Chen et al. 2020 — Seventeen elements in five Chinese milk types by ICP-MS
Chen and colleagues at the Shandong Academy of Agricultural Sciences analyzed 17 major, minor, and trace elements in 350 raw milk samples spanning five species — goat, cow, buffalo, yak, and camel — collected from five Chinese provinces. The analytical platform was ICP-MS (Thermo iCAP Q) operated in kinetic energy discrimination mode after microwave HNO3/H2O2 digestion, with two milk-matrix CRMs (GBW10011 wheat, GBW10017 milk powder) used for accuracy verification. The paper’s primary aim was to use multi-element profiles plus chemometrics (PCA, factor analysis, LDA) to discriminate species; the heavy-metals subset (Pb, Cd, tAs, Cr, Ni, Al, Sn) is a secondary outcome that is nevertheless one of the few cross-species ICP-MS occurrence datasets for raw milk in China. Mean Pb in cow milk (23.4 ng/g) exceeded the Codex CXS 193-1995 maximum level for raw milk (20 ng/g, 20 ppb); all other toxic trace elements were below the corresponding limits.
Key numbers
All concentrations from Table 4. Footnote in source: starred elements (As*, Cd*, Cr*, Ni*, Pb*, Se*, Sn*) are reported in ng/g (= µg/kg = ppb); unstarred elements (Al, Ca, Cu, Fe, K, Mg, Mn, Na, Sr, Zn) are reported in µg/g (= mg/kg, so ppb = µg/g × 1000). Values are mean ± SD; different superscript letters in the source table indicate statistically significant differences (Duncan’s multiple range test, p < 0.05).
Toxic and HMTc-relevant trace elements, by milk type:
| Element | Goat (n=100) | Cow (n=100) | Buffalo (n=50) | Camel (n=50) | Yak (n=50) |
|---|---|---|---|---|---|
| Pb (ng/g = ppb) | 7.97 ± 7.50 | 23.4 ± 13.8 | 17.3 ± 14.8 | 18.2 ± 7.1 | 4.31 ± 2.45 |
| Cd (ng/g = ppb) | 0.425 ± 0.305 | 0.767 ± 0.558 | 0.676 ± 0.725 | 0.786 ± 1.191 | 0.254 ± 0.221 |
| tAs (ng/g = ppb) | 4.27 ± 3.93 | 4.61 ± 2.20 | 3.81 ± 2.26 | 8.06 ± 6.57 | 1.12 ± 6.57 |
| Cr (ng/g = ppb) | 11.7 ± 5.2 | 15.0 ± 9.2 | 7.94 ± 5.80 | 13.6 ± 5.3 | 1.85 ± 0.63 |
| Ni (ng/g = ppb) | 38.3 ± 26.3 | 81.9 ± 68.0 | 62.4 ± 47.4 | 131 ± 148 | 66.8 ± 65.3 |
| Sn (ng/g = ppb) | 66.9 ± 28.3 | 97.6 ± 61.9 | 52.5 ± 16.7 | 50.6 ± 29.0 | 14.2 ± 30.5 |
| Al (µg/g, = ppb × 10⁻³) | 0.277 ± 0.155 | 0.493 ± 0.196 | 0.391 ± 0.176 | 0.455 ± 0.287 | 0.379 ± 0.181 |
| Se (ng/g = ppb) | 28.1 ± 10.4 | 37.2 ± 14.2 | 32.4 ± 11.0 | 29.4 ± 18.0 | 14.0 ± 5.3 |
Converted to consistent ppb (µg/kg) for the seven HMTc-relevant analytes measured:
- Pb (total): goat 7.97, cow 23.4, buffalo 17.3, camel 18.2, yak 4.31 ppb. Cow milk mean exceeds Codex CXS 193-1995 raw-milk maximum level of 20 ppb; the authors flag this as a “human health risk according to CAC standards.”
- Cd: 0.25–0.79 ppb across species; all well below typical milk Cd benchmarks.
- tAs (total arsenic, no speciation): 1.12–8.06 ppb. Camel highest, yak lowest. Note: the yak SD of ±6.57 ppb on a 1.12 ppb mean is anomalously large in the source table and is reproduced verbatim here; the authors do not flag it.
- Cr (total chromium, no Cr-VI speciation): 1.85–15.0 ppb. Cow highest, yak lowest. Well below China GB2762-2017 milk Cr limit of 300 ppb.
- Ni: 38.3–131 ppb. Camel highest (with very large SD of ±148, indicating high inter-sample variance).
- Sn: 14.2–97.6 ppb. Cow highest, yak lowest. No Codex MRL for inorganic Sn in milk applies here (CXS 193-1995 Sn limit is for canned foods).
