HMTc Category 5 — Beverages — Master Summary
Staff content — not public
This page is the staff Standards Workbench cross-product view of HMT&C Category 5. It carries threshold-selection arithmetic (clean / dirty designation, p30, p90, p100, n_a_tier, CC eligibility) for every (subcategory × analyte) cell. Per OPERATING.md Part 2 and CLAUDE.md Part 19, this content is HMT&C certification work product and does not appear on the public per-product pages at heavymetalindex.com; those pages carry only the literature-native view. The page is marked noindex: true for that reason.
Public readers landing here from a bookmark or stale link should go to the wiki index for the public navigation or to a specific beverage product page in the covers list above for the literature-native view of that subcategory.
This master summary mirrors the HMTc Category 1 master structure but for the locked Category 5 beverage architecture (16 subcategories from category-5-beverages.md Step 0 lock plus the regulatory-crosswalk-field-findings synthesis page). The locked rows preserve the contamination-platform vs clean-benchmark splits established in the Step 0 process and are not renumbered here.
The 16 product subcategories × 10 HMT&C analytes = 160 product cells. The 17th page (regulatory-crosswalk-field-findings) is a synthesis page rather than a per-cell readiness row and is not included in the percentile-readiness math; it routes regulatory cross-walk evidence (FDA juice action levels, EU 2023/915, Codex CCCF) across the 16 subcategories as a horizontal layer.
The readiness bar is the same as Cat 1 per CLAUDE.md Part 6 + Part 19: each cell is at-bar (Path A with n_a_tier ≥ 2 and confidence ≥ medium), approaching, Path A thin (n_a_tier=1 sample-level), Path B candidate (5×LOQ available), or explicit data gap with rationale.
Subcategory readiness roll-up across the 160 cells
| Subcategory | At bar | Approaching | Path A thin (n_a_tier=1) | Data gap |
|---|---|---|---|---|
| Fruit juices, non-apple | Pb (publishable: dirty p30=0 ppb, p90=14.9 ppb sample-level n=993 from FDA 2018) | iAs (grape p30=7.8 ppb sample-level Path A thin; partial-fit caveat) | Cd, tAs, Ni | MeHg, tHg, Al, Cr-VI, Sn |
| Fruit juices, apple-containing | iAs (publishable: dirty p30=3.0 ppb), Pb (publishable: clean p90=6.0 ppb sample-level n=650 from FDA 2018) | — | Cd, tAs, Ni | MeHg, tHg, Al, Cr-VI, Sn |
| Vegetable juices, non-root | — | — | — | Pb, Cd, tAs, iAs, MeHg, tHg, Ni, Al, Cr-VI, Sn (all 10) |
| Vegetable juices, root-vegetable-containing | — | — | — | Pb, Cd, tAs, iAs, MeHg, tHg, Ni, Al, Cr-VI, Sn (all 10) |
| Plant milks, non-soy-non-rice | — | — | Pb, tHg, Ni | Cd, tAs, iAs, MeHg, Al, Cr-VI, Sn |
| Plant milks, soy-based | — | Pb, Ni, Al | tHg | Cd, tAs (Path B), iAs, MeHg, Cr-VI, Sn (Path B) |
| Plant milks, rice-based | — | iAs (summary-level p30≈12 ppb < EU 30 cap), Pb, tHg, tAs | Ni | Cd, MeHg, Al, Cr-VI, Sn |
| Flavored waters | — | — | Sn (canned context) | Pb, Cd, tAs, iAs, MeHg, tHg, Ni, Al, Cr-VI |
| Sports/energy drinks | — | — | — | All 10 |
| Herbal/botanical infusions | — | — | — | All 10 |
| True tea (Camellia sinensis) | — | — | — | All 10 |
| Matcha | — | — | — | All 10 |
| Coffee | — | — | — | All 10 |
| Soft drinks/carbonated beverages | — | — | Sn (canned context) | Pb, Cd, tAs, iAs, MeHg, tHg, Ni, Al, Cr-VI |
| Fermented beverages, non-tea-based | — | — | — | All 10 |
| Kombucha (tea-based fermented) | — | — | — | All 10 |
Counts (160 cells total):
- At readiness bar: 3 cells. (cycle 21 update: +2 cells via FDA 2018 longitudinal Pb sample-level extraction)
- iAs in fruit-juices-apple-containing — Path A sample-level dirty p30 = 3.0 ppb (FDA 2011 n=94 + FDA 2016 grape cross-juice context). Cycle 13 build (commit b789bc1).
