Lacto-N-difucohexaitol I

• Molecular Formula: C38H65NO29
• Molecular Weight: 999.9 g/mol
• SMILES: CC1C(C(C(C(O1)OC2C(OC(C(C2OC3C(C(C(C(O3)CO)O)O)OC4C(C(C(C(O4)C)O)O)O)NC(=O)C)OC5C(C(OC(C5O)OC6C(OC(C(C6O)O)O)CO)CO)O)CO)O)O)O
• InChI: InChI=1S/C38H65NO29/c1-8-16(45)20(49)25(54)35(58-8)65-29-14(7-43)63-34(67-31-19(48)12(5-41)61-37(27(31)56)64-28-13(6-42)60-33(57)24(53)23(28)52)15(39-10(3)44)30(29)66-38-32(22(51)18(47)11(4-40)62-38)68-36-26(55)21(50)17(46)9(2)59-36/h8-9,11-38,40-43,45-57H,4-7H2,1-3H3,(H,39,44)/t8-,9-,11+,12+,13+,14+,15+,16+,17+,18-,19-,20+,21+,22-,23+,24+,25-,26-,27+,28+,29+,30+,31-,32+,33?,34-,35-,36-,37-,38-/m0/s1
• InChIKey: RQNFGIWYOACERD-OCQMRBNYSA-N

Lacto-N-difucohexaitol I is the sodium borohydride-reduced alditol form of lacto-N-difucohexaose I (LNDFH I; CAS 16789-38-1), a neutral, di-fucosylated hexasaccharide belonging to the lacto-N-tetraose (LNT) series of human milk oligosaccharides (HMOs). Its core structure, Fucα1-2Galβ1-3[Fucα1-4]GlcNAcβ1-3Galβ1-4Glc-ol, incorporates the Lewis b (Leᵇ) blood group tetrasaccharide epitope and terminates in glucitol rather than free reducing glucose, conferring resistance to alkaline degradation and making it a defined substrate for α-L-fucosidase enzymes.

Lacto-N-difucohexaitol I (Reduced LNDFH I) — Structural Identity and Procurement-Relevant Classification

Lacto-N-difucohexaitol I is the sodium borohydride-reduced alditol form of lacto-N-difucohexaose I (LNDFH I; CAS 16789-38-1), a neutral, di-fucosylated hexasaccharide belonging to the lacto-N-tetraose (LNT) series of human milk oligosaccharides (HMOs) [1]. Its core structure, Fucα1-2Galβ1-3[Fucα1-4]GlcNAcβ1-3Galβ1-4Glc-ol, incorporates the Lewis b (Leᵇ) blood group tetrasaccharide epitope and terminates in glucitol rather than free reducing glucose, conferring resistance to alkaline degradation and making it a defined substrate for α-L-fucosidase enzymes [2]. The compound is exclusively produced in the milk of secretor-positive, Lewis-positive mothers and is absent from non-secretor milk, establishing it as a genotype-dependent biomarker and a functionally non-interchangeable glycan for research and formulation applications [3].

• [1] National Center for Biotechnology Information. PubChem Compound Summary for CID 16219584, Lacto-N-difucohexaose I. Available at: https://pubchem.ncbi.nlm.nih.gov/compound/16219584 View Source
• [2] Immune Epitope Database (IEDB). Epitope: alpha-L-Fucp-(1->4)-[alpha-L-Fucp-(1->2)-beta-D-Galp-(1->3)]-beta-D-GlcpNAc-(1->3)-beta-D-Galp-(1->4)-D-Glcp (Lacto-N-difucohexaitol I). Available at: https://iedb.org/ View Source
• [3] Gu F. Page 218, Table 7.4: Concentrations of 17 human milk oligosaccharides (g/L) in full-term human milk collected around 1 month postpartum, from KOALA cohort study (n=121). In: Human Milk Oligosaccharides: Occurrence, Structures and Functions. Global Academic Press. View Source

Why 2′-FL, LNFP I, or LNT Cannot Substitute for Lacto-N-difucohexaitol I in Research and Formulation

