Heavy Metal Levels in Milk and Cheese Produced in the Kvemo Kartli Region, Georgia

This study characterized 11 elements (Cr, Mn, Fe, Co, Ni, Cu, Zn, Se, Mo, Cd, Pb) in raw cow’s milk (n=195, 9 villages) and traditional Georgian cheeses (n=25; 16 Imeruli, 9 Sulguni) from the Kvemo Kartli region of Georgia, an agricultural area along the Mashavera River with documented soil, water, and pasture contamination from the Madneuli polymetallic mining plant (operating since 1975) and the RMG gold/copper mine (since 2014). Quantification was by ICP-MS (Agilent 7800) on samples ashed with concentrated HNO3 (65%) in a muffle furnace, with quality control via ISO 17034 traceable CRMs and the GOST 26929-94 Georgian national method. Mean Cd in milk was below LOD (<0.001 mg/L) in all villages. Mean Pb in milk ranged 0.004–0.048 mg/L across villages, with seven individual samples exceeding the EU limit of 0.020 mg/L in Kvemo Bolnisi, Bolnisi, Mtskneti, and Ratawani. Cu means in Sabereti and Vanati (0.568 and 0.592 mg/L) exceeded the EU limit of 0.4 mg/L. In cheese, Pb in both Imeruli (mean 0.121 mg/kg ww) and Sulguni (mean 0.258 mg/kg ww) exceeded the EU/Codex maximum level of 0.020 mg/kg but were below the Georgian regulation MPC of 0.500 mg/kg; Mn, Cr, Se, Mo, Zn, and Fe in both cheeses also exceeded the cited permissible limits. A t-test showed Sulguni cheese had significantly higher Cr, Se, Mn, Pb, Co, Cu, and Ni than Imeruli cheese, consistent with the fact that Sulguni is made by re-processing Imeruli curd and that heavy metals partition preferentially into the casein/curd fraction. A second t-test on milk samples grouped by river basin showed significantly higher Fe, Mn, Cu, Pb, and Cr in the Mashavera-region villages than in the Khrami-region villages.

Key numbers

All milk values in mg/L; cheese values in mg/kg wet weight (ww). Method: ICP-MS (Agilent 7800). n=195 milk samples; n=25 cheese samples (16 Imeruli, 9 Sulguni).

Table 1 — Raw cow’s milk by village (mean ± SD, mg/L):

Permissible limits cited (milk, mg/L): Cr 0.02 [Flynn 1992]; Mn 0.02–0.05 [Knowles et al. 2006]; Fe 0.7 [Storelli et al. 2007; Safonov 2020]; Co 0.006 [Flynn 1992; Storelli et al. 2007; Hurley 1997]; Ni 0.027 [Flynn 1992; European Commission 2001]; Cu 0.4 [European Commission, EU [68]]; Zn 3–5 [WHO 1996], 2–6 [Pechovà et al. 2008]; Se 0.5 [EFSA 2011]; Mo 0.05 [Flynn 1992]; Cd 0.2 [Georgian Ministry of Labour and Health, Social Affairs 2001]; Pb 0.020 [Codex Alimentarius 1995; European Commission 2006] and 0.500 [Georgian regulation 2001].

LOD, milk (mg/L): Cr 0.00013; Mn 0.00010; Fe 0.00041; Co 0.00001; Ni 0.00007; Cu 0.00010; Zn 0.00033; Se 0.00533; Mo 0.00005; Cd 0.00001; Pb 0.00006.

Table 2 — Cheese (mean ± SD, mg/kg wet weight):

Lead exceedance in milk. Seven individual milk samples exceeded the EU/Codex Pb limit of 0.020 mg/L (Section 4 Conclusions), in the villages Kvemo Bolnisi, Bolnisi, Mtskneti, and Ratawani. The Kvemo Bolnisi village mean (0.048 ± 0.204 mg/L) carries an SD several times the mean, indicating that the exceedance is driven by a small number of high-end samples within an otherwise low-Pb local distribution. The abstract reports an individual-sample Pb range of “0.027 to 1003 mg L⁻¹” in the exceedance subset; the 1003 mg L⁻¹ upper bound is two to four orders of magnitude above any per-village mean reported in Table 1 and is treated as a typographical error in the abstract (likely intended 1.003 mg L⁻¹). The per-village means in Table 1 remain the authoritative summary of the milk Pb distribution.

Copper exceedance in milk. Mean Cu in Sabereti (0.568 mg/L) and Vanati (0.592 mg/L) exceeded the EU permissible limit of 0.4 mg/L (European Commission [68]); the abstract’s per-village range for Cu in these villages is given as 0.42–1.28 mg/L (individual samples). Cu was below the EU limit at the village-mean level in the other seven villages.

