Lin et al. 2024 — Twelve heavy metals across 24 edible-oil types in Taiwan

Lin and colleagues measured 12 heavy metals (As, Pb, Cd, Ni, V, Cr, Co, Cu, Fe, Zn, Mn, Ba — mercury was not analysed) in 50 commercial edible-oil samples representing 24 oil types (12 refined, 12 unrefined) sold in Taiwan, using ICP-MS after microwave digestion. None of the sampled oils exceeded Taiwan, CODEX, IOC, or EU limits for any analyte. Unrefined (cold-pressed and virgin) oils carried significantly higher concentrations than refined counterparts for 8 of 12 metals, with the largest fold-changes in Mn (468-fold), Ba (23-fold), and Zn (11-fold). Olive oil was the consistent exception: per the source text (page 177), refined olive oil carried higher metal concentrations than unrefined extra-virgin olive oil for every measured metal except Pb, Cr, and Co (note: Table 1 vs Table 2 values in fact show refined > unrefined for Pb and Cr in olive; this paper-internal text-vs-table inconsistency is flagged in Verification notes rather than silently re-synthesised here). Incremental lifetime cancer risk from arsenic exceeded 10⁻⁶ for unrefined oils in every age group except 19–65 years, with the unrefined-sesame-oil ILCR in the 0–3 year group also exceeding 10⁻⁶.

Key numbers

Refined oils, concentration ranges across the 12 oil types (µg/kg, page 177):

• As: 0.325 ~ 2.811
• Pb: <LOQ ~ 3.338
• Cd: <LOQ ~ 0.115
• Ni: <LOQ ~ 14.072
• V: <LOQ ~ 0.553
• Cr: <LOQ ~ 7.038
• Co: <LOQ ~ 0.329
• Cu: <LOQ ~ 90.123
• Fe: <LOQ ~ 191.054
• Zn: 40.506 ~ 128.143
• Mn: 0.892 ~ 26.315
• Ba: <LOQ ~ 9.905

Unrefined oils, concentration ranges across the 12 oil types (µg/kg, page 177):

• As: 0.262 ~ 12.262
• Pb: 1.516 ~ 45.226
• Cd: <LOQ ~ 2.512
• Ni: 0.721 ~ 40.173
• V: <LOQ ~ 1.449
• Cr: <LOQ ~ 8.316
• Co: <LOQ ~ 0.495
• Cu: 2.075 ~ 111.840
• Fe: <LOQ ~ 1367.566
• Zn: 66.768 ~ 128.143
• Mn: 2.038 ~ 15,105.418
• Ba: 5.090 ~ 799.567

Refined-oil values for the ingredients with HMI ingredient pages (mean ± SD, µg/kg, Table 1):

• Corn oil (n=2): As 1.261 ± 0.021; Pb 3.198 ± 0.984; Cd 0.115 ± 0.051; Ni 0.833 ± 0.341; Fe 32.789 ± 20.047; Zn 69.612 ± 16.344; Mn 2.743 ± 0.191; Ba 9.905 ± 0.329
• Rice bran oil (n=2): As 2.469 ± 0.470; Pb 2.578 ± 0.186; Cd 0.079 ± 0.001; Ni 0.973 ± 0.371; Fe 30.742 ± 2.404; Zn 101.367 ± 39.665; Mn 2.446 ± 0.931; Ba 9.526 ± 3.212; Cu 21.080 ± 3.638
• Olive oil refined (n=3): As 2.115 ± 0.755; Pb 3.338 ± 3.611; Ni 14.072 ± 18.474; Cr 7.038 ± 5.888; Co 0.329 ± 0.399; Cu 10.566 ± 5.483; Fe 56.122 ± 10.077; Zn 88.288 ± 20.525; Mn 6.122 ± 3.087; Ba 8.310 ± 5.176
• Sunflower oil refined (n=3): As 0.551 ± 0.288; Pb 0.675 ± 0.368; Cd 0.079 ± 0.116; Ni 1.201 ± 0.388; V 0.155 ± 0.116; Cu 3.223 ± 1.361; Fe 90.123 ± 6.878; Zn 69.607 ± 25.511; Mn 2.296 ± 1.008; Ba 6.470 ± 0.447
• Canola oil refined (n=3): As 0.678 ± 0.084; Pb 1.000 ± 0.420; Ni 0.273 ± 0.321; Cu 9.821 ± 10.686; Fe 65.364 ± 29.503; Zn 77.171 ± 6.464; Mn 0.892 ± 0.419; Ba 9.092 ± 1.162 (routed to rapeseed-oil slug)
• Soybean oil refined (n=3): As 1.589 ± 0.870; Pb 1.126 ± 0.698; Ni 0.967 ± 0.167; Cu 4.442 ± 3.991; Fe 55.404 ± 8.561; Zn 95.038 ± 26.591; Mn 1.964 ± 0.441; Ba 6.470 ± 1.793

