Potential mechanisms and modifications of dietary antioxidants on the associations between co-exposure to plastic additives and diabetes

Plastics are ubiquitous in modern life and contribute substantial environmental contamination due to large production volumes, low recycling rates, and additive leaching. Phthalate esters (PAEs) and organophosphate esters (OPEs) are widely used as plasticizers and flame retardants and are not chemically bound to products, leading to human co-exposure via ingestion, inhalation, and dermal routes. Prior studies link individual PAE or OPE exposures with diabetes, insulin resistance, and other adverse outcomes, but combined exposure effects on diabetes and underlying mechanisms remain unclear. Oxidative stress, a proposed pathway for diabetogenic effects of PAEs/OPEs, may be mitigated by dietary antioxidants. This study used NHANES 2011–2018 to: (1) examine associations between combined exposure to PAEs and OPEs and diabetes; (2) assess modification by dietary antioxidant intake via the composite dietary antioxidant index (CDAI); and (3) explore potential mechanisms using an adverse outcome pathway (AOP) approach integrating CTD, DisGeNET, and MalaCards.

Design and population: Cross-sectional analysis of NHANES 2011–2018 combining four cycles to enhance stability of PAE/OPE measurements. Of 5434 participants with exposure and covariate data, exclusions were age <18 (n=2046), pregnancy (n=29), missing outcome (n=3), or missing exposure/covariates (n=514), yielding 2824 adults. Diabetes status was self-reported (“Did you have diabetes or sugar diabetes?”). Two outcome scenarios were defined: (1) diabetes including borderline high blood sugar; (2) diabetes excluding borderline cases (borderline treated as non-diabetic). Exposure assessment: Urinary metabolites of 10 PAEs and 5 OPEs were quantified using enzymatic hydrolysis, automated SPE, RP-HPLC, and isotope-dilution LC–MS/MS per NHANES protocols. Metabolites with <30% missing were included: PAE metabolites (MEP, MnBP, MIBP, MBzP, MEHHP, MEOHP, MECPP, MCPP, MCNP, MCOP) and OPE metabolites (BCPP, BCEP, BDCP, DBUP, DPHP). Values below LOD were imputed as LOD/√2. Concentrations were creatinine-adjusted (µg/g creatinine). Spearman correlations among metabolites were summarized. Covariates: Sampling time, sex, age, race/ethnicity, marital status, education, BMI, poverty-to-income ratio, smoking (≥100 cigarettes lifetime), alcohol intake (≥12 drinks/year), hypertension, gout, family history of diabetes, energy, protein, carbohydrate, total fat intakes, liver function (ALT, AST), kidney function (BUN, serum creatinine). Dietary antioxidants were quantified via the composite dietary antioxidant index (CDAI), summing normalized intakes (food only) of vitamins A, C, E, selenium, zinc, and carotenoids from 24-h recall; CDAI dichotomized at 75th percentile (high vs low). Statistical analysis: Survey design and weights were accounted for using R (v4.3.0) survey package. Single-metabolite associations (log-transformed exposures) with diabetes were estimated by survey-weighted logistic regression with two adjustment sets: Model 1 (demographics and health variables) and Model 2 (Model 1 plus total nutrient intake and liver/kidney function). Mixture analyses used: (a) Environmental Risk Score (ERS), defined as weighted sum of metabolites with weights (β) from adaptive elastic net (adENET) logistic regression (tuning via 5-fold cross-validation optimizing prediction error; lambda1 for sparsity, lambda2 for stability). ERS evaluated continuously (per 1-SD) and by tertiles; trend tested by modeling tertile ordinal term. To address potential bias from internally derived weights, a cross-validated ERS (ERS_CV) was constructed by 5-fold cross-fitting (weights learned in training folds, applied to held-out fold). (b) Quantile g-computation (qg-computation) to estimate the overall mixture effect (psi1) of increasing all exposures by one quantile, in crude and covariate-adjusted (Model 2) forms. Effect modification by antioxidants was assessed by adding the interaction term (scaled ERS × CDAI high/low) in weighted logistic regression and, separately, by including CDAI components in qg-computation to test attenuation of the mixture effect. Significance threshold P<0.05. Mechanistic analysis (AOP): Identified key OPE parent chemicals from mixture weighting (TCEP, TCPP). Retrieved chemical–gene and chemical–phenotype associations and T2DM-related genes/phenotypes from Comparative Toxicogenomics Database (CTD), DisGeNET, and MalaCards (as of Jan 14, 2024). Intersected chemical-related genes with T2DM genes; performed GO and KEGG enrichment (clusterProfiler) for Homo sapiens; intersected enriched terms with T2DM phenotypes to define target phenotype set; merged with CTD-derived chemical–T2DM candidate genes to form target gene sets. Visualized gene–phenotype networks in Cytoscape and proposed putative AOPs linking TCPP/TCEP to T2DM via molecular initiating events and key events (e.g., fatty acid beta-oxidation, insulin signaling, glucose metabolism).

