CFS 2012 — First Hong Kong Total Diet Study: Inorganic Arsenic

The Centre for Food Safety (CFS) of Hong Kong’s Food and Environmental Hygiene Department conducted the First Hong Kong Total Diet Study (1st HKTDS) to estimate dietary exposure of the Hong Kong population to inorganic arsenic, specifically using direct iAs speciation rather than total-arsenic conversion factors. A total of 600 composite samples (from 150 food items, 3 purchases per item across 4 sampling occasions March 2010 to February 2011) were tested using hydride generation ICP-MS, with an LOD of 3 µg/kg in food and 1.5 µg/kg in water. Dietary exposure estimates were 0.22 µg/kg bw/day for average consumers and 0.38 µg/kg bw/day for high consumers (95th percentile), both below the JECFA BMDL0.5 of 3.0 µg/kg bw/day (range 2-7), with margins of exposure (MOEs) of 9-32 and 5-18 respectively. Rice dominates the iAs exposure picture: cooked white rice alone contributed 45.2% of total dietary exposure, with cereals and their products accounting for 53.5% overall.

Note on series: this report is No. 2 in the 1st HKTDS series (published February 2012). The companion report covering the full metallic contaminants panel (Al, Sb, Cd, Pb, MeHg, Ni, Sn, V) is Report No. 5 (January 2013), ingested as cfs2013-hktds-metallic-contaminants.

Key numbers

Food group mean iAs concentrations (µg/kg, composite samples, wet weight as consumed; ND = below LOD of 3 µg/kg, half-LOD substituted for non-detects per WHO GEMS/Food-EURO guidance):

Selected food-item highlights (from Appendix 1):

Condiments and sauces group (20 composite samples, 40% below LOD, group mean 8 µg/kg, range ND–65 µg/kg): Soya sauce was the highest-iAs condiment (mean 21 µg/kg, range 3–65 µg/kg); oyster sauce was lower (mean 12 µg/kg, range 6–17 µg/kg); tomato paste/ketchup was mean 7 µg/kg (range ND–14); cornstarch 2 µg/kg (range ND–4); table salt at half-LOD (ND). The 65 µg/kg upper bound that drives the group range comes from soya sauce, not from oyster sauce. The source does not explain mechanistically why soya sauce iAs was elevated.

Dietary exposure estimates:

Age-gender subgroup averages ranged from 0.19 µg/kg bw/day (female 20-29) to 0.26 µg/kg bw/day (male 60-69) for average consumers, and from 0.33 to 0.46 µg/kg bw/day for high consumers across the same span.

Dietary exposure by food group (average adult population, total 0.22 µg/kg bw/day):

Cooked white rice alone contributed 45.2% of total dietary exposure (84.5% of the cereals group contribution). Brown/unpolished rice had nearly double the iAs concentration (mean 43 µg/kg vs 22 µg/kg for white rice) but contributed only 1.3% of total population exposure due to its minor role in Hong Kong diet (consumed by approximately 5% of the population; among unpolished-rice consumers the exposure contribution from unpolished rice alone was 0.06 µg/kg bw/day average and 0.18 µg/kg bw/day for high consumers of that food item).

Water spinach (Ipomoea aquatica) cultivar variance (Table 5): The four occasions’ composite samples were: occasion 1 = 2 “green stem” + 1 “white stem” mixed at 98 µg/kg; occasion 2 = 3 “white stem” only at 120 µg/kg; occasion 3 = 3 “green stem” only at 43 µg/kg; occasion 4 = 3 “green stem” only at 35 µg/kg. Composites containing “white stem” samples (occasions 1 and 2) had higher iAs than composites containing only “green stem” samples (occasions 3 and 4); the source proposes growing condition (aquatic vs moist soil) as the driver, analogous to the rice/water-uptake mechanism. Despite the elevated concentration, water spinach contributed only 3% of total dietary exposure due to limited consumption frequency.

Reference benchmark: JECFA 2010 withdrew the former PTWI of 15 µg/kg bw/week (2.1 µg/kg bw/day) for iAs because it falls within the BMDL0.5 range of 2-7 µg/kg bw/day for a 0.5% increased incidence of lung cancer. Hong Kong population average exposure (0.22 µg/kg bw/day) represented approximately 10% of the former PTWI and produced MOEs of 9-32 against the BMDL0.5; high-consumer exposure (0.38 µg/kg bw/day) represented approximately 18% of the former PTWI with MOEs of 5-18.

International comparison (Table 6, adult dietary iAs exposure µg/kg bw/day):

Direct cross-country comparison requires caution because of differences in methods, time, food consumption survey design, and treatment of below-LOD values; studies marked as conversion-factor based estimate iAs from total arsenic measurements, not direct speciation.

