Sipahi et al. 2014 — Toxic and essential metals in Turkish infant formulas
This A-tier peer-reviewed analytical study determines six metals (Pb, Cd, Al, Mn, Cr, Co) in 63 infant foods and formulas from 21 manufacturers acquired in Ankara, Turkey in 2006. The samples span three groups: cereal-based (n=23, single or combination of wheat/rice/corn/oats), milk-based (n=28, single or combination of milk forms — not split by soy vs non-soy in the paper), and mixed (n=12, cereals + milk + fruit + vegetables). Author scope is partial on matrix axis (groups distinguished but soy/non-soy split absent for milk-based; rice/non-rice composition not enumerated per sample for cereal-based) and partial on format axis (powder vs liquid vs reconstituted not consistently labeled). Per CLAUDE.md Part 6, this routes to multiple HMTc Category 1 subcategories at summary-level with explicit row-fit caveats. The paper concludes that metal levels in infant formulas are higher than in breast milk, supporting the recommendation that breast milk be preferred when feasible and that infant foods be regularly monitored.
Key Numbers (Table II — group means ± SD)
Note on units: Table II header reads “Al (ng/g)” but the discussion text refers to the same Al values as “µg/g”. Cross-checking the paper’s PTDI calculations suggests Al values are likely intended as ng/g per the table header (8.02 ng/g = 8.02 ppb total, which is implausibly low for infant formula compared to other studies reporting 400-1700 ppb), OR µg/g per the discussion (8.02 µg/g = 8020 ppb, plausible but on the high end). The discussion’s daily-intake calculation (1008.3 µg/kg bw/day from highest-Al sample at 42.35 unit/g for a 4.2 kg infant consuming 75 g/day) only yields 1008 with Al in µg/g, so the µg/g interpretation appears more consistent with the paper’s own arithmetic. Reported here as published in Table II with the unit ambiguity flagged for downstream use.
| Group | n | Pb (ng/g) | Cd (ng/g) | Al (ng/g per table; possibly µg/g per discussion) | Mn (µg/g) | Cr (ng/g) | Co (ng/g) |
|---|---|---|---|---|---|---|---|
| Milk-based | 28 | 7.68 ± 4.08 (range 3.75-24.90) | 0.96 ± 1.46 (range ND-7.49) | 8.02 ± 8.61 (range 2.40-35.55) | 5.73 ± 13.52 (range 0.49-60.36) | 40.76 ± 59.10 (range 10.15-240.03) | 120.93 ± 80.07 (range 53.65-380.17) |
| Cereal-based | 23 | 6.96 ± 4.31 (range 0.55-17.45) | 8.88 ± 8.79 (range 0.04-27.23) | 7.94 ± 10.33 (range 1.70-42.35) | 7.18 ± 5.87 (range ND-21.69) | 42.57 ± 44.95 (range 12.55-218.48) | 28.29 ± 29.62 (range 0.55-104.95) |
| Mixed | 12 | 6.22 ± 3.26 (range 1.80-12.20) | 5.50 ± 4.29 (range 0.46-13.16) | 8.19 ± 5.35 (range 2.16-19.46) | 2.91 ± 2.40 (range 0.02-6.49) | 39.44 ± 19.67 (range 10.90-80.65) | 58.82 ± 30.15 (range 23.65-127.15) |
| Total (all 63) | 63 | 7.14 ± 4.00 | 4.72 ± 6.67 | 8.02 ± 8.67 | 5.72 ± 9.75 | 41.17 ± 48.04 | 75.28 ± 71.26 |
Cd is significantly higher in cereal-based and mixed than in milk-based (p<0.001), consistent with the broader pattern that grain-based foods accumulate more Cd than milk-based products.
Pb mean across all 63 samples: 7.14 ng/g = 7.14 ppb, consistent with FDA 2026 special-survey milk-formula values (mean 0.4 ppb but max 0.6 ppb for non-soy powder; Sipahi 2014 Turkish samples are notably higher) and with Igweze 2020 Nigerian formula context. Geographic-context flag applies.
Cd mean cereal-based (8.88 ppb) is comparable to FDA 2024 baby-food rice-cereal Cd p50 (10.20 ppb) and within the FDA 2024 non-rice cereal Cd p50 (15.00 ppb) range — Turkish cereal-based infant foods sit between U.S. rice and non-rice cereal Cd levels.
