Background
Thyroid cancer (TC) is the most common endocrine malignancy, with differentiated thyroid carcinomas (DTC) accounting for 90% of cases. While most DTC patients have a favourable prognosis, some develop aggressive disease requiring extensive treatment. Identifying high-risk individuals earlier would improve treatment stratification and prevent patients with indolent TC from receiving unnecessarily intensive therapy. Metabolomics, the systematic analysis of small-molecule metabolites in a biological system, offers a powerful approach to discover early biomarkers of TC development.
Objectives
To apply untargeted metabolomics to prospectively collected plasma from subjects with indolent T1 DTC (tumours ≤2 cm in greatest dimension, limited to the thyroid gland), higher stages T2-T4 DTC, and matched controls from the European Prospective Investigation into Cancer and nutrition (EPIC) cohort, aiming to identify biomarkers specifically associated with each TC subtype.
Methods
Prospectively collected serum samples were analysed at IARC using untargeted high-resolution LC-MS, employing positive and negative electrospray ionization modes. Our study included a total of 327 DTC cases (291 women/36 men) and 682 controls (577 women/105 men) matched on appropriate confounders (centre, sex, blood collection date, fasting status, menopausal status, menstrual cycle, and exogenous hormone use). Of the cases, 141 were T1 and 78 T2–T4 tumours. After excluding features with more than 25% missing values across samples and removing samples with atypical multivariate metabolomic profiles, 1,662 features in positive mode and 2,266 in negative mode were retained, log-transformed, imputed, and normalised to correct for technical variability across samples. Associations between metabolomic features and TC risk, including analyses by tumour subtype, were assessed using conditional logistic regression, conducted separately according to ionization modes and corrected for multiple comparison.
Results
In analyses of positive ionization mode, Model 1 (stratified by matching factors) identified one feature inversely associated with TC risk (FDR<0.05). In overall TC and subtype analyses, additional adjustment for education, alcohol, and BMI (Model 2) did not alter the findings. In female-specific analysis, Model 3 (Model 2 further adjusted for reproductive factors) identified 3 features inversely associated with TC risk, 1 of which already identified in the sex-combined analysis. Additionally, 6 features showed stronger inverse association with T1 than with T2–T4 cases (FDR<0.05), highlighting potential subtype-specific effects, though formal tests of heterogeneity were not performed.
Conclusions
Preliminary results suggest that specific metabolomic features may be related to reduced risk of TC, specifically with early-stage (T1) TC. Ongoing analyses, including analysis of RP-negative features and validation in independent cohorts (E3N and Sister Study), will be available at the time of presentation. These analyses may help highlight candidate biomarkers and metabolic pathways linked to thyroid cancer, potentially improving understanding of its aetiology and informing patient stratification. In the long term, they could contribute to strategies to reduce overtreatment of very early-stage cancers and mitigate overdiagnosis.