Chiger & Lynch (Abt/HBBF) 2017 — IQ loss and monetized benefits of reducing iAs in infant rice cereal

Abt Associates prepared this December 2017 report for Healthy Babies Bright Futures to quantify avoided IQ loss and the monetized economic benefit of reducing inorganic arsenic exposure from infant rice cereal and broader rice consumption in U.S. children aged 0–6. The report does not generate primary measurements; it pools published distributions of iAs in infant rice cereal (FDA 2016a/c, FDA Total Diet Study, Consumer Reports 2012/2014, Juskelis et al. 2013, Brockman & Brown 2012, Karagas et al. 2016, Carignan et al. 2016, Shibata et al. 2016) with consumption data (NHANES/WWEIA, FITS 2002/2008, Karagas et al. 2016) and applies five concentration-response functions (Hamadani 2011 log-linear and linear; Wasserman 2004 quadratic and linear-extrapolation; Wasserman 2007 log-linear) drawn from Bangladeshi child cohorts to estimate IQ point loss and dollar value across the U.S. cohort. Headline findings: replacing all rice and rice products with arsenic-free alternatives in U.S. children 0–6 would avoid more than 9 million IQ points per year and yield $12–18billioninadditionalannualearnings;replacinginfantricecerealspecificallywouldavoidalmost1millionIQpointsperyearandyield$1.2–1.8 billion in additional earnings. Both headline ranges in the Executive Summary are drawn from the Wasserman et al. (2004) quadratic / linear-extrapolation CRFs at the 0 ppb (full replacement) scenario; the report’s wider per-CRF sensitivity envelope is much larger (see Key numbers).

Key numbers

Mean iAs concentrations in infant rice cereal and alternatives (Table 1, p. 8-10)

HBBF independent testing (Figure 1, p. 11) reports approximate mean iAs concentrations of ~96 ppb (brown rice), ~73 ppb (rice), ~23 ppb (multigrain with rice), ~17 ppb (multigrain without rice), ~17 ppb (wheat), ~13 ppb (oatmeal), ~10 ppb (quinoa), ~7 ppb (barley), ~5 ppb (buckwheat).

Daily iAs exposure from infant rice cereal across published estimates (Table 5, p. 18)

*Carignan et al. (2016) Table 3 in the original paper reports total-As exposures only; the iAs values in this row were derived by the Abt authors using the inorganic-arsenic concentrations Carignan drew from FDA’s 2013 infant-rice-cereal analysis (median iAs 0.20 µg/g for average scenarios; maximum iAs 0.25 µg/g for high scenarios), not a fixed-fraction conversion from total As. The Xue et al. 2010 row separately uses the Shibata 2016 assumption that infant rice cereal contributes 55% of total dietary iAs exposure to back-out infant-rice-cereal-specific exposure from Xue’s total-iAs estimate.

Shibata et al. (2016) Monte Carlo estimates (Table 4, p. 17): daily iAs from infant rice cereal at the 25th/50th/75th percentile is 0.051/0.23/0.46 µg/kg-bw/day for 4–5 month-olds, 0.12/0.29/0.53 for 6–8 months, and 0.11/0.30/0.53 for 9–11 months. Xue et al. (2010) SHEDS modelling (Table 2): inorganic As exposure from all food in infants <1 yr has mean 0.23 ± 0.19 µg/kg-bw/day; 5th–99th percentile 0.01–0.80.

Health-based limits compared (Section 4.2)

Concentration-response functions selected (Table 11, p. 40)

Nationwide avoided IQ loss and monetized benefit — infant rice cereal only (Tables 16-17, p. 49)

For U.S. children 0–6 (n = 27,989,207 per Census Bureau 2015), the avoided IQ loss per year and annual benefit at each scenario:

Per-IQ-point dollar value: $9,421(low)to$13,943 (high), based on U.S. EPA 2011b in 2015 dollars at a 3% discount rate. The Executive Summary’s “almost 1 million IQ points” and “$1.2–1.8 billion” headline numbers track the Wasserman 2004 quadratic and linear-extrapolation estimates at the 0 ppb scenario.