Unrefined-oil values for the ingredients with HMI ingredient pages (mean ± SD, µg/kg, Table 2):

• Olive oil unrefined (extra virgin, n=5): As 0.467 ± 0.455; Pb 1.516 ± 0.509; Ni 0.721 ± 0.285; Cr 2.530 ± 0.547; Cu 6.328 ± 2.381; Zn 66.768 ± 17.405; Mn 2.038 ± 0.849; Ba 5.090 ± 1.722
• Sunflower oil unrefined (n=1): Pb 13.641 ± 9.368; Ni 6.550 ± 3.884; V 0.386 ± 0.205; Cr 4.812 ± 4.661; Co 0.495 ± 0.340; Cu 53.349 ± 31.253; Fe 271.430 ± 49.252; Zn 143.134 ± 23.400; Mn 14.079 ± 2.470; Ba 10.895 ± 6.012
• Peanut oil unrefined (n=3): As 4.919 ± 1.237; Pb 17.739 ± 4.047; Cd 1.408 ± 1.097; Ni 40.173 ± 62.106; V 0.209 ± 0.126; Cu 101.181 ± 7.608; Fe 910.367 ± 600.046; Zn 513.946 ± 464.134; Mn 350.365 ± 179.116; Ba 23.479 ± 13.708
• Sesame oil unrefined (n=5): As 12.262 ± 14.574; Pb 6.670 ± 1.689; Cd 0.129 ± 0.079; Ni 2.907 ± 1.256; V 1.449 ± 2.791; Cr 2.906 ± 0.981; Co 0.069 ± 0.025; Cu 106.273 ± 4.700; Fe 265.255 ± 363.774; Zn 204.854 ± 211.290; Mn 26.946 ± 55.366; Ba 83.162 ± 172.522

Pb-vs-Taiwan-upper-limit (0.1 mg/kg = 100 µg/kg) ratios in unrefined oils (page 181): camellia 45.2 %, peanut 17.7 %, sunflower 16.2 %. The authors flag camellia oil specifically for ongoing Pb monitoring.

Refined-vs-unrefined comparison, same oil type (page 177):

• Refined oils contained significantly lower metal concentrations than their unrefined counterparts for 8 of 12 metals (p < 0.05).
• Largest fold-decreases in sesame oil: Mn 468.4-fold, Ba 23.2-fold, Zn 11.2-fold lower in refined than unrefined.
• Olive oil reverses the pattern as the source describes it (page 177): “Concentrations of all of the heavy metals in refined olive oil were higher than those in the unrefined oil, except for Pb, Cr, and Co.” (Table 1 vs Table 2 olive rows in fact show refined > unrefined for Pb and Cr; this paper-internal contradiction is preserved as-written and flagged in Verification notes rather than silently re-synthesised.)

Principal-component / cluster analysis (page 181): PC1 (52.24 % variance) was driven by Cd, Ba, Fe, Pb, Co, Zn, and Mn; PC2 (15.80 %) by As and V. Cluster analysis grouped the 19 oil types into five clusters: (1) corn + safflower + olive + coconut + rice bran (higher Ba, Fe, Zn); (2) canola + cottonseed + soybean + avocado + linseed (lower Mn, As, Ni); (3) palm + peanut (higher Co, As, Ni); (4) sunflower + sesame (lower Mn, Ni); (5) camellia + tea seed (higher Ba, Fe, Mn, Cu).

Health-risk metrics (page 181):

• All non-carcinogenic hazard quotients < 1 across all age groups for both refined and unrefined oils.
• Incremental lifetime cancer risk (ILCR, using US EPA oral slope factor 1.5 (kg/day)/mg for inorganic As) from arsenic across both refined and unrefined oils ranged 1.718 × 10⁻⁶ to 1.770 × 10⁻⁸ across age groups; unrefined-sesame ILCR in the 0–3 year group and unrefined oils in all age groups except 19–65 years exceeded the 10⁻⁶ threshold.
• All other heavy-metal exposure proportions fell well below JECFA PTWI (Cu, Cd, Fe, Zn) and EPA PTDI (As, Pb, Cu, Cd, Ni, Fe, Zn, Mn) reference levels (figure 3, all age groups).

