Houlihan and Brody 2019 — HBBF: What’s in My Baby’s Food?

Healthy Babies Bright Futures (HBBF) commissioned Brooks Applied Labs (Bothell, Washington) to test 168 commercially purchased baby food containers for four toxic heavy metals — arsenic, lead, cadmium, and mercury — across 61 brands and 13 food types purchased from 14 US metropolitan areas. Testing found detectable levels of at least one metal in 95 percent of baby foods sampled, with one in four containers positive for all four metals simultaneously. The report also commissioned Abt Associates to estimate IQ loss attributable to lead and arsenic in the diets of children aged 0 to 24 months, finding a total population-level loss exceeding 11 million IQ points from dietary exposure, with rice-based foods alone accounting for nearly one-fifth of that total.

Note on evidence tier: This is a B-tier nonprofit report, not a peer-reviewed primary study. The laboratory results are from a single commercially accredited lab without published QC data in the main report body; they are presented without uncertainty estimates or statistical comparison across food types. The IQ-loss estimates are modeled by Abt Associates using federal dietary survey data rather than directly measured. Use with appropriate attribution.

Key numbers

Detection rates across 168 baby foods (tAs/total arsenic unless noted; concentrations reported as product-as-tested/as-sold, including dry foods):

• Arsenic detected in 73% of baby foods (inorganic arsenic specifically measured in the 25 highest-arsenic foods)
• Lead detected in 94% of baby foods
• Cadmium detected in 75% of baby foods
• Mercury (tHg) detected in 32% of baby foods
• 95% had at least one metal detected; only 9 of 168 were free of all four metals
• 26% of containers had all four metals simultaneously

Detection by food type (selected categories):

• Puffs and other snacks (n=21): arsenic 19/21, lead 21/21, cadmium 19/21, mercury 14/21
• Teething biscuits and rice rusks (n=10): all four metals 10/10
• Infant rice cereal (n=7): all four metals detected in 7/7; inorganic arsenic exceeded FDA’s proposed 100 ppb guidance level in 4 of 7 cereals tested
• Non-rice infant cereals (n=11): arsenic 11/11, lead 10/11, cadmium 11/11, mercury 2/11
• Vegetables (n=38): arsenic 38/38, lead 34/38

Average heavy metal concentrations for selected food types (ppb, product-as-tested/as-sold), from HBBF analysis combining own tests with FDA Total Diet Study 2014-2017 data:

• Rice puff snacks: 98 ppb average (Appendix F sample-weighted total metals, HBBF+FDA n=38: 97.51)
• Infant rice cereal: 85 ppb average inorganic arsenic only (Appendix F HBBF 2017 iAs n=42: 85.00); sample-weighted total metals (HBBF 2019 n=7 + FDA n=76 = pooled basis): 140.07 ppb (HBBF 2019 alone)
• Teething biscuits/rice rusks: 64 ppb average total metals (Appendix F sample-weighted HBBF+FDA n=37: 64.09; HBBF 2019 alone n=10: 54.61; FDA 2013-14 alone n=27: 67.60)
• Carrot baby food: 28 ppb average (Appendix F sample-weighted total metals n=26: 27.62)
• Sweet potato baby food: 19 ppb average (Appendix F sample-weighted total metals n=31: 18.76)
• Non-rice infant cereals: 14 ppb average inorganic arsenic only (Appendix F HBBF 2017 iAs n=63: 14.00); HBBF 2019 total metals (n=11): 40.91 ppb
• Rice-free baby snacks: 5 ppb average total metals (Appendix F sample-weighted HBBF+FDA n=136: 5.07)
• Other fruits and vegetables baby food: 7.4 ppb average (Appendix F sample-weighted total metals n=39: 7.42)

For most food-type comparisons these averages represent the sum of four metals (Pb + Cd + Hg + tAs, with iAs used in place of tAs only for the cereal comparison per the report’s stated exception). All values above are taken from Appendix F’s sample-weighted “Total metals” column unless explicitly labeled “inorganic arsenic only.” Arsenic dominates the total for rice-based products; lead and cadmium dominate for root vegetables.

