Song et al. 2017 — Dietary cadmium exposure assessment for the Chinese population (PLOS ONE)

A national semi-probabilistic dietary cadmium exposure assessment for China combining 228,687 food samples measured by graphite furnace atomic absorption spectrometry (GFAAS) between 2011 and 2015 with food consumption data from the 2002 China National Nutrition and Health Survey (CNNHS; n=67,608). Exposures are estimated for the general population and for age-sex sub-groups (children 4–11 years, young people 12–17 years male and female, adults ≥18 years male and female), and compared north of the Huai-River/Qinling-Mountains line versus south of it. The reference health-based guidance value is the JECFA Provisional Tolerable Monthly Intake (PTMI) of 25 µg/kg body weight per month.

Mean cadmium exposure of the general Chinese population was 15.3 µg/kg BW/month (61.2% of PTMI). High-consumer (P95) exposure for the general population was 33.0 µg/kg BW/month (1.3× PTMI). For children 4–11 years the P95 reached 48.2 µg/kg BW/month (1.93× PTMI). Southern-population mean exposure was 1.7–1.8× higher than the northern-population mean across all age-sex groups. Rice was the single largest contributor (55.8% of intake in the general population, 58.6% in the high-exposure sub-population, 65.1% in the southern population, 37.8% in the northern population, and 90% of intake reported by Vietnam in a comparator study cited by the authors).

Key numbers

Cadmium concentrations by food group (Table 1, n samples per group; mean ± SD in mg/kg fresh weight; mean daily consumption g/day where reported). Reported with middle-bound treatment of left-censored results (½ LOD substitution) except for milk and fresh eggs (>60% <LOD), where upper-bound (LOD substitution) was used.

Purple seaweed had the single highest mean concentration (1.034 mg/kg), followed by sea crab (0.544), molluscs (0.377), dry mushroom (0.366), dry kelp (0.252), kidney (0.222), kelp (0.168), and freshwater crab (0.101). Among cereals, rice (0.062 mg/kg) was the highest. Leafy vegetables (0.021 mg/kg) were the highest among vegetables excluding edible fungi. Overall mean across all foods: 0.049 mg/kg.

Dietary cadmium exposure by age-sex sub-group (Table 2; µg/kg BW/month).

All mean values are at or below the JECFA PTMI of 25 µg/kg BW/month; all P95 values exceed it. The P95 ranges from 27.4 (adult male) to 48.2 (children 4–11 y) µg/kg BW/month, equating to 1.1–1.93× PTMI. Children 4–11 y had significantly higher mean exposure than young people and adults (p<0.01) while still below the PTMI on average.

North versus south comparison (Table 3; µg/kg BW/month, mean ± SD). All south-versus-north factors are statistically significant at p<0.01.

Food-group contributions to dietary cadmium intake (Fig 2; %).

Rice was the dominant contributor in both the general and high-exposure sub-populations. Wheat flour was the second-largest contributor in the general population (11.8%) but dropped to 2.0% in the P95 sub-population, while shellfish rose from 4.8% to 13.2% and seaweed from 2.4% to 6.4%. Foods with high cadmium concentrations (purple seaweed, sea crab, molluscs, dry mushroom, kelp, kidney) contributed modestly to total exposure because consumption volumes are small.

Authors’ general-population intake comparison (µg/d basis where stated in the discussion). China this study: 30.6 µg/d (back-calculated by the authors as 15.3 µg/kg BW/month for a 60 kg adult). Previous Chinese total diet studies: 13.8 µg/d (1990), 19.4 µg/d (1992), 22.2 µg/d (2000). Bangladesh: 34.6 µg/d. Bangkok adult males: 50–56 µg/d. Japan: 22.8 µg/d. Lebanon: 15.8 µg/d. South Korea: 14.5 µg/d. EU population (EFSA): 7.6 (6.4–9.6) µg/kg BW/month. Australia: 2.2–6.9 µg/kg BW/month. United States: 4.6 µg/kg BW/month. Chile: 9 µg/kg BW/month. (The Chinese general-population mean of 15.3 versus the EU mean of 7.6 µg/kg BW/month is a ~2× ratio by wiki arithmetic on the source-stated numbers; the source does not assert this ratio explicitly.)

Methods

Food sampling. 228,687 individual food samples collected from supermarkets, local markets, and field harvest sites across 31 provinces, autonomous regions, and municipalities of China between 2011 and 2015. Samples aggregated into 32 food groups including rice, wheat flour, nuts, fish, molluscs, shrimp, crab, vegetables (root/stalk, leafy, cucurbits/fruit, legume), mushrooms (fresh, dry), fungi, purple seaweed, kelp (fresh, dry), meat, liver, kidney, egg (fresh, preserved), milk, milk powder, fruits, juice, and tea.