- Al: 277–493 ppb (0.277–0.493 µg/g). Cow highest, goat lowest.
- Se: 14.0–37.2 ppb. Cow highest, yak lowest.
Major elements (µg/g, = ppm wet weight): K 698 (buffalo) to 1377 (goat); Ca 516 (cow) to 888 (camel); Na 253 (goat) to 428 (camel); Mg 58.5 (buffalo) to 96.7 (yak). Zn 3.11 (goat) to 5.81 (camel) µg/g. Cu 0.165 (cow) to 0.522 (yak) µg/g. Fe 1.08 (goat) to 2.54 (yak) µg/g. Sr 0.695 (buffalo) to 3.05 (camel) µg/g. Mn 0.156 (goat) to 0.256 µg/g (yak; cow 0.187, buffalo 0.169, camel 0.188).
Mercury (Hg, tHg, MeHg) was NOT measured. Inorganic As (iAs) and hexavalent Cr (Cr-VI) speciation was NOT performed. Antimony (Sb), uranium (U), and methylmercury (MeHg) were NOT measured.
Regulatory context cited by the authors:
- Codex Alimentarius Commission, Codex Stan 193-1995: Pb maximum residue level in raw milk = 20 ng/g (20 ppb). Cow milk in this study (23.4 ng/g) exceeds this limit; goat, buffalo, camel, and yak means are below.
- China GB2762-2017 (Ministry of Health of the People’s Republic of China, 2017): Pb 50 ng/g, Cr 300 ng/g, As 100 ng/g for milk. All species means in this study are below all three Chinese limits.
Methods (brief)
Sample collection: 350 raw untreated milk samples from small farm cooperatives and large-scale farms. After stirring in holding tanks, 100 mL composites were drawn from upper, middle, and lower thirds (300 mL per tank). Samples stored at −20 °C in polyethylene bottles until analysis. Provincial allocation: goat (50 Shandong + 50 Shaanxi), cow (100 Shandong), buffalo (50 Guangxi), camel (50 Xinjiang), yak (50 Sichuan).
Digestion: 2 g milk + 6 mL HNO3 (65%, trace metal grade) + 2 mL H2O2 (30%, analytical reagent) in tetrafluoroethylene containers, overnight pre-soak then microwave digestion (CEM MARS5) in three stages (120 °C/5 min, 150 °C/10 min, 190 °C/30 min). Digestate diluted to 25 mL with ultrapure water (Milli-Q, 18 MΩ·cm).
Analysis: ICP-MS (Thermo iCAP Q) with autosampler, 1550 W RF power, kinetic energy discrimination mode, dwell time 20 ms. Isotopes measured: 27Al, 75As, 44Ca, 111Cd, 52Cr, 65Cu, 57Fe, 39K, 24Mg, 55Mn, 23Na, 60Ni, 208Pb, 82Se, 118Sn, 88Sr, 66Zn. Blank correction applied; each sample measured three times and averaged.
Quality assurance: Two CRMs (GBW10011 wheat powder; GBW10017 milk powder; Institute of Geophysical and Geochemical Exploration, China) digested and analyzed alongside samples. t-test (95% CI) showed no significant difference between measured and certified values for any of 17 elements. Accuracy ranged 88.2% (Zn) to 114.3% (Pb) across the two CRMs. Precision (RSD on three replicates): 0.7–12.5%. LODs (mg/kg, from Table 3): Pb 0.003, Cd 0.0001, As 0.0008, Cr 0.001, Ni 0.008, Al 0.033, Sn 0.009, Se 0.008.
Statistics: One-way ANOVA with Duncan’s multiple range test (p < 0.05) for species comparisons. Pearson correlation analysis. PCA, factor analysis, and stepwise LDA (calibration set n=280, external validation set n=70) for species discrimination, all in SPSS 23.0.
Implications
Certification: One of the few cross-species occurrence datasets for raw milk in China at ICP-MS sensitivity, covering all of Pb, Cd, tAs, Cr (total), Ni, Al, Sn — seven of the ten HMTc analytes. Cow milk mean Pb at 23.4 ppb is a notable above-Codex-MRL finding from a 100-sample Shandong dataset; relevant when assessing cow-milk-derived ingredient priors and when benchmarking certification thresholds for any dairy-bearing product category. Buffalo and camel milk Pb means (17.3 and 18.2 ppb) sit near the 20 ppb Codex limit. The Cd, tAs, Cr, Ni, Al, Sn means across all species sit well below both Codex and Chinese regulatory ceilings, providing a useful occurrence baseline for the lower-priority HMTc analytes in non-cow dairy. No Hg, MeHg, iAs, or Cr-VI data — these gaps remain in the camel/yak/buffalo dairy literature.