Routing to HMTc subcategories
| Subcategory | Route | n_a_tier impact |
|---|---|---|
| infant-formula-powder-non-soy | Indirect: milk-based n=28 not split by soy/non-soy or by powder/RTF format. Per Part 6, the milk-based group routes to non-soy by default (soy is the comparator and would have been called out separately if measured). Format axis unresolved. | Adds n_a_tier=1 (already at n_a_tier=5+ for Pb/Cd/Al cells; Sipahi brings to n_a_tier=6+). Confidence stays medium. |
| infant-formula-rtf-liquid-non-soy | Same as above; Sipahi did not distinguish powder vs liquid formula format. | Adds n_a_tier=1 to existing n_a_tier=2 (Hernandez 2019 + Saraiva 2021); brings to n_a_tier=3 (medium confidence per Part 6). Approaches the readiness bar for Pb/Cd cells. |
| baby-cereals-dry-rice-based | Indirect: cereal-based n=23 includes rice in single or combination compositions; specific rice-containing subset count not enumerated. Per FDA 2016 multigrain-with-rice convention, the n=23 cereal-based group is broad-cereal context for both rice and non-rice cereal cells. | Adds n_a_tier=1 to existing Pb/Cd/Al cells. |
| baby-cereals-dry-non-rice | Same as above; n=23 cereal-based covers non-rice grains too. | Adds n_a_tier=1. |
| mixed-meals-rice-containing | Indirect: mixed n=12 (cereals + milk + fruit + vegetables); rice-containing status not enumerated. Broad mixed-meal context. | Adds n_a_tier=1 to existing Pb/Cd/tAs (was Path A thin n_a_tier=1; now n_a_tier=2). |
| mixed-meals-non-rice | Same as above. | Adds n_a_tier=1. |
Geographic-context flag
Sipahi 2014 covers Turkish-market products from 2006. Pb in milk-based (mean 7.68 ppb) is substantially higher than U.S. FDA 2026 special-survey milk-formula Pb (mean 0.4 ppb p90, p100 0.6 ppb), more than 10× difference. This is consistent with the broader pattern that Turkish, Nigerian, Brazilian, and South Asian markets show higher infant formula Pb values than U.S./EU regulated markets. The 2006 sample year also predates EU’s 2013 Pb action level for infant formula (10 ppb), so post-2013 Turkish products may be lower. Per Part 6, jurisdiction composition must be labeled when this source contributes to a global aggregate.
Methods (brief)
Total metals: 0.5 g sample digested in 6 mL HNO₃ (65%, Merck) + 1 mL H₂O₂ (30%, Merck) in Milestone MLS-1200 MEGA microwave digestion unit. Analysis on Varian 30/40 atomic absorption spectrophotometer with Varian GTA96 graphite tube atomizer and DS-15 data station. Wavelengths 309.3 nm (Al), 242.5 (Co), 283.3 (Pb), 357.9 (Cr), 228.8 (Cd), 279.5 (Mn). Pyrolysis temperatures 500-2400°C; atomization 1500-2450°C. Standard concentrations 0.5-50 ng/mL depending on metal. ANOVA (Kruskal-Wallis) for between-group differences; Mann-Whitney U for pairwise.
Evidence Fitness
EF-3 limited evidence — group means ± SD with min/max range, but no per-sample tabulation. Sample-level percentile pooling not feasible from this paper alone. The Al unit-labeling inconsistency between table header (ng/g) and discussion text (µg/g) is a notable QC concern; downstream HMTc-aggregate use should treat Al values with explicit caveat until a follow-up paper or correspondence resolves it.
Limitations
- 2006 sample collection — older than ideal; Turkish market regulations may have evolved since.
- Milk-based group (n=28) not split by soy vs non-soy or by powder vs liquid format.
- Cereal-based group (n=23) not enumerated by rice presence.
- Mixed group (n=12) not enumerated by rice presence.
- Al unit-labeling inconsistency between table and discussion (see “Note on units” above).
- GFAAS method is older than ICP-MS; LOD/LOQ values not explicitly published in the paper.
- Cobalt and manganese reported but are not HMT&C analytes (HMT&C tracks Pb, tAs, iAs, Cd, MeHg, tHg, Ni, Al, Cr-VI, Sn).
- Total Cr reported but Cr(VI) not speciated; per Part 14 do not interpret total Cr as Cr-VI. Hernandez 2019 + Saraiva 2021 confirm Cr(VI) is essentially absent in food matrices, so total Cr in Sipahi 2014 is likely all Cr(III).
Implications
Certification: Adds n_a_tier breadth to multiple Pb, Cd, Al cells across milk-based formula, cereal-based, and mixed-meal subcategories. Particularly impactful for the milk-based RTF formula cells (was n_a_tier=2 with Hernandez + Saraiva for Cr-VI specifically; Sipahi adds Pb/Cd/Al at n_a_tier=3, satisfying medium-confidence readiness for those cells). Geographic context (Turkish 2006 market) provides triangulation against the dominant U.S./EU/Brazilian sources but should not be silently merged without jurisdiction labeling.
Courses: Useful for teaching the geographic-variation-in-infant-formula story (Turkish > U.S. on Pb by 10× at this sample year) and the importance of metal speciation (Cr total reported here, but Cr(VI) speciation needs Hernandez 2019 / Saraiva 2021 to interpret).
App: Supports milk-based formula and cereal-based infant food contamination_profile values for Pb, Cd, Al, Mn (Mn for app-layer infant nutrition tracking, not HMT&C threshold).
Microbiome: No direct microbiome endpoint.
Provenance Notes
Karen externally fetched this paper on 2026-05-09 and dropped it at raw/external-fetch/2014_Sipahi_TurkishJPediatric.pdf. Published in The Turkish Journal of Pediatrics (open online journal but article-level access varies).