Nationwide avoided IQ loss — all rice + rice products in 0–6 year-olds (Tables 19-20, p. 51)

The Executive Summary headline “more than 9 million IQ points / $12–18 billion” tracks the Wasserman 2004 quadratic / linear-extrapolation estimates at the 0 ppb scenario.

Exposure scenarios analyzed (Table 6, p. 23) and resulting iAs intake changes (Tables 7-8)

Six scenarios: (1) implement FDA 100 ppb (initial 103.9 ppb → final 83.5 ppb in white infant rice cereal); (2) implement alternate 50 ppb (initial 103.9 ppb → final 20.8 ppb); (3) reduce consumption from 3 to 1 serving/day at 83.5 ppb; (4) replace infant rice cereal with alternate grain at 15 ppb (3 → 0 servings of rice cereal). Using FDA 2016c serving assumptions, the proposed FDA standard reduces high-consumer (3 srv/day) intake from 0.35 → 0.28 µg/kg-bw/day (change 0.07); the 50 ppb alternate reduces it 0.35 → 0.07 (change 0.28); reducing consumption to 1 srv/day reduces it 0.28 → 0.13 (change 0.15); full replacement reduces it 0.28 → 0.05 (change 0.23). Using Shibata 2016 serving assumptions (4.6 g/tablespoon vs FDA’s 2.5 g), all values approximately double.

Methods (brief)

This is a literature-synthesis, exposure-modelling, and benefit-cost report. No primary measurement of arsenic concentrations, biomarkers, or neurodevelopmental outcomes was performed.

Section 3 compiles published iAs concentration distributions in infant rice cereal, infant non-rice cereals, and alternative grains (FDA 2014, FDA 2016c, FDA Total Diet Study 2006-2011, Juskelis et al. 2013, Brockman & Brown 2012, Carbonell-Barrachina et al. 2012, Consumer Reports 2012/2014, plus pre-publication HBBF 2017 testing of 105 cereals). Consumption data are drawn from NHANES/WWEIA 2003-2010, FITS 2002 and 2008 (Fox et al. 2006; Briefel et al. 2010; Siega-Riz et al. 2010), and Karagas et al. 2016 (n=954 New Hampshire birth cohort). Body-weight assumptions from EPA Exposure Factors Handbook 2011a.

Section 4 estimates dose changes under six exposure scenarios using both FDA 2016c serving assumptions (2.5 g/tablespoon) and Shibata 2016 (4.6 g/tablespoon) to span the assumption envelope. Health-based limits compared: USEPA IRIS 1991 RfD, ATSDR 2007 MRLs (acute and chronic), Tsuji et al. 2015 RfD (and Abt’s modified-uncertainty-factor re-derivation), Shibata et al. 2016 MCLs.

Section 5 reviews 13 epidemiological studies of arsenic-IQ associations (Table 10, p. 32-35) drawn from three prior reviews (Bellinger 2013; Rodríguez-Barranco et al. 2013; Tsuji et al. 2015) plus a primary-literature search through Wasserman et al. 2016 and Rodrigues et al. 2016. Inclusion: full-scale IQ endpoint, quantifiable concentration-response, sufficient quality rating in prior reviews. Exclusions: studies using biomarkers other than urinary or water iAs (Wasserman 2011/2016 blood and creatinine-adjusted urinary); Wasserman 2014 (U.S. but no dose-response); Wang 2007 (categorical only); Rocha-Amador 2007 (uncontrolled fluoride co-exposure); von Ehrenstein 2009 (high-exposure outlier attenuation); Rodríguez-Barranco 2013 meta-analysis (no per-study betas reported).

Section 6 applies five CRFs (Hamadani 2011 log-linear and linear; Wasserman 2004 quadratic and linear-extrapolation; Wasserman 2007 log-linear) to U.S. exposure distributions. Urinary-based CRFs are converted to dietary-dose CRFs via Tsuji et al. 2015’s dose-conversion equation (Section 4.2.2). Monetization uses U.S. EPA 2011b dollar-per-IQ-point estimates updated to 2015 dollars at a 3% discount rate. The IQ analysis layers infant rice cereal exposure (ages 0-1) on top of broader rice exposure (ages 1-6 held at average levels).