Generic substitution with more abundant HMOs such as 2′-fucosyllactose (2′-FL), lacto-N-fucopentaose I (LNFP I), or the non-fucosylated core lacto-N-tetraose (LNT) fails because none of these compounds carries the Lewis b (Leᵇ) di-fucosylated epitope — the α1,2-fucose on the terminal galactose plus the α1,4-fucose on the subterminal GlcNAc — that defines LNDFH I's molecular recognition properties [1]. This dual fucosylation pattern is the structural prerequisite for binding to the Helicobacter pylori blood group antigen-binding adhesin (BabA) and for recognition by human rotavirus VP8* domains, functions that 2′-FL (single α1,2-fucose on lactose), LNFP I (single α1,2-fucose on LNT), and LNT (no fucose) cannot replicate [2]. Furthermore, the reduced alditol terminus of lacto-N-difucohexaitol I eliminates the anomeric equilibrium present in the reducing form, providing a chemically homogeneous single species essential for quantitative enzymatic assays (e.g., EC 3.2.1.51 α-L-fucosidase) where variable anomer ratios would introduce measurement error [3]. Procurement of the correct compound is therefore dictated by the specific epitope requirement (Leᵇ vs. H-type-1, Leᵃ, or non-fucosylated) and by the assay chemistry (reduced vs. reducing terminus).

• [1] Yang L, Zhu Y, Zhang W, Mu W. Recent progress in fucosylated derivatives of lacto-N-tetraose and lacto-N-neotetraose. Crit Rev Food Sci Nutr. 2024;64(28):10384-10396. doi:10.1080/10408398.2023.2224431. PMID: 37341681. View Source
• [2] Miyazaki T, Kato H, Ajisaka K. Enzymatic synthesis of lacto-N-difucohexaose I which binds to Helicobacter pylori. Methods Enzymol. 2010;480:511-524. doi:10.1016/S0076-6879(10)80023-0. PMID: 20816225. View Source
• [3] BRENDA Enzyme Database. EC 3.2.1.51 — alpha-L-fucosidase: Substrate lacto-N-difucohexaitol I (NaBH4-reduced oligosaccharide). Available at: https://www.brenda-enzymes.org/ View Source

Quantitative Differentiation Evidence for Lacto-N-difucohexaitol I Against Its Closest Analogs

Lewis b Blood Group Epitope: Structural Determinant Absent in 2′-FL, LNFP I, and LNT

Lacto-N-difucohexaitol I (LNDFH I) is the only major human milk oligosaccharide that presents the complete Lewis b (Leᵇ) tetrasaccharide epitope — Fucα1-2Galβ1-3[Fucα1-4]GlcNAc — requiring both an α1,2-linked fucose on galactose and an α1,4-linked fucose on N-acetylglucosamine [1]. The closest analogs lack one or both of these fucosyl linkages: 2′-FL (Fucα1-2Galβ1-4Glc) contains only the α1,2-fucose on a lactose core; LNFP I (Fucα1-2Galβ1-3GlcNAcβ1-3Galβ1-4Glc) carries only the α1,2-fucose on the LNT backbone; and LNT (Galβ1-3GlcNAcβ1-3Galβ1-4Glc) is entirely non-fucosylated [1]. The functional consequence is that LNDFH I specifically binds the H. pylori BabA adhesin, a property confirmed by direct bacterial binding assays, whereas 2′-FL and LNT show no detectable BabA interaction [2].

Evidence Dimension	Fucosylation pattern and Lewis b epitope presence
Target Compound Data	Fucα1-2Galβ1-3[Fucα1-4]GlcNAcβ1-3Galβ1-4Glc-ol (di-fucosylated; Leᵇ-positive; two fucose residues: α1,2 + α1,4)
Comparator Or Baseline	2′-FL: Fucα1-2Galβ1-4Glc (mono-fucosylated; no Leᵇ). LNFP I: Fucα1-2Galβ1-3GlcNAcβ1-3Galβ1-4Glc (mono-fucosylated; no Leᵇ). LNT: Galβ1-3GlcNAcβ1-3Galβ1-4Glc (zero fucose; no Leᵇ)
Quantified Difference	LNDFH I is the only member of this set possessing dual fucosylation (α1,2 + α1,4) and the Leᵇ epitope; comparators carry 0 or 1 fucose and are Leᵇ-negative. Functional binding to BabA: positive for LNDFH I, negative for 2′-FL and LNT.
Conditions	Structural comparison based on established HMO sequences; BabA binding verified by H. pylori whole-cell adhesion assay (Miyazaki et al., 2010)

Why This Matters

For any application requiring Leᵇ-specific molecular recognition (H. pylori anti-adhesion studies, rotavirus VP8* binding assays, Leᵇ antibody development), only a compound bearing the di-fucosylated Leᵇ epitope is functionally competent; procurement of 2′-FL, LNFP I, or LNT would yield null results.