Courses: Useful for teaching multi-element ICP-MS dairy methodology (CRM selection, KED mode rationale for ArCl-on-As interference, microwave digestion staging, LOD vs LOQ derivation), and for teaching chemometric species discrimination as a tool for authenticity verification (LDA classifying 95.7% of held-out samples correctly). Also a good worked example of why total-element ICP-MS without speciation has limited utility for As and Cr risk assessment — both elements would need HPLC-ICP-MS coupling to support a defensible risk-characterization claim.
App: Provides occurrence priors for raw cow, goat, buffalo, camel, and yak milk in a Chinese supply-chain context for seven HMTc analytes. The cow-milk Pb prior (mean ≈23 ppb, max not reported but inferable from SD as approaching 60 ppb) is meaningful for any consumer-facing assessment of Chinese-origin raw cow milk or dairy ingredients sourced from Shandong. The Cd, tAs, Cr, Ni, Al, Sn priors are all “low ppb” across all species and would land in the low-risk band for an app-layer summary.
Wiki pages this source may touch
Verification notes
The cite-key uses milk-types-elements-china rather than the raw_handle’s camel-milk-heavy-metals because the paper analyzes five milk species and 17 elements — camel milk is one of five matrices, not the primary subject. The raw_handle is preserved verbatim from the discover skill’s auto-fetch.
Table 4 footnote states “concentration is expressed as ng g−1” but the starred-vs-unstarred convention applies only to the starred columns (As*, Cd*, Cr*, Ni*, Pb*, Se*, Sn*). Unstarred elements (Al, Ca, Cu, Fe, K, Mg, Mn, Na, Sr, Zn) are in µg/g. This dual-unit convention is unusual; the Key numbers section converts everything to ppb (µg/kg) to align with HMI conventions.
The yak tAs value of 1.12 ± 6.57 ng/g in Table 4 carries an SD nearly 6× the mean, which is internally inconsistent for a 50-sample dataset. The authors do not comment on this; one possibility is a typo in the SD digit (e.g., 0.657 transposed to 6.57). The value is reproduced verbatim here per HMI policy of preserving source values; downstream synthesis should treat the yak tAs SD as suspect and weight by the other four species’ tAs values.
The paper does not report individual sample maxima, P90, or P95 — only mean ± SD per species per element. Distribution shape cannot be characterized from the published data; range information is qualitative (“ranged from X to Y”) in the prose for some major elements and absent for the toxic trace elements except via mean and SD.
No brand names appear in the source; only farm-level aggregation by province. Brand-firewall (Part 12) compliant. Method-vendor names (Thermo iCAP Q, CEM MARS5, Millipore Milli-Q, SPSS, Sinopharm, IBM) are reproduced under the methods-vendor exception.
Sample-year range is not stated in the paper; the manuscript was received 14 January 2020 and published 11 February 2020, so sample collection was likely 2018–2019 but is not confirmed in text.
Speciation absence: tAs and Cr values are total-element only. Any future use of these values in iAs or Cr-VI synthesis would require an explicit assumption about speciation fractions (e.g., iAs ≈ 70–90% of tAs in dairy is a commonly cited heuristic but should be flagged as inferred, not measured).
The Codex Stan 193-1995 raw-milk Pb maximum residue level cited by the authors (20 ng/g) is the General Standard for Contaminants and Toxins in Food and Feed limit, not the canned-foods Sn standard. The HMI regulation slug codex-cxs-193-1995-tin-canned-foods is the canned-tin standard and is not the appropriate wikilink target for the Pb finding here; the prose summary states the limit numerically without wikilinking.
Audit subagent (general-purpose, fresh context, 2026-05-30) flagged the Mn parenthetical in Key numbers as garbled — “Mn 0.156 (yak) to 0.256 µg/g (yak; cow 0.187, goat 0.156, …)” was internally inconsistent. Verified against Table 4: lowest Mn is goat (0.156 µg/g), highest is yak (0.256 µg/g). Corrected to “Mn 0.156 (goat) to 0.256 µg/g (yak; cow 0.187, buffalo 0.169, camel 0.188).” Audit also raised a concern that the milk-species matrices (cow-milk, goat-milk, buffalo-milk, yak-milk, camel-milk, raw-milk) might need to be flagged as proposed additions to the closed matrices vocabulary — verified against current usage: cow-milk is established in akhmetsadykova2013, camel-milk and raw-milk in yazdanian2025; these are existing matrix slugs in active use, not proposed additions. Finding rejected as false positive.
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.