Methods (brief)

The 1st HKTDS purchased 150 commonly consumed food items (based on the Hong Kong Population-Based Food Consumption Survey 2005-2007), prepared them as consumed, combined into composites (3 purchases per item per occasion), and collected on four occasions March 2010 to February 2011 for a total of 1,800 individual samples reduced to 600 composites. Laboratory analysis for inorganic arsenic was conducted by CFS’s Food Research Laboratory using solubilisation in concentrated HCl, arsenite extraction into chloroform, back-extraction to diluted HCl, dry-ashing destruction of organic matter, redissolution in HCl, and quantification by hydride generation ICP-MS. The method specifically measures the sum of arsenite (As(III)) plus arsenate (As(V)); the extraction procedure is also known to capture a small amount of monomethylarsenic (MMA). LOD = 3 µg/kg food, 1.5 µg/kg water; LOQ = 10 µg/kg food, 5 µg/kg water. Non-detects assigned ½ LOD for exposure calculations per WHO GEMS/Food-EURO guidance. About 51% of composite samples were detected above LOD. Dietary exposures combined analytical results with food consumption data via the EASY (Exposure Assessment System) web-based tool; mean and 95th percentile of the exposure distribution represent average and high consumer respectively. Margins of exposure (MOE) calculated as BMDL0.5 / dietary exposure, with the range reflecting the BMDL0.5 confidence interval (2-7 µg/kg bw/day).

Analytical note: Direct iAs speciation, not total arsenic with conversion factor. This is the critical methodological distinction from the previous 2002 FEHD secondary-school study, which used a 10% conversion factor applied to total arsenic from surveillance data with LOD of 76 µg/kg; that approach systematically overestimated iAs in seafood (where total As is high but inorganic fraction is typically <10–15%) and underestimated the rice contribution (where the inorganic fraction varies from 17% to 100% depending on cultivar and growing conditions). The 2002 study found “seafood other than fish” as the largest contributor (51%); the present direct-speciation study finds cereals/rice as the largest contributor (53.5%, with rice alone 45.2%).

Implications

Certification. Rice is confirmed as the dominant contributor to dietary iAs exposure in rice-consuming populations; the 22 µg/kg (cooked white rice) and 43 µg/kg (cooked unpolished rice) concentration values from a government TDS contribute occurrence data to the wiki’s rice iAs profile. The unpolished-vs-white rice differential (43 vs 22 µg/kg, approximately 2x) supports the processing-effect evidence on the rice ingredient page. Among condiments, soya sauce shows mean 21 µg/kg with range 3–65 µg/kg — relevant for any product category using soya sauce as an ingredient — and oyster sauce shows mean 12 µg/kg (range 6–17 µg/kg). Egg products show a wide spread driven by salted (58 µg/kg) and lime-preserved (10 µg/kg) eggs while chicken eggs were ND; salt-curing using plant ash or loess is identified by the source as a plausible iAs source for processed eggs.

Courses. Illustrates the methodological superiority of direct iAs speciation over total-arsenic conversion factors. Using a universal 10% conversion factor (as in the 2002 FEHD study) shifted the apparent major contributor from seafood (because total As is high in seafood) to cereals (because the inorganic fraction is much higher in rice than in seafood). Provides concrete washing-and-cooking mitigation data: rinse washing removes approximately 10% of both total and inorganic arsenic; cooking with high water volume (rice-to-water ratio 1:6) and discarding excess water removes 35% of total As and 45% of iAs versus uncooked rice (both figures relative to the same uncooked baseline, not cumulative with the rinse-washing reduction). Local Hong Kong practice (rice-to-water ratio approximately 1:1.5–2.5, no excess water discarded) achieves only the rinse-only removal, approximately 10%.

App. TDS-level iAs data for 150 food items as consumed (cooked) rather than raw commodity. Condiments group mean (8 µg/kg) with soya sauce specifically at 21 µg/kg (up to 65 µg/kg) and oyster sauce at 12 µg/kg (up to 17 µg/kg) are relevant for any ingredient-list-based exposure model. Table salt at ND, cornstarch at 2 µg/kg, tomato paste/ketchup at 7 µg/kg.

Limitations noted by the source

Seaweeds — which can contain very high inorganic arsenic levels (hijiki 30,000–130,000 µg/kg per literature; other seaweed typically below 2,000 µg/kg) — were not included in the TDS food list because the Hong Kong Population-Based Food Consumption Survey 2005-2007 captured only limited consumption data for them. This may underestimate iAs exposure in subpopulations who consume seaweed regularly (e.g., sushi consumers). Conversely, treatment of non-detects with ½ LOD substitution may overestimate exposure from the non-alcoholic beverages group (95% of which were below LOD); if a zero-substitution rule were applied, the beverage contribution would fall from 13.0% to approximately 1% of total exposure.

