Metabolic dysregulation and biological age acceleration in Hashimoto's thyroiditis: a cross-sectional study based on clinical biomarker aging indices and metabolomics.

Background: Hashimoto's thyroiditis (HT) is a common autoimmune thyroid disease characterized by thyroid autoantibody positivity, chronic lymphocytic inflammation, and varying degrees of thyroid dysfunction. Whether HT is associated with quantifiable biological age acceleration and metabolic age-related remodeling remains insufficiently examined. Methods: This cross-sectional study included two clinical discovery cohorts, an NHANES 2007-2012 validation cohort, and a metabolomics cohort. Klemera-Doubal method biological age (KDM biological age), Phenotypic Age (PhenoAge), and corresponding age acceleration metrics were calculated from clinical biomarkers, and the proportions of participants with age acceleration were compared. In the metabolomics cohort, a random forest metabolic age model was trained in healthy controls and internally evaluated by five-fold cross-validation. Metabolic age acceleration (MAA) was defined as predicted metabolic age minus chronological age. Metabolites associated with age, FT3, FT4, and TSH, together with differential metabolites in euthyroid HT (EHT) and thyroid dysfunction HT (DHT) relative to controls, were integrated to screen candidate metabolites and construct exploratory prediction models. Results: In the two discovery cohorts, the HT group showed higher KDM biological age and/or PhenoAge indices and higher proportions of age acceleration than healthy controls. In the NHANES validation cohort, after adjustment for age, sex, socioeconomic factors, and lifestyle factors, overall HT was associated with higher KDM biological age (beta = 3.16 years, 95% CI 1.77-4.56) and PhenoAge (beta = 1.54 years, 95% CI 1.05-2.03). Stage-specific analyses suggested relatively stable associations for euthyroid HT and subclinical hypothyroid HT, whereas results for overt hypothyroid HT should be interpreted cautiously because of the small sample size and wide confidence intervals. In the random forest metabolic age model, the mean test-set performance across five-fold cross-validation was RMSE 8.75 years, MAE 7.11 years, R 0.634, and R2 0.427. MAA differed among CON, EHT, and DHT groups (Kruskal-Wallis P = 6.6 x 10^-5), and both EHT and DHT had higher MAA than CON after adjustment for age and sex. Integrated analysis identified 18 candidate metabolites; the HT classification model based on these metabolites had an AUC of 0.980, and predicted values from the FT3, FT4, and TSH models were significantly correlated with the measured values. Considering clinical relevance, between-group expression trends, and model-intersection evidence, citric acid, LPC 20:0 sn-1, and SM 34:2 were prioritized as core candidate metabolites. Conclusions: In the cross-sectional cohorts included in this study, HT was associated with higher biological age and metabolic age acceleration. Metabolomics results suggest that citric acid-related energy metabolism and lipid molecules such as LPC and SM may reflect HT- and thyroid-function-related metabolic features.