• [1] Yang L, Zhu Y, Zhang W, Mu W. Recent progress in fucosylated derivatives of lacto-N-tetraose and lacto-N-neotetraose. Crit Rev Food Sci Nutr. 2024;64(28):10384-10396. doi:10.1080/10408398.2023.2224431. PMID: 37341681. View Source
• [2] Miyazaki T, Kato H, Ajisaka K. Enzymatic synthesis of lacto-N-difucohexaose I which binds to Helicobacter pylori. Methods Enzymol. 2010;480:511-524. doi:10.1016/S0076-6879(10)80023-0. PMID: 20816225. View Source

Human Milk Concentration: LNDFH I Is 5.7-Fold More Abundant Than Its Isomer LNDFH II

A systematic review aggregating data from 18 term-milk studies (353 mothers, 556 samples for 2′-FL; 61 mothers, 214 samples for LNDFH I) reports mean neutral HMO concentrations with 95% confidence intervals: LNDFH I at 0.80 g/L (95% CI 0.66–0.94), compared with 2′-FL at 2.74 g/L (2.43–3.04), LNFP I at 1.31 g/L (1.08–1.53), LNT at 0.79 g/L (0.59–0.98), and LNDFH II at 0.14 g/L (0.10–0.18) [1]. In the KOALA cohort (n=121, Le⁺Se⁺ mothers, ~1 month postpartum), LNDFH I concentration was 0.70 g/L (SD 0.27), ranking it as the most abundant single HMO in that phenotype after LNFP I (0.61 g/L) and 2′-FL (0.49 g/L) [2]. Critically, LNDFH I is 5.7-fold more abundant than its structural isomer LNDFH II (0.80 vs. 0.14 g/L), reflecting the distinct biosynthetic regulation of α1,2- versus α1,3/4-fucosylation in the mammary gland [1].

Evidence Dimension	Mean concentration in term human milk (secretor mothers)
Target Compound Data	LNDFH I: 0.80 g/L (95% CI 0.66–0.94); KOALA Le⁺Se⁺: 0.70 g/L (SD 0.27)
Comparator Or Baseline	2′-FL: 2.74 g/L (2.43–3.04). LNFP I: 1.31 g/L (1.08–1.53). LNT: 0.79 g/L (0.59–0.98). LNDFH II: 0.14 g/L (0.10–0.18)
Quantified Difference	LNDFH I is 3.4× less abundant than 2′-FL, 1.6× less than LNFP I, approximately equal to LNT (+1.3%), and 5.7× more abundant than LNDFH II
Conditions	Systematic review of term milk from secretor mothers, various lactation stages (0 to >100 days), multiple quantification methods; KOALA cohort: 1 month postpartum, HPLC with fluorescence detection

Why This Matters

In infant formula development aiming to mimic secretor-milk HMO profiles, LNDFH I must be dosed at approximately 0.7–0.8 g/L, significantly higher than LNDFH II (~0.14 g/L) but lower than 2′-FL (~2.7 g/L); substituting one for another without adjusting concentration would produce a non-physiological HMO profile.

• [1] Soyyılmaz B, Mikš MH, Röhrig CH, Matwiejuk M, Meszaros-Matwiejuk A, Vigsnæs LK. The Mean of Milk: A Review of Human Milk Oligosaccharide Concentrations throughout Lactation. Nutrients. 2021;13(8):2737. Data extracted from Table 3 of systematic review available via Oxford Journals. View Source
• [2] Gu F. Page 218, Table 7.4: Concentrations of 17 human milk oligosaccharides (g/L) in full-term human milk collected around 1 month postpartum, from KOALA cohort study (n=121). In: Human Milk Oligosaccharides: Occurrence, Structures and Functions. Global Academic Press. View Source

Structural Isomer Functional Divergence: LNDFH I (Leᵇ) vs. LNDFH II (Leᵃ) — Distinct Lectin and Viral Binding Profiles