Regulatory limits referenced (mg/kg, page 177): Taiwan: Pb 0.1, As 0.1. CODEX: Pb 0.1, As 0.1, Fe 0.4, Cu 0.4. IOC (olive oils): Pb 0.1, As 0.1, Cu 0.1. EU: Pb 0.1. None of the 50 sampled oils exceeded any of these caps.

Methods (brief)

ICP-MS (Agilent 7800, Tokyo, Japan) after microwave digestion (Multiwave Pro, Anton Paar, Graz, Austria). 0.4-g aliquot of each oil sample digested with 2 mL of 65 % HNO₃ (ultrapure, J.T. Baker) at 800 W for 15 min then 1100 W for 40 min, diluted to 10 mL with deionised water, filtered through 0.45-µm ANPEL filters. Calibration range 0.05 ~ 100 µg/kg from a 100 g/L multi-element AccuStandard stock in 2 % HNO₃; correlation coefficients all > 0.995. Quality control by certified reference material (Merck 109460 Multi-element Standard II Oil Dissolved) analysed at 4 µg/kg in triplicate: recoveries 87 ~ 117 %, coefficients of variation 1.332 ~ 3.128 %. Plasma 1450 W; argon 18 L/min; auxiliary 1.2 L/min; nebuliser 0.96 L/min. Statistical analysis in SPSS 19.0 and Minitab 17; one-way ANOVA with p < 0.05 significance, plus principal component analysis and cluster analysis.

Arsenic is reported as total arsenic (no speciation between inorganic and organic forms); the ILCR calculation conservatively treats the total-As measurement as inorganic-As for risk purposes using the EPA oral slope factor of 1.5 (kg·day)/mg. Mercury (any species) was not analysed in this study because of analytical-method incompatibility with the other 11 metals on the same instrumental run. Chromium is reported as total chromium with no speciation between Cr(III) and Cr(VI).

Implications

Certification: Provides primary measurements for Pb, tAs (total As), Cd, Ni, Cr (total), and Cu in 24 commercial edible oil types sold in Taiwan, with both refined and unrefined samples for each oil type where market availability permitted. The dataset is directly routeable to products/olive-oil (HMTc Cat 7 Row 1) and products/cooking-oils-other (HMTc Cat 7 Row 2) as primary literature evidence for refined-vs-unrefined heavy-metal occurrence in commercial edible oils.

Courses: The dataset is a strong teaching example of how refining (degumming, deacidification, bleaching, deodorisation) removes heavy metals from edible oils for most oils but not for olive oil, and of how PCA can cluster oils by contamination profile. The 468-fold Mn reduction in refined sesame oil and the camellia-oil Pb monitoring flag are useful single-finding hooks.

App: Unrefined oils carry materially higher heavy-metal loads than refined counterparts for most edible-oil types, with camellia oil, peanut oil, and sunflower oil showing the highest unrefined Pb levels (45.2 %, 17.7 %, and 16.2 % of the Taiwan 0.1 mg/kg cap respectively, page 181). For consumer-facing risk context the app can convey that “cold-pressed/virgin” oils — generally marketed as a premium attribute — also carry higher concentrations of multiple heavy metals than their refined counterparts.

Speciation caution: Arsenic figures are total arsenic. The ILCR calculation treats total As as inorganic As, which is an over-estimating assumption when organic-As fractions are non-trivial; the authors flag the lack of speciation as a limitation. Chromium figures are total chromium with no Cr(VI) discrimination.

Wiki pages this source may touch

• olive-oil
• soybean-oil
• sunflower-oil
• rice-bran-oil
• peanut-oil
• sesame-oil
• corn-oil
• rapeseed-oil
• vegetable-oils
• olive-oil
• cooking-oils-other
• lead
• cadmium
• arsenic-total
• nickel
• chromium
• cobalt
• copper
• iron
• zinc
• manganese
• barium
• vanadium

