de Paiva et al. 2020 — Aluminum in cereal-based baby foods (Brazil)
This A-tier peer-reviewed analytical study determines total aluminum and its in vitro bioaccessibility in 35 cereal-based infant cereal samples from the Brazilian market across six product compositions and three brands. Total Al varied from 0.92 mg/kg (corn-flour cereal, brand A) to 8.82 mg/kg (multicereal, brand-aggregated), with multicereals and fruit-plum cereals presenting the highest concentrations and corn-flour and rice-flour cereals the lowest. Bioaccessibility ranged 1.5–10.4% across compositions. Author scope is exact on matrix axis (six compositions explicitly labeled by ingredient mix) and exact on format axis (dry infant cereal as sold; bioaccessibility separately reported with whole-milk reconstitution per manufacturer instructions). Per the corrected row-fit rule (CLAUDE.md Part 6), de Paiva 2020 routes to both baby-cereals-dry-rice-based (the five rice-containing compositions: multicereals, fruit-banana-apple, fruit-plum, rice-and-oat, rice-flour) and baby-cereals-dry-non-rice (corn-flour-only composition).
Key Numbers
Table 2 reports brand × composition group means (mean ± SD across 2-3 batches per cell):
Rice-based subcategory pool (compositions containing rice flour)
| Composition (HMTc rice-based pool) | Brand | n | Al mean ± SD (mg/kg) | Al mean (ppb) |
|---|---|---|---|---|
| Multicereals (rice + wheat + corn + oat + barley) | A | 3 | 3.46 ± 0.19 | 3460 |
| Multicereals | B | 3 | 7.13 ± 0.85 | 7130 |
| Multicereals | C | 2 | 6.59 ± 0.66 | 6590 |
| Fruit-based banana+apple (rice flour + corn starch + fruit) | A | 3 | 2.45 ± 0.19 | 2450 |
| Fruit-based banana+apple | B | 3 | 3.18 ± 0.23 | 3180 |
| Rice flour + oat | A | 3 | 1.99 ± 0.48 | 1990 |
| Rice flour + oat | B | 3 | 4.95 ± 0.43 | 4950 |
| Fruit-based plum (rice flour + oatmeal + malt + plum) | B | 3 | 5.30 ± 1.67 | 5300 |
| Rice flour | A | 3 | 1.30 ± 0.21 | 1300 |
| Rice flour | B | 3 | 3.80 ± 0.13 | 3800 |
| Rice flour | C | 2 | 3.22 ± 0.14 | 3220 |
Rice-based pool n=29 across 11 brand × composition group means. Group-mean range: 1300 – 7130 ppb. Aggregate weighted mean across the 11 group means: ~3920 ppb (treating each group mean as a single data point with its n weighting).
Non-rice subcategory pool (corn flour only)
| Composition | Brand | n | Al mean ± SD (mg/kg) | Al mean (ppb) |
|---|---|---|---|---|
| Corn flour | A | 3 | 1.17 ± 0.16 | 1170 |
| Corn flour | B | 3 | 4.65 ± 0.59 | 4650 |
Non-rice pool n=6 across 2 brand × composition group means. Range: 1170 – 4650 ppb. Below 10-sample defensibility floor; use as triangulation against Chekri 2019 broad cereal Al (mean 630 max 3810 ppb).
Bioaccessibility
In vitro bioaccessible Al fractions (post gastrointestinal digestion model with 15 mL whole milk reconstitution): 1.5–10.4% of total Al across the 35 samples. Higher bioaccessibility in single-flour compositions (rice flour, corn flour) than in multicereals — author hypothesizes phytate and mineral chelators in plant seeds and bran reduce Al bioavailability. The 21-g serving size assumed for risk assessment yields up to 10.48% of the FAO/WHO PTWI of 2 mg/kg b.w. for a 2-6 year-old child consuming three portions daily of the highest-Al multicereal brand B product.
Routing to HMTc subcategories
| Subcategory | Route | n_a_tier impact |
|---|---|---|
| baby-cereals-dry-rice-based | Direct: 5 of 6 product compositions contain rice flour (multicereals, fruit-banana-apple, fruit-plum, rice-flour-oat, rice-flour). 11 brand×composition group means aggregate to n=29 samples. | Closes Al-in-rice-cereal cell from data gap to Path A; combined with Chekri 2019 broad cereal Al (n=17 mean 630 max 3810 ppb) gives n_a_tier=2 summary-level evidence. |
| baby-cereals-dry-non-rice | Direct: corn-flour composition (n=6, 2 brand group means). | Closes Al-in-non-rice-cereal cell from data gap to Path A thin (n_a_tier=1 with Chekri 2019; n=6 below defensibility floor for absolute readiness). |
Geographic-context flag
de Paiva 2020 is Brazilian-market data. Brazilian samples (multicereals 7.13 mg/kg brand B, rice-and-oat 4.95 mg/kg brand B) are higher than Chekri 2019’s French TDS broad cereal Al mean (630 ppb = 0.63 mg/kg). The de Paiva authors note this discrepancy is consistent with prior Brazilian-market data (Souza et al. 2019 reporting up to 17.1 mg/kg for multicereal infant cereal in Brazil); they speculate that open-block sample-prep methodology may facilitate sample contamination from airborne Al but their own closed-microwave-vessel results still show 7+ mg/kg multicereal Al, suggesting the elevated Brazilian values reflect actual ingredient/processing rather than sample-prep artifact. Per Part 6, the geographic mix should be labeled when Brazilian + French + German values are aggregated.
