Carey et al. 2015 — Percolating cooking water removes up to 85% of inorganic arsenic from rice

This PLOS ONE study tested three percolating-cooking approaches against the conventional absorption method for removing inorganic arsenic (iAs) from rice: (1) a laboratory cooking experiment at water-to-rice ratios of 3:1, 6:1 and 12:1 under reflux to capture evaporative losses; (2) a Soxhlet apparatus where cooking water was continuously recycled through distilled steam; and (3) an off-the-shelf domestic coffee percolator running near-boiling water through the rice. The 12:1 ratio removed an average 57±5% of iAs across all rice types; the Soxhlet method removed 59±8% (polished) and 69±10% (wholegrain); the coffee percolator removed 49±7% on average and up to 85% in the best individual sample. Dimethylarsinic acid (DMA) was not significantly reduced by any cooking method tested. Potassium was the only major mineral lost in substantial quantity (53% on coffee-percolator cooking).

Key numbers

Values are in mg/kg dry-weight unless noted; arsenic species are inorganic arsenic (iAs = arsenite + arsenate) and dimethylarsinic acid (DMA), measured by IC-ICP-MS.

Cooking volume experiment (n=6 wholegrain + 6 polished; Fig 1) — rice cooked at 3:1, 6:1 and 12:1 water:rice ratios in a sealed-flask reflux setup that returned evaporated steam to the cooking vessel (no water discarded; rice freeze-dried whole).

• 12:1 ratio: 57±5% iAs removed across all rice; 53±5% for polished, 61±3% for wholegrain.
• 3:1 ratio: 29±5% iAs removed across all rice.
• 6:1 ratio: 29±4% iAs removed across all rice.
• Uncooked wholegrain iAs roughly two-fold higher than uncooked polished on average; wholegrain showed a steeper removal slope at higher cooking volumes.
• DMA was not significantly reduced by cooking; only the rice-type term was significant (P<0.001), with wholegrain having higher DMA than polished.

Soxhlet recycled-condensation experiment (all 41 samples; Fig 2) — rice (2 g) held in a Soxhlet thimble while cooking water cycled through three reflux passes; receiving flask started with 200 mL deionized water.

• iAs removal: 59±8% for polished (n=27) and 69±10% for wholegrain (n=13).
• Individual samples reached >80% iAs removal in some cases.
• For polished rice, higher initial iAs correlated with higher percentage iAs removed; for wholegrain the relationship was slightly inverse (saturation at high initial iAs).
• DMA removal was poor; the rice-type and initial-DMA terms were not significant in the percentage-removal GLM.

Coffee percolator experiment (home-cooking trial, n=6 wholegrain + 6 polished; Fig 3, Fig 4) — 500 g rice in a Bravilor Bonamat catering coffee percolator with paper filter; 2 L of tap water passed through twice (polished; 20 min total) or three times (wholegrain; 30 min total); inlet water temperature 98 °C. Each rice type cooked in triplicate.

• iAs removal: 49±7% across all 12 samples; up to 85% in the best individual sample.
• Rice type (wholegrain vs polished) not a significant factor for percentage iAs removal.
• iAs in the tap water was 0.050±0.02 µg/L (LOD 0.036 µg/L).
• Mineral co-losses (paired t-test cooked vs uncooked): potassium 53% loss (P<0.001); phosphorus 7% loss (P=0.016); calcium, copper, iron, manganese, sulphur and zinc not significantly changed.

Uncooked rice iAs ranges (read across Figs 1, 2 and 3):

• Polished: approximately 0.01 to 0.23 mg/kg (highest values in the coffee-percolator subset, Fig 3).
• Wholegrain: approximately 0.085 to 0.46 mg/kg (highest values in the coffee-percolator subset, Fig 3).

Quality control:

• Certified reference material: NIST SRM 1568b Rice flour, certified for both iAs (0.092 mg/kg) and DMA (0.182 mg/kg); recovery (n=11) was 88.2±3.3% for iAs and 91.6±4.4% for DMA, sum-of-species recovery 90.3%.
• LOD for iAs and DMA in rice: 0.0028±0.001 mg/kg dry-weight (n=5 calibration).
• All samples were above LOD for iAs; two were below LOD for DMA (½ LOD substitution used).

Methods

Rice was purchased from major UK retailers in Belfast and via UK online retail; 41 packets in total (13 wholegrain, 28 polished), with the country-of-origin labels listed under sample_population above.

Three cooking systems were compared:

1. Standardised laboratory cooking at controlled water:rice ratios. 20 g rice in a 250 mL round-bottomed Pyrex flask heated in an electrothermal mantle; the flask neck connected to a 25 cm dimpled Vigreux condensing tube (42 dimples) that returned steam-as-water to the flask. Double-distilled, deionized water added to give 3:1, 6:1, or 12:1 water:rice ratios. After cooking, samples were freeze-dried and powdered in a Retsch PM100 ball mill using a zirconium-oxide-lined vessel and zirconium-oxide grinding balls.
2. Soxhlet recycled-condensation cooking. 2 g rice was held in a VWR Soxhlet thimble inside a 25 cm × 3.5 cm Quickfit Soxhlet attached to a 250 mL receiving flask containing 200 mL of double-distilled, deionized water; cooking timed by Soxhlet reflux cycles, with three cycles sufficient for full cook.
3. Home coffee-percolator cooking. Off-the-shelf Bravilor Bonamat catering coffee percolator (no hotplate; filter unit only); 500 g rice packet placed in the manufacturer-supplied paper filter; the 2 L reservoir took 10 min to discharge per cycle. Polished rice cooked with 2 × 2 L (4 L total, 20 min); wholegrain cooked with 3 × 2 L (6 L total, 30 min). Inlet water entered the filter at 98 °C. Cooking experiments were run in triplicate for each rice type.