Verification notes

• 2026-06-01 ingest (Claude Opus, MFK_jsfa-reports-2024-lin-determination-distribution-a): fresh page from the Manual Fetch Kimi / June 1 batch. Three identity checks (DOI 10.1002/jsf2.186, raw_handle MFK_jsfa-reports-2024-lin-determination-distribution-a, cite-key lin2024-edible-oils-taiwan-12-metals) all negative against wiki/sources/; no merge-enhance target. Mercury (any species) deliberately omitted from metals: per page 182 acknowledgement that Hg was not analysed in this study. Arsenic recorded as tAs (no speciation); chromium recorded as Cr (total). All 12 measured analytes appear in metals: including Co, Cu, Fe, Zn, Mn, Ba, V (non-HMTc panel metals; included for routing to corresponding metals/ pages).
• Missing-ingredient slugs the paper provides primary data for (FLAGGED, not created — per skill scope): palm oil, coconut oil, medium-chain triglyceride oil, grapeseed oil, cotton seed oil, safflower oil, avocado oil, camellia oil, linseed oil, perilla seed oil, tea seed oil. Canola oil routed to existing rapeseed-oil slug (canola is the LEAR rapeseed cultivar). When these accumulate freq-2 across the corpus the auto-stub script will create scaffolds.
• Part 12 brand-firewall check: the paper sources oils from supermarkets and online platforms and names no individual brand in body text or tables; rows in Tables 1 and 2 are oil-type aggregates with sample size n=1 to n=5 per oil. No brand-firewall edits required. Method-section vendor names (Agilent 7800, Anton Paar Multiwave Pro, Merck 109460 CRM, ANPEL filters, AccuStandard calibration stock, J.T. Baker HNO₃, Millipore Milli-Q, SPSS 19.0, Minitab 17) are scientific-method names per the 2026-05-17 carve-out and are retained.
• Mn unrefined range upper bound “15,105.418 µg/kg” reported on page 177 is anchored in camellia oil Mn per Table 2 (camellia row mean ± SD = 15,105.418 ± 4,455.119, range 11,953.173–18,255.663; tea seed oil row mean ± SD = 732.451 ± 608.060, range 302.488–1,162.415). The page-181 text-narrative agrees: “Pb, Mn, and Ba in camellia oil … Cu and Fe in tea seed oil.” No transposition; the Mn anchor is camellia, not tea seed.
• Refined-olive-oil-higher-than-unrefined-olive-oil finding is a paper-internal observation that the authors highlight (page 177); the source text explicitly excludes Pb, Cr, and Co from the “refined higher” pattern (“Concentrations of all of the heavy metals in refined olive oil were higher than those in the unrefined oil, except for Pb, Cr, and Co.”). Note: Table 1 vs Table 2 olive rows in fact show refined > unrefined for Pb (3.338 vs 1.516 µg/kg) and Cr (7.038 vs 2.530 µg/kg), so the source’s text/table internal contradiction is preserved as-written rather than silently re-synthesised. Co cannot be compared because Table 2 unrefined-olive Co reads ND.
• 2026-06-01 audit application (Claude Opus, this session, applying findings from a fresh-context general-purpose audit subagent): subagent verdict REVISE, with three ❌ findings on Check 1 (numerical fidelity) and four ⚠️ concerns. Findings applied: (a) Verification-notes Mn anchor corrected from “tea seed oil” to “camellia oil” with Table 2 verification (above); (b) App-section Pb-percentile sentence corrected from a four-oil list (“camellia 45 %, tea seed 28 %, peanut 18 %, sesame 17 %”, two values fabricated) to the source’s actual three-oil list (camellia 45.2 %, peanut 17.7 %, sunflower 16.2 % per page 181); (c) “Olive oil reverses the pattern: refined > unrefined for every analyte” replaced in the body intro and in the Refined-vs-unrefined comparison block with the source’s stated exception (“except for Pb, Cr, and Co”, page 177), with a flag of the source’s internal text/table contradiction; (d) sunflower-refined Cd SD corrected from ”± 0.388” to ”± 0.116” — the ”± 0.388” value matched the Ni column ± 0.388 on the same row and produced a mean ± SD range “0.079 ± 0.388” mathematically inconsistent with the printed range “ND~0.289”; ± 0.116 mirrors the adjacent V-column SD and is consistent with the printed range; (e) PC1 driver list expanded from five metals (Cd, Ba, Fe, Pb, Co) to the seven listed on page 181 (Cd, Ba, Fe, Pb, Co, Zn, Mn); (f) cluster 2 oil list expanded to include soybean oil (page 181); (g) ILCR-range attribution scope corrected from “unrefined oils” to “refined and unrefined oils across age groups”. Findings rejected (false positives): none. The audit subagent’s ⚠️ on sesame-unrefined Ni “2.907 vs 2.906” was a last-digit transcription whose materiality is below the 0.001 µg/kg threshold; left unchanged.

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.