Cheese vs milk: Pb and Cd partitioning. Pb in Imeruli (mean 0.121 mg/kg ww) and Sulguni (mean 0.258 mg/kg ww) exceeded the EU/Codex Pb ML of 0.020 mg/kg ww but were below the Georgian MPC of 0.500 mg/kg. Cd in cheese (Imeruli 0.002, Sulguni 0.007 mg/kg ww) was below the Georgian MPC of 0.2 mg/kg. The authors note that during cheesemaking κ-casein hydrolysis partitions Pb and Cd preferentially into the casein/curd fraction (citing Maas et al. 2011: 60% of Cd to curd, 14% to whey), concentrating these metals in the finished cheese relative to the source milk.

Other element exceedances in cheese (paper’s own classification, Section 4 Conclusions): Mn, Cr, Se, Mo, Zn, and Fe in both Imeruli and Sulguni exceeded the cited permissible limits. Co, Ni, and Cu in both cheeses were below their cited permissible limits.

Table 3 — t-test, Imeruli vs Sulguni cheese (alt. hypothesis: Imeruli < Sulguni; df=23):

The authors emphasise Cr (Cohen’s d 1.209) and Se (Cohen’s d 1.095) as showing the largest effect sizes; Pb, Mn, and Co also showed large effects.

Table 4 — t-test, milk Mashavera vs Khrami river basin (alt. hypothesis: Khrami < Mashavera; df=189; means in mg/L; n=146 Mashavera, n=45 Khrami):

Fe shows the largest river-basin effect (Cohen’s d 0.702); Mn second (0.490). Pb, Cr, and Cu also significantly higher in Mashavera-basin milk despite small effect sizes.

Methods (brief)

ICP-MS quantification on an Agilent 7800 (Agilent Technologies, Santa Clara, CA, USA). Sample preparation: milk samples 50.0 mL in duplicate placed in open porcelain crucibles treated with 1 mL concentrated HNO3 (65%) and heated on a hot plate (IKA, Germany) until fumes stopped, then dry-ashed in a muffle furnace (Naberthem L15/11, Germany) ramped from 250 °C to 450 °C (50 °C every 30 min); residue dissolved in dilute HNO3, dried at 140 °C, and re-ashed at 300 °C for 30 min until white ash was obtained; final dissolution in 5 mL concentrated HNO3 and dilution to 25 mL. Cheese samples 10.0 g in duplicate were treated with 1 mL HNO3 (65%) diluted 1:1 with deionized water, ashed by the same protocol, and diluted to 50 mL. Two reagent blanks were run with each batch.

Quality control: 10-point calibration curves per element (nine standards + one blank) with R ≥ 0.9999 before analysis; daily internal control materials; duplicate analyses every 20 samples; participation in inter-laboratory proficiency testing; CRMs prepared per ISO 17034 and traceable to NIST. Method conformed to GOST 26929-94 (Georgian national standard for analysis of toxic elements in food). Measurement uncertainty calculated per the EURACHEM/CITAC guide. The state research institute laboratory participates in proficiency programs but is not independently accredited. Statistics: SPSS 27.0 (IBM, Armonk, NY, USA); independent-samples t-test; Cohen’s d for effect size; Vovk–Sellke maximum p-ratio also reported.

LOD (milk, mg/L) and LOD (cheese, mg/kg ww) are listed in ## Key numbers above.

Implications

Certification: Kvemo Kartli is a documented heavy-metal-contaminated agricultural zone driven by adjacent active mining (Madneuli polymetallic plant since 1975; RMG gold/copper mine since 2014) and Mashavera River pollution. Pb in the finished cheese exceeded the EU/Codex ML of 0.020 mg/kg ww in both traditional cheese types by 6× (Imeruli) and 13× (Sulguni); milk Pb passed the EU limit on a village-mean basis in all 9 villages but had 7 individual-sample exceedances. The cheese-vs-milk Pb concentration ratio (Sulguni 0.258 mg/kg ww vs Kvemo Bolnisi milk 0.048 mg/L mean) demonstrates that fermentation/curd-formation concentrates Pb from milk by a factor of several × on a wet-weight basis, consistent with the casein-binding mechanism the authors cite. Sourcing dairy from mining-adjacent regions for HMTc-relevant categories requires sample-level Pb and Cd testing at the finished-cheese stage, not only at the milk stage, because the cheese matrix is the binding endpoint.

The river-basin t-test (Table 4) shows that the milk supply along the Mashavera River basin carries significantly higher Pb, Cu, Fe, Mn, and Cr loads than the Khrami basin within the same Kvemo Kartli region. This is an explicit geographic-supply-chain stratification within a single national jurisdiction.

Courses: Demonstrates the soil → pasture → animal → milk → cheese contamination pathway in an active mining region, and the casein-mediated metal partitioning that concentrates Pb in cheese during processing. The Madneuli/RMG mining context provides a well-documented point-source case study. The paper also documents that Imeruli and Sulguni are traditional Georgian cheeses that cannot be exported (do not meet international standards) and are consumed only domestically, which is the relevant exposure framing.

App: Contributes per-village raw cow’s milk and traditional Georgian cheese occurrence data under a mining-adjacent contamination scenario to the milk-and-dairy contamination profile. Per-element and per-village means with SD are usable as a regional reference for the Caucasus mining belt.