- Pb in fruit-juices-apple-containing — Path A sample-level clean p90 = 6.0 ppb (FDA 2018 longitudinal n=650 + FDA 2022 TDS n=3). p90 40% below FDA 2022 draft 10 ppb apple-juice cap. Cycle 21 build (this commit).
- Pb in fruit-juices-non-apple — Path A sample-level dirty p30 = 0 ppb, p90 = 14.9 ppb (FDA 2018 longitudinal n=993 + FDA 2022 TDS grapefruit n=3). p95=20.7 right at FDA 2022 draft 20 ppb other-juice cap. Cycle 21 build (this commit).
- Approaching the bar: 15 cells. iAs in plant-milks-rice-based (Da Mato 2026 summary p30≈12 ppb < EU 30 cap; awaits sample-level extraction or 3rd source for at-bar). tAs and tHg in plant-milks-rice-based (Da Mato 2026 + Marques 2021). Fruit-juice Pb across both subcategories (FDA 2018 longitudinal compliance + FDA 2022 draft action levels + Weldegebriel 2025 Ethiopia). Plant-milk-soy Pb/Cd/tAs/Ni/Al/Sn (Milani 2023 multi-element + Marques 2021). Plant-milk-non-soy-non-rice Pb/tHg/Ni (Marques 2021 ICP-MS).
- Path A thin (n_a_tier=1): 7 cells. Mostly Cd/tAs/iAs/Ni in fruit juices and tAs in rice plant-milk; plus Sn in flavored-water and soft-drink-canned subcategories from canned-beverage Sn-leach evidence (Tarigan 2016 + Benoy 1971 + ATSDR 2005).
- Data gap: 139 cells. Dominant gap drivers: (a) Tea, matcha, coffee, kombucha, sports/energy drinks, soft drinks, fermented beverages, herbal infusions — 8 subcategories × 10 analytes = 80 cells with no direct in-corpus evidence in
wiki/sources/. (b) Vegetable juices (both splits) — 20 cells with no direct evidence. (c) MeHg and Cr-VI systematically gap across all 16 subcategories (Cr-VI by chemistry-mechanism inference from Hernandez 2019 + Saraiva 2021 carries forward to milk-like beverages but has not been extended to acidic/aqueous beverage matrices in the corpus).
Per Karen’s readiness bar, every gap cell carries a documented rationale below; the data-gap-wishlist items will populate as ingest cycles continue.
Per-analyte tables
The per-cell percentile values, n, n_a_tier, confidence, and CC eligibility for each of the 160 cells will populate in subsequent cycles as the per-page CC blocks are built. This skeleton commits the readiness roll-up + Source Legend now so subsequent cycles can fill in cell-level numbers without re-introducing schema drift.
Pb (lead) — Path A summary
| Subcategory | Cell state | Best evidence | n_a_tier | Notes |
|---|---|---|---|---|
| Fruit juices, non-apple | Approaching | FDA 2018 longitudinal Pb in juice FY2005-FY2018 + FDA 2022 draft action level (20 ppb other juice) + FDA 2022 TDS + Weldegebriel 2025 (n=80 Ethiopia) | 3 | Sample-level distribution feasible from FDA 2018 + Weldegebriel 2025; FDA 2022 draft 20 ppb cap applies. |
| Fruit juices, apple-containing | Approaching | FDA 2018 + FDA 2022 draft (10 ppb apple juice cap) + Weldegebriel 2025 | 3 | Apple subset has a stricter 10-ppb draft cap; sample-level distribution feasible. |
| Vegetable juices, non-root | data gap | — | 0 | Veg-juice-specific Pb data not in corpus. Closing requires a vegetable-juice multi-element survey. |
| Vegetable juices, root-vegetable-containing | data gap | — | 0 | Same gap; root-veg-containing subset would inherit ingredient-cascade Pb context from root-vegetable analyses. |
| Plant milks, non-soy-non-rice | Path A thin | Marques 2021 (Pb detected in 1 non-organic oat drink composite among small subset; numeric value pending source-table review per extraction-quality caveat) | 1 | Single-source clean-benchmark cell with composite-sample design. Closing requires sample-level Pb extraction from 2nd source. |
| Plant milks, soy-based | Approaching | Milani 2023 (Pb mean 2.2 ppb in soybean group, <10.9 ppb LOQ in isolate-protein and hydrosoluble groups; soybean max 13 ppb) + Marques 2021 | 2 | Path A summary-level. Closing requires sample-level Pb from lower-LOD method. |
