Synthesis

This page is the index of the Heavy Metal Index’s synthesis layer. Each linked subpage is a self-contained synthesis findings page: a thesis statement built on at least three corroborating A-tier sources, presenting a non-obvious finding that emerges from the corpus and would not be evident from any single source page. Synthesis pages sit one layer above the source and product pages and one layer below the strategic / certification documents, and the wiki’s epistemic authority rests on them remaining provisional, source-traceable, and explicit about their own uncertainty.

Synthesis findings pages

• Climate adaptation creates new heavy-metal trade-offs — established 2026-05-13. Rice irrigation drives an iAs/Cd inverse correlation that climate-smart practices are accelerating; climate change is projected to increase methylmercury in freshwater wild fish by approximately 60 percent by 2031-2060. Three anchor studies (Seyfferth 2025, Limmer 2024, Wu 2025). Implications for certification standards calibrated to historical baselines.
• Subsistence fishing methylmercury is a global vulnerable-population pattern — established 2026-05-13. Regulatory fish-consumption advice is calibrated against typical Western market consumer patterns while subsistence fishing populations globally face MeHg exposures routinely exceeding regulatory thresholds, with climate change amplifying the disparity. Three anchor studies (Li 2024 PNAS, Wu 2025 PNAS, Auzier-Guimarães 2025). Implications for HMTc seafood standard population-of-reference framing.
• Single-metal regulatory limits systematically underprotect against dietary heavy-metal exposure — established 2026-05-13. Three independent assessments of cumulative or single-metal dietary exposure (Suomi 2023 Finnish HI greater than 1; JECFA 91st cocoa-driven Cd at 96 percent of PTMI in under-12 children; EFSA 2024 iAs MOEs below 1 across adult European range) converge on the architectural finding that single-metal regulatory floors do not aggregate to a health-protective outcome. Scientific basis for HMTc’s combined-exposure framing.
• US chocolate lead has measurably declined since 2015; cadmium has not — established 2026-05-13. Three anchor studies (Abt 2018 US baseline; Hands 2024 multi-year trajectory; Zhao 2024 global mechanistic context) document a Prop-65-aligned decline in US chocolate Pb while cadmium has not moved equivalently. Mechanistic asymmetry: lead is processing-side and intervention-responsive; cadmium is soil-side and responds only to upstream agronomic intervention or sourcing-region shifts.

When to promote a synthesis findings page

A synthesis findings page is warranted when a thematic cluster in the corpus meets all of the following criteria.

The finding rests on at least three independent A-tier or high-quality B-tier sources. Two sources are an interesting coincidence; three is the floor for treating the convergence as evidence of a real pattern rather than an artifact of any single research group’s framing. The sources should be independent in the sense of distinct authorship, distinct institutional context, and ideally distinct geographies.

The finding spans multiple metals, ingredients, or products. Findings that fit cleanly inside one existing page belong on that page, not in synthesis. Synthesis exists for claims whose support and implications span boundaries the existing taxonomy was not built to handle: a finding about climate effects on multiple food matrices, a finding about how regulatory bodies disagree across multiple metals, a finding about a vulnerable-population exposure pattern that aggregates across products and ingredients.

The finding is not obvious from any single source. A synthesis page is doing work that no source page can do on its own. If the finding is just “rice contains arsenic” or “cocoa contains cadmium,” that belongs in the relevant ingredient page’s body, not in synthesis. Synthesis findings are the connections, contradictions, and trajectories that emerge when the corpus is read as a whole.

The finding has strategic relevance to the wiki’s audiences. The wiki’s brand-legal, regulatory, and educational audiences are different from each other, but a synthesis finding should serve at least one of them in a load-bearing way: it should change how a defense counsel reads the wiki, how a regulator weighs the evidence, or how a course module is built. Findings that are intellectually interesting but operationally inert can live as notes on source pages or in the log; synthesis pages are reserved for findings that have downstream consequences.

When all four criteria are met, the candidate thread becomes a synthesis proposal. Proposals are surfaced from three places in the workflow.

First, batch reports. Every batch close commit produced by an ingest session should include a Synthesis threads note section listing thematic clusters the batch contributed to or established. When the same thread appears in three consecutive batch reports, it is automatically a promotion candidate.

Second, orchestrator scans. The orchestrator at tools/orchestrator/analyze-gaps.mjs includes a PROPOSE-SYNTHESIS-PAGES signal (in development) that periodically scans the routing audit and recent log entries for thematic clusters. The signal does not write the synthesis page; it surfaces the candidate to Karen’s queue at data/evidence/synthesis-proposals.md.

Third, ad-hoc curator decisions. Karen or any session reading the corpus may identify a promotion candidate at any time. The threshold is the four criteria above, not the surfacing mechanism.