Inorganic arsenic (iAs) specifically: 25 of the 168 baby foods (those with highest total arsenic) were separately tested for iAs speciation. Four of seven infant rice cereals exceeded FDA’s proposed 100 ppb iAs action level.

IQ loss estimate (Abt Associates 2019b, commissioned model): Children aged 0 to 24 months collectively lose an estimated 11+ million IQ points from dietary exposure to arsenic and lead combined. Rice-based foods account for roughly 20% of that total. The 15 foods responsible for 55% of estimated total risk, ranked by share of total IQ harm, include: rice dishes with beans and vegetables (10.0%), whole milk (8.4%, due to high consumption volume rather than high metal concentration), white and brown rice (7.0%), apple juice (6.1%), infant formula (5.3%, driven by volume), 100% fruit juice blend (4.1%), infant rice cereal (2.7%), grape juice (2.0%), oat ring cereals (1.6%), sweet potato baby food (1.6%), soft cereal bars and oatmeal cookies (1.4%), macaroni and cheese (1.4%), puffs and teething biscuits (1.3%), bottled drinking water (1.2%, volume-driven), and fruit yogurt (1.2%). The primary driver is arsenic for most items; lead dominates for infant formula, grape juice, sweet potato, macaroni and cheese, puffs, and fruit yogurt.

83% of baby foods tested had more lead than the 1 ppb limit endorsed by EDF and other public health organizations. Of the 13 food types tested, 10 types had no FDA enforceable standard or guidance for any of the four metals.

Safer-choice substitution effect (HBBF analysis): HBBF estimated that rice-free snacks had 93% lower average toxic-metal levels than rice puff snacks; other infant cereals had 84% lower inorganic arsenic than infant rice cereal; and tap water had 68% lower average toxic-metal levels than fruit juice.

Perchlorate (additional neurotoxin, separate test, n=25 foods): 19 of 25 foods positive; maximum levels ranged from 4.6 ppb (snacks) to 19.8 ppb (fruits and vegetables). All 19 perchlorate-positive foods also contained heavy metals; 12 had all four heavy metals.

Methods (brief)

Heavy metals testing: Brooks Applied Labs, Bothell WA. Total recoverable arsenic, cadmium, and lead were analyzed using a modified AOAC 2015.01 ICP-QQQ-MS method; total mercury was analyzed using EPA Method 1631 with CVAFS detection. Total arsenic (tAs) measured in all 168 samples; inorganic arsenic (iAs, sum of As(III) + As(V)) by arsenic speciation in the 25 highest-arsenic samples using TFA digestion and IC-ICP-CRC-MS. Total mercury (tHg) measured, not speciated as MeHg. Method detection limits are not summarized in the main narrative; Appendix C provides lab-method and QA/QC detail. Results are reported as product-as-tested/as-sold and include dry foods, so do not treat the full table as uniformly wet-weight. The report pools HBBF test results with FDA Total Diet Study 2014-2017 data for some averages; these pooled averages are not always separated by source in the main report text. Perchlorate testing performed by Southwest Research Institute, separate subsample.

LOD/LOQ: Appendix C contains lab-specific values; these were not extracted for this source page.

Key limitations: (1) Single-lab, single-timepoint, no replication or peer review of results. (2) IQ-loss estimates are modeled and rely on assumptions about consumption rates and dose-response relationships derived from external studies; the model does not account for MeHg or Cd contributions. (3) Concentration averages pool small sample sizes across heterogeneous food types; within-type variance is not reported in the main text. (4) No statistical significance testing across food categories.

Implications

Certification: This report documents category-level field contamination for 13 commercial baby food types and provides a ranked risk framework prioritizing rice-based foods, root vegetable purees, and teething products. The 4 of 7 infant rice cereals exceeding FDA’s proposed 100 ppb iAs level, and the universal detection of all four metals in puffs and teething biscuits, are useful HMT&C category-risk context. The 83% exceedance of the 1 ppb advocacy lead limit across all baby foods benchmarks current market performance, but it should not be treated as a standalone HMT&C threshold-setting source.

