Diviš et al. 2020 — Total mercury in fish sauces by DGT-TDA-AAS

This methods-development and application paper validates a Diffusive Gradients in Thin Films (DGT) preconcentration approach combined with thermal decomposition gold amalgamation atomic absorption spectroscopy (TDA-AAS) for total dissolved mercury measurement in fish sauce — a high-salt fermented condiment matrix that otherwise corrodes the metal components of TDA-AAS instruments. The validated method is then applied to ten commercial fish sauces from Vietnam (three) and Thailand (seven). Total Hg ranged from 0.8 to 42.8 µg/kg across the ten samples; every sample sat well below the EU Commission Regulation (EC) No 1881/2006 maximum of 500 µg/kg for fish products. The DGT preconcentration step achieved a ten-fold lower LOD than direct TDA-AAS and was unaffected by NaCl concentrations up to 50 g/L or pH values in the range 3–6 typical of fish sauce.

Mercury speciation note: the study measures total dissolved Hg only. It does not speciate methylmercury (MeHg) or distinguish inorganic from organic species. The companion paper Reichstädter et al. 2020 (Talanta 217, 121059), cited as reference [30], reports simultaneous Hg, Cd, and Pb determination in fish sauce by DGT and shares first-author overlap with this work.

Key numbers

Total Hg concentrations in undiluted fish sauce (n = 10 samples), measured both via DGT preconcentration (cDGT) followed by TDA-AAS analysis of the binding gels and via direct TDA-AAS of the sample (cTDA-AAS), with the calculated mass concentration in undiluted sauce (c, µg/kg) from Table 2 of the paper:

Range across the ten samples: 0.8–42.8 µg/kg. EU maximum for fish products: 500 µg/kg (Commission Regulation (EC) No 1881/2006, cited as reference [5]); every sample sat well below this limit. The three highest-Hg samples (FS7, FS9, FS10) were all Thailand-origin. Two Thailand samples (FS4, FS6) fell below the direct TDA-AAS LOD and were only quantifiable through the DGT preconcentration step. Differences between cDGT and cTDA-AAS were not statistically significant for any sample (p > 0.05). The literature comparison cited by the authors — Mabesa et al. 1985 (Philipp. J. Sci.) and Funatsu et al. 2001 (Nippon. Suisan Gakkaishi) — reported 1–90 µg/kg in various fish sauces, consistent with these findings.

Method performance (Table 1 of the paper):

• TDA-AAS: LOD 0.68 µg/L (0.068 ng Hg); LOQ 1.3 µg/L (0.13 ng Hg); relative bias 2.0%; repeatability RSDr 2.9%; reproducibility RSDR 4.3%; expanded uncertainty 8.6% (k = 2, 95% CL).
• DGT (with Purolite S924 binding gel, 24 h deployment): LOD 0.071 µg/L; LOQ 0.116 µg/L; relative bias 5.4%; repeatability RSDr 5.6%; reproducibility RSDR 8.2%; expanded uncertainty 16% (k = 2, 95% CL).

Instrument: AMA 254 (Altec, Prague, Czech Republic). Certified reference material ERM-BB422 (fish muscle, certified Hg = 0.601 ± 0.030 mg/kg, n = 6); determined value 0.589 ± 0.017 mg/kg, p = 0.3808, α = 0.05; coefficient of variation 2.9%. A second CRM, ERM-CC580 (sediment, certified Hg = 132 ± 3 mg/kg, n = 53), was run daily for long-term control charting; determined mean 130.8 ± 2.6 mg/kg, p = 0.0880. Interlaboratory wastewater proficiency exercise (PT/CHA/2/2018, CSlab Prague) achieved Z-score 0.15, well within the |Z| ≤ 2 satisfactory threshold (Commission Decision 2002/657/EC). DGT recovery in spiked 5-fold diluted fish sauce: 95% (background 2.5 ± 0.3 µg/L; spiked 5.6 ± 0.3 µg/L; DGT-determined 5.3 ± 0.4 µg/L, n = 6). Effective diffusion coefficient of Hg in 5-fold diluted fish sauce De = (3.42 ± 0.23) × 10⁻⁶ cm²/s, compared with De = 6.23 × 10⁻⁶ cm²/s in the basic model solution (the difference attributed to Hg binding to proteins and amino acids in the sauce). Diffusive boundary layer thickness 50 µm, included in DGT calculations as an extended Δg.

