Lee et al. 2023 — Seven potentially toxic metals in fresh fruits consumed in South Korea

This is the first health risk assessment of dietary exposure to seven metals (As, Ba, Cd, Cr, Ni, Pb, and Sb) through fresh fruit consumption in South Korea, using ICP-MS to analyze 207 samples of 14 commonly consumed fruit species collected from supermarkets across six regions in 2019. Mean concentrations (mg/kg fresh weight) across all samples were: As <0.0021, Ba 0.3675, Cd <0.0022, Cr 0.0307, Ni 0.0815, Pb 0.0236, and Sb <0.0021. Non-carcinogenic hazard indices (HI = 0.0275) and carcinogenic risks for As and Pb (4.62×10⁻⁷ and 5.05×10⁻⁷ respectively) were all below threshold values for the Korean population as a whole, though Pb hazard quotient and cancer risk in apples were the highest among children aged 1–2 years.

Key numbers

Seven metals quantified in 207 fresh fruit samples from 14 species across six South Korean regions.

Overall means (mg/kg fresh weight, n=207): As <0.0021, Ba 0.3675, Cd <0.0022, Cr 0.0307, Ni 0.0815, Pb 0.0236, Sb <0.0021.

Detection rates across the 207 samples: Ba 207/207 (100%), Cr 207/207 (100%), Ni 207/207 (100%), Pb 193/207 (93.2%), Sb 42/207 (20.3%), As 36/207 (17.4%), Cd 23/207 (11.1%).

Ba was the most abundant metal in every fruit group. Fruit-group Ba means (mg/kg fw): citrus fruits 0.5357 (n=30), tropical fruits 0.5005 (n=45), pome fruits 0.3315 (n=45), berries 0.2945 (n=48), stone fruits 0.2161 (n=39). Among species, Ba mean was highest in kiwi 0.9745 ± 0.4812 (range 0.2285–1.8392, n=15) and lowest in grape 0.0710 ± 0.0359 (range 0.0309–0.1805, n=15). Individual species means for the other higher-Ba fruits (mg/kg fw, n=15 each): orange 0.5475 ± 0.2882 (0.1142–1.1670), mandarin 0.5239 ± 0.2782 (0.1136–0.9161), persimmon 0.4444 ± 0.1732 (0.2319–0.8842), strawberry 0.4204 ± 0.2275 (0.1086–1.0728), pear 0.3404 ± 0.2174 (0.1040–0.9955), apple 0.2076 ± 0.1296 (0.0034–0.4914).

Ni mean (mg/kg fw) across all samples 0.0815; highest in mandarin 0.1947 and lowest in grape 0.0287. Cr mean 0.0307; highest in apple 0.0567 and lowest in persimmon 0.0155.

Pb mean (mg/kg fw) across all samples 0.0236; significantly higher in apples (0.0464) and lower in strawberries (0.0060). The Cd mean was <0.0022 in all species except pears (0.0026).

As, Cd, and Sb were below the LOD in most samples and most species. Maximum single-sample concentrations (mg/kg fw): As in cherry 0.040, Ba in kiwi 1.8392, Cd in strawberry 0.0098, Cr in apple 0.1262, Ni in mandarin 0.4779, Pb in pineapple 0.0968, Sb in blueberry 0.0089. All measurements were below the Codex Alimentarius, Korean Ministry of Food and Drug Safety, EU Commission Regulation (EC) No. 1881/2006, and Japanese MHLW maximum levels reviewed by the authors.

Hazard quotient ranking (all-ages Korean population, per Table 3 and page 14 body text): Pb (0.0148) > Ni (0.0081) > Ba (0.0031) > As (0.0010) > Cd (0.0004) > Cr (0.0001) > Sb (NA). Sum hazard index HI = 0.0275 (below 1). Sb HQ is NA because all Sb concentrations in all fruit species were below the LOD. Percentage contributions to HI: Pb 54.0% > Ni 29.4% > Ba 11.3% > As 3.7% > Cd 1.4% > Cr 0.2%.

The paper’s abstract reports a different HQ ranking (Pb 0.0149 > As 0.0086 > Ni 0.0081 > Sb 0.0080 > Ba 0.0031 > Cd 0.0027 > Cr 0.0001) whose values do not sum to the stated HI 0.0275; the body-and-Table-3 values (which do sum to 0.0275) are reported here as the rigorous figures.

Estimated daily intakes (EDI, µg/kg b.w./day, all-ages Korean population): Ba 0.6213 > Ni 0.1612 > Cr 0.0759 > Pb 0.0594 > Cd 0.0004 > As 0.0003 > Sb NA. Apple intake (44.59 g/day, the most-consumed species) contributed the most to EDI of Pb (0.0338, 56.9% of total Pb EDI) and Ba (0.1511, the largest single-species Ba EDI). Mandarin contributed the most to Ni EDI (0.0474, 29.4% of total Ni EDI).