| Plant milks, rice-based | Approaching | Marques 2021 + Da Mato 2026 (Pb co-measured) | 2 | Sample-level distribution feasible from Marques 2021 subset + Da Mato 2026 co-measured Pb. |
| Flavored waters | data gap | — | 0 | Sn context from canned-beverage corpus does not extend to Pb in this subcategory. |
| Sports/energy drinks | data gap | — | 0 | No sports/energy drink Pb data in corpus. |
| Herbal/botanical infusions | data gap | — | 0 | Closing requires an herbal-tea multi-element survey. |
| True tea (Camellia sinensis) | data gap | — | 0 | Tea-leaf Al/Pb data exists in broader literature (e.g., Karak & Bhagat) but not yet ingested into corpus. Wishlist item. |
| Matcha | data gap | — | 0 | Matcha-specific Pb data sparse; whole-leaf consumption changes the exposure model vs steeped tea. Wishlist item. |
| Coffee | data gap | — | 0 | Coffee Pb data not in corpus. Brewed coffee typically shows lower Pb than the bean (extraction efficiency varies). Wishlist item. |
| Soft drinks/carbonated beverages | data gap | — | 0 | Pb in soft drinks expected to be very low (formulated beverages with treated water + sugar/syrup); historical lead-soldered-can era pre-1990 only. |
| Fermented beverages, non-tea-based | data gap | — | 0 | Includes water kefir, jun, beverage-tibet — niche corpus. Wishlist item. |
| Kombucha (tea-based fermented) | data gap | — | 0 | Tea-base contamination cascades through; needs kombucha-specific Pb measurement to close cell. |
Cd (cadmium) — Path A summary
| Subcategory | Cell state | Best evidence | n_a_tier | Notes |
|---|---|---|---|---|
| Fruit juices, non-apple | Path A thin | FDA 2022 TDS Cd in juice subset + Weldegebriel 2025 | 1-2 | Sample-level Cd distribution achievable. |
| Fruit juices, apple-containing | Path A thin | FDA 2022 TDS apple-juice subset + Weldegebriel 2025 | 1 | One A-tier sample-level source plus FDA TDS coverage; second source needed. |
| Vegetable juices, non-root + root | data gap | — | 0 | Root-veg Cd ingredient cascade exists at HMTc Cat 1 (root-veg purees) but not yet extended to root-veg juice subcategory. |
| Plant milks, non-soy-non-rice | data gap | — | 0 | Marques 2021 did not measure Cd. |
| Plant milks, soy-based | Path B candidate | Milani 2023 reports all 18 samples <LOQ (3.8 ppb ICP-OES); Path B 5×LOQ = 19 ppb upper bound | 1 (Path B) | Closing requires sample-level Cd from a lower-LOD method (ICP-MS sub-ppb LOQ). |
| Plant milks, rice-based | data gap | — | 0 | Da Mato 2026 measured iAs only; corpus lacks rice-plant-milk Cd sample-level. |
| Flavored waters, sports/energy, herbal infusions, tea, matcha, coffee, soft drinks, fermented, kombucha | data gap | — | 0 | All 9 subcategories lack direct Cd evidence in corpus. |
tAs (total arsenic) — Path A summary
| Subcategory | Cell state | Best evidence | n_a_tier | Notes |
|---|---|---|---|---|
| Fruit juices, non-apple | Path A thin | Da Mato 2026 (covers tAs alongside iAs) + FDA 2022 TDS | 1 | tAs co-measured. |
| Fruit juices, apple-containing | Path A thin | FDA 2011 apple juice tAs co-measurement + FDA 2016 Juice-Grape | 2 | Strong As corpus for apple-containing juice. |
| Plant milks, soy-based | Path B candidate | Milani 2023 reports all 18 samples <LOQ (38.2 ppb ICP-OES — high vs ICP-MS); Path B 5×LOQ = 191 ppb | 1 (Path B) | Closing requires lower-LOD method. |
| Plant milks, rice-based | Path A thin | Da Mato 2026 + Gu 2020 rice-milk-powder | 2 | tAs co-measured with iAs across two A-tier sources. |
| All other subcategories | data gap | — | 0 | — |
iAs (inorganic arsenic) — Path A summary