- Sample: 2824 adults; 15 urinary metabolites (10 PAEs, 5 OPEs) analyzed. - Mixture effects (ERS): First scenario (including borderline): per 1-SD increase in ERS associated with higher diabetes odds (OR 1.25, 95% CI 1.13–1.39). Second scenario (excluding borderline): OR 1.21 (95% CI 1.09–1.34). Tertiles vs T1: First scenario T2 OR 1.17 (0.87–1.58), T3 OR 1.77 (1.32–2.37), P-trend <0.001; Second scenario T2 OR 1.41 (1.01–1.97), T3 OR 1.97 (1.43–2.73), P-trend <0.001. ERS_CV tertiles showed similar positive trends (e.g., T3 OR 1.49 [1.12–1.99] first scenario; 1.48 [1.08–2.02] second scenario). - Key contributors (ERS weights): Highest positive weights from OPE metabolites BCEP and BCPP; other high-weight contributors included DBUP, BDCP, MCNP, and MEOHP. - Effect modification by antioxidants (CDAI): Significant interaction between ERS and CDAI. First scenario: OR per 1-SD ERS in low CDAI 1.83 (1.37–2.55) vs high CDAI 1.28 (1.15–1.45), Pinteraction=0.038. Second scenario: low CDAI 1.90 (1.41–2.68) vs high CDAI 1.21 (1.09–1.36), Pinteraction=0.009. ERS_CV showed consistent attenuation in high CDAI subgroup. - Quantile g-computation: Mixture effect (psi1) positive without CDAI components. First scenario: crude OR 1.274 (95% CI 1.053–2.867), adjusted OR 1.285 (1.016–2.763). Second scenario: crude OR 1.360 (1.107–3.026), adjusted OR 1.393 (1.083–2.954). Including CDAI components attenuated associations to null (e.g., first scenario crude OR 1.086 [0.849–2.337]; adjusted OR 1.324 [0.916–2.500]). In qg models, MECPP and MCNP had largest positive weights; vitamin C and zinc had strongest protective (negative) weights. - Mechanistic AOP findings: TCPP implicated via CPT1A and PPARA activation leading to response to xenobiotic stimulus, upregulated fatty acid beta-oxidation, alterations in glucose metabolism, culminating in T2DM. TCEP implicated via AKT1 and HNF4A with downstream effects on insulin signaling (KEGG hsa04910), cellular response to insulin stimulus, insulin secretion, glucose metabolism, glucose homeostasis, and pancreas development, leading to T2DM. - Single-metabolite models: Several PAE metabolites and BDCP showed positive crude associations with diabetes that attenuated after adjustment; marginal associations re-emerged after adding CDAI components.

This study demonstrates that co-exposure to urinary phthalate and OPE metabolites is positively associated with self-reported diabetes in US adults, addressing a gap where prior research focused largely on individual chemicals. The consistent positive mixture effects across ERS, ERS_CV tertiles, and quantile g-computation support a robust association. Importantly, dietary antioxidant intake modifies this relationship: higher CDAI levels attenuate the strength of association between plastic additive mixtures and diabetes, and inclusion of antioxidant components in mixture models reduces the overall effect to null, suggesting a potential mitigating role for antioxidants (notably vitamin C and zinc). Mechanistic analyses via AOPs provide biological plausibility by implicating TCPP and TCEP in pathways affecting fatty acid beta-oxidation, glucose metabolism, and insulin signaling (via AKT1 and HNF4A), as well as processes related to insulin secretion and glucose homeostasis, aligning with hypotheses about oxidative stress and inflammation in diabetogenesis. These findings are consistent with and extend previous observations linking PAEs/OPEs with insulin resistance and glycometabolic disturbances, underscoring the public health relevance of reducing exposure to plastic additive mixtures and considering dietary antioxidant status when evaluating metabolic risk.

Co-exposure to phthalate and OPE metabolites is associated with increased odds of diabetes among US adults. Antioxidant-rich diets, reflected by higher CDAI, appear to mitigate this association, with vitamin C and zinc showing notable protective contributions in mixture analyses. AOP-based bioinformatics implicate TCPP and TCEP in perturbing fatty acid oxidation, glucose metabolism, and insulin signaling, offering mechanistic insight. Public health strategies should prioritize reducing exposures to key parent chemicals (e.g., TCEP, TCPP, DEHP, DCNP) and promoting antioxidant-rich dietary patterns. Future research should use longitudinal designs, broaden chemical coverage beyond PAEs/OPEs, refine diabetes phenotyping (type-specific outcomes), and validate mechanistic pathways experimentally to inform targeted interventions and policy.

- Cross-sectional design limits causal inference and temporal ordering between exposure and diabetes. - Potential unmeasured or residual confounding (e.g., detailed diet patterns, lifestyle factors, genetic predisposition) may influence associations. - Chemical scope restricted to selected PAE and OPE metabolites; other plastic-associated chemicals (polymer matrices, degradation products, adsorbed pollutants) were not assessed. - Diabetes outcome based on self-report and not type-specific (type 1 vs type 2), limiting etiologic specificity. - Differences between mixture modeling approaches (ERS vs qg-computation) yielded slight discrepancies in key contributors, complicating precise intervention targeting.