Analytical method. Cadmium in foods was analysed by graphite furnace atomic absorption spectrometry (GFAAS) per the protocol in the China National Health Monitoring Handbook of Food Safety and China National Food Safety Standard GB/T 5009.15–2003. Samples analysed in local laboratories across the 31 PLADs, all using the same procedure with prior training.

Limit of detection (LOD). LOD ranged from 0.00001 mg/kg to 0.1 mg/kg across food groups. Overall 35.7% of samples were below LOD. Censored-value treatment per GEMS/Food: samples below LOD assigned LOD or ½ LOD depending on the recommended approach for the food group. For milk and fresh eggs (>60% <LOD), upper-bound (UB; LOD substitution) was used; otherwise middle-bound (MB; ½ LOD substitution) was applied.

Food consumption data. 2002 China National Nutrition and Health Survey (CNNHS), n=67,608 subjects, stratified multi-stage cluster sampling from 31 PLADs. Three consecutive 24-hour dietary recalls per subject, with individual body weight recorded. Within-category daily consumption summed per investigated person and matched to the cadmium occurrence data of each food category.

Exposure calculation. Daily intake = Σⱼ (Fⱼ × Cⱼ) ÷ Wᵢ, multiplied by 30 to obtain monthly exposure: Expᵢ = Σⱼ (Fⱼ × Cⱼ ÷ Wᵢ) × 30, where Fⱼ is consumer i’s consumption of food category j (g/day), Cⱼ is the cadmium content of food category j (mg/kg), Wᵢ is body weight (kg), and n is the number of food categories. Mean and 95th-percentile exposure derived for the general population and each age-sex group (children 4–11 y, young people 12–17 y, adults ≥18 y).

Statistical analysis. SPSS 19.0. Means ± SD reported. Between-group comparisons by t-test or one-way ANOVA with Bonferroni post-hoc when equal variances assumed and Dunnett’s T3 post-hoc when not. p<0.01 deemed significant.

Speciation

No speciation. Cadmium is reported as total Cd. Inorganic cadmium is the toxicologically relevant species in food matrices, and dietary exposure assessment routinely treats food Cd as total without sub-speciation; the source does not address oxidation state or organic-Cd species.

Quality control

Laboratory standardisation: all 31 PLAD laboratories used the same analytical protocol per GB/T 5009.15–2003, with training provided before analysis. The source does not report per-laboratory recovery against certified reference materials, inter-laboratory comparison, calibration curve quality, or replicate RSD. Quality control is asserted via shared protocol rather than reported per-sample.

Uncertainties acknowledged by the authors

• Beans (legumes other than vegetable legume) were not included in the 32-category structure; absent foods may underestimate true exposure.
• Food consumption data from CNNHS 2002 are more than 10 years older than the 2011–2015 concentration data; dietary pattern changes from China’s rapid economic development since 2002 are not captured and may further underestimate exposure for high-consumption sub-groups.
• The high <LOD rate for several categories means choice of censoring strategy (0, LOD, ½ LOD) materially affects the estimate; the authors used the LOD-substitution upper bound for milk/eggs and the ½-LOD middle bound elsewhere, a conservative-but-not-maximally-conservative convention.
• The assessment is for food-borne exposure only; inhalation and smoking pathways, which can be material for cadmium body burden, are excluded.

Implications

This page records what Song et al. 2017 reports about cadmium exposure across 32 Chinese food categories. It does not draw category-level synthesis conclusions about ingredient contamination percentiles — that is Part 9 synthesis work that should be done together with Sun et al. 2022 (the China PTMI derivation) and the prior Chinese total diet studies (1990 / 1992 / 2000) the authors cite. The single load-bearing observation is that rice contributes 56–65% of cadmium dietary intake to Chinese adults depending on region, and that the southern subpopulation’s mean exposure is 1.7–1.8× higher than the northern subpopulation’s across all age-sex strata at p<0.01.

The 32-category coverage means this paper is also a routing fan-out source for ingredient pages well beyond the originating folder’s condiment focus: rice, wheat, leafy vegetables, root vegetables, vegetables (generic), mushrooms, seaweed, fish, freshwater fish, molluscs, shellfish, seafood, meat, poultry, organ meats, eggs, milk and dairy, fruits, tea, and legumes are all touched by the source per the Table 1 measurements.

