Integral assessment of the environmental safety of plant-based milk alternatives based on heavy metal analysis

Overview

Begday and colleagues at North-Caucasus Federal University in Stavropol’ measured lead, cadmium, zinc, and copper in eight plant-based milk beverages distributed in or producible from Russian-market raw materials: four commercial ready-to-drink beverages (one almond, one rice, one coconut, one oat) purchased at a Russian retail chain, and four laboratory-prepared beverages produced from mature buckwheat groats, green buckwheat groats, oat flakes, and long-grain rice via a standardized soak-and-filter protocol. All eight samples fell well below Russian SanPiN 42-123-4089-86 maximum permissible concentrations (MPCs) for the four metals, with the highest reported value (lead at 0.053 mg/kg in mature-buckwheat-groat milk) reaching only 10.6 percent of the 0.5 mg/kg lead MPC. Cadmium was non-detected in all four laboratory-prepared samples but was quantified in all four commercial beverages (0.0068 to 0.0108 mg/kg). The authors computed hazard quotients (HQ) for adults (70 kg, 120 kg beverage/year) and children (31 kg, 137.4 kg beverage/year) against EPA 2021 reference doses; per-metal HQs ranged from 0.000004 to 0.00874 and total HQ sums per beverage ranged from 0.00061 to 0.0062, all well below the threshold of 1 that would indicate non-carcinogenic concern.

Key numbers

All concentrations below are in milligrams per kilogram (mg/kg) of the beverage as analyzed and as consumed; the authors report a single basis throughout and do not distinguish wet versus dry weight. Values are arithmetic means of two parallel determinations rounded to four decimal places, per the source’s Materials and Methods. “n.d.” denotes non-detect (below the source’s reported detection limit calculated as 3 tin (Sn) / (p√n)).

Concentration by sample (Table 1, source p. 63)

The authors report that across all eight samples, concentrations remained within 4.8 to 36.7 percent of the corresponding MPC depending on metal and sample. Specifically: maximum lead (Pb) of 0.0530 mg/kg (mature buckwheat-groat milk) is 10.6 percent of the 0.5 mg/kg MPC; maximum cadmium (Cd) of 0.0108 mg/kg (commercial coconut beverage) is 36.0 percent of the 0.03 mg/kg MPC; maximum Zn of 0.4409 mg/kg (green buckwheat-groat milk) is 14.7 percent of the 3 mg/kg MPC; maximum Cu of 0.0233 mg/kg (oat-flake milk) is 0.23 percent of the 10 mg/kg MPC. The authors highlight three cross-sample trends: zinc consistently highest in green buckwheat (0.4409 mg/kg); copper at moderate levels (0.0097 to 0.0233 mg/kg) across lab-prepared samples; cadmium below detection in every laboratory-prepared sample but detectable in every commercial sample.

Hazard quotients for commercial samples (Table 2, source p. 64)

Hazard quotients for laboratory-prepared samples (Table 3, source p. 65)

Cadmium HQ rows are absent from Table 3 for all four laboratory-prepared samples because Cd was non-detected. The authors note the maximum HQ across all eight samples is 0.00874 (cadmium, child consumer of commercial coconut beverage); every per-metal HQ and every total HQ sum sits at least two orders of magnitude below the HQ = 1 threshold that would indicate non-carcinogenic concern under the Russian R 2.1.10.1920-04 health risk assessment guidance.

Exposure assumptions used in the HQ calculation

The hazard quotient is HQi = Di ÷ RfD, where Di = (Ci × M × Tp) / (P × T) and Ci is the measured concentration, M is the annual beverage consumption, Tp is exposure duration in years, P is body weight, and T is the averaging time. The authors use: adult P = 70 kg, M = 120 kg/year; child P = 31 kg, M = 137.4 kg/year; Tp = 1 year for both; T = 30 years for both. EPA 2021 reference doses applied: Pb 0.0035 mg/kg/day, Cd 0.0005 mg/kg/day, Cu 0.04 mg/kg/day, Zn 0.3 mg/kg/day. MPCs taken from SanPiN 42-123-4089-86.

