Rodriguez, Raber & Goessler 2009 — Arsenic speciation in fish sauce by HPLC-ICPMS

Six commercial fish sauces from Thailand (n=4) and Vietnam (n=2), purchased at local stores in Graz, Austria, were analysed for total arsenic by flow injection ICP-MS and for arsenic speciation by HPLC-ICP-MS with HPLC/ESI–MS confirmation on the three highest-concentration samples. Total arsenic ranged from 0.69 to 2.75 mg/L across the six samples. Speciation identified arsenobetaine as the dominant species (82–94% of total arsenic), with arsenocholine (4.9–7.7%), trimethylarsine oxide (TMAO, 0.7–7.8%), and trimethylarsoniopropionate (TMAP, 0.5–2.1%) as minor components. Anion-exchange chromatography showed that the highly toxic anionic species — arsenite, arsenate, methylarsonic acid (MA), and dimethylarsinic acid (DMA) — were all below the 0.01 mg/L detection limit in every sample. The authors note that their results contradict an earlier study (Kato, Nagashima & Shiomi 2004, Fisheries Science 70, 695–702), which reported DMA as the major arsenical in Japanese fish sauces; the authors attribute the divergence either to matrix-influenced detection by the Kato et al. instrumentation or to genuine difference in fish-sauce arsenic profile between Japanese and Thai/Vietnamese product. The conclusion drawn by the authors is that the fermentation process does not transform marine non-toxic arsenicals into toxic forms, but they recommend broader sampling to resolve the Kato et al. discrepancy before fish sauce is endorsed as an iron-fortification vehicle in iron-deficient populations.

Key numbers

Total arsenic in fish sauce, undiluted (n=6 samples; each value mean ± SD, n=3 replicates per sample; units mg/L = mg/L of sauce as-purchased):

Range across the six samples: 0.69–2.75 mg/L total arsenic. Mass-balance check: FI-ICPMS total and HPLC-ICPMS sum of species agreed within reported uncertainties for all six samples (no sample showed >5% difference between the two methods).

Arsenic species in fish sauce (n=6; mean ± SD, n=3 per sample; units mg/L; Table 2 of the source):

Species fractional composition (calculated by the authors from the data above; reported as ranges across the six samples):

• Arsenobetaine: 82–94% of total arsenic.
• Arsenocholine: 4.9–7.7%.
• TMAO: 0.7–7.8%.
• TMAP: 0.5–2.1%.

Toxic anionic species below detection (n=6; LOD 0.01 mg/L; anion-exchange HPLC-ICPMS with 20 mM phosphate buffer at pH 6.0):

• Arsenite (As(III)): <0.01 mg/L in all six samples.
• Arsenate (As(V)): <0.01 mg/L in all six samples.
• Methylarsonic acid (MA): <0.01 mg/L in all six samples.
• Dimethylarsinic acid (DMA): <0.01 mg/L in all six samples. A spike-recovery confirmation on fish sauce 4 (spiked with 5.0 µg/L DMA) showed a clearly resolved DMA peak in the spiked sample and no DMA peak in the unspiked sample (Fig. 2), supporting the absence finding.

Thio-arsenical screen (n=6; chromatographic conditions per Raml, Goessler & Francesconi 2006):

• No sulphur-containing arsenic species detected in any sample, despite the characteristic odour of fish sauce that the authors note could have suggested their presence (the odour is attributed by the cited fermentation literature to amino acids, nucleotides, peptides, ammonia, and urea rather than thio-arsenicals).

Quality control against DORM-2 (NRCC dogfish muscle CRM; mean ± SD; certified values from NRCC):

• Total As, conventional ICPMS of extract: 17.2 ± 0.2 mg/kg.
• Total As, FI-ICPMS of extract: 17.2 ± 0.3 mg/kg.
• Arsenobetaine, HPLC-ICPMS: 16.6 ± 0.4 mg/kg (NRCC certified value: 16.4 ± 1.1 mg/kg).
• Tetramethylarsonium ion, HPLC-ICPMS: 0.24 ± 0.09 mg/kg (NRCC certified value: 0.248 ± 0.054 mg/kg).