Carcinogenic risks (all-ages): As CR = 4.62×10⁻⁷ (highest single-species contribution from strawberry 2.28×10⁻⁷, 49.4%; then cherry 1.20×10⁻⁷ and blueberry 1.14×10⁻⁷); Pb CR = 5.05×10⁻⁷ (highest single-species contribution from apple 2.87×10⁻⁷, 56.9%). Both totals are below the 1×10⁻⁴ threshold.

Age-group fruit consumption (g fw/day): 1–2 y 115.94, 3–6 y 129.79, 7–12 y 117.50, 13–19 y 73.22, 20–64 y 126.32, ≥65 y 148.75, all ages 125.44. Females 138.19, males 112.74. Apple was the most-consumed species at 44.59 g/day across all ages.

Children aged 1–2 years carried the highest exposure of any age group because their assumed body weight (11.94 kg) is the lowest. For this age group: HI across all 14 fruits = 0.1283 (still below 1); HQ Pb across all fruits = 0.0611, of which apple intake alone contributed 0.0307 (50.3%); HQ Ba = 0.0150; HQ Ni = 0.0403; HQ As = 0.0094; HQ Cd = 0.0023; HQ Cr ≈ low. Pb CR across all fruits = 2.08×10⁻⁶, of which apple intake alone contributed 1.05×10⁻⁶ (50.3%). Strawberry-attributable As CR for 1–2 y was 2.57×10⁻⁶, 60.7% of the all-fruits As CR (4.23×10⁻⁶) for this age group. The paper concludes that targeted Pb monitoring of apples for the 1–2 y group, and washing/peeling, are warranted; no risk threshold (HQ > 1 or CR > 10⁻⁴) was crossed.

Reference dose (RfD, mg/kg b.w./day) values used: As (inorganic) 0.0003, Ba 0.2, Cd 0.001, Cr (III) 1.5, Ni 0.02, Sb 0.0004, Pb 0.004. Cancer slope factor (CSF, (mg/kg b.w./day)⁻¹) for inorganic As 1.5; for Pb 0.0085.

LOD: 2.098 (Sb)–3.009 (Ba) µg/kg; LOQ: 6.923 (Sb)–9.930 (Ba) µg/kg for the seven metals.

Spearman correlation analysis: only Ba showed a significant correlation with another metal — a negative correlation with Pb (ρ = −0.5385) across all fruit samples at the 95% confidence level. All other metal pairs showed low-to-moderate positive correlations (except As–Ba, As–Ni, As–Pb, Ba–Cd, Ba–Cr, Ba–Pb, Ba–Sb, Cd–Pb, Ni–Pb, Ni–Sb) that did not reach significance.

PCA: three components with eigenvalue > 1 explained 68.14% of variance (PC1 30.08%, PC2 20.71%, PC3 17.35%). HCA clustered Ba and Ni as distinct from the As/Cd/Cr/Pb/Sb group; among species, kiwi was distinct from all other fruits at linkage distance 8.

Methods (brief)

207 fresh fruit samples (14 species, n=15 per species except blueberry n=16, strawberry n=17, cherry n=9) were collected from local supermarkets in six South Korean regions (Seoul, Gyeonggi-do, Chungcheong-do, Jeolla-do, Kyungsang-do, Gangwon-do) between April and October 2019, sampled to mirror the 2018 population census distribution. Species selection followed the Korean National Health and Nutrition Examination Survey (KNHANES) 2016–2020 ranking of most-consumed fruits.

Edible portions were homogenized (Facilcic 7000, Tefal, Veenendaal, Netherlands) after washing, peeling, and removing inedible parts, then stored at −20 °C. Approximately 0.25 g of homogenate was digested in 0.5 mL 70% HNO₃, 2.5 mL 30% H₂O₂, and 5 mL ultrapure H₂O using a Milestone UltraWAVE single-reaction-chamber microwave digester (15 PTFE vessels; three-step program — (1) ramp to 210 °C over 10 min, (2) warm-up at 250 °C for 20 min, (3) maintain at 250 °C for 20 min; oven at 1500 W full power), then diluted to 10 mL. Each sample was processed in triplicate alongside triplicate analytical blanks.