LNDFH I (Lewis b, type 1 chain, α1,2 + α1,4 fucosylation) and LNDFH II (Lewis a, type 1 chain, α1,3 + α1,4 fucosylation) are constitutional isomers with divergent biorecognition properties. The YesU lectin from Vulcanisaeta distributa binds LNDFH II with a dissociation constant KD = 0.34 ± 0.01 μM, whereas LNDFH I shows no detectable binding in the same assay [1]. Conversely, LNDFH I is co-crystallized with human rotavirus VP8* domains from P[8] (PDB 7JHZ, resolution 2.68 Å) and P[4] (PDB 7KHU) genotypes, demonstrating a specific histo-blood group antigen recognition mechanism that LNDFH II does not support for these genotypes [2]. The di-fucosylated isomer LNnDFH I (type 2 chain, α1,2 + α1,3 fucosylation) binds YesU with KD = 1.14 ± 0.03 μM, 3.4-fold weaker than LNDFH II [1].

Evidence Dimension	Lectin binding affinity and viral VP8* domain co-crystallization
Target Compound Data	LNDFH I: no YesU binding detected; co-crystallized with P[8] VP8* (PDB 7JHZ, 2.68 Å) and P[4] VP8* (PDB 7KHU)
Comparator Or Baseline	LNDFH II: YesU KD = 0.34 ± 0.01 μM. LNnDFH I: YesU KD = 1.14 ± 0.03 μM. LNFP I: YesU KD = 1.23 ± 0.02 μM
Quantified Difference	LNDFH II binds YesU with sub-micromolar affinity; LNDFH I shows no binding. LNDFH I binds P[8]/P[4] rotavirus VP8*; LNDFH II does not (genotype-dependent). LNnDFH I binds YesU 3.4× weaker than LNDFH II.
Conditions	YesU: SPR on NTA chip, 50 μg/mL His-tagged lectin, pH 7.4. Rotavirus VP8*: X-ray crystallography, recombinant VP8* domains, hanging-drop vapor diffusion.

Why This Matters

Researchers investigating rotavirus–host glycan interactions or developing anti-adhesion therapeutics against specific rotavirus genotypes must procure LNDFH I (Leᵇ) rather than LNDFH II (Leᵃ); using the incorrect isomer would produce false negatives in VP8* binding screens.

• [1] Table 1: Dissociation constants (KD) of YesU (50 μg/mL) His tagged to NTA chip for selected glycans. Sci Rep. 2018;8:Article 31241. doi:10.1038/s41598-018-31241-8. View Source
• [2] Xu S, Stuckert MR, McGinnis KR, Jiang X, Kennedy MA. Crystal structure of the carbohydrate-binding domain VP8* of human P[8] rotavirus strain BM13851 in complex with LNDFH I. RCSB PDB 7JHZ, deposited 2020-07-21, released 2021-07-28. doi:10.2210/pdb7jhz/pdb. View Source

Enzymatic Synthesis Complexity: LNDFH I Requires Four Glycosyltransferase Steps with 6% Overall Yield

The enzymatic synthesis of LNDFH I proceeds through four sequential steps, each mediated by a distinct glycosyltransferase or glycosidase: (1) β-1,3-N-acetylglucosaminyltransferase adds GlcNAc to lactose at 44% yield, producing lacto-N-triose II; (2) recombinant β-1,3-galactosidase from Bacillus circulans attaches galactose via transglycosylation at 22% yield, forming LNT; (3) recombinant human FUT1 (α1,2-fucosyltransferase) attaches the first fucose to LNT at 71% yield, yielding LNFP I; and (4) commercial FUT3 (α1,3/4-fucosyltransferase) attaches the second fucose to LNFP I at 85% yield, producing 1.7 mg of purified LNDFH I with an overall yield of 6% [1]. In contrast, the monofucosylated intermediate LNFP I is obtained after only three steps, and the core LNT after two steps — both achieved at substantially higher cumulative yields because fewer enzymatic conversions are required [1].