Both totals and the two speciated CRM analytes recover within the certified uncertainty bands; the FI-ICPMS and conventional ICPMS methods agreed on DORM-2 total As to within their reported uncertainties.

HPLC/ESI–MS confirmation (subset of three samples, fish sauce 4, 5, 6):

• AB confirmed as the major component in each of fish sauces 4–6 by ESI–MS. The HPLC/ESI–MS chromatographic profiles of fish sauces 1–6 were qualitatively similar under both anionic and cationic chromatographic conditions; ESI–MS was applied only to the three higher-As samples because of ESI sensitivity considerations.

Methods (brief)

Six commercial fish sauces (four Thailand-origin brands, two Vietnam-origin brands) purchased at local stores in Graz, Austria. Samples were diluted 1+99 with Milli-Q water (18.2 MΩ·cm, Millipore Bedford, MA, USA) prior to FI-ICPMS analysis and speciation analysis. For quality control, the extract of DORM-2 (dogfish muscle, National Research Council Canada, Ontario): 250 mg of CRM weighed into polyethylene vials with 10 mL water/methanol (1+1, v/v), shaken top-over-bottom for 18 h, centrifuged at 4500 rpm for 15 min, and 1 mL aliquots digested alongside the fish sauce samples.

Flow injection ICPMS used the Rodriguez, Francesconi & Goessler 2008 method (J Anal At Spectrom 23, 235–239). Hewlett–Packard 1100 series HPLC system (binary pump, vacuum degasser, thermostatted autosampler with 100 mm³ injection loop) connected by PEEK capillary directly to an Agilent 7500ce ICPMS (Waldbronn, Germany) with PFA microconcentric nebulizer, Scott double-pass spray chamber, and octopole reaction cell. Helium at 3.0 mL/min as collision gas. Eluent: 0.3% HNO₃ with 10% methanol (v/v). Injection volume 20 µL. Quantification by normalisation to ⁷⁴Ge internal standard (spike level 400 µg/L in standards and samples) and against an external arsenic calibration (1.0–500 µg As/L in 1% HNO₃).

HPLC/ICPMS speciation under both anion- and cation-exchange conditions. Anion exchange: PRP-X100 column (4.1 × 250 mm, 10 µm particles; Hamilton, Reno, Nevada), 40 °C, mobile phase 20 mM phosphate buffer pH 6.0 (adjusted with aqueous NH₃), 1.5 mL/min. Cation exchange: Zorbax 300 SCX column (4.6 × 250 mm; Hewlett–Packard, Waldbronn), 30 °C, mobile phase 10 mM pyridine pH 2.3 (adjusted with formic acid), 1.5 mL/min. Injection volume 20 µL. Ion intensities monitored at m/z 75 and 77; quantification by peak area. TMAP was quantified by the AB correlation equation because no TMAP standard was used.

HPLC/ESI–MS confirmation on fish sauces 4–6 in positive mode using an Agilent LC/MSD 1100 series single-quadrupole MS (SL type). Separation on a Shodex RSpak NN-614 column (6 × 150 mm) with 5 mM ammonium formate buffer pH 3.0 at 0.4 mL/min, 30 °C, 5 µL injection.

Thio-arsenical screen used the chromatographic conditions of Raml, Goessler & Francesconi 2006 (J Chromatogr A 1128, 164–170). A polyatomic interference at m/z 75 from ⁴⁰Ar³⁵Cl appeared at ~2 min in the cation-exchange chromatograms because of the 20–30% NaCl content typical of fish sauce; the AB, TMAO, AC, and Tetra peaks were chromatographically resolved from this interference.

All reagents analytical grade. Arsenic and germanium 1000 mg/L stock standards from CPI (Santa Rosa, USA). Nitric acid from Merck (Germany). Methanol from Carl Roth GmbH (Karlsruhe, Germany).