Determination by ICP-MS (Agilent Technologies JP-7900 model; RF power 1550 W; sampling depth 8.0 mm; Ar carrier 0.82, makeup 0.33, nebulizer 0.85 L/min; He kinetic-energy discrimination 5.0 mL/min). Isotopes monitored: ⁷⁵As, ¹³⁷Ba, ¹¹¹Cd, ⁵²Cr, ⁶⁰Ni, ²⁰⁸Pb, ¹²¹Sb. NIST 1573a tomato leaves used as the certified reference material; recoveries were 99.8% (Pb) to 114.1% (Ba), precision < 3.2% (Ba). Ba and Pb (not certified in NIST 1573a) were additionally validated by triplicate spike recovery on grape at three concentrations (Ba: 0.4, 2.0, 4.0 mg/kg; Pb: 0.1, 0.3, 0.6 mg/kg). LOD and LOQ defined as 3× and 10× the SD of the method blank signal (n=7).

Concentrations reported on fresh-weight basis. Non-detect samples reported at or below LOD/LOQ. Cr assumed trivalent for health-risk calculations (RfD for Cr-III); As assumed inorganic for cancer-risk calculations (CSF for iAs). Fruit consumption inputs derived from 24-h recall records of ~32,407 KNHANES participants (2016–2020), stratified by age (1–2, 3–6, 7–12, 13–19, 20–64, ≥65 y) and sex.

Statistics: nonparametric Kruskal–Wallis test with Dunn’s post-hoc (p<0.05) for between-group differences; Spearman rank correlation; PCA and HCA (Ward’s method, Euclidean distance) on mean species concentrations. Software: Microsoft Excel 2022, GraphPad Prism 9.4.1, SPSS 28, RStudio 4.2.2 (with survey, reshape2, dplyr).

Limitation: As measured as total arsenic; paper assumes 100% inorganic for cancer-risk calculation, which overstates iAs exposure if organic As fractions are significant in fruit matrices. Frontmatter therefore lists tAs rather than iAs.

Implications

Certification: Pb levels in apples, though low on average, are the dominant risk signal for young children (1–2 years) in this dataset. Supports monitoring of Pb in apple-based products for infant and toddler populations.

Courses: Demonstrates that fresh fruits as a category generally pose low heavy metal risk in a well-monitored market (South Korea), with Ba as the most prominent quantified metal and Pb as the primary regulatory concern.

App: Mean fresh weight values usable as baseline estimates for fresh fruit matrices in Korea; Ba values are well above detection for app purposes; As, Cd, Sb concentrations largely at or below LOD.

Wiki pages this source may touch

• apple — apple Pb mean 0.0464 mg/kg fw (highest of 14 species); Cr 0.0567 (highest); Ba 0.2076.
• pears — pear Cd mean 0.0026 mg/kg fw (only species with mean Cd > LOD).
• oranges / mandarins — citrus Ba mean 0.5357 (group rank 1); mandarin Ni mean 0.1947 (highest of 14 species).
• grapes — grape Ba mean 0.0710 (lowest of 14 species); grape Ni 0.0287 (lowest).
• strawberries — strawberry Pb mean 0.0060 (lowest); strawberry Cd single-sample max 0.0098.
• kiwi — kiwi Ba mean 0.9745 (highest of 14 species) and maximum single sample 1.8392 mg/kg fw.
• banana, pineapple — tropical-fruit Ba group mean 0.5005 (rank 2); pineapple max Pb 0.0968 mg/kg fw.
• peaches, cherries, plums, persimmon, blueberries — stone- and pome-fruit context; cherry single-sample As 0.040 mg/kg fw is the dataset maximum.
• lead — Pb dominant HQ and CR contributor; apples flagged for children 1–2 y.
• barium — most-abundant analyte in every fruit group; novel dataset for Ba in fruit (paper notes no prior fruit-Ba literature for comparison).
• nickel — second-highest detected metal across the dataset.
• fresh-fruit — direct evidence row for the locked HMTc fresh-fruit category, Korea jurisdiction.

Verification notes

2026-05-20 — merge-enhance pass against the source PDF (Lee et al. 2023, Applied Biological Chemistry 66:40, DOI 10.1186/s13765-023-00799-x). Corrections applied:

