Year: 2026
Type: preprint
Study design: Observational Study
Abstract: Background: The incidence of papillary thyroid carcinoma (PTC) is continuously increasing, and conventional examinations have obvious limitations in early PTC screening. Although the triglyceride-glucose (TyG) index is associated with multiple malignancies, its predictive value for PTC risk is still undefined.This study aimed to identify independent PTC risk factors, construct effective machine learning predictive models, and develop a novel composite indicator to provide evidence for early screening and risk stratification of PTC. Methods: A total of 869 patients (447 PTC and 422 benign thyroid nodule cases) from 2022 to 2025 were included for retrospective analysis. Logistic regression analyses were used to screen PTC risk factors. Nine machine learning models were constructed and evaluated by AUC and sensitivity, and SHAP analysis was performed for model interpretation. A novel composite indicator TyG-Th was established and validated. Results: Age and thyroglobulin (Tg) were protective factors for PTC, while TSH, BMI, and TyG index were independent risk factors. The ANN model exhibited optimal predictive performance with Tg as the core feature. This study innovatively constructed a novel composite indicator, TyG-Th, which achieved a standalone AUC of 0.754 and outperformed conventional single indicators. The combination of TyG-Th with core variables further increased the AUC to 0.804. Stratified analyses validated its stable predictive ability across age and gender subgroups, and higher TyG-Th was correlated with PTC clinical characteristics, demonstrating its promising predictive value. Conclusions: TyG-Th is a reliable PTC biomarker across all ages and genders. Age and Tg were protective factors against PTC, while TSH, BMI and TyG index were independent risk factors. The models performed well, with Tg as the core ANN feature.
Source: Research Square
Institution: First Affiliated Hospital of Zhengzhou University
Language: English
Subfield: Endocrinology, Diabetes and Metabolism
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7237214601