Wiki pages this source may touch

• milk-and-dairy
• lead
• cadmium
• chromium
• nickel
• copper
• manganese
• iron
• cobalt
• zinc
• molybdenum

Verification notes

2026-05-19 (Claude Code, merge-enhance from prior 2026-05-14 page). The prior revision had several defects corrected here: (a) raw_path: was truncated to …Kvemo Kartl.pdf and pointed to a non-existent file; corrected to the full filename Heavy Metal Levels in Milk and Cheese Produced in the Kvemo Kartli Region, Georgia.pdf (verified the real file exists at that path). (b) metals: listed only [Pb, Cd, Cr, Ni]; the paper measures 11 elements (Cr, Mn, Fe, Co, Ni, Cu, Zn, Se, Mo, Cd, Pb) and the field is now complete. (c) Frontmatter previously listed ingredients: [milk, cheese]; neither slug exists in the current taxonomy. Replaced with the only valid umbrella slug [[ingredients/milk-and-dairy]]. No cheese ingredient slug exists in the taxonomy and the live cheese ingredient pages (cheddar-cheese, processed-american-cheese, fromage-frais, macaroni-and-cheese) do not match Imeruli/Sulguni; the matrices field carries the traditional-cheese specificity instead. (d) The prior body text stated the Cu EU limit was “~0.05 mg/L” — incorrect. The paper cites the European Commission Cu limit for milk as 0.4 mg/L (reference [68]); corrected in ## Key numbers and ## Implications. (e) The prior body misspelled the village name “Ratewani”; corrected to “Ratawani” per Table 1 and Section 4. (f) The legacy heading ## Wiki pages updated on ingest was renamed to ## Wiki pages this source may touch per current Part 6 template. (g) ## Key numbers previously gave only abstract-level summaries; expanded to reproduce Tables 1, 2, 3, and 4 in full so the per-village, per-element, and t-test detail is on the page rather than only in the source PDF.

Sample size. sample_n = 220 = 195 milk + 25 cheese, consistent with Section 2.1.

Abstract typographical anomaly preserved. The abstract reports the Pb-exceedance individual-sample range as “0.027 to 1003 mg L⁻¹”. The 1003 mg L⁻¹ upper bound is incompatible with every village mean (max 0.048 ± 0.204 mg/L in Kvemo Bolnisi) and with the cheese maxima, and is treated as a typographical error in the published abstract (likely intended 1.003 mg L⁻¹, or possibly 0.103 mg L⁻¹). Per-village means in Table 1 are authoritative. This caveat is recorded here rather than silently corrected.

Audit subagent (2026-05-19) flagged source-internal Table 2 vs Table A1 disagreement on Se Imeruli mean. Table 2 (p.8) reports Imeruli Se mean = 1.003 ± 0.901 mg/kg ww; Appendix A Table A1 (p.15) reports Imeruli Se mean = 0.878, SD 0.906 mg/kg ww. The Section 3.3 narrative (p.12) and the Table 3 t-test mean-difference of 2.182 are computed against the 0.878 figure (Imeruli) and 3.06 (Sulguni). Verified against the source — finding correct. The wiki page reproduces both values faithfully (Table 2 Imeruli Se 1.003 in ## Key numbers; Table 3 t-test mean diff 2.182 derived from 0.878) because both numbers appear in the published paper and the page’s purpose is to reflect what the paper reports. Downstream synthesis passes should treat this as a source-internal inconsistency to resolve rather than as a wiki defect; the Section 3.3 narrative (0.878) is the more internally consistent value because it is the one the paper’s own statistical analysis is computed from. Minor source-side Mn rounding (Table 2 Imeruli 0.886 vs Table A1 0.887; Table 2 Sulguni 2.348 vs Table A1 2.35) is the same class of finding and is similarly preserved verbatim.

Brand firewall. No consumer-brand names appear in the source paper. The methods section names instrument and equipment vendors (Agilent 7800; Naberthem L15/11 muffle furnace; IKA hot plate; IBM SPSS 27.0) and the GOST 26929-94 / ISO 17034 / NIST references; these are retained per the 2026-05-17 scientific-method exception (Check 4 of docs/gpt-collaboration/audit-prompt.md). No brand-by-brand contamination ranking exists in the source or this page.

Wiki/HMTc firewall. No HMTc threshold proposals, consumer risk advisories, or cross-source synthesis claims are made. The ## Implications section limits itself to what this single paper contributes to threshold work and explicitly does not propose certification values.

Speciation. Paper reports total chromium without hexavalent speciation; metals: uses Cr not Cr-VI. No As or Hg analyses are reported in the paper.

Products field. products: [] is retained because no raw-cow-milk or traditional-cheese product slug exists in the current taxonomy; the Category 5 (Dairy/Eggs/Alternatives) Step 0 Lock has not yet been authored. The matrices field carries the product-form specificity (raw-cow-milk, cheese, imeruli-cheese, sulguni-cheese).

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.