Courses: The 15-food IQ-loss ranking provides a usable pedagogical structure for baby food contamination risk. The substitution table (93%, 84%, 68%, and 73% reduction scenarios) is useful for teaching how an advocacy report converts contamination data into consumer-facing recommendations, but the wiki should preserve that framing as source-described advocacy rather than issue independent guidance. The 4× greater IQ impact from arsenic vs. lead in the first two years of life (for dietary exposure) is a key teaching point.

App: Contamination averages by food type (rice puff snacks 98 ppb, infant rice cereal 85 ppb, carrot baby food 28 ppb, etc.) can inform qualitative/category prioritization for product-level app logic, subject to the usual B-tier weighting. Because these are pooled averages combining HBBF and FDA TDS data, and because the cereal comparison is iAs-only while most other comparisons sum four metals, they are best used as order-of-magnitude orientation rather than as primary quantitative inputs; pair with A-tier per-analyte speciated values where available.

Microbiome: Not addressed in this source.

Wiki pages this source may touch

• baby-cereals-dry-rice-based
• baby-cereals-dry-non-rice
• root-vegetable-purees
• non-root-vegetable-purees
• fruit-purees
• teething-and-snacks-rice-based
• teething-and-snacks-non-rice
• snacks-crackers-biscuits
• infant-formula-powder
• infant-formula-rtf-liquid-non-soy
• fruit-juice-not-canned
• rice
• carrots
• sweet-potato
• apple
• lead
• arsenic-inorganic
• arsenic-total
• cadmium
• mercury-total