Estimated dietary intake (Section 3.4 of the paper): assuming 4 kg/year per-capita fish sauce consumption (the Asia–Pacific upper estimate cited by the authors), a 60 kg body weight, and the mean Hg of 11.4 µg/kg measured across the ten samples, EDI = 0.0021 µg/kg-bw/day and EWI = 0.014 µg/kg-bw/week, equivalent to 0.35% of the JECFA Provisional Tolerable Weekly Intake (PTWI) of 4 µg Hg/kg-bw/week. Using the highest measured sample (42.8 µg/kg) in the same intake assumption gives EWI = 0.055 µg/kg-bw/week. The authors note these PTWI comparisons should be read against the limitation of a single-purchase 10-sample cross-section.

Methods (brief)

Ten commercial fish sauces purchased at Asian markets in Brno (Czech Republic) and Brussels (Belgium). DGT pistons (DGT Research Ltd., Lancaster, UK) assembled per Docekalova and Diviš (2005) with a novel binding gel: 0.35 g Purolite S924 thiol-functional polystyrene resin (Purolite, Paris, France), milled from native 0.5–0.8 mm particles down to ~0.1 mm and sieved through a 0.1 mm sieve, mixed with 4 mL hot 1.5% agarose (Merck, Darmstadt, Germany) and pipetted between glass plates separated by 0.05 cm Teflon foil. Diffusive gel: 1.5% agarose prepared identically; alternative thicknesses 0.025, 0.075, 0.1 cm tested. Membrane filter (Pall, 2.5 cm, 0.014 cm thick, 0.45 µm pore) overlaid the binding/diffusive gels. Fish sauces diluted 5-fold with ultrapure water (Elga Purelab, 18 MΩ·cm), stirred 1 h in 2 L glass beakers, and four DGT pistons exposed 24 h. Binding gels analysed directly by TDA-AAS. Parallel direct TDA-AAS on undiluted sauce was run in triplicate; results converted from µg/L to mg/kg using fish sauce density 1.2 g/mL. Effective diffusion coefficient determined by time-series deployment (2, 4, 6, 8, 24 h) using Equation 2 of the paper; DBL thickness determined by the inverse-plot method across diffusive gel + membrane thicknesses 0.039, 0.064, 0.089, 0.114 cm in duplicate. pH influence tested at pH 3, 4, 5, 6; salt influence tested at 2–50 g/L NaCl. Statistical processing: XLstat (Addinsoft, New York). Limitations stated by the authors: only 10 fish sauce samples; single-purchase cross-section with no temporal replication; total Hg only — no MeHg / inorganic-Hg speciation; no certified reference material exists for fish sauce, so DGT trueness was inferred from a single spike-recovery test rather than from a matrix-matched CRM.

Implications

Regulatory comparison: every fish sauce in this cross-section sat at least an order of magnitude below the EU 500 µg/kg maximum for fish products. The wide within-set range (0.8–42.8 µg/kg, ~50× span) and the clustering of the three high-outlier values in Thailand-origin samples suggest origin- and supplier-specific variability that a small ten-sample cross-section cannot fully resolve.

Analytical methods: the paper is methodologically important for demonstrating that DGT with a Purolite S924 thiol-functional binding gel preconcentrates Hg from a high-salt fermented matrix without the corrosion penalty that direct TDA-AAS of fish sauce imposes on the instrument, and that the ten-fold LOD improvement allows quantification of samples (FS4, FS6) that direct TDA-AAS could only flag as below detection.

Speciation gap: the analyte is total dissolved Hg only. MeHg fraction in fermented fish products is reported elsewhere in the literature to be the dominant Hg species, but this paper does not measure it; downstream uses of these numbers should be flagged as tHg, not as MeHg.

