Andrade et al. 2023 - Metals in Portuguese cow milk and soy beverages

Andrade and colleagues measured lead, cadmium, and manganese in 28 retail beverages purchased on the Portuguese market in Lisbon between February and May 2019, split evenly between 14 cow-milk (CM) brands and 14 soy-beverage (SB) brands. The basis is liquid as-purchased, measured in micrograms per liter; the authors do not convert to per-kilogram or apply a reconstitution factor. Soy beverages carried significantly higher cadmium (mean 5.6 μg/L vs 2.15 μg/L, p < 0.05) and roughly twenty-fold higher manganese (mean 117.4 μg/L vs 5.93 μg/L, p < 0.05) than cow milk; lead was numerically higher in cow milk (19.3 vs 13.4 μg/L) but the difference was not statistically significant (p > 0.05). The cow-milk Pb mean sat just under the EU 2023/915 maximum level of 20 μg/kg, and one cow-milk brand’s cadmium value exceeded the 5 μg/kg EU ceiling that applies to infant-formula products manufactured from cow’s milk proteins.

Key numbers

All concentrations below are in micrograms per liter (μg/L) of the beverage as purchased and as analyzed. The authors did not apply density conversion to express values per kilogram in the results tables; they cite a cow-milk density of 1.030 kg/L when comparing their lead value to the EU 20 μg/kg ceiling.

Global means by beverage class

Within-class ranges across brands

Within the 14 cow-milk brands, lead ranged from 5.6 to 39.9 μg/L, cadmium from 0.6 to 6.8 μg/L, and manganese from 4.6 to 8.9 μg/L. Within the 14 soy-beverage brands, lead ranged from 5.4 to 41.7 μg/L, cadmium from 0.9 to 15.4 μg/L, and manganese from 72.3 to 177.9 μg/L. Per-brand values are reported in Tables 10 and 11 of the source under anonymized codes CM1-CM14 and SB1-SB14; this page does not reproduce the per-brand table under the brand-firewall rule.

Regulatory benchmarks the authors cite

The Commission Regulation (EU) 2023/915 maximum level for lead in milk is 0.020 mg/kg, or 20 μg/kg. Converting the cow-milk lead mean of 19.3 μg/L using the cited density of 1.030 kg/L gives roughly 18.7 μg/kg, close to but below the ceiling. The same regulation sets a 5 μg/kg ceiling on cadmium in infant formula, follow-on formula, foods for special medical purposes intended for infants and young children, and formulas for young children placed on the market in liquid form and manufactured from cow’s milk proteins or cow’s milk protein hydrolysates. The mean cow-milk cadmium of 2.15 μg/L falls below that 5 μg/kg ceiling, but one cow-milk brand reached 6.8 μg/L of cadmium, which is above the 5 μg/kg ceiling that would apply if that milk were used as the cow’s-milk protein source in a liquid infant formula. The paper notes that European legislation has not set maximum levels for the metals studied here in soy beverages specifically.

Method validation

Calibration coefficients of determination ranged from 0.995 to 0.999 across the three metals. Limits of detection by the calibration-plot method were 0.767 μg/L for Pb, 0.294 μg/L for Cd, and 1.375 μg/L for Mn; limits of quantification were 2.325 μg/L for Pb, 0.892 μg/L for Cd, and 4.166 μg/L for Mn. Intra-day precision relative standard deviation ranged from 2.63 percent for lead to 5.80 percent for cadmium. Accuracy was assessed against certified reference material NRCC TM-24.3 (fortified lake water) and yielded Z-scores below 2 for all three metals (Pb 1.47, Cd 0.68, Mn 1.47), which the authors take as evidence that the method is valid given the absence of a milk-matrix or soy-beverage-matrix CRM.

Context against other published cow-milk lead values

The discussion places the Portuguese cow-milk lead mean (19.3 μg/L) into context against prior surveys: Solis et al. 2009 reported 65 μg/L in Mexico, Amer et al. 2021 reported 45.06 μg/L in Egypt, Derakhshesh and Rahimi 2012 reported 13.45 μg/L in Iran, Oliveira et al. 2017 reported 2.12 to 37.36 μg/L in Brazil, Zhou et al. 2019 reported 0.46 to 2.96 μg/L in China, and Freschi et al. 2011 reported values below an unspecified limit of detection (1.49 μg/L) in Brazil. The Portuguese mean sits in the middle of this distribution.

Methods (brief)

Sample preparation digested 0.5 mL of beverage (CM or SB) in a polytetrafluoroethylene (PTFE-Teflon) digestion vessel with 6 mL of 2N nitric acid and 1 mL of 30 percent hydrogen peroxide; the mixture stood at room temperature for three hours with the vessels closed, then proceeded through a five-step microwave digestion program in a Berghof Speedwave Two apparatus reaching 170-200 °C at 90 percent power. Vessels cooled overnight, were transferred to a 15 mL volumetric flask with three ultrapure-water rinses, and were brought to volume before analysis. Quantification used a PerkinElmer AAnalyst 700 graphite furnace atomic absorption spectrometer with deuterium background correction and single-element hollow-cathode lamps. Magnesium nitrate at 2 g/L was used as a chemical modifier for manganese; no modifier was added for cadmium or lead because the authors found it had no effective benefit. Selected wavelengths were 228.8 nm (Cd), 279.5 nm (Mn), and 283.3 nm (Pb). Statistical comparison between CM and SB groups used the Mann-Whitney U test after the Kolmogorov-Smirnov test rejected normality, with significance set at p < 0.05.