Verification notes

• This page was enhanced 2026-05-17 against the prior 2026-05-13 revision. Slug-vocabulary fixes only — the prior page’s substantive content (Key numbers, Methods, Implications) was already strong and was preserved as-is. Defects corrected: (a) invalid product slugs infant-formula and infant-rice-cereal replaced with taxonomy-valid infant-formula-powder, infant-formula-rtf-liquid-non-soy, baby-cereals-dry-rice-based, teething-and-snacks-rice-based; (b) added missing valid slugs (non-root-vegetable-purees, teething-and-snacks-non-rice, snacks-crackers-biscuits, fruit ingredient apple); (c) matrices vocabulary replaced with system-prompt valid terms; (d) legacy ## Wiki pages updated on ingest heading modernized; (e) metals page references expanded to include arsenic-inorganic, arsenic-total, mercury-total rather than the generic arsenic / mercury umbrella slugs (the report measures and reports speciated iAs separately for the 25 highest-arsenic foods).
• Cross-vendor audit 2026-05-17 corrected the local raw_path, clarified that results are product-as-tested/as-sold rather than uniformly wet-weight, added Appendix C analytical-method detail, noted that the cereal safer-choice comparison is iAs-only, and reframed substitution/HMT&C/app implications to avoid wiki-side consumer guidance or threshold-setting claims.
• 2026-05-31 near-duplicate registration (manual-fetch v2.0): the 6-page Executive Summary (03_2019_BabyFoodReport_EXEC-SUMM-ENGLISH_R6.pdf, version R6 March 2020) was encountered in a Kimi-corruption-recovery extract of 01_Bottles_Nipples_SippyCups. Full-PDF read confirmed it is a strict content subset of the already-ingested full report: same authors, same 168-baby-foods Brooks Applied Labs dataset, same detection percentages (As 73%, Pb 94%, Cd 75%, Hg 32%; 95% one-or-more; 26% all-four), same Abt Associates IQ-loss commission (11+ million IQ points; rice-based foods ≈ one-fifth share; 15 foods drive 55% of risk), same 5 safer-choice substitution table (snacks 93%, teething 91%, cereal 84%, drinks 68%, fruits/veggies up to 73%), same FDA/Baby Food Council framing. No numerical findings, methods detail, or implications appear in the exec summary that are absent from the full report. Registered as a near_duplicates entry rather than ingested as a parallel source page, to preserve the one-study-equals-one-evidence-bearing-page invariant.
• 2026-05-31 near-duplicate registration (manual-fetch v2.0, second pass): the 50-page full report file 07_2019_BabyFoodReport_FULLREPORT_ENGLISH_R5b.pdf was encountered in the same Kimi-corruption-recovery extract of 01_Bottles_Nipples_SippyCups. Full-PDF read of the cover, ToC, executive summary, summary findings, IQ-analysis table, and recommendations sections confirmed it is the earlier R5b revision of the same October 2019 HBBF report that the page already cites (the R6_0 revision under raw/Manual Fetch Kimi /06_Infant_Foods_Formula/). Same authors Houlihan & Brody, same Brooks Applied Labs lab, same 168-baby-foods/61-brands/13-food-types dataset, same 95%/26%-all-four/88%-lack-federal-limits/4-of-7-rice-cereals-above-100-ppb-iAs/83%-above-1-ppb-Pb statistics, same Abt Associates 11+ million IQ-points commission with identical 15-foods table including primary-toxic-metal assignments, same 5 safer-choice substitution table with identical percentages, same FDA-rice-and-juice-guidance framing, same Appendix A-F structure. Registered as a near_duplicates entry so future manual-fetch ticks skip the PDF cleanly; not re-ingested as a parallel source page, preserving the one-study-equals-one-evidence-bearing-page invariant.
• 2026-06-01 byte-identical duplicate registration (manual-fetch v2.0, regulatory-folder pass): the file 16_What's_in_My_Baby's_Food__2019.pdf was encountered in the Kimi-corruption-recovery extract of 05_Regulatory_CPSIA_CPSC_FDA. SHA-256 (a4e1fa75ce93e056d6a864e4a6825cf7537579dd558aa8b121209b2f4a1625b7) and byte length (3,156,916) match the canonical R6_0 PDF at the existing raw_path exactly. This is not even a separate revision — it is the same file bytes filed under a different folder during Kimi corruption recovery. Cover, ToC, and executive-summary pages re-read to confirm identity (Houlihan & Brody October 2019, 168 baby foods, 95% / 26% / 88% / 83% framing, same 5-safer-choices table 93%/91%/84%/68%/up-to-73%). Registered as a near_duplicates entry so future manual-fetch ticks skip the PDF cleanly.