Verification notes

• PDF source provenance: file resides in raw/manual-fetch/Kimi_Agent_Download Corruption Issue/condiments2_papers/03_Condiments/12_China_dietary_cadmium_exposure_2017.pdf; the parent folder name identifies this batch as a Kimi-agent download corruption issue. The PDF is the complete 12-page PLOS ONE published article (DOI 10.1371/journal.pone.0177978; PLOS Article ID e0177978; published 18 May 2017; received 23 Dec 2016; accepted 5 May 2017) with all tables, figures, and references intact. No corruption observed in this particular file.
• Cite-key construction: song2017-china-dietary-cadmium-exposure — first-author surname (Song) + year + jurisdiction + analyte + scope. The handle KADC_12-china-dietary-cadmium-exposure-2017 is the raw folder filing handle, preserved in raw_handle.
• Title placement in folder is misleading: this paper is filed in 03_Condiments under the Kimi corruption-issue tree, but the paper itself is a national dietary exposure assessment covering 32 food categories (rice, wheat, leafy vegetables, etc.), not a condiment-focused study. The frontmatter reflects the paper’s actual scope, not the folder name; products: [] because no packaged product category is the focus of the paper. The condiment-folder placement is a quirk of the raw-folder organization and does not constrain routing.
• Authors as printed on page 1: Yan Song, Yibana Wang, Weifeng Mao, Haixia Sui, Ling Yong, Dajin Yang, Dingguo Jiang, Lei Zhang*, Yunyun Gong (CFSA / China National Center for Food Safety Risk Assessment, Beijing, China; Y. Gong at University of Leeds, UK). Lei Zhang is the corresponding author (zhanglei@cfsa.net.cn).
• License: CC-BY (Creative Commons Attribution License) per the published front-matter copyright statement on page 1. CC-BY-4.0 recorded conservatively (PLOS ONE uses CC-BY 4.0 by default; the paper is from 2017 so this is the applicable version).
• Brand firewall (Part 12): the source does not name any brand. No redaction required.
• Wiki/HMTc firewall (Part 2): no HMTc threshold proposals. The JECFA PTMI of 25 µg/kg BW/month is reported as the source uses it — as the toxicological reference value the authors compare exposures to — not as a wiki-endorsed threshold. The 1.7–1.8× south-north exposure factor and the 1.93× children-P95-versus-PTMI exceedance are reported as the source states them.
• Taxonomy choices: ingredients use only slugs that exist in the current taxonomy. Items the source measures but for which no current slug exists (purple seaweed specifically, crab freshwater/sea separately, shrimp, kidney, liver, fungi, dry vs fresh mushroom, milk powder vs fluid milk, juice) are folded into the closest existing umbrella slug (seaweed, shellfish, organ-meats, mushrooms, milk-and-dairy, fruits) rather than overspecifying. The Key numbers table preserves the source’s full granularity for downstream synthesis.
• Sample size: sample_n: 228687 is the food-sample count from Table 1 (“Total: 228,687”). The 67,608 figure is the consumption-survey subject count from CNNHS 2002, recorded in sample_population.
• Sampling year range: food sampling 2011–2015 (Materials and Methods §1); consumption data 2002 (CNNHS, §2). Both recorded in sampling_year_range.
• Sampling locations: recorded as [CN-mainland] umbrella; the 31 PLADs are described in sample_population.
• Censoring methodology: middle-bound (½ LOD) for most groups, upper-bound (LOD) for milk and fresh eggs (>60% <LOD). This is faithfully reproduced in the Key numbers narrative and Methods section.
• Evidence tier: A (peer-reviewed national exposure assessment from CFSA, published in PLOS ONE 2017, n=228,687 food samples × n=67,608 dietary survey subjects, GFAAS per national standard GB/T 5009.15–2003). The 24-hour-recall consumption data from 2002 is the most-cited limitation the authors raise; it is a known constraint of the assessment, not a methodological flaw warranting a tier downgrade.
• Audit subagent (2026-06-02, fresh general-purpose context) flagged “The Chinese general-population mean exceeds the EU mean by approximately 2×” as wiki arithmetic (15.3/7.6 = 2.01) on source-stated numbers, not as a claim the source itself asserts. Verified correct on re-read of the Discussion (p.8): the authors give the Chinese 15.3 and EU 7.6 µg/kg BW/month separately without asserting a ratio. Sentence reworded to flag the ratio as wiki-derived arithmetic on source numbers per Cochrane-quality fidelity discipline.

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.