| Subcategory | Cell state | Best evidence | n_a_tier | Notes |
|---|---|---|---|---|
| Fruit juices, non-apple | Path A thin (sample-level computable) | FDA 2016 grape and grape-containing juice category n=58 quantified | 1 | Path A thin (n_a_tier=1, sample-level). Computed grape-category sample-level distribution: p30=7.8, p50=11.2, p90=22.6, p100=49.6 ppb. Empirical override of variant_type label: despite being clean_benchmark per locked architecture, iAs is empirically dirty in non-apple juice — grape-category iAs (mean 12.9 ppb) is 2-5× apple-juice iAs (mean 4.1 ppb). p30=7.8 ppb is 61% below EU 20 ppb cap. Partial-fit caveat (FDA 2016 covers grape only; broader non-apple-juice population includes lower-iAs orange/grapefruit/cranberry). Closing to at-bar requires 2nd non-apple-juice iAs source covering more juice types. |
| Fruit juices, apple-containing | At bar | FDA 2011 single-strength apple juice (n=94) + FDA 2016 (grape-and-grape-containing subset for cross-juice context) + FDA CTZ 10 ppb cap | 2 | At bar (cycle 13). Per Part 19 dirty designation: dirty p30 = 3.0 ppb (FDA 2011 lower-bound deterministic; TR values treated as 0). Sample-level CSV at data/evidence/category5_apple_juice_arsenic_speciation_*.csv. p100 = 9.8 ppb. p30 is 70% below FDA CTZ 10 ppb cap. Per-page CC block rebuilt at fruit-juices-apple-containing. |
| Plant milks, rice-based | Approaching | Da Mato 2026 (n=25 EU rice-based drinks; summary-level mean=15, median=15, range 7-24 ppb) + Gu 2020 (rice-milk-powder n=3 triangulation, below 10-sample floor) + EU 2023 30 ppb cap | 2 | Path A summary-level (no per-sample distribution from Da Mato 2026 published). Estimated dirty p30 ≈ 12 ppb (range-based); p100 = 24 ppb. Cell is approaching, not at-bar — confidence is low (2 studies; Part 6 medium = 3+). Closing requires either sample-level extraction from Da Mato 2026 supplementary tables or a 3rd direct rice-plant-milk iAs source. |
| All other subcategories | data gap | — | 0 | iAs in non-rice plant-milks, vegetable juices, tea/coffee/soft-drinks/etc. not in corpus. |
MeHg (methylmercury)
All 16 subcategories: data gap. MeHg is rarely measured in beverage matrices — most exposure routes through fish-containing foods (Cat 1 fish baby foods n_a_tier=1 from Tatsuta 2024). The Brombach 2017 ingredient-cascade evidence for rice MeHg could route to plant-milks-rice-based via the rice-ingredient cascade, but rice-drink-specific MeHg measurement not yet in corpus. Closing requires MeHg-in-beverage speciation papers.
tHg (total mercury)
| Subcategory | Cell state | Best evidence | n_a_tier | Notes |
|---|---|---|---|---|
| Plant milks, non-soy-non-rice | Path A thin | Marques 2021 reports tHg not detected across all milks and plant-based drinks (clean baseline) | 1 | Path A thin clean baseline; cleanest Hg evidence in Cat 5 plant milks. |
| Plant milks, rice-based | Approaching | Marques 2021 + Brombach 2017 ingredient cascade | 2 | Marques 2021 measured tHg directly; Brombach 2017 provides rice-ingredient tHg context. |
| All other subcategories | data gap | — | 0 | tHg in fruit juices, soft drinks, tea, coffee not in corpus. |
Ni (nickel)
| Subcategory | Cell state | Best evidence | n_a_tier | Notes |
|---|---|---|---|---|
| Fruit juices, non-apple + apple-containing | Path A thin | Weldegebriel 2025 Ethiopia n=80 Ni | 1 | One A-tier source; second needed. |
| Plant milks, non-soy-non-rice | Path A thin | Marques 2021 Ni measured in plant-drinks panel; almond+oat composite subset values pending source-table review | 1 | Path A thin single-source. |
| Plant milks, soy-based | Approaching | Milani 2023 (group means; soybean group mean 29 ppb, max 46 ppb; isolate/hydrosoluble groups <25.7 ppb LOQ) + Marques 2021 | 2 | Path A summary-level approaching. Closing requires sample-level extraction or 3rd source. |