Evidence Dimension	Number of enzymatic steps and cumulative synthesis yield
Target Compound Data	LNDFH I: 4 enzymatic steps; step yields 44%, 22%, 71%, 85%; overall yield 6%; final purified amount 1.7 mg
Comparator Or Baseline	LNFP I (monofucosylated intermediate): 3 enzymatic steps; fewer steps, higher cumulative yield (not explicitly calculated but necessarily >6% as it is an intermediate). LNT (core): 2 enzymatic steps; cumulative yield ~9.7% (44% × 22%)
Quantified Difference	LNDFH I requires 1 additional fucosyltransferase step beyond LNFP I, reducing overall yield by the factor of that step (85% per-step yield). LNDFH I overall yield (6%) is approximately 1.6× lower than LNT cumulative yield (~9.7%).
Conditions	In vitro multi-enzyme cascade; partially purified β-1,3-GnT from bovine blood; recombinant β-1,3-galactosidase from B. circulans; recombinant human FUT1 expressed in baculovirus; commercial FUT3; activated carbon column purification

Why This Matters

Procurement decisions for synthetic LNDFH I must account for the substantially higher production cost and lower availability relative to LNFP I or LNT, driven by the additional enzymatic step and cumulative yield penalty; this directly impacts bulk pricing and lead times for large-scale orders.

• [1] Miyazaki T, Kato H, Ajisaka K. Enzymatic synthesis of lacto-N-difucohexaose I which binds to Helicobacter pylori. Methods Enzymol. 2010;480:511-524. doi:10.1016/S0076-6879(10)80023-0. PMID: 20816225. View Source

Infant Growth Parameter Association: LNDFH I Positively Correlated with Head Circumference and Length, in Contrast to 3FL and MFLNH-III

In a Philippine mother-infant dyad study (n=41, exclusively breastfed, first 4 months of lactation), seven HMOs showed statistically significant associations with infant growth parameters at 6 months. Among these, LNDFH I was positively associated with at least one growth parameter — specifically, the area under the concentration-time curve (AUC) for LNDFH I over the first 4 months correlated positively with head circumference and head circumference-for-age z-score at 6 months [1]. In contrast, 3-fucosyllactose (3FL) AUC was inversely associated with length gain and length gain-for-age z-score, and monofucosyllacto-N-hexaose-III (MFLNH-III) AUC was inversely associated with head circumference, weight, and length metrics [1]. Lactodifucotetraose (LDFT), LNnDFH, and 2′-FL also showed positive associations, but the specific parameter pattern for LNDFH I (head circumference and length) distinguishes it from the broader growth associations of LDFT and 2′-FL [1].

Evidence Dimension	Association between HMO concentration AUC (0–4 months) and infant growth parameters at 6 months
Target Compound Data	LNDFH I: positive association with head circumference and head circumference-for-age z-score at 6 months (AUC over first 4 months)
Comparator Or Baseline	3FL: inverse association with length gain and length gain-for-age z-score. MFLNH-III: inverse association with head circumference, weight, length, and their z-scores. 2′-FL: positive association with length gain and length gain-for-age z-score. LDFT: positive association with head circumference and length.
Quantified Difference	LNDFH I positive association direction contrasts with 3FL (inverse) and MFLNH-III (inverse). Effect magnitude described as 'significant but weak' (exploratory analysis); exact β coefficients not reported in the accessible summary
Conditions	Prospective observational cohort; 41 Filipino mother-infant dyads; exclusively breastfed; HMOs quantified by HPLC-MS at 0.75, 1.5, 2.5, and 4 months; growth measured at 6 months

Why This Matters

For infant formula manufacturers selecting HMOs to support neurodevelopment (head circumference as a proxy), LNDFH I offers a distinct growth association profile that is not replicated by 3FL or MFLNH-III; substituting LNDFH I with a structurally unrelated HMO that shows inverse growth associations could produce unintended developmental outcomes.

• [1] Manalac T. Human milk oligosaccharides tied to infant growth. MIMS. 2022 Oct 19. Reporting on: Sci Rep. 2022;12:17304. doi:10.1038/s41598-022-22244-7. View Source

Procurement-Driven Application Scenarios for Lacto-N-difucohexaitol I Based on Quantitative Differentiation Evidence

Helicobacter pylori BabA Anti-Adhesion Drug Discovery

LNDFH I is the minimal soluble oligosaccharide containing the complete Lewis b epitope recognized by the H. pylori BabA adhesin. As confirmed by direct bacterial binding assays and the established structural requirement for dual α1,2/α1,4 fucosylation, neither 2′-FL, LNFP I, nor LNT can serve as functional surrogates in competitive adhesion inhibition screens [1]. Researchers developing BabA antagonists should procure LNDFH I (or its reduced alditol form for enhanced chemical stability) as the reference ligand for surface plasmon resonance (SPR) competition assays and for co-crystallization with recombinant BabA. The reduced form, lacto-N-difucohexaitol I, is preferred for prolonged incubation experiments due to its resistance to alkaline peeling and reducing-end degradation [2].