• Ba values for kiwi, apple, orange, and mandarin in the prior Key numbers paragraph were transposed between adjacent rows of Table 2. The prior page assigned the tropical-fruit GROUP mean (0.5005 ± 0.4822, n=45) to kiwi; the actual kiwi species mean is 0.9745 ± 0.4812 (n=15). The prior page assigned orange’s Ba value (0.5475 ± 0.2882, n=15) to apple; the actual apple Ba mean is 0.2076 ± 0.1296 (n=15). The prior page assigned mandarin’s Ba value (0.5239 ± 0.2782, n=15) to orange. All four values are now corrected against Table 2 (PDF page 7) and cross-checked against the discussion text on PDF page 8 (“the mean Ba concentrations were the highest in kiwi fruits (0.9745) and lowest in grapes (0.0710)”).
• The prior page stated “Pb was detected above LOD in some apple and pear samples.” This understated detection: Pb was detected in 193/207 samples (93.2%) per PDF page 9. Corrected to give the actual detection rate and to add the species means flagged in the discussion (apple 0.0464, strawberry 0.0060).
• Methods section expanded to record the Milestone UltraWAVE digester, the Agilent JP-7900 ICP-MS parameters and monitored isotopes, the spike-recovery validation for Ba and Pb (which are not certified in NIST 1573a), and the RfD/CSF inputs. Prior version was thin enough to be ambiguous about which arsenic species the cancer-risk calculation used.
• Detection rates added for each metal.
• Maximum single-sample values added for every analyte (per the dataset’s explicit per-species maxima on PDF page 8).
• Frontmatter ingredient slugs corrected: apples → apple and bananas → banana (singular form is the existing ingredient page); blueberries added (the paper analyzed blueberries explicitly, n=16). The remaining slugs (kiwi, cherries, mandarins, persimmon) are correctly named for the auto-stub pipeline (CLAUDE.md Part 10): if they do not yet exist as pages, the freq>=2 auto-stub or single-slug provisional-ingredient-scaffold tool will create them; this ingest does not create ingredient pages by hand.
• raw_path corrected — prior value was truncated (...arsenic, bariu.pdf) and did not match a real file on disk.
• sample_population annotated with the six region names.
• ## Wiki pages updated on ingest heading renamed to ## Wiki pages this source may touch per current page-template convention (the ingest pass does not edit those downstream pages directly; the routing audit and synthesis pass do).

Brand firewall: no brand names appear in source-page content. The Milestone UltraWAVE, Agilent JP-7900, NIST 1573a, GraphPad Prism, SPSS, and RStudio names in Methods are scientific-method vendor/material identifiers (CLAUDE.md Part 12, Exception 2 locked 2026-05-17).

Wiki/HMTc firewall: no synthesis claims, no HMTc threshold proposals, no consumer advisories. The Implications section reports what this single paper contributes to threshold work without proposing values.

2026-05-20 — Fresh-context audit subagent (per skill v2.0 Phase 2) returned REVISE with three ❌ and one ⚠️. Verified each against the source PDF and applied:

• ❌ HQ ranking incorrect (verified correct): the prior page had copied the abstract’s HQ list (Pb 0.0149 > As 0.0086 > Ni 0.0081 > Sb 0.0080 > Ba 0.0031 > Cd 0.0027 > Cr 0.0001), whose values do not sum to the stated HI of 0.0275 and which the body of the paper does not reproduce. Corrected to the body-and-Table-3 ranking (Pb 0.0148 > Ni 0.0081 > Ba 0.0031 > As 0.0010 > Cd 0.0004 > Cr 0.0001 > Sb NA) which does sum to HI 0.0275. The discrepancy between the abstract and body of the paper is noted in-line in the Key numbers section. Source: PDF page 14 + Table 3 (page 12).
• ❌ Apple Pb HQ for 1–2 y misattributed (verified correct): the prior page reported “apple-derived Pb produced the highest per-fruit hazard quotient in the dataset (≈0.06 from Fig 2c)“. The 0.06 (precisely 0.0611) is the all-fruits combined Pb HQ for the 1–2 y age group; the apple-specific Pb HQ for that age group is 0.0307 (50.3% of 0.0611). Corrected and expanded with the apple-specific CR (1.05×10⁻⁶) and the strawberry-As CR contribution for the same age group. Source: PDF page 15.
• ⚠️ Digestion heating program incomplete (verified correct): the prior page collapsed the three-step program into two steps, omitting the 250 °C/20 min warm-up phase. Method reproducibility now records all three steps. Source: PDF page 3.
• ❌ Pineapple flag (verified false positive): the audit subagent flagged pineapple as a missing ingredient slug not present in the taxonomy snapshot. Verified against the live wiki — wiki/ingredients/pineapple.md exists. The taxonomy snapshot (docs/gpt-collaboration/taxonomy-snapshot.md, last touched 2026-05-19) is stale relative to the live taxonomy. No change made to the wiki page on this finding; the slug as written is valid. (Stale-snapshot drift is a tooling issue, not a source-page issue.)

The audit subagent’s positive findings: numerical fidelity in the Ba transposition fixes was independently re-verified clean; species means for grape, persimmon, strawberry, pear, group means, single-sample maxima, RfD/CSF values, LOD/LOQ ranges, PCA/HCA percentages, Spearman ρ, age-group consumption values, sample sizes per species, methods detail (ICP-MS parameters, isotopes monitored, recoveries, spike validation, software) — all clean.

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.