A promoted candidate becomes a synthesis page through one of two paths. The fast path is a Claude session writing the page directly using the anchor sources, the routing audit, and CLAUDE.md Part 15 prose conventions. The slower path is a dedicated subprocess dispatch (SYNTHESIZE-<topic-slug>) which the orchestrator can run autonomously with the anchor sources as input.

Either path produces the page at wiki/synthesis/<topic-slug>.md with the synthesis-page frontmatter and an entry in the index above. The original Gaps and load-bearing questions surfaced during ingest section below preserves the older gap-tracking pattern; gaps that mature into synthesis-page-worthy findings should migrate to dedicated subpages over time. The page is provisional until at least two corroborating studies enter the corpus after the initial promotion, at which point it should be re-synthesized.

The cadmium synthesis (in flight)

The first substantive synthesis entry will cover cadmium in food, integrating the regulatory tox documents currently in ingest (EFSA 2009, JECFA 2010, EPA IRIS, ATSDR, OEHHA Prop 65, Codex CCCF17) with the textbook chapters that provide the canonical toxicological frame (Handbook on the Toxicology of Metals chapter 32 by Nordberg, Nogawa, and Nordberg 2015, Casarett & Doull’s Essentials chapter 23 by Ufelle and Barchowsky 2021, and Patty’s Toxicology chapter 7 by Jakubowski 2012). The full cadmium synthesis will be posted after the regulatory-tox ingests complete and the textbook chapters are integrated. Readers should treat any claim here as provisional and trace it upward to the source page before citing.

Gaps and load-bearing questions surfaced during ingest

Issues raised by completed ingests that must be resolved or explicitly acknowledged in the eventual synthesis.

EFSA TWI versus JECFA PTMI for cadmium

Raised during ingest of EFSA Cd 2009 on 2026-04-24; updated during ingest of JECFA 91st 2022 on 2026-04-24.

The EFSA Panel on Contaminants in the Food Chain established a tolerable weekly intake for cadmium of 2.5 µg/kg body weight per week in January 2009, anchored on renal tubular dysfunction biomarkered by urinary beta-2-microglobulin, with a reference point of 1.0 µg cadmium per gram creatinine in urine at age 50 (EFSA 2009). The Joint FAO/WHO Expert Committee on Food Additives adopted a provisional tolerable monthly intake of 25 µg/kg b.w./month at its 73rd meeting in 2010 (report published 2011 as WHO Technical Report Series No. 960) (JECFA 73rd). Expressed on a weekly basis, the JECFA PTMI is approximately 5.83 µg/kg b.w./week, a factor of approximately 2.3 higher than the EFSA TWI. EFSA issued a reaffirming statement in 2011 explaining the derivation differences between the two bodies; that 2011 statement is not yet in the corpus and is flagged for future ingest. JECFA’s 91st meeting (November 2020, published 2022) carried the 25 PTMI forward without revision while updating dietary exposure assessments; the 91st meeting monograph is ingested (JECFA 91st).

The two bodies drew from overlapping primary literature. The divergence reflects different methodological choices including the selection of pivotal studies, the treatment of the urinary-cadmium-to-dietary-cadmium translation, and the choice of averaging window (EFSA uses a weekly window rounded from a 2.52 µg/kg b.w./week derivation; JECFA uses a monthly window specifically to reflect cadmium’s long biological half-life). The synthesis cannot average or pick one; it must present both with derivation transparency and let readers see where the methodological choices led to the gap. HMT&C certification-threshold decisions that calibrate to one reference value or the other need to name the rationale per the CLAUDE.md firewall rule.

A key 91st meeting finding that sharpens the EFSA-JECFA reconciliation: updated 2019 occurrence data for cadmium in cocoa products pushed total dietary cadmium exposure in children aged 0.5 to 12 to as high as 96 percent of the JECFA PTMI, indicating that even the less stringent of the two reference values is approached or saturated in subpopulations (JECFA 91st). A fortiori, these subpopulations exceed the EFSA TWI substantially. The regulatory gap between the two values therefore matters most precisely where population risk is highest, which argues for HMT&C infant and child product tiers to calibrate to the EFSA TWI rather than the JECFA PTMI.

