Gestational Polyphenol Levels and Risk of Atopic and Respiratory Outcomes in Early-Life: Insights from The LiNA StudyTo the editor,Maternal diet during pregnancy may potentially influence immune development and subsequent atopic risk in children1. Dietary antioxidant and anti-inflammatory compounds, such as polyphenols, have been associated with a reduced risk of allergies and respiratory diseases2. However, most available evidence on their health effects comes from self-reported food-frequency questionnaires, which are intrinsically susceptible to recall and reporting biases3. Thus, there is a need for objective biomarkers of polyphenol internal exposure during gestation. Here, we investigated the link between maternal polyphenol markers in late pregnancy, individually and as mixtures, with the lifetime prevalence of atopic dermatitis (AD), food sensitization, wheezing, and bronchitis in 3-year-olds (Figure-1A and Figure-S1 ), and assessed whether the children’s Th2 blood cytokines (Table-S1 ) may mediate their association with AD.Our study included 581 mother-child pairs of the prospective birth-cohort LiNA (n=622 at baseline)4 with semiquantitative LC-HRMS-based measurements of 46 food markers in gestational urine samples5 (Table-S2 ). To capture consistent patterns of internal exposure and reduce sparse-data bias, only polyphenol markers with a detection rate >70% (n=14) were selected for analysis (Table-S3 ). Some polyphenols were strongly positively correlated, e.g., isosakuranetin and homoeriodictyol (r =0.97) or negatively correlated, e.g., naringenin and enterolactone (r =-0.83) (Figure-S2B ). All statistical models were fitted using the annotated LC-HRMS signal intensities categorized into tertiles (Ts) and adjusted for environmental and lifestyle variables to control for potential confounding (details in Supporting Information).In the 3-year follow-up subcohort (n=478, Figure-1B ), adjusted logistic regression models showed that children whose mothers ranked the upper tertile (T3) of the flavonoids isosakuranetin (aOR=0.45, 95% CI: 0.22-0.88), norwogonin-glucuronide (aOR=0.42, 95% CI: 0.20-0.84), and pinocembrin (aOR=0.41, 95% CI: 0.20-0.81) had lower odds of AD compared with those in the lowest tertile (T1, used as reference). The upper tertiles of enterolactone and hippuric acid, two microbiota-derived compounds, were also associated with reduced odds of wheezing (aOR=0.50, 95% CI: 0.31-0.80) and bronchitis (aOR=0.60, 95% CI: 0.37-0.95) (Figure-1B ). Next, we examined whether mixtures of major polyphenol groups detected in pregnancy may be associated with health outcomes at age 3. In the quantile g-computation analyses, a one-tertile increase in the mixture of flavonoids was associated with reduced odds of AD (aOR=0.50, 95% CI: 0.27-0.92; p =0.027), with pinocembrin and norwogonin-glucuronide having the largest weights (33% and 23%) in the overall estimate (Figure-2A ). Among microbiota-derived compounds, hippuric acid and enterolactone contributed equally (weights: 44%) to the mixture model linked with lower wheezing risk (aOR=0.68, 95% CI: 0.48-0.96; p =0.029) (Figure-S3 ).Lastly, we performed mediation analyses to assess whether the children’s Th2 cytokine profile (n=268) was consistent with the observed inverse relationship between isosakuranetin, norwogonin-glucuronide, and pinocembrin and AD. The indirect association via IL-5 was statistically significant (-0.022, 95% CI: -0.055, -0.001, p =0.042), whereas the direct association in the outcome model reached only borderline significance (-0.101, 95% CI: -0.201, 0.009, p =0.079) (Figure 2B ). This trend aligns with in vitro andin vivo studies on certain phytochemicals’ capacity to modulate Th2-signaling, involving IL-5 and NF-κB pathways6.Altogether, we identified five polyphenol markers (isosakuranetin, norwogonin-glucuronide, pinocembrin, enterolactone, and hippuric acid) whose relative levels were associated with lower risk of selected childhood outcomes in the LiNA cohort and provided exploratory evidence consistent with IL-5-related immunomodulation in AD. Nonetheless, several limitations must be acknowledged: detection of metabolites in a single urine sample with semiquantification, a relatively small subcohort with blood cytokines, potential parental misreporting, lack of microbiome data, and generalizability limited to Central European populations. Our findings are hypothesis-generating and require replication in larger cohorts with quantitative metabolomics analysis of repeated exposure assessment. This study underscores a gap in approaches that integrate prenatal maternal diet, gut microbiome function, and immune programming to strengthen causal inference.