Methods (brief)

Sample preparation followed GOST 26929-94 microwave-assisted acid mineralization. Concentrations were quantified by atomic absorption spectrometry (AAS) on a Thermo Scientific iCE 3300 instrument. Results were expressed in mg/kg and compared against the Russian SanPiN 42-123-4089-86 maximum permissible concentrations for toxic elements in foodstuffs. Each sample was measured in two parallel determinations; results were averaged and rounded to two decimal places (Table 1 values appear to four decimals after author rounding from intermediate calculations). The authors did not perform speciation; all reported values are total lead, total cadmium, total zinc, and total copper. The reported detection-limit treatment follows m_max < (3 Sn × Y × K) / (p × √n), where Sn is the residual standard deviation, Y is the original solution volume, K is the dilution coefficient, p is the sample mass, and n is the number of parallel absorbance measurements. The four laboratory-prepared beverages were produced per Russian patent No. 2756071: 100 g of cereal raw material soaked in 500 mL water for 8 hours at 20 ± 2 °C, then ground and filtered through double cheesecloth; the resulting beverage was analyzed the same day. Statistical comparison between commercial and laboratory-prepared samples is descriptive only; no formal hypothesis test is reported.

Implications

Certification: This is a Russian-market plant-beverage survey with n = 1 per botanical type for commercial samples (almond, rice, coconut, oat) and n = 1 per cereal raw material for laboratory-prepared samples (mature buckwheat, green buckwheat, oat flakes, long-grain rice), peer-reviewed in Izvestiya KGTU. It is direct evidence for plant-milk beverages on the Russian retail market in 2025-2026 against Russian SanPiN 42-123-4089-86 MPCs; it is not direct evidence against EU 2023/915, Codex, or FDA Closer to Zero thresholds because the authors do not compare to those frameworks. The very small per-product-type sample size (n = 1 per commercial brand category) limits the value of any single reported concentration as a population estimator. The strongest cross-cutting finding is the cadmium contrast between commercial and laboratory-prepared samples: detectable Cd (0.0068 to 0.0108 mg/kg) in all four commercial beverages, non-detect in all four laboratory-prepared beverages from the same nominal botanical category (rice in particular). The authors attribute this to commercial supply-chain factors and raw material origin but do not test that hypothesis directly.

App: Route to plant-milk, rice-milk, almond, coconut, oat, rice, and non-rice-grains (buckwheat) ingredient pages, and to plant-milks-non-soy-non-rice, plant-milks-rice-based, and rice-beverages-rice-milk product pages, and to lead and cadmium metal pages. The single-sample-per-botanical structure means these values function as plausibility anchors rather than centrals for population pooling; downstream pooling should treat each row as a single observation, not as a percentile estimate.

Courses: Useful for illustrating (a) the structural difference between commercial supply-chain plant beverages and home/laboratory-prepared analogues from the same raw material category, (b) how Russian-jurisdiction MPCs compare in magnitude to other regional frameworks for plant beverages, (c) the hazard-quotient framework for non-carcinogenic risk assessment of low-level food contamination, and (d) the role of cereal-raw-material agronomy (Pb in mature buckwheat groats accumulating from soil) in shaping the contamination profile of the derived beverage.

Related evidence

• Plant milk base
• Rice Milk
• Almond
• Coconut
• Oat
• Rice
• Non Rice Grains
• Plant Milks (Almond, Oat, Coconut, Other Non-Soy/Non-Rice)
• Plant Milks, Rice-Based — RETIRED
• Rice beverages (rice milk)
• Lead
• Cadmium
• Zinc
• Copper

Update history

The five most recent substantive edits to this page, classified major (evidence or structure moved), correction (a published value or statement was wrong and has been fixed), or minor (narrative rewritten without changing the underlying evidence). Each description is derived from what the edit did to this page; the linked commit is the authoritative record, routine regeneration passes are excluded, and the full version history lives in git.