The ATSDR 2012 Toxicological Profile, ingested on 2026-04-24, makes this a three-way (and eventually four-way) divergence rather than a two-way gap. ATSDR’s chronic oral MRL of 0.1 µg/kg/day is approximately 3.6 times tighter than the EFSA TWI and approximately 8 times tighter than the JECFA PTMI when all three are expressed on a daily per-body-weight basis (ATSDR 2012). The ATSDR derivation uses the same endpoint class as EFSA (renal tubular dysfunction biomarkered by low-molecular-weight proteinuria) but arrives at a tighter value through a different methodological path: UCDL10 of 0.5 µg/g creatinine as the point of departure (versus EFSA’s BMDL5-to-reference-point chain that ends at 1.0 µg/g creatinine), a pharmacokinetic model fit to peak renal cortex concentration at age 55 (versus EFSA’s one-compartment fit to non-smoking Swedish women age 58-70), and an uncertainty factor of 3 for human variability (versus EFSA’s decision that no additional UF was needed beyond the CSAF) (ATSDR 2012, EFSA 2009). The operative US EPA IRIS oral RfD, when ingested from the 1985 primary record, is understood from secondary citations to be approximately 10 times more permissive than ATSDR, or roughly comparable to JECFA in daily-equivalent terms (EPA IRIS Cadmium). The ordering of the four reference values from tightest to most permissive is ATSDR → EFSA → JECFA → EPA IRIS; HMT&C calibration to any of the four requires naming which and why.

ATSDR explicitly notes in the 2012 profile that its chronic oral MRL is below typical US dietary cadmium intake (approximately 0.3 µg/kg/day per Choudhury et al. 2001), and that the UCDL10 point of departure is only twofold above the CDC 2011 US adult geometric mean urinary cadmium concentration (0.247 µg/g creatinine) (ATSDR 2012). This is itself a load-bearing fact for the wiki and for any certification defensibility story: the most conservative major reference value for dietary cadmium is derived by a US public health agency that acknowledges its own reference value is below typical US exposure, not above it.

Codex CCCF17 (April 2024) did not produce a new health-based value but documents two developments that matter for the synthesis (Codex CCCF17 2024). First, the Committee forwarded a new Codex ML for cadmium in quinoa at 0.15 mg/kg for CAC47 adoption, selected under ALARA with one delegation citing that “in their region, the tolerable weekly intake (TWI) was exceeded for many consumers” (paragraph 117) (Codex CCCF17 2024). This is a Codex-internal acknowledgment of the EFSA TWI as the operative regional reference value that exposure is already exceeding, not the JECFA PTMI. It is a small but meaningful signal that Codex CCCF deliberations are not uniformly anchored to the PTMI even though the PTMI is formally the health-based guidance value Codex MLs are derived against. Second, CCCF17 initiated new work on a broader Code of Practice for the Prevention and Reduction of Cadmium Contamination in Foods, extending CXC 81-2022 (the existing Cocoa Beans CoP) to potential annexes for rice, cereals and cereal products, vegetables, fish, and seafood (Codex CCCF17 2024). The commodity list aligns with the major population-level contributors identified across the EFSA, JECFA, and ATSDR ingests, confirming cross-body convergence on what drives dietary cadmium exposure.

Cadmium cardiovascular endpoint: framing shift between the 2015 Handbook and the 2021 Essentials

Raised during ingest of Ufelle & Barchowsky 2021 on 2026-04-24.

The Handbook on the Toxicology of Metals chapter 32 (Nordberg, Nogawa, Nordberg 2015) characterizes cadmium’s cardiovascular endpoints as contested, with positive and null associations both reported across animal and human studies, and cautions that residual confounding by tobacco smoke cannot be excluded in the positive literature (Nordberg 2015). Casarett & Doull’s Essentials of Toxicology chapter 23 (Ufelle and Barchowsky 2021), published six years later, describes a “strong association” between cadmium and peripheral vascular disease and raises the possibility that cadmium partially mediates the negative cardiovascular effect of smoking (Ufelle and Barchowsky 2021). The two textbook positions are not in direct contradiction (new epidemiology can accumulate over six years in a direction that would justify stronger language), but they are materially different framings of the same endpoint class.

The wiki records both positions on cadmium without resolution. This is the correct posture under the CLAUDE.md drift-protection rule: strengthening or softening either position to reconcile them would corrupt the record, whereas presenting the gap honestly preserves the wiki’s epistemic authority. The synthesis pass should resolve this by searching the 2015-to-present cardiovascular cadmium literature and determining whether the 2021 stronger framing reflects a real consensus shift or an authorial choice that has not been broadly adopted.

To resolve before the cadmium synthesis is written:

• Search the 2015-to-present primary literature on cadmium cardiovascular effects, particularly the Tellez-Plaza, Fagerberg, and Navas-Acien cohorts that appear in both source chapters with somewhat different framing.
• Identify whether any intervening major review, regulatory reassessment, or meta-analysis has explicitly updated the consensus framing.
• If the literature base has not resolved the question, the wiki should continue to present both positions and explicitly state that the endpoint remains under active investigation.