Evidence Fitness

This is a small (n=6) convenience-sample methods-and-application paper that pairs a flow-injection ICPMS total-As measurement with HPLC-ICPMS speciation across all four cationic organoarsenicals plus an HPLC-anion-exchange screen for the toxic anionic species. The peer-reviewed quantitative findings — total As 0.69–2.75 mg/L, arsenobetaine 82–94%, arsenite/arsenate/MA/DMA all <0.01 mg/L — are first-look speciation data for Thai and Vietnamese fish sauces in the Graz retail channel, not occurrence-distribution estimates for the global fish-sauce category. The CRM-anchored speciation discipline (DORM-2 arsenobetaine recovered at 16.6 ± 0.4 mg/kg vs certified 16.4 ± 1.1 mg/kg; tetramethylarsonium recovered at 0.24 ± 0.09 mg/kg vs certified 0.248 ± 0.054 mg/kg) supports the methodological strength of the paper at this single-laboratory level.

Best uses:

• Speciation evidence that the dominant arsenical in Thai and Vietnamese fish sauces in this sample is non-toxic arsenobetaine, not the toxic dimethylarsinic acid (DMA) reported earlier by Kato et al. 2004 for Japanese fish sauces. This is a substantive open question in the fish-sauce arsenic literature.
• Occurrence indicator that fermented fish-based condiments can carry total arsenic in the ~0.7–2.8 mg/L range while remaining below detection for all four toxic anionic species at a 0.01 mg/L LOD.
• Methods reference for HPLC-ICPMS speciation in high-salt (20–30% NaCl) matrices, including the ⁴⁰Ar³⁵Cl polyatomic-interference handling.
• Speciation-vs-totals teaching example: in this matrix the total-arsenic value alone would be misleadingly alarming (the high end exceeds the EU 0.5 mg/kg general food-contaminant ceiling for total As if naively compared), whereas the speciation shows that the bulk is non-toxic AB and the toxic fraction is below LOD.

Not appropriate for:

• Setting a percentile-based HMTc occurrence value for fish sauce (n=6 is too small; convenience sample; single retail channel in Graz; 2007-or-earlier vintage; no temporal replication).
• Brand-level claims (the source does not name brands and the wiki does not propagate brand-level values per Part 12).
• Generalising the speciation finding to all fermented fish products globally — the paper’s own contrast with Kato et al. 2004 (Japanese fish sauces, DMA-dominated) indicates that the speciation profile may be supplier- or fermentation-process-specific.
• Direct cross-product extrapolation to soy sauce, shrimp paste, or other Asian fermented condiments without independent speciation data.

Implications

• Total arsenic in fish sauce is high but is overwhelmingly non-toxic arsenobetaine in this sample. The 0.69–2.75 mg/L range looks alarming when compared with general food-contaminant arsenic ceilings, but ≥82% is arsenobetaine and all four toxic anionic species (As(III), As(V), MA, DMA) are below 0.01 mg/L. This is the canonical “speciation changes the regulatory picture” case for fermented fish products, and downstream wiki pages on fish sauce, fermented condiments, and seafood-derived arsenic should propagate the speciation breakdown, not the total-As number alone.
• The literature on arsenic species in fish sauce is unresolved. Rodriguez et al. 2009 (Thai/Vietnamese fish sauces, AB-dominated) directly contradicts Kato et al. 2004 (Japanese fish sauces, DMA-dominated). The authors propose that the divergence is either matrix-driven HPLC/ESI–MS misidentification in the earlier study or a genuine regional difference in fermentation-substrate arsenic profile. Wiki synthesis on fish sauce arsenic should flag this discrepancy as a paired-citation open question rather than treating either single paper as the consensus.
• Iron-fortification programs using fish sauce as the vehicle require resolved speciation evidence. The authors note that fish sauce is being studied as an iron-fortification vehicle (Fidler et al. 2003 AJCN 78, 274–278; Thuy et al. 2003 AJCN 78, 284–290) in iron-deficient populations and that the arsenic speciation question is load-bearing for that public-health intervention. This is a relevant context for any future synthesis on iron fortification or on fish-sauce consumption in Southeast Asian populations.
• Speciation discipline (tAs vs iAs vs species) is non-negotiable for this matrix. Reporting only total arsenic would misrepresent the public-health signal for fish sauce. Any wiki page that propagates these numbers must carry the species breakdown alongside the total. The metals: [tAs, iAs] frontmatter records both: tAs is what FI-ICPMS measured; iAs is what HPLC anion exchange was characterising and found below LOD.
• Fermentation does not appear to convert non-toxic marine arsenicals into toxic forms in this sample. The paper’s main biological finding — that arsenobetaine, the dominant non-toxic species in marine animals, is preserved through fermentation rather than being demethylated or otherwise transformed into more toxic species — is meaningful for the broader question of how processing alters arsenic speciation in marine-derived foods. The finding is consistent with the absence of thio-arsenicals in the screen, but the small n=6 and single-laboratory design do not establish this as a general rule.

