilian (0.30 mg/kg cocoa paste) and EU (0.60 mg/kg cocoa powder) limits. Linear association between cocoa-solids percentage and finished-chocolate Cd and Pb (semi-sweet/dark > milk > white) is reported in a Brazilian e-commerce-sampled bean-to-bar market in 2020–2022. The cobalt maximum (1.88 mg/kg in a Pará-origin 70–100%-cocoa dark chocolate) modelled to 123% of the EFSA upper level for children at 15.75 g/day; Co is not currently tracked under the HMTc ten-analyte panel.
- Courses: Useful worked example of the bean-to-finished-product Cd attenuation pathway (Amazonas beans 0.68 → Amazonas dark chocolate 0.145 mean, max 0.74). Also useful for the post-harvest ochratoxin A risk pathway, with the drying stage identified as the critical control point as water activity drops.
- App: Brazilian regional Cd-in-cocoa-bean and Cd-in-chocolate geographic-breakdown data for the
[[ingredients/cocoa]]and[[ingredients/chocolate]]contamination_profilesynthesis pass. Per-region maxima and means supplied; not a market basket.
Wiki pages updated on ingest
Verification notes
- Source-internal inconsistencies in Tabela 9 (p.97): the Amazonas bean-to-bar chocolate row reports “0,50 (<0,04–0,16)” — mean 0.50 µg/kg above the stated maximum 0.16 µg/kg, which is mathematically impossible; the Bahia bean-to-bar chocolate row reports “0,01 (<0,04–1,11)” — mean 0.01 µg/kg appears inconsistent with 6 of 34 positive samples reaching 1.11 µg/kg. Values are reproduced verbatim from the thesis table; downstream synthesis should treat Tabela 9’s mean column as unreliable and use the range column plus positive-sample counts instead.
- The thesis Resumo (p.viii) reports ochratoxin A “em 20% e 30% das amostras” while the Chapter 3 results (Tabela 9, §3.3) document 13/43 = 30% bean detection and 12/62 ≈ 19% chocolate detection. The wiki page uses the table-derived figures (30% beans, ~19% chocolates) rather than the abstract figures.
- Cocoa-bean ICP-OES LOD for Pb is 0.057 mg/kg and for Cd is 0.171 mg/kg; several bean values are left-censored at “<0,17” or “<0,06” in source tables and the per-region “mean” lines include these censored values. The single Bahia bean Pb maximum of 0.313 mg/kg is the only quantified value driving the all-region Pb mean. Treat bean Pb/Cd quantification as left-censored-dominant.
- Audit subagent (2026-05-30) flagged matrices field as overspecified (used
cocoa-beans, dark-chocolate, milk-chocolate, white-chocolate, bean-to-bar-chocolate); corrected to the broad-scope canonical vocabulary[cocoa, chocolate]per docs/gpt-collaboration/system-prompt.md matrices guidance. - Audit subagent (2026-05-30) flagged Implications/Certification “flag for cobalt to enter HMTc cocoa-derivative analyte review” as crossing toward HMTc governance proposal; softened to a factual note that Co is not currently in the HMTc ten-analyte panel.
Page history
The five most recent substantive edits to this page. The full version history lives in git; when DOI minting comes online (see schema docs), each entry below will also link to a version-pinned DataCite DOI.