Appendix A reviews the limited human evidence on DMA health effects (renal, urinary, developmental effects in animals at higher-than-dietary doses; insufficient human data). Appendix B summarises additional arsenic-IQ studies excluded from CRF derivation.

Speciation: report focuses on inorganic As (As^III arsenite + As^V arsenate) only; DMA and MMA discussed contextually but not used in CRFs.

Implications

Reference and exposure-modelling: Chiger & Lynch 2017 is the most comprehensive single document compiling U.S. infant rice cereal iAs concentrations and consumption-derived exposure estimates across the FDA, Xue, Karagas, Carignan, and Shibata frameworks. Researchers building U.S. cumulative exposure models for iAs from cereal-grain foods can pull pooled concentration and exposure distributions directly from Tables 1-5 rather than reconstructing the inputs from primary sources.

Regulatory comparison: the report makes the IRIS 1991 vs Tsuji 2015 RfD distinction explicit. The IRIS RfD of 0.3 µg/kg-bw/day is derived from Tseng 1977/1968 skin-lesion endpoints (with a UF of 3); the Tsuji 2015 RfD of 0.4–1.0 µg/kg-bw/day uses a 1-IQ-point LOAEL from Hamadani 2011 with author-selected UFs of 1–3. Abt argues those UFs are too small and re-derives the Tsuji RfD at 0.01–0.05 µg/kg-bw/day with UFs of 30–100. The 100-fold spread between published Tsuji and Abt-re-derived Tsuji values is the operative uncertainty in any regulatory exercise that anchors on neurodevelopmental endpoints rather than the IRIS skin-lesion endpoint.

CRF selection sensitivity: the per-CRF IQ-point estimates (Tables 16, 19) span almost two orders of magnitude. Hamadani 2011 log-linear and Wasserman 2007 produce the largest avoided-loss estimates (because log-linear functions have a steep slope in the low-dose region where U.S. dietary exposures sit). Hamadani 2011 linear produces the smallest. The Wasserman 2004 quadratic and linear-extrapolation CRFs sit in the middle and are the basis of the Executive Summary headline numbers. Section 6.4 of the report notes the report’s own uncertainty about which functional form is most appropriate in the low-dose region.

Audience translation: at the per-child level (Table 18), switching from infant rice cereal at U.S. average concentrations to an alternate grain (15 ppb iAs) avoids 0.02–2.60 IQ points per child depending on CRF; the median across CRFs is around 0.28 IQ points (Wasserman 2004 quadratic / linear extrapolation). Per-individual effects are small; population-scale effects are large because the U.S. cohort is large. For a parent: replacing infant rice cereal in your child’s diet with oatmeal, quinoa, barley, or other non-rice infant cereal reduces inorganic arsenic exposure substantially (the non-rice options carry roughly 6–17 ppb iAs vs about 100–120 ppb in infant rice cereal). The IQ benefit for any single child is likely too small to detect, but at population scale it is meaningful, and arsenic exposure during the first year of life appears to have neurodevelopmental effects that do not reverse if exposure later decreases (Wasserman et al. 2016 follow-up).

App: the per-IQ-point dollar valuations and the CRF-by-CRF avoided-IQ-loss distributions can drive a per-child exposure calculator. Concentration inputs for infant rice cereal (mean ~100–120 ppb iAs depending on white vs brown) and for alternate grains (5–17 ppb) are pre-pooled in Section 3.1.

Microbiome: not addressed; this is exposure + dose-response + economic valuation, not mechanism. Arsenic-microbiome interactions are not covered by this source.

Wiki pages this source may touch

• arsenic-inorganic — primary iAs neurodevelopmental endpoint reference; pooled U.S. exposure distributions for ages 0-6
• rice — broader rice-products exposure scenario (Section 6.3)
• rice-cereal — primary subject; concentration and consumption pooled distributions
• rice-flour — secondary, via the “rice and rice products” aggregate (per-capita 0.556 g/kg bw/d in 0-6 yr-olds, Section 6.3)
• baby-cereals-dry-rice-based — primary product target
• infant-cereal — broader infant-cereal context
• mixed-meals-rice-containing — secondary, in the “all rice products” scenarios
• teething-and-snacks-rice-based — secondary, in the “all rice products” scenarios
• fda-iAs-rice-cereal-100ppb — proposed FDA action level discussed Section 2.2
• eu-2015-1006-iAs-rice — EU 100 ppb standard for rice in infant food, discussed Section 2.1
• epa-iris-inorganic-arsenic-rfd — IRIS 1991 chronic oral RfD, Section 4.2.1