Codex standards provenance gaps

Raised during ingest of Codex CCCF17 2024 on 2026-04-24.

CCCF17 is a session report, not a standards text. The operative matrix-level Codex cadmium MLs live in Codex Standard CXS 193-1995 (General Standard for Contaminants and Toxins in Food and Feed), ingested on 2026-04-30 with the full matrix-level Cd ML table transcribed (rice 0.4 mg/kg polished, cocoa powder 2.0 mg/kg, dark chocolate ≥70 percent solids 0.9 mg/kg, leafy vegetables 0.2 mg/kg, marine bivalve molluscs 2 mg/kg, and others). The cocoa-specific Code of Practice CXC 81-2022 (Code of Practice for the Prevention and Reduction of Cadmium Contamination in Cocoa Beans) was ingested on 2026-05-03 directly from FAO and supplies the operative agronomic, processing, and partial formulation cocoa-cadmium mitigation specifics referenced from the mitigation strategy pages. Codex CXS 193-1995 was ingested via WebFetch from FAO’s Codex Standards web service; CXC 81-2022 was ingested via the same pathway with the full text transcribed into the source page.

Resolution status: closed for both CXS 193-1995 and CXC 81-2022. The remaining open items in this section are the EFSA 2011 reaffirming statement (raw PDF still pending) and historical JECFA derivation documents.

To resolve before the cadmium synthesis is written:

• Ingest the EFSA 2011 reaffirming statement (pending raw PDF availability).
• Ingest the JECFA 73rd meeting primary derivation documents (2010 monograph and WHO TRS 960, pending).
• Compare the critical study selections side by side and identify the specific methodological choices that drive the numeric gap.

Dietary cadmium carcinogenicity dose-response

Raised during ingest of EFSA Cd 2009 on 2026-04-24.

The International Agency for Research on Cancer classifies cadmium as a Group 1 human carcinogen on the basis of occupational inhalation studies (EFSA 2009, Nordberg 2015). General-population epidemiology has produced statistical associations between cadmium exposure and cancers of the lung, endometrium, bladder, and breast (EFSA 2009, Nordberg 2015). EFSA judged these data insufficient for quantitative dose-response modeling and anchored the TWI on the renal endpoint instead (EFSA 2009). The open question is whether sufficient data have since accumulated to support a quantitative carcinogenic risk assessment for dietary cadmium at general-population exposure levels, or whether the renal endpoint remains the defensible anchor.

The synthesis will need to state clearly what IARC says, what EFSA concluded in 2009, and what the more recent primary literature shows, without overclaiming on cancer risk (because the dose-response is not yet quantitative) and without underclaiming (because the Group 1 classification is durable and the general-population associations are replicating). Consumer-facing sections require particular care given the dose-and-population specificity rule in CLAUDE.md.

To resolve before the cadmium synthesis is written:

• Ingest the operative 1985 EPA IRIS chemical record 0141 for cadmium, which carries the US-side oral RfD and its critical endpoint. The 1999 reassessment draft in raw/reports/ is a draft that never became Agency position and is deferred for a later historical ingest.
• Ingest the ATSDR toxicological profile, which synthesizes the carcinogenicity literature through approximately 2012.
• Note that the OEHHA Prop 65 documents ingested on 2026-04-24 address reproductive toxicity, not carcinogenicity; the separate Prop 65 carcinogenicity listing for cadmium and its No Significant Risk Level are pending ingest.
• Scan recent general-population studies flagged during textbook-chapter ingest for any that have since produced credible quantitative dose-response estimates.

EPA IRIS Cd assessment provenance gap

Raised during attempted ingest of EPA_IRIS_Cadmium_ToxicologicalReview.pdf on 2026-04-24.

The raw file in raw/reports/ with this name is a March 4, 1999 external review draft marked “DO NOT QUOTE” and does not represent Agency position. The operative EPA IRIS cadmium assessment is the 1985 record (IRIS chemical record 0141), which is understood from secondary citations to establish oral RfD values of roughly 5 × 10⁻⁴ mg/kg/day (water) and 1 × 10⁻³ mg/kg/day (food), anchored to a kidney-cortex concentration of 200 µg/g. The wiki cannot state those values as verified until the primary 1985 IRIS record is in the corpus with SHA-256 provenance on its own source page; recording them from secondary citations would violate CLAUDE.md’s numeric-verification rule.

To resolve:

• Obtain the primary 1985 IRIS chemical record 0141 document (likely accessible via EPA’s IRIS portal at epa.gov/iris) and ingest it as a source page with full provenance.
• Separately ingest the 1999 reassessment draft as historical context, with explicit “DRAFT, NOT ADOPTED” flagging throughout its source page.