Rotavirus P[8]/P[4] Genotype VP8* Glycan Receptor Structural Biology

Co-crystal structures of LNDFH I bound to human rotavirus P[8] VP8* (PDB 7JHZ, 2.68 Å) and P[4] VP8* (PDB 7KHU) establish this glycan as a defined histo-blood group antigen ligand for the dominant P[II] genogroup rotaviruses that cause the majority of severe pediatric gastroenteritis worldwide [3]. LNDFH II and LNnDFH I do not support the same VP8* binding mode for these genotypes. Structural biologists and antiviral researchers should procure LNDFH I (CAS 16789-38-1, or its alditol form) for VP8* binding studies, competitive displacement assays with candidate entry inhibitors, and as a positive control in glycan microarray screens. The reduced alditol form eliminates anomeric heterogeneity in NMR-based binding studies [2].

Infant Formula HMO Profile Design for Secretor-Positive Milk Biomimicry

In secretor-positive, Lewis-positive mothers (Le⁺Se⁺), LNDFH I is present at a mean concentration of 0.70–0.80 g/L, constituting one of the five dominant HMOs alongside 2′-FL (~2.74 g/L), LNFP I (~1.31 g/L), LNT (~0.79 g/L), and LNnT (~0.74 g/L) [4]. The 5.7-fold concentration ratio between LNDFH I (0.80 g/L) and its isomer LNDFH II (0.14 g/L) reflects natural mammary gland biosynthetic regulation and must be replicated in any formula claiming to approximate the HMO composition of secretor milk. Infant formula developers should incorporate LNDFH I at approximately 0.7–0.8 g/L, using analytically validated reference standards, rather than attempting to substitute this component with LNDFH II or additional 2′-FL, which would distort the physiological HMO profile [4].

α-L-Fucosidase Substrate for Enzyme Characterization and Inhibitor Screening

Lacto-N-difucohexaitol I (the NaBH4-reduced form) is a registered substrate for α-L-fucosidase (EC 3.2.1.51) in the BRENDA enzyme database, where it is specifically hydrolyzed to yield α-L-fucose and the monofucosylated product Galβ1-3[Fucα1-4]GlcNAcβ1-3Galβ1-4Glc-ol [2]. The reduced glucitol terminus prevents both alkaline peeling and anomerization, providing a chemically homogeneous substrate essential for reproducible kinetic measurements (Km, Vmax, kcat). Enzyme biochemists characterizing novel fucosidases from gut microbiota (e.g., Bifidobacterium longum subsp. infantis, Pomacea canaliculata) or screening fucosidase inhibitors should use the reduced alditol form to eliminate the confounding variable of mutarotation that plagues the reducing form. This specific application is uniquely served by the reduced compound and cannot be addressed using non-reduced LNDFH I.

• [1] Miyazaki T, Kato H, Ajisaka K. Enzymatic synthesis of lacto-N-difucohexaose I which binds to Helicobacter pylori. Methods Enzymol. 2010;480:511-524. doi:10.1016/S0076-6879(10)80023-0. PMID: 20816225. View Source
• [2] BRENDA Enzyme Database. EC 3.2.1.51 — alpha-L-fucosidase: Substrate lacto-N-difucohexaitol I (NaBH4-reduced oligosaccharide). Available at: https://www.brenda-enzymes.org/ View Source
• [3] Xu S, Stuckert MR, McGinnis KR, Jiang X, Kennedy MA. Crystal structure of the carbohydrate-binding domain VP8* of human P[8] rotavirus strain BM13851 in complex with LNDFH I. RCSB PDB 7JHZ. Also PDB 7KHU for P[4] VP8* with LNDFH I. View Source
• [4] Soyyılmaz B, Mikš MH, Röhrig CH, Matwiejuk M, Meszaros-Matwiejuk A, Vigsnæs LK. The Mean of Milk: A Review of Human Milk Oligosaccharide Concentrations throughout Lactation. Nutrients. 2021;13(8):2737. And Gu F, Table 7.4 (KOALA cohort), Global Academic Press. View Source