Verification notes

• 2026-05-29: Manual-fetch re-ingest pass against the source PDF (raw/Manual Fetch Kimi /2017_Effects-of-Inorganic-Arsenic-in-Infant-Rice-Cereal-on-Children-s-Neurodevelopment.pdf). Page was originally created 2026-05-14 by claude_opus_47_session under the pre-v2 manual_phase1_priority workflow; this pass enhances per the v2 manual-fetch-pdf skill. Changes:
  • Frontmatter audience: [regulator, educator, consumer, app] added (was missing). sample_population corrected: prior text said “~24M children” but Section 6.2.2 specifies 27,989,207 per U.S. Census Bureau 2015 — corrected. sampling_locations: [] and sampling_year_range: null retained since this is a literature-synthesis report with no primary sampling. ingest_method updated to manual_fetch_v2. cite_key, raw_handle, raw_path, license, access_url, near_duplicates preserved per skill’s EXISTING path.
  • Headline numbers (“$1.2-1.8B"infantricecerealreplacementand"$12-18B” all-rice replacement, plus “~1 million IQ points” and “more than 9 million IQ points”) now explicitly attributed to the Wasserman 2004 quadratic / linear-extrapolation CRFs at the 0 ppb scenario. Previously the page stated the headline numbers without naming the CRF, leaving readers without a way to interpret the much-wider per-CRF envelope (Tables 16-17, 19-20). The full per-CRF tables are now in Key numbers so the reader can see the sensitivity.
  • Key numbers expanded from 5 short rows to a full reconstruction of Tables 1, 5, 9, 11, 16-17, and 19-20 with page references. Health-based limits table populated with values, endpoints, and original-derivation sources (Tseng 1977/1968 for IRIS; Mizuta 1956 for acute MRL; Hamadani 2011 derivation for Tsuji RfD).
  • Methods section expanded: previously thin and missing the CRF inclusion/exclusion criteria (Section 6.1), the Tsuji 2015 dose-conversion equation usage, and the age-layering approach (infant rice cereal exposures for 0-1; rice exposures held at average for 1-6). All now included with section pointers.
  • ”## Implications” section retained (template requirement) but rewritten to drop HMTc-promotional language. Previously stated: “this report is the canonical advocacy/policy reference that establishes the $1.2-1.8B-per-year-in-avoidable-IQ-loss claim that HBBF uses in advocacy. HMTc certification of low-iAs infant rice cereal is a partial market intervention against this loss; brand-legal positioning around HMTc should explicitly cite Chiger 2017 as the cost-of-not-acting reference” and “HMTc Cat 1 infant rice cereal thresholds should anchor on the Tsuji RfD for the toxicology basis” — both removed per Part 2 (wiki/HMTc firewall): the wiki reports what the source says, not how HMTc should position around it. New Implications text frames the report’s role as a pooled exposure-modelling and CRF-sensitivity reference for researchers, with explicit notes on the IRIS 1991 vs Tsuji 2015 vs Abt-re-derived Tsuji envelope.
  • Legacy ”## Wiki pages updated on ingest” heading renamed to ”## Wiki pages this source may touch” per current template (the page does not itself update the listed pages; routing audit and synthesis pass handle propagation).
  • Wiki-pages-may-touch list corrected for taxonomy drift: regulations/fda-iAs-rice-cereal-2020 → [[regulations/fda-iAs-rice-cereal-100ppb]] (the correct extant slug; verified ls wiki/regulations/). regulations/usepa-iris-1991-iAs-rfd (does not exist) → [[regulations/epa-iris-inorganic-arsenic-rfd]] (the correct extant slug). Removed regulations/tsuji-2015-iAs-neurodevelopmental-rfd — Tsuji et al. 2015 is a peer-reviewed study, not a regulation; should not have a regulation page. Removed regulations/atsdr-2007-iAs-mrl — no such page exists in current taxonomy; the ATSDR 2007 MRLs are discussed within the iAs metal page and at the existing ATSDR pages (atsdr-aluminum-mrls, atsdr-cadmium-mrls). Added [[regulations/eu-2015-1006-iAs-rice]] (extant slug for the EU 100 ppb standard discussed in Section 2.1).
