Development, validation, and web implementation of a glucose-to-lymphocyte ratio-clinical-ultrasonographic model for central lymph node metastasis in papillary thyroid carcinoma

ObjectiveTo evaluate whether the preoperative glucose-to-lymphocyte ratio (GLR), combined with clinical and ultrasonographic variables, is associated with central lymph node metastasis (CLNM) in papillary thyroid carcinoma and to externally validate and implement a preoperative risk calculator.MethodsThis retrospective prediction-model study included 814 adults with pathologically confirmed papillary thyroid carcinoma for model development and internal validation and an independent external validation cohort of 178 patients. GLR was calculated using preoperative fasting plasma glucose level and lymphocyte count. Logistic regression models integrating GLR, clinical factors, and ultrasonographic features were evaluated using discrimination, calibration, threshold-based classification, and decision curve analysis. The locked model was implemented as an online calculator.ResultsIn the internal cohort, CLNM was identified in 369 of 814 patients (45.33%). Median GLR was lower in patients with CLNM than in those without CLNM (2.51 (2.08-3.04) vs 2.75 (2.27-3.32), p < 0.001). GLR ≥3.09 was associated with lower odds of CLNM after adjustment (odds ratio: 0.68, 95% confidence interval: 0.46-0.95, p = 0.021). The fusion model had an internal validation area under the receiver operating characteristic curve of 0.751 (95% confidence interval: 0.703-0.795). In external validation, the area under the receiver operating characteristic curve was 0.748 (95% confidence interval: 0.670-0.819), sensitivity was 0.725, specificity was 0.622, Brier score was 0.202, calibration slope was 1.110, and Hosmer-Lemeshow p-value was 0.741.ConclusionsPreoperative GLR was inversely associated with pathological CLNM in papillary thyroid carcinoma. A GLR-clinical-ultrasonographic model, implemented as an online calculator, demonstrated moderate validation performance and may support adjunctive preoperative risk stratification.