• 2026-06-01 audit-subagent finding applied (Check 1 numerical fidelity, ❌): the prior Key-numbers list quoted “Teething biscuits/rice rusks: 85 ppb average (approximate, from comparative figure)” which did not reconcile with the source. Independent verification against Appendix F page 44 (re-read 2026-06-01) shows the actual sample-weighted total-metals value is 64.09 ppb (HBBF 2019 n=10: 54.61; FDA 2013-14 n=27: 67.60). The 85 ppb figure appears nowhere in Appendix F for this food category — the closest 85 ppb value in the appendix is the HBBF 2017 inorganic-arsenic average for infant rice cereal (not teething biscuits), which suggests the prior page had transposed two adjacent rows. Corrected the teething-biscuit value to 64 ppb and added the Appendix-F provenance anchors. While re-verifying, the rest of the Key-numbers averages were re-grounded to Appendix F’s sample-weighted total-metals column explicitly; the rice-free-baby-snacks value 7.1 ppb (which the audit flagged as ⚠️ unverifiable) was reconciled to Appendix F’s 5.07 ppb sample-weighted HBBF+FDA total-metals value (n=136); the iAs-only basis for the cereal comparison is now labeled in-line on the relevant cereal rows rather than only in the trailing paragraph. The teething-biscuit transposition is the same numerical-fidelity failure mode the GPT-collaboration audit-prompt anticipates (transposed values across adjacent table rows); recording it here so future audits can spot the pattern.
• 2026-06-01 audit-subagent ⚠️ findings noted, not applied as corrections: (a) the executive-summary per-food-type detection counts (puffs 19/21/19/14, non-rice cereals 11/10/11/2, vegetables 38/38/34/38) trace to Appendix A’s per-sample tables rather than the executive summary I cross-checked; the audit could not verify them from the visible appendix pages but they remain consistent with the per-category sample-size framing of the report and are retained without change. (b) The matrices field contains baby-snack, vegetable-puree, fruit-puree, etc., which the public taxonomy snapshot does not enumerate (the snapshot covers ingredients, products, metals, and regulations only); these matrices are retained as the system-prompt’s controlled vocabulary for matrices remains the authority, and these slugs were taxonomy-corrected during the 2026-05-17 audit pass.
• 2026-06-09 byte-identical duplicate path registration (manual-fetch v2.0, June 8 Kimi-corruption-recovery pass): the file 03_2019_BabyFoodReport_EXEC-SUMM-ENGLISH_R6.pdf was encountered a third and fourth time under raw/Manual Fetch Kimi /June 8/Kimi_Agent_Download Corruption Issue/_extracted_infantcontact_01_Bottles_Nipples_SippyCups/01_Bottles_Nipples_SippyCups/ and raw/Manual Fetch Kimi /June 8/Kimi_Agent_Download Corruption Issue/batch1_contact_products/01_Bottles_Nipples_SippyCups/. SHA-256 (4098c155087a5d8a7485707b59249d053ed4f78fd4e149c25aabf11d71218d72) and byte length (896,304) match the two May 21 copies exactly — all four are bit-for-bit the same exec-summary file already registered as a strict content subset of the canonical R6_0 full report on 2026-05-31. Cover, pages 1-5, and acknowledgements page re-read 2026-06-09 to confirm identity (Houlihan & Brody October 2019, version R6 March 2020, 168 baby foods, 95% / 26% / 88% / 83% framing, 4-of-7-rice-cereals iAs above FDA’s 100 ppb action level, 5-safer-choices table with 93%/91%/84%/68%/up-to-73% reductions, healthybabyfood.org URL, Brooks Applied Labs Bothell WA, Abt Associates IQ-loss model). All four byte-identical paths now consolidated into a single near_duplicates entry so future manual-fetch ticks skip the PDF cleanly under either corruption-recovery parent date folder.
• 2026-06-09 byte-identical duplicate path registration (manual-fetch v2.0, June 8 Kimi-corruption-recovery pass, full-report regulatory-folder copy): the file 16_What's_in_My_Baby's_Food__2019.pdf was encountered a second time under raw/Manual Fetch Kimi /June 8/Kimi_Agent_Download Corruption Issue/_extracted_infantcontact_05_Regulatory/05_Regulatory_CPSIA_CPSC_FDA/. SHA-256 (a4e1fa75ce93e056d6a864e4a6825cf7537579dd558aa8b121209b2f4a1625b7) and byte length (3,156,916) match the canonical R6_0 full report at the existing raw_path and the May 21 regulatory-folder copy already registered on 2026-06-01 — all three are bit-for-bit the same 50-page full report file, not a separate revision. Cover, ToC, executive-summary cover, and “What’s new about this study?” pages re-read 2026-06-09 to confirm identity (Houlihan & Brody October 2019, 168 baby foods / 61 brands / 13 food types, 95% / 26% / 88% / 83% framing, 4-metals detection rates 73%/94%/75%/32% per executive-summary test-results panel page 1, 5-safer-choices table with 93%/91%/84%/68%/up-to-73% reductions, healthybabyfood.org URL, Brooks Applied Labs Bothell WA, Abt Associates IQ-loss model, 15-foods-account-for-55%-of-risk framing, Appendix A-F structure). Both byte-identical regulatory-folder paths now consolidated into the existing HBBF-2019-FULLREPORT-ENGLISH-R6_0-regulatory-folder-copy near_duplicates entry so future manual-fetch ticks skip the PDF cleanly under either corruption-recovery parent date folder. No substantive content change to the page; this is path-registration housekeeping only, so no audit-subagent spawn was needed for this revision.

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.