| All other subcategories | data gap | — | 0 | — |
Al (aluminum)
| Subcategory | Cell state | Best evidence | n_a_tier | Notes |
|---|---|---|---|---|
| Plant milks, soy-based | Path A thin | Milani 2023 group means by soy-source: isolate-protein 758 ppb, hydrosoluble-extract 609 ppb, soybean 176 ppb; range 45-1822 ppb across all groups | 1 | Path A thin (n_a_tier=1, single source). Group-mean structure shows clear soy-source dependence: isolate-protein and hydrosoluble extract beverages have higher Al than whole-soybean beverages, consistent with the processing step concentrating Al. Closing to approaching requires a 2nd direct soy-beverage Al source. |
| All other subcategories | data gap | — | 0 | True tea is a known Al accumulator (~hundreds of mg/kg in dry leaf, 1-10 mg/L in steeped tea per literature) but tea-Al data is not yet in corpus — wishlist item with high regulatory pull. |
Cr-VI (hexavalent chromium)
All 16 subcategories: data gap by direct measurement. Per CLAUDE.md Part 14 + Hernandez 2019 + Saraiva 2021 chemistry-mechanism finding, food-matrix Cr is essentially Cr(III); the milk-like-beverage subset (plant milks soy/rice/non-soy) inherits the Cr(VI) <LOD chemistry-mechanism inference (n_a_tier=1). For acidic-aqueous-matrix beverages (juices, soft drinks, sports/energy drinks, kombucha), the Cr(VI)-stability question is less settled — acidic conditions can stabilize Cr(VI) more than neutral-alkaline matrices. Closing requires beverage-matrix-specific Cr speciation. Wishlist: Mathebula 2019 (PMID 30931809; OA via PMC) covers milk + dairy + cereal but not beverages.
Sn (tin)
| Subcategory | Cell state | Best evidence | n_a_tier | Notes |
|---|---|---|---|---|
| Fruit juices (canned subset) | Path A thin | Benoy 1971 (canned fruit juice Sn toxicity 1971) + ATSDR 2005 Sn toxprofile + Tarigan 2016 (factors-tin-released-canned-beverages) | 1 | Historical canned-juice Sn corpus relevant to canned-juice subset only. Modern aluminum/aseptic packaging has largely supplanted Sn-lined cans for juice. |
| Plant milks, soy-based | Path B candidate | Milani 2023 isolate-protein group mean 4.3 ppb (LOD-zero substitution), max 26 ppb; other 2 groups <18 ppb LOQ; Path B 5×LOQ = 90 ppb | 1 (Path B) | Modern soy plant-milk is aseptic-pack; canned-Sn-leach context not applicable. |
| Soft drinks (canned) + Flavored waters (canned) | Path A thin | Tarigan 2016 + Benoy 1971 cascade | 1 | Same canned-beverage Sn-leach context. |
| All other subcategories | data gap | — | 0 | — |
Source Legend (Cat 5 primary corpus)
The numbered citations route here in IEEE-style. Per-cell numbering is omitted in this skeleton commit and will populate as cycle 13+ work fills in per-page CC blocks.
Primary A-tier sources contributing to multiple cells in this Cat 5 master:
- fda2011-apple-juice-arsenic-speciation — FDA 2011 single-strength apple juice; sample-level iAs n=94.
- fda2016-infant-toddler-foods-inorganic-arsenic — FDA 2016 inorganic arsenic in infant/toddler foods; Juice-Grape subset n=61 routes to fruit-juices-apple-containing (grape product) and fruit-juices-non-apple subsets.
- fda2018-lead-in-juice-fy2005-fy2018 — FDA 2018 longitudinal Pb in juice FY2005-FY2018; covers fruit-juices-non-apple, fruit-juices-apple-containing, fruit-juice-not-canned subsets.
- fda2022-tds-elements-fy2018-fy2020 — FDA TDS FY2018-FY2020 elements; Pb/Cd/tAs/Ni in juice subset.
- fda2022-draft-lead-juice — FDA 2022 draft action level Pb in juice (10 ppb apple, 20 ppb other juice).
- fda2004-juice-haccp-lead — FDA 2004 juice HACCP rule (50 ppb Pb historical).