Implications

Certification: This is a Portuguese-retail-market liquid-beverage survey with N = 14 per class, peer-reviewed in Toxics and CRM-validated against TM-24.3. It is direct evidence for cow-milk-as-purchased and soy-beverage-as-purchased lead, cadmium, and manganese occurrence on a single EU national market in 2019. It is not direct evidence for cow-milk-derived infant formula, soy-protein infant formula, or any powdered or reconstituted product; the authors note that European maximum levels for these metals apply to infant formula and similar products manufactured from these milks but did not themselves test such reconstituted products. Pooling these values into clean-or-dirty subcategory percentiles requires recognizing that the basis is liquid as-purchased in μg/L, not as-consumed per kilogram, and that the publication date is recent enough that a 2030-era pool covering 2019-2030 retail sampling can include these rows without geographic-bias concerns specific to outdated literature.

App: Route to cow milk, soy beverage, milk and dairy, plant-milks-soy-based, lead, cadmium, and manganese. The CM mean lead value sitting one μg/L under the EU 20 μg/kg ceiling is a useful retail-market reference for app-level “how close to the regulatory limit” framing for European-market cow milk. Do not route as cow-milk-derived infant formula or soy-protein infant formula evidence; the authors did not test those products.

Courses: Useful for teaching the cow-milk versus plant-based beverage contamination comparison, the manganese-in-soy issue (soy beverages here carried Mn at roughly twenty times cow-milk levels, in line with prior reports that soy as a botanical accumulates Mn from soil and water), the distinction between metals with set EU ceilings in milk-based products (Pb, Cd) and metals without (Mn), and the discipline of comparing a retail-market mean against a regulatory ceiling using an explicit density conversion rather than treating μg/L and μg/kg as interchangeable.

Wiki pages this source may touch

• milk-and-dairy
• soy-milk
• milk-and-dairy
• plant-milks-soy-based
• lead
• cadmium
• manganese
• eu-2023-915-cadmium
• eu-2023-915-lead-infant-young-child-foods

Verification notes

Per-brand concentration values are reported in the source’s Tables 10 and 11 under anonymized codes CM1-CM14 and SB1-SB14. This page reports global means, standard deviations, and within-class ranges only, consistent with the Part 12 brand-firewall rule. The single brand-level observation preserved on this page - that one cow-milk brand reached 6.8 μg/L of cadmium, above the 5 μg/kg EU infant-formula ceiling - is noted without the CM code attribution because the source provides no brand-identifying mapping for those anonymized codes and the regulatory-comparison point is the load-bearing observation, not the brand attribution.

The basis throughout this page is μg/L of liquid beverage as-purchased and as-analyzed, matching the source’s reporting basis. The source does not convert its values to μg/kg in the results tables; the single μg/L-to-μg/kg conversion the authors perform is the comparison of their CM lead mean (19.3 μg/L) to the EU 20 μg/kg ceiling using a cited cow-milk density of 1.030 kg/L. Downstream pooling that requires per-kilogram bases should apply that density conversion explicitly.

The source provides no speciation for any of the three metals; all values are total Pb, total Cd, and total Mn by GFAAS on acid-digested samples. The species labels in this page’s metals frontmatter therefore stay as the totals (Pb, Cd, Mn).

The source’s Mann-Whitney p-value for cadmium is reported in the abstract as p < 0.05, in Figure 2 as p < 0.05 with an asterisk, and in the Results text consistently as p < 0.05. The source’s Mann-Whitney p-value for manganese is reported identically (p < 0.05 in abstract, Figure 3, and Results text). The source’s Mann-Whitney p-value for lead is reported in Figure 1 and in the Results text as p > 0.05 (not significant). This page preserves all three p-value reportings verbatim.

The paper’s Conclusions characterize the SB mean Mn (117.4 μg/L) as “about 20 times higher than the Mn concentrations in the CM brands (5.9 μg/L)”; the actual ratio is 117.4 / 5.93 = 19.8, which rounds to 20, so this page reports the paper’s “twenty-fold” framing in the lead paragraph and the exact arithmetic ratio in the Implications section.

The source reports its cow-milk Cd and Mn global means at two different precisions: the abstract and Discussion narrative give CM Cd as 2.15 ± 1.84 μg/L and CM Mn as 5.93 ± 1.21 μg/L, while Table 10’s “Global mean / Global SD” rows round these to 2.2 ± 1.8 and 5.9 ± 1.2 respectively. This page uses the higher-precision abstract values throughout because they are what downstream readers will most often see cited. Two minor source-internal precision mismatches are preserved without reconciliation: the abstract reports SB Mn SD as 30.3 while Table 11 reports 30.4, and the abstract reports SB Pb SD as 9.6 while Table 11 reports 9.7; in both cases this page follows Table 11 because the table is the primary data record.

Audit subagent (2026-06-08) flagged products/milk-and-dairy as absent from docs/gpt-collaboration/taxonomy-snapshot.md; verified independently — finding was a false positive because the taxonomy snapshot was generated 2026-05-18 and is stale relative to the live wiki. wiki/products/milk-and-dairy.md exists as a provisional scaffold created 2026-06-01 (provisional_scaffold: true, hmtc_category: pending_taxonomy_review), and the routing audit successfully resolves this source to that target with 0 unresolved and 0 malformed. No frontmatter change applied. Audit subagent (2026-06-08) also flagged the original 2.2 ± 1.8 and 5.9 ± 1.2 cow-milk Cd and Mn values as rounded relative to the abstract’s 2.15 ± 1.84 and 5.93 ± 1.21; verified against PDF page 1 abstract — finding was correct; values updated to the abstract precision.

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.