Provenance notes

• DOI is 10.1016/j.foodchem.2008.06.054. The article is Food Chemistry 112 (2009) 1084–1087.
• Publisher: Elsevier Ltd. License is publisher copyright (Elsevier); redistribution is restricted by the publisher’s terms. license: "publisher-copyright-elsevier".
• Corresponding author: Walter Goessler, Institute of Chemistry, Analytical Chemistry, Karl-Franzens University, Universitätsplatz 1, 8010 Graz, Austria. Tel +43 316 380 5302; email walter.goessler@uni-graz.at (per the paper’s masthead).
• Acknowledgments name the Austrian Academic Exchange Service (scholarship to I. B. Rodriguez) and the FWF project P16088 – N03 as funders.
• Article received 26 December 2007; accepted 25 June 2008; published in 2009 (volume 112). The fish sauces were purchased at Graz retail at an unstated date prior to submission; sampling_year_range: "2007 or earlier" is an inferred bound.

Verification notes

• Brand-firewall posture (Part 12, strict). The source paper does not name any of the six fish sauces by brand. Samples are coded “fish sauce 1” through “fish sauce 6” throughout the source text and tables. No brand redaction was required.
• Methods-vendor names (Part 12 Exception 2). Vendor names are retained where they identify the instrument, column, reference material, or reagent: Hewlett–Packard 1100 series HPLC, Agilent 7500ce ICPMS, Hamilton PRP-X100 column, Zorbax 300 SCX column (Hewlett–Packard), Shodex RSpak NN-614 column, Agilent LC/MSD 1100 series single-quadrupole MS, CPI arsenic and germanium standards, Merck nitric acid, Carl Roth methanol, Millipore Milli-Q water, NRCC DORM-2 dogfish muscle reference material. These are scientific-reproducibility identifiers, not contamination rankings, and do not violate Part 12.
• Speciation: tAs and iAs both in metals. The paper measures total arsenic (FI-ICPMS) and explicitly characterises the four toxic anionic species (arsenite, arsenate, MA, DMA) via anion-exchange HPLC-ICPMS, finding all of them below the 0.01 mg/L LOD. Both tAs (the totals measurement) and iAs (the inorganic-species characterisation, even where the result is below LOD) are within scope, so metals: [tAs, iAs] rather than metals: [tAs] alone. The organoarsenicals (AB, AC, TMAO, TMAP) are reported in the Key numbers tables but are not analyte categories in the HMI metal taxonomy.
• Ingredient routing. ingredients: ["[[ingredients/fish]]", "[[ingredients/seafood]]"] uses two existing taxonomy slugs that cover the fermentation substrates the source describes (fish — anchovy, tuna, salmon trimmings; seafood — shellfish, oyster, shrimp). There is no fish-sauce ingredient slug and no anchovy ingredient slug in the current docs/gpt-collaboration/taxonomy-snapshot.md (generated 2026-05-18). A precedent source (divis2020-mercury-fish-sauce-dgt-tda-aas.md) has used [[ingredients/fish-sauce]] and [[ingredients/anchovy]] slugs that do not yet have corresponding pages; this page follows the more conservative Berger Ritchie 2013 precedent and uses only existing taxonomy slugs, with the gap surfaced for Karen’s taxonomy review.
• Product routing. products: ["[[products/condiments-general]]"] uses the single existing taxonomy slug that covers fish sauce. There is no dedicated fish-sauce product slug in the current taxonomy snapshot; the condiments-general row is the routing destination consistent with the Berger Ritchie 2013 hot-sauce precedent. The Divis 2020 page declares a [[products/fish-sauce]] slug that does not yet correspond to a page; this page does not propagate that slug pending taxonomy review.