Methods (brief)
Total Al: 0.5 g cereal + 8 mL HNO₃ + 2 mL H₂O₂ digested in closed microwave (Milestone Start E Sorisole) using 4 ramps to 170°C for 25 min; final volume 25 mL, ICP-OES at 396.152 nm with U-series Ezylok nebulizer (Agilent 5100 VDV). LOD 53 µg/kg, LOQ 89 µg/kg. Calibration 2-200 µg/L. Method validation against NIST SRM 1548a (Typical Diet) and EGGS-1 egg-powder CRM (NRCC).
In vitro digestion (Perales, Barberá, Largada, Farré 2006 method, adapted): 2.5 g cereal + 15 mL whole milk + saliva (5 min, 37°C) → pepsin gastric (pH 2.0, 2 h, 37°C) → pancreatin-bile salt intestinal (pH 7.0-7.5, 2 h, 37°C) → centrifugation, dialysate filtration, ICP-OES of bioaccessible fraction.
Evidence Fitness
EF-2 reconstructable A-tier evidence at the brand × composition group-mean level (mean ± SD per cell, n=2-3 per cell). Per-batch sample-level values are not in the published Table 2 (only group means); the authors note batch-level testing was performed (3 batches per brand-composition combination for most cells) but only group statistics are reported. Sample-level extraction would require contacting the authors for the underlying per-batch data; deferred.
Limitations
- Only total Al reported; no per-sample distribution. Sample-level percentile pooling for rice-based and non-rice subcategories not feasible from this paper alone.
- 35-sample total reflects 3 brands × 6 compositions × 2-3 batches per cell; brand-bias possible if one brand dominates a composition.
- “Multicereals” composition includes wheat, rice, corn, oatmeal, barley flours — routing to rice-based subcategory uses the rice-presence rule per HMTc Category 1 architecture; some readers may prefer routing to a “mixed-grain” subcategory if HMTc later adds one.
- Bioaccessibility 1.5-10.4% is much lower than the unbioaccessible-corrected total; if HMTc adopts bioaccessibility as a normalization, the actual ingested Al dose is 10-100× lower than the total-Al value.
- Brazilian-market focus; geographic applicability to U.S./EU markets requires explicit jurisdiction labeling.
Implications
Certification: Direct A-tier summary-level evidence for Al in rice-based cereal (n=29 across 11 brand × composition group means; weighted mean ~3920 ppb) and non-rice cereal (n=6 across 2 group means, range 1170-4650 ppb). Combined with Chekri 2019 (French TDS broad cereal Al), the Al-rice-cereal cell now has n_a_tier=2 summary-level evidence; Al-non-rice-cereal cell n_a_tier=2 summary. Rice-based aggregate Al p90 estimate (informal) ~6500-7000 ppb based on the upper group means; cleaner percentile pooling awaits per-batch sample-level data. Bioaccessibility 1.5-10.4% is a critical risk-modifier for HMTc threshold-setting — total-Al-only thresholds may overstate health risk by 10-50×.
Courses: Useful illustration of bioaccessibility’s modulating effect on heavy-metal exposure assessment, and of geographic variation in baby-cereal Al (Brazilian multicereals 7-8 mg/kg vs French TDS broad cereal mean 0.63 mg/kg).
App: Supports rice-based and non-rice baby-cereal Al contamination_profile values; flag bioaccessibility as a future modeling input.
Microbiome: No direct microbiome endpoint, though the authors discuss antinutritional factors (phytates, fiber) reducing Al bioaccessibility — a potential gut-microbiome-mediated mechanism.
Provenance Notes
Karen externally fetched this paper on 2026-05-09 and dropped it at raw/external-fetch/depaiva2020.pdf. The paper was substituted from the wishlist’s Al-baby-cereal slot (originally Filippini 2020 PMID 32247442 was queued; de Paiva 2020 covers the same gap with Brazilian instead of European data). Published in Food Research International under standard Elsevier paywall; author-shared copy used internally only. Wiki cites the article record (DOI 10.1016/j.foodres.2019.108965).