Abstract / Summary
To develop and validate a nomogram based on LASSO and multivariable logistic regression for predicting preoperative central lymph node metastasis (CLNM) in patients with cT1–2N0M0 papillary thyroid carcinoma (PTC), and to assess its clinical utility. This retrospective study included 652 patients with cT1–2N0M0 PTC who underwent thyroidectomy with central lymph node dissection at the Department of Thyroid Surgery, Guangxi Zhuang Autonomous Region People’s Hospital, between January 2024 and June 2025. Demographic characteristics, Hashimoto’s thyroiditis status, laboratory parameters, preoperative fine-needle aspiration findings, including BRAFV600E mutation status, and preoperative ultrasonographic features were collected. For consistency with the subsequent statistical analyses and model presentation, capsular invasion, extrathyroidal extension (ETE), multifocality, and bilaterality used throughout this manuscript refer to the corresponding findings suspected or identified on preoperative ultrasonography, unless otherwise specified. These terms represent preoperative ultrasound-based assessments and should not be interpreted as postoperative pathological diagnoses. Postoperative pathological findings were used solely to determine CLNM status and were not used as model predictors. Patients were classified as CLNM-positive or CLNM-negative according to postoperative pathology. The cohort was randomly divided into training and internally held-out test cohorts at a 6:4 ratio. Least absolute shrinkage and selection operator regression and multivariable logistic regression were performed in the training cohort to identify independent predictors of CLNM and develop a nomogram. Model performance was evaluated in both cohorts using receiver operating characteristic curves, the area under the curve (AUC), calibration curves, and decision curve analysis (DCA). Multivariable analysis and feature importance ranking identified tumor size, age, extrathyroidal extension, and multifocality as the major predictors of CLNM. The AUC was 0.7335 (95% CI: 0.6838–0.7831) in the training cohort and 0.685 (95% CI: 0.619–0.7511) in the internally held-out validation cohort. DCA indicated that the nomogram had favorable clinical utility. Tumor size, age, ETE, and multifocality were the main factors associated with CLNM in patients with cT1–2N0M0 PTC. The nomogram shows moderate predictive performance and may assist in preoperative risk stratification and individualized surgical decision-making, pending external validation. Not applicable.