Analytical chemistry: cooked and uncooked rice was freeze-dried and milled, then 0.1 g powder digested overnight in 2 mL of 1% Aristar HNO3, microwave-digested in a CEM MARS 6 (30 min staged ramp to 95 °C), diluted to 10 mL, and treated with 10 µL H2O2 to convert arsenite to arsenate prior to ion-chromatography. Arsenic speciation by IC-ICP-MS used a Thermo Scientific IC5000 ion-chromatography system with a Thermo AS7 (2 × 250 mm) column and AG7 (2 × 50 mm) guard, a 20–200 mM ammonium-carbonate linear gradient over 15 min, interfaced to a Thermo iCAP Q ICP-MS monitoring m/z 75 with He collision-cell mode; calibration species DMA, iAs, MMA, tetramethyl arsonium and arsenobetaine. Total elements were measured by the same iCAP Q in direct-solution mode with Rh as internal standard; additional elements were measured by bench-top XRF (Rigaku CG). Multi-element digestion used a more aggressive program to 180 °C with 2 mL concentrated HNO3 plus 2 mL H2O2. Blanks and NIST SRM 1568b rice-flour CRM were included in each batch of 40 samples.

Statistics: General Linear Modeling (GLM) using Minitab v.16; residuals tested for normality with Anderson–Darling and found normal so no transformation was applied except for percentage data, which were rank-transformed before analysis.

Implications

The source supports rice mitigation context for iAs reduction by cooking-water management. Findings should be reported with the cooking-water regime (water:rice ratio, percolating vs static, distilled vs tap) preserved, because both percentage-removal and final cooked iAs vary substantially with the regime. The paper measures rice as a raw grain commodity cooked in a laboratory or domestic setting; it does not measure or test commercially manufactured infant rice cereal or rice-based plant milk, so any application of these removal percentages to manufactured rice products is downstream extrapolation, not direct evidence.

Wiki pages this source may touch

• rice
• rice-bulk-grain
• arsenic-inorganic
• arsenic-total

Verification notes

Page rebuilt 2026-05-28 from the source PDF (raw_sha256 5529a95f… matches the file currently at the raw_path above; this is the same PDF that backed the 2026-05-12 page). Defects in the prior 2026-05-12 revision corrected in this pass, with the source citations that justified each correction:

• Title was a fabricated descriptive line (“Evaluating the effectiveness of filters and cooking procedures for reducing inorganic arsenic (iAs) in rice in the context of providing healthy food for infants”). Source p. 1 gives the actual title “Rethinking Rice Preparation for Highly Efficient Removal of Inorganic Arsenic Using Percolating Cooking Water.” Corrected.
• Authors list invented a fifth author (“Price AH”) who does not appear on this paper and used initials (“Carey AM”) that do not match the source. Source p. 1 byline: Manus Carey, Xiao Jiujin, Júlia Gomes Farias, Andrew A. Meharg (four authors). Corrected.
• Geographic origins in the prior Key numbers listed “UK, US, India, Pakistan, Bangladesh” as sample origins. Source p. 2 Materials and Methods → Rice sourcing explicitly enumerates Spain (11), Italy (6), Thailand (5), USA (5), generic EU (2), France (2), Egypt (1), Japan (1), Australia (1), Lebanon (1), Pakistan (1), Turkey (1). India and Bangladesh are mentioned only as a literature comparison in the Introduction, not as sample origins. Corrected; jurisdictions now lists UK as market plus the actual country-of-origin codes.
• Method-effect attributions in the prior Key numbers stated “Conventional absorption cooking: retains approximately 90% of raw grain iAs” and “High-volume water drain method: retains approximately 70% (~30% reduction).” This paper did not test a static-absorption protocol or a drain-the-water protocol; its three experiments were the controlled-ratio reflux experiment (3:1/6:1/12:1, no water discarded), the Soxhlet recycled-distillation experiment, and the coffee-percolator experiment, all percolating-water designs. The 30% and 45% reductions for traditional Asian “rinse-and-discard” cooking are cited from prior literature in the Introduction (refs 8 and 10), not measured here. Corrected; the prior absorption/drain claims removed and the actual three-experiment results substituted with their source-reported percentages and ranges.
• Analytical method was given as “HPLC-ICP-MS.” The paper uses ion-chromatography ICP-MS (Thermo IC5000 IC system + iCAP Q ICP-MS, AS7 column, ammonium-carbonate gradient), not reversed-phase HPLC. Corrected.
• Frontmatter products was [[products/baby-cereals-dry-rice-based]]. This paper measures rice as a raw grain commodity prepared by various cooking methods; it does not measure commercially manufactured baby cereal. The Discussion section (p. 10) speculates that the percolating-cooking approach could be applied in baby-rice manufacture, but no baby-cereal product is tested. Changed to [[products/rice-bulk-grain]], which matches what the paper actually measures and matches the routing used by pogoson2020-rice-cooking-cadmium-arsenic (same Belfast retail-rice cohort, same lab).
• Frontmatter metals now includes DMA because DMA was speciated alongside iAs and reported (cooking does not significantly remove DMA — a finding-bearing result for downstream synthesis). tAs retained for total arsenic.
• Frontmatter raw_path was raw/studies/carey2015-rice-arsenic-percolating-cooking.pdf (a path that does not exist). Updated to the verified location at raw/manual-fetch/seasonal-geographic-variance/auto-fetched/wl-0007_2015_10-1371-journal-pone-0131608.pdf. SHA-256 unchanged and matches the file at the new path.
• Frontmatter raw_handle was missing; added auto-fetched to match the discovery-pipeline convention used by sibling auto-fetched pages.
• Frontmatter field rename sha256 → raw_sha256 to match the current v2 schema (e.g., dehghani2020-food-lead-cadmium-mspe).
• Legacy section ## Wiki pages updated on ingest renamed to ## Wiki pages this source may touch per v2 schema; the prior body of that section listed [[mitigation/cooking-methods-arsenic]], which is not in the current taxonomy and was removed.
• Audience-segmented “Implications” lines from the prior version that mapped findings into Certification / Courses / App / consumer-advice slots have been removed because they violate Part 2 of CLAUDE.md (source page reports what one paper found, not how it should drive HMTc thresholds, courses, or app translations). The retained ## Implications section now describes the routing/mitigation-context role of this source only.