Wiki pages updated on ingest

• mercury-total
• fish-sauce
• anchovy
• fish-sauce
• condiments-general

Verification notes

Merge-enhanced 2026-05-28 from prior 2026-05-14 revision (cite-key, raw_handle, products/ingredients/matrices/jurisdictions arrays, near_duplicates, and license preserved verbatim). Changes:

1. raw_path corrected — prior revision truncated the filename to “…Thermal Decomposition .pdf”; actual on-disk filename is “Determination of Mercury in Fish Sauces by Thermal Decomposition Amalgamation Atomic Absorption Spec.pdf”. The truncated path would not resolve.
2. access_url added — present-schema sources carry a resolvable URL alongside the DOI.
3. sample_population reworded to surface the Vietnam-3 / Thailand-7 split and the per-sample anchovy fraction (range 55–97% for nine samples, n.a. for FS8) that the prior revision elided.
4. Key numbers table expanded to include cDGT and cTDA-AAS columns from Table 2 of the paper, not just the calculated mass-concentration column c (µg/kg). Both methods are reported in the source; reproducing only c discards the methods-comparison information the paper is built around.
5. Method-performance metrics expanded from LOD-only to the full Table 1 row (LOD, LOQ, RSDr, RSDR, relative bias, expanded uncertainty for both TDA-AAS and DGT) and added the ERM-CC580 sediment CRM daily-control numbers and the interlaboratory PT/CHA/2/2018 Z-score that the prior revision omitted.
6. Effective diffusion coefficient De = (3.42 ± 0.23) × 10⁻⁶ cm²/s and DBL = 50 µm added to Key numbers; these are the DGT-specific validation outputs that future methods-development reads will care about.
7. EDI and highest-sample EWI numbers added from Section 3.4. Prior revision included only the mean-based EWI.
8. Implications section trimmed to remove an HMT&C-program scope speculation (“For HMT&C, fish sauce is not currently a primary product category but is relevant to condiment and fermented-food modules”) that crossed the wiki/HMT&C firewall per CLAUDE.md Part 2. Also removed prior ## Implications → Courses and → App audience translations, which are audience-rendering decisions made downstream of the source page, not source-page content.
9. Methods (brief) expanded with the Purolite S924 milling protocol (0.5–0.8 mm milled to ~0.1 mm), agarose gel preparation, DGT piston / membrane filter vendor detail, 5-fold dilution and 24 h deployment timing, the diffusive-gel thicknesses tested for DBL determination, the XLstat statistical package, and the four limitations stated by the authors (sample size, single-purchase, no speciation, no fish-sauce CRM). Vendor-name retention follows the methods-vendor carve-out in docs/gpt-collaboration/verification-checklist.md §4.
10. Wiki pages updated on ingest expanded to include [[metals/mercury-total]] and the two product slugs, both of which the source’s frontmatter declares but the prior list did not surface.

No brand names appear in either the existing page or the source paper itself — fish sauces are coded FS1…FS10 throughout the paper with only origin country and label-stated anchovy fraction, so no Part 12 redaction was required.

Audit subagent (2026-05-28, fresh-context general-purpose) returned PROMOTE: numerical fidelity ✅ (full Table 1 / Table 2 / Section 3.4 cross-check passed), speciation and methods ✅, Part 12 brand firewall ✅, Part 2 wiki/HMTc firewall ✅. The one ⚠️ concern was that ingredients/fish-sauce, ingredients/anchovy, and products/fish-sauce are not yet entries in docs/gpt-collaboration/taxonomy-snapshot.md. Verified: the routing audit (npm run evidence:source-routes) resolves all three under the provisional-scaffold mechanism per CLAUDE.md Part 10, with zero unresolved and zero malformed entries for this source. Slugs are descriptively accurate to the paper (fish sauce as the analyte matrix; anchovy as the dominant fish source in 9 of 10 samples; fish-sauce as the product form distinct from the umbrella condiments-general row). Retained without change.

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.