- weldegebriel2025-ethiopia-packaged-fruit-juice-metals — Ethiopia n=80 packaged fruit juices; Pb/Cd/Cr/Ni sample-level.
- damato2026-inorganic-arsenic-rice-based-beverages — Italy n=20 EU rice-based drinks; iAs/tAs sample-level. Direct primary source for plant-milks-rice-based iAs at-bar status.
- marques2021-trace-elements-milks-plant-based-drinks — Spain n=42 cow/goat milk + soy/almond/rice/oat plant-based drinks + infant formulas; ICP-MS Pb, tHg, Ni, U sample-level. Primary source for plant-milks-soy/rice/non-soy-non-rice across multiple cells.
- milani2023-trace-elements-soy-based-beverages — Brazil n=18 soy beverages; Al/tAs/Cd/Cr/Ni/Pb/Sb/Sn sample-level multi-element. Primary source for plant-milks-soy-based across most analytes.
- gu2020-arsenic-rice-infant-food-australia — Australia n=39 rice-based foods including rice-milk-powder subset; tAs/iAs. Plant-milks-rice-based ingredient-cascade evidence.
- brombach2017-methylmercury-european-commercial-rice — European rice n=87 incl. rice-product subsets; tHg/MeHg ingredient cascade for rice-plant-milk subcategory.
- signes-pastor2016-inorganic-arsenic-rice-products-infants — Rice products iAs ingredient cascade.
- benoy1971-tin-toxicity-canned-fruit-juices — Historical Sn corpus for canned-juice subset.
- atsdr2005-tin-toxicological-profile — Sn toxicology reference; canned-beverage Sn-leach context.
- tarigan2016-factors-tin-released-canned-beverages — Factors driving Sn release from canned beverages; n_a_tier=1 across canned subcategories (canned juice, canned soft drinks, flavored-water canned subset).
Regulatory references (Cat 5):
- fda2023-ias-apple-juice-10 — FDA CTZ apple-juice iAs 10 ppb.
- fda2022-draft-pb-apple-juice-10 — FDA draft Pb apple-juice 10 ppb.
- fda2022-draft-pb-other-juice-20 — FDA draft Pb other-juice 20 ppb.
- fda2004-juice-haccp-pb-50 — FDA 2004 historical HACCP 50 ppb.
- eu2023-arsenic-rice-based-drinks — EU 2023/915 iAs 30 ppb in rice-based drinks.
Key data gaps and what would close them
| Gap pattern | Cells affected | What would close it |
|---|---|---|
| Tea (true tea, matcha, herbal infusions, kombucha tea-base) | 40 cells (4 subcategories × 10 analytes) | Tea-leaf and steeped-tea multi-element surveys. Tea-leaf Al is well-known (~hundreds mg/kg dry leaf); high regulatory pull. Karak & Bhagat 2010 type review + primary occurrence papers. |
| Coffee | 10 cells | Coffee multi-element survey. Brewed coffee Pb/Cd/As typically lower than ground coffee; needs both. |
| Soft drinks + sports/energy drinks | 20 cells | Multi-element soft-drink/sports-drink survey. FDA TDS may have partial coverage. |
| Vegetable juices (non-root + root) | 20 cells | Vegetable-juice multi-element survey. Root-veg ingredient cascade exists at Cat 1 root-veg purees but not yet extended to juice form. |
| Flavored waters (beyond canned-Sn context) | 9 cells | Flavored-water multi-element survey. |
| Fermented beverages (non-tea) | 10 cells | Water kefir, jun, beverage-tibet multi-element survey. Niche corpus. |
| MeHg in any Cat 5 subcategory | 16 cells | MeHg-in-beverage speciation paper. Brombach 2017 covers rice-ingredient MeHg only; needs beverage-matrix MeHg measurement. |
| Cr-VI in acidic-aqueous beverages | 16 cells (chemistry-mechanism inference for milk-like beverages partially carries; acidic-aqueous beverages need direct measurement) | Cr speciation in juice/soft-drink/kombucha matrices. Acidic conditions can stabilize Cr(VI) more than neutral-alkaline matrices. |
Geographic-context flag
The Cat 5 corpus skews heavily U.S.+EU+Australia: FDA 2011/2016/2018/2022 are U.S.; Da Mato 2026 + Marques 2021 are EU; Milani 2023 is Brazilian; Gu 2020 is Australian; Weldegebriel 2025 is Ethiopian. Per CLAUDE.md Part 6 jurisdiction-mix labeling, any Cat 5 mixed-jurisdiction p90 must document composition. The Ethiopian + Brazilian sources expand beyond the U.S./EU baseline and are particularly valuable for the global-applicability dimension.