Audit subagent (2026-05-28) flagged the following; verified against source and applied or rejected:

• 3:1 / 6:1 ratio attributions were transposed in the initial 2026-05-28 ingest. Source p. 6 ¶1: “This figure for all rice was 29±5% and 29±4% at ratios of 3:1 and 6:1 respectively.” Verified — corrected so 3:1 = 29±5% and 6:1 = 29±4%.
• Soxhlet thimble dimensions (“8 cm × 3.5 cm”) were fabricated in the initial 2026-05-28 ingest. Source p. 3 (Cooking… recycled through condensation) states “accurately weighed (2g) into a VWR Soxhlet thimble and was then placed into an 25cm long and 3.5cm diameter Quickfit Soxhlet” — only the Quickfit Soxhlet has stated dimensions, not the thimble. Verified — invented thimble dimensions removed; the 25 cm × 3.5 cm dimensions now correctly attached to the Quickfit Soxhlet body.
• Uncooked rice iAs range under-stated. Subagent observed that Fig 3 (coffee-percolator subset) shows higher uncooked iAs than the Fig 1/Fig 2 subsets cited in the original ingest — polished reaches ~0.23 mg/kg and wholegrain ~0.46 mg/kg in Fig 3. Verified — ranges extended and Fig 3 added to the citation.
• AG7 guard column dimensions omitted. Source p. 4 specifies “AG7, 2x50mm guard column”; verified and added.
• Source-internal sample count inconsistency. The PDF p. 2 country-of-origin enumeration sums to 37 (2 EU + 11 ES + 6 IT + 5 TH + 2 FR + 1 EG + 1 JP + 1 AU + 1 LB + 1 PK + 1 TR + 5 US = 37), yet the paper repeatedly states the total is 41 (with 13 wholegrain / 28 polished). The discrepancy is internal to the source — the inline enumeration omits 4 packets — and is presumably reconciled in the S1 Table sample description. Flagged here; the wiki transcribes the paper’s stated total of 41 and the paper’s inline country list, without attempting to silently reconcile.
• DMA metals-vocabulary extension. The system-prompt metal abbreviation list (Pb, Cd, iAs, tAs, iHg, MeHg, tHg, Ni, Al, Cr, Cr-VI, Sn, Sb, U) does not include DMA. DMA is included here because the paper centrally speciates DMA and reports a finding-bearing result on it (cooking does not significantly remove DMA). Flagged for taxonomy-vocab review.
• matrices sub-variants (polished-rice, wholegrain-rice, cooked-rice) flagged as possibly off-vocab by the audit subagent; verified against pogoson2020-rice-cooking-cadmium-arsenic which uses the same lab’s polished-rice, cooked-rice and pre-soaked-rice matrices entries — finding is a false positive in light of established same-lab precedent. No correction applied.
• jurisdictions mixed market/origin semantics flagged by the audit subagent; the intent is already explained in the sample_population field and in the existing Verification-notes geographic-origins entry — finding is a false positive against existing documentation. No correction applied.
• tAs in metals frontmatter flagged as borderline by the audit subagent; retained because total arsenic is implicitly available as the sum of iAs + DMA + minor species and is reportable to downstream routing as a derived total. No correction applied.

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.