  • Consumer-audience translation added inside Implications with <!-- audience: consumer --> ... <!-- /audience --> tagging per Part 7 audience layering. Anchored on quantitative concentration values (~100-120 ppb infant rice cereal vs 5-17 ppb non-rice alternatives) rather than undefined “trace”/“safe” qualifiers, per conventions.
• Brand firewall (Part 12): the report itself names “Nestlé Feeding Infants and Toddlers Study” — that is a scientific-instrument vendor-style reference (the FITS study is the named data source, not a contamination value attribution), which is the Exception-2 carve-out per the 2026-05-17 lock. Retained as “FITS 2002 and 2008” and “Nestlé FITS” without brand-attribution-to-values. No brand-by-brand contamination listings exist anywhere in the source for this paper, so no Part 12 redactions were needed.
• HMTc firewall (Part 2): the report itself takes a clear advocacy position (avoidable IQ loss → tighter standard). The wiki summary preserves the report’s quantitative findings and reasoning chain but does not endorse the advocacy direction in the wiki voice. The Implications section names the regulatory-uncertainty envelope (IRIS vs Tsuji vs Abt-re-derived Tsuji) without recommending which value HMTc or any other certifier should adopt.
• 2026-05-29 (audit application): Fresh-context Agent subagent (general-purpose) ran the five-check audit per docs/gpt-collaboration/audit-prompt.md. Verdict: REVISE with 3 ❌ findings and 2 ⚠️ concerns. Findings verified against source and applied:
  • ❌ Table 11 Hamadani 2011 log-linear and linear rows were both labeled “Verbal IQ” — corrected to “Full-scale IQ” per source p. 41 (“verbal IQ: β = −2.4 (95% CI: −3.8, −1.1); full-scale IQ: β = −1.4 (95% CI: −2.7, −0.1)” and Section 6.1.1 confirming Abt selected the full-scale β=−1.4 log-linear slope and the full-scale 0.9 IQ-pts/100 µg/L linear slope; the verbal linear slope of 2.6 IQ-pts/100 µg/L is what Tsuji 2015 used for the RfD). Both rows now also annotate the parallel verbal-IQ slope for cross-reference.
  • ❌ Carignan Table 5 footnote previously stated the 55% Shibata assumption was applied to convert Carignan total-As to iAs. Source p. 19 Table 5 footnotes show the 55% assumption is the Xue (*) footnote, not Carignan (**); Section 3.3.4 (p. 16) confirms Carignan derived iAs using FDA 2013 inorganic-arsenic concentrations directly (median 0.20 µg/g; max 0.25 µg/g). Footnote rewritten to reflect what Carignan actually did.
  • ⚠️ Bare [[microbiome/]] directory wikilink in Implications section dropped (Microbiome line now simply states the source does not cover this topic).
  • ⚠️ Matrices vocabulary ([policy-cost-benefit, exposure-modeling, neurodevelopment]) flagged as non-standard but left in place: routing audit accepts these values (source not in routing_malformed.csv); the values are descriptive of a literature-synthesis report that has no primary sampling matrix. If a future taxonomy migration introduces a report-type field, these should move there.
  • ⚠️ FITS / Nestlé Exception-2 rationale in prior Verification notes is rationale-only and was not edited; correct rationale is that “FITS” is a standard scholarly study identifier (like NHANES, WWEIA) and the body does not name brand products against contamination values.
  • ⚠️ Consumer-audience block (“For a parent: replacing infant rice cereal in your child’s diet with oatmeal, quinoa, barley, or other non-rice infant cereal reduces inorganic arsenic exposure substantially…”) left as-is: it is quantitatively anchored (6–17 ppb non-rice vs ~100–120 ppb infant rice cereal), properly hedged on per-child IQ effect, and faithfully echoes the source’s own explicit recommendation (p. 30). Defensible at the boundary of Part 2 consumer-translation discipline. No findings were rejected as false positives. Routing audit re-run clean after edits.

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.