• Matrix vocabulary. matrices: [fish-sauce, fermented-fish-condiment, condiment] captures the as-tested matrix (fish sauce specifically), the fermented-fish-condiment process category (which links to other fish-sauce and fermented-condiment analyses), and the broader condiment grouping that maps onto condiments-general as the routing destination. Matrices vocabulary is bare strings, not wikilinks.
• Jurisdictions. jurisdictions: [AT, TH, VN] reflects: Austria as the retail purchase point (Graz); Thailand as the country of origin for four of the six samples; Vietnam as the country of origin for two of the six samples. The per-sample country-of-origin mapping is not stated in the source text; only the aggregated 4-Thai / 2-Vietnamese split is reported.
• Evidence-tier B rationale. Peer-reviewed primary research with CRM-anchored QC (DORM-2 total As recovered at 17.2 ± 0.2 mg/kg by conventional ICPMS and 17.2 ± 0.3 mg/kg by FI-ICPMS; AB recovered at 16.6 ± 0.4 mg/kg vs certified 16.4 ± 1.1; tetramethylarsonium recovered at 0.24 ± 0.09 vs certified 0.248 ± 0.054). The methodological strength is high (HPLC-ICPMS with HPLC/ESI–MS confirmation; explicit anion-exchange screen for the toxic species). The primary tier-B limitation is the convenience sample of n=6 from a single retail channel in Graz, Austria, and the single-purchase cross-section without temporal replication.
• Disagreement with Kato et al. 2004. The authors note their results contradict Kato, Nagashima & Shiomi 2004 (Fisheries Science 70, 695–702), which reported DMA as the major arsenical in Japanese fish sauces. The Implications section records this as an open paired-citation question rather than as a Kato-superseded claim. This is a literature-level observation, not a HMTc-level pivot, and stays within the Part 2 firewall.
• Iron-fortification framing. The authors cite the iron-fortification context (Fidler et al. 2003, Thuy et al. 2003, Mannar & Gallego 2002) as motivation for resolving the speciation. The wiki Implications section retains this as a downstream-synthesis flag, not as a wiki-side recommendation about fortification programs.
• fish-sauce ingredient and product slugs absent from taxonomy. Surfaced for Karen’s future taxonomy-review attention. Current routing uses [[ingredients/fish]], [[ingredients/seafood]], and [[products/condiments-general]]. Per CLAUDE.md, this ingest skill does not create ingredient or product pages; the gap is logged for the taxonomy review pass.
• Folder placement. PDF is filed under raw/manual-fetch/Kimi_Agent_Download Corruption Issue/condiments2_papers/03_Condiments/. The folder is the upstream sorter’s category and is preserved (immutable per Part 4).

Wiki pages this source may touch

• arsenic-total
• arsenic-inorganic
• fish
• seafood
• condiments-general
• Future fish-sauce (does not yet exist; condiments-general is the current routing destination)
• Future fish-sauce (does not yet exist; fish + seafood are the current routing destinations)
• Future anchovy (does not yet exist; fish is the current routing destination)

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.