Phase / cycle reference
This Cat 5 master skeleton is committed at autonomous-loop cycle 12 (commit pending). Subsequent cycles will:
- Cycle 13+: Begin per-page CC candidate block builds for the at-bar and approaching subcategories (fruit-juices-apple-containing, plant-milks-rice-based, plant-milks-soy-based, plant-milks-non-soy-non-rice, fruit-juices-non-apple). Then move to the wishlist papers for the tea/coffee/soft-drink subcategories as ingest opportunities arise.
- Per CLAUDE.md Part 19: each per-page CC block carries clean/dirty designation per analyte, p90 (clean) or p30 (dirty) values, and explicit data-gap rationale where applicable.
Coordination payload for Cowork (Heavy Metals in Beverages Standards Briefing)
Refreshed 2026-05-11 (cycle 34). Cowork’s hmtc-standards skill builds the Standards Briefing as a .docx. 3 Cat 5 cells at the readiness bar (up from 2 at cycle-20 baseline; +1 from cycle 21 FDA 2018 juice Pb extraction):
| # | Subcategory | Analyte | Standard target | n | n_a_tier | Regulatory cap | Headroom |
|---|---|---|---|---|---|---|---|
| 1 | Fruit juices, apple-containing | iAs (dirty) | dirty p30 = 3.0 ppb (FDA 2011 lower-bound; TR=0) | 94 | 2 | 10 ppb FDA CTZ apple juice; 20 ppb EU 2023/915 | 70% below FDA cap |
| 2 | Fruit juices, apple-containing | Pb (clean) | clean p90 = 6.0 ppb (FDA 2018 longitudinal extraction cycle 21) | 650 | 2 | 10 ppb FDA 2022 draft apple juice | 40% below cap |
| 3 | Fruit juices, non-apple | Pb (dirty) | dirty p30 = 0 ppb / p90 = 14.9 ppb / p95 = 20.7 ppb (FDA 2018 longitudinal n=993) | 993 | 2 | 20 ppb FDA 2022 draft other-juice; 30 ppb EU 2023/915 | p95 right at FDA cap; passion-fruit + pomegranate + prune + beet drive upper tail |
Approaching cells (12) — next readiness targets
| Subcategory | Analytes approaching | Closing requirement |
|---|---|---|
| Plant milks, rice-based | iAs (summary p30≈12 ppb < EU 30 cap), Pb, tHg, tAs | Sample-level extraction from Da Mato 2026 SI; or 3rd direct rice-plant-milk iAs source |
| Plant milks, soy-based | Pb, Ni, Al (n=18 sample-level Milani 2023 with Al p100=1822 ppb) | 2nd direct soy-plant-milk multi-element source |
| Plant milks, non-soy-non-rice | Pb, tHg, Ni (Marques 2021 composites with extraction caveat) | Sample-level extraction or 2nd direct source |
| Fruit juices, non-apple | iAs (FDA 2016 grape p30=7.8 ppb sample-level Path A thin) | 2nd direct non-apple-juice iAs source (orange, cranberry, pomegranate) |
Recommended Briefing structure
- Lead with the fruit-juice Pb pair (apple clean p90=6.0; non-apple dirty p30=0/p95=20.7): the empirical apple-vs-non-apple variance is the strongest contamination-platform comparison in the Cat 5 publishable set; Part 19 empirical override of variant_type label is the structural finding.
- Apple-juice iAs as the cleanest cell: 3.0 ppb 70% below FDA CTZ — most regulatory headroom.
- Non-apple Pb upper-tail story: passion fruit (134.8 ppb max), pomegranate (115 ppb), beet (79 ppb) — practical retail-monitoring guidance.
- Plant-milk Al concern: Milani 2023 n=18 soy beverages mean 758 ppb (isolate-protein); regulatory-pull data point for soy-formula-substitute consumers.
Wiki Pages Updated On Ingest
- category-5-beverages (existing hub)
- All 16 Cat 5 subcategory pages (per-page CC block work deferred to subsequent cycles)
- regulatory-crosswalk-field-findings (synthesis page)