Predictive models for contralateral occult thyroid carcinoma and central lymph node metastasis in unilateral multifocal papillary thyroid carcinoma based on ultrasound.

Background: Patients with unilateral multifocal papillary thyroid carcinoma (MPTC) are more likely to undergo total thyroidectomy (TT), which may represent overtreatment. This study aimed to develop clinical models that accurately predict the risk of contralateral occult thyroid carcinoma (OTC) and central lymph node metastasis (CLNM) (number ≥5) in these patients using preoperative clinical variables, providing clearer guidance for surgical decision-making. Methods: The clinical data of papillary thyroid carcinoma (PTC) patients who underwent initial surgery in Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, from May 2020 to May 2024 were collected. This study utilized 12 machine learning (ML) algorithms and 113 combinations to calculate the area under the receiver operating characteristic (ROC) curve (AUC) index for each model across training and validation datasets. Additionally, nine common algorithms were employed for hyperparameter optimization through cross-validation grid search. The ROC curve, calibration curve, and decision curve analysis (DCA) curve were plotted to identify the optimal model. SHapley Additive exPlanations (SHAP) interpretation was used for personalized risk assessments. Results: Age, gender, body mass index (BMI), and ultrasonographic findings, including the sum of the longest diameter (SLD) of unilateral malignant nodules and capsular disruption, constructed the model to predict the risk of CLNM (number ≥5). The clinical model for contralateral OTC incorporated independent variables reported by ultrasound, such as capsular disruption, the number and the SLD of malignant nodules, and malignant isthmus nodules. ROC curves indicated excellent discrimination for both datasets (AUC index ≥0.75). Calibration curves showed satisfactory fitting effects; meanwhile, DCA curves demonstrated high practical value. Based on the prediction models, thyroid lobectomy (TL) may be feasible for patients classified as low or medium risk; meanwhile, the surgical method should be determined according to recurrence risks and patient thoughts in high-risk cases. Conclusions: This study developed models for predicting contralateral OTC and CLNM (number ≥5) in unilateral MPTC based on ultrasound to mitigate the risk of overtreatment while considering the potential for recurrence, thereby providing guidance for the precise implementation of individualized surgical strategies.