Wiki pages updated on ingest

• rice
• cereals
• leafy-vegetables
• vegetables
• seafood
• eggs
• meat
• fruits
• legumes
• beverages
• tomato-paste
• ketchup
• seafood
• condiments-general
• fruit-juices-non-apple
• tomato-paste
• arsenic-inorganic
• jecfa-arsenic-bmdl

Verification notes

Merge-enhanced 2026-05-28 from prior 2026-05-14 revision (cite_key, raw_handle, raw_path, license, access_url, near_duplicates preserved verbatim). Re-read the full source PDF (39 pages, including Executive Summary §§1-11, body §§1-46, References, Appendix 1 food-item table, Appendix 2 age-gender table). Changes:

1. Soya sauce / oyster sauce values transposition corrected (Check 1, ❌). Prior revision had Oyster sauce 21 µg/kg with range 3–65 µg/kg and Soya sauce at half-LOD (ND). The source (Appendix 1, p. 34) reports the opposite: Soya sauce mean 21 µg/kg [3–65], Oyster sauce mean 12 µg/kg [6–17]. The prior revision’s narrative “the outlier driving this range is oyster sauce … derived from shellfish and thus carries elevated iAs consistent with seafood origin” is therefore factually incorrect — soya sauce is the outlier, and the source does not propose a shellfish-derivation mechanism (which would not apply to soya sauce anyway). Corrected both the food-item table and the condiments narrative; reframed the Implications-Certification paragraph to refer to soya sauce as the high-condiment outlier without inventing a mechanism the source does not provide.
2. metals: [iAs, tAs] → metals: [iAs]. This study measures inorganic arsenic directly by hydride generation ICP-MS speciation; total arsenic is referenced contextually (e.g., Table 1’s iAs/tAs ratios drawn from earlier literature) but is not an analyte of this study. Per CLAUDE.md Part 14 and the audit-prompt Check 3 rule, the metals: array should list what was actually measured.
3. Invalid ingredient slug removed. Prior revision included [[ingredients/condiments-sauces]], which is not in the current ingredients taxonomy. Replaced with [[ingredients/tomato-paste]] and [[ingredients/ketchup]] (both in taxonomy, both with concrete µg/kg values in Appendix 1 of the source). Added [[products/tomato-paste]] to product routing.
4. Key numbers expanded. Added full age-gender breakdown (15 rows from Appendix 2 instead of selected rows), expanded food-item highlights to include yellow croaker (30 µg/kg, the highest fish), crab (27), shrimp (23), scallop (20), Chinese pastries (17), lime-preserved egg (10), table salt and chicken egg at ND. Added “All food groups” row to food-group table (n = 600, 49% below LOD). Added the full food-group exposure table (all 13 categories including the four below 1%), the international comparison table (Table 6), and the water-spinach cultivar breakdown (Table 5). The prior page covered the same study but elided several distributional details that downstream synthesis will need.
5. Wiki pages updated section expanded. Added the two new ingredient routes (tomato-paste, ketchup), one product route (tomato-paste), the iAs metal slug (using the taxonomy entry arsenic-inorganic), and the JECFA arsenic BMDL regulation page (the study explicitly anchors its risk characterisation to JECFA 2010’s BMDL0.5).
6. Added Limitations section capturing the two source-stated limitations (seaweed exclusion, ½ LOD substitution effect on beverages contribution) that the prior revision omitted.
7. Methods (brief) tightened to spell out the LOD/LOQ pair for water (1.5/5) alongside food (3/10), the explicit detection rate (51% above LOD), and the MOE = BMDL0.5 / exposure calculation. The prior revision had the LOD pair only for food.
8. Brand-firewall scan (Part 12): the source uses commodity descriptors only (water spinach, oyster, salted egg, tomato paste/ketchup, soya sauce) — no manufacturer brand names anywhere in the report. No redactions required.
9. Wiki/HMTc-firewall scan (Part 2): removed the prior “support the wiki’s rice iAs profile” phrasing in favour of the literature-baseline phrasing “contribute occurrence data to the wiki’s rice iAs profile”; no HMTc threshold proposals or synthesis-across-papers claims appear in the rewritten Implications.

Audit subagent (2026-05-28, fresh-context general-purpose) returned REVISE with all five checks otherwise clean (Check 1 ⚠️ x2; Checks 2-5 ✅). Both Check 1 ⚠️ findings independently verified against the source PDF and applied:

• Water spinach Table 5 attribution (⚠️ → applied). Auditor noted the prior wording attributed the 98 µg/kg occasion-1 value purely to the white-stem cultivar; PDF Table 5 (p.22) shows occasion 1 was a mixed composite of 2 green-stem + 1 white-stem samples at 98 µg/kg, occasion 2 was 3 white-stem samples only at 120 µg/kg, occasions 3 and 4 were 3 green-stem each at 43 and 35 µg/kg. The wiki paragraph now reports the composite composition per occasion accurately and uses comparative language (“composites containing white-stem samples … had higher iAs than composites containing only green-stem samples”) rather than attributing per-cultivar concentrations the composite design cannot isolate.
• Cooking-water removal “additional” framing (⚠️ → applied). Auditor noted that PDF para 34 reports cooking-removal as 35%/45% “as compared with the uncooked rice” — i.e. cooking vs uncooked, not cumulative with the rinse-washing reduction. The Courses paragraph dropped the word “additional” and now explicitly clarifies that both reductions are referenced to the same uncooked baseline (not cumulative).

No false positives; no findings rejected. Subagent verdict: REVISE (now resolved). Updated audit-queue.csv status to audited-revised.

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.