Development and validation of a predictive nomogram for cervical lymph node metastasis in aspect ratio ≥1 papillary...

Preoperative prediction of cervical lymph node metastasis (CLNM) remains challenging in patients with papillary thyroid carcinoma (PTC) exhibiting an aspect ratio ≥1, a subgroup in which ultrasound alone has limited sensitivity (approximately 50%) and no dedicated predictive model currently exists. This study aimed to develop and internally validate a predictive nomogram combining ultrasonic features and serum calcitonin for CLNM in this specific population.

This retrospective study consecutively enrolled 156 patients with aspect ratio ≥1 PTC who underwent thyroidectomy and neck lymph node dissection at a single center (November 2023 to June 2025). The gold standard outcome was pathologically confirmed CLNM. Forty-six candidate predictors (clinical, ultrasonographic, and serological variables) were collected. Predictor selection was performed using least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation, followed by multivariate logistic regression. A nomogram was constructed and internally validated using 1,000 bootstrap resamples. Model performance was assessed by area under the curve (AUC), calibration curve, sensitivity, specificity, and decision curve analysis (DCA).

Among 156 patients [116 females, 40 males; median age 52 years, interquartile range (IQR), 45-57 years], CLNM was present in 48 patients (30.8%). Five independent predictors were identified: multifocality [odds ratio (OR) =5.87; 95% confidence interval (CI): 1.85-18.66], maximum diameter ≥1 cm (OR =11.57; 95% CI: 2.26-59.16), ill-defined margin (OR =7.43; 95% CI: 2.06-26.76), microcalcifications (OR =3.43; 95% CI: 1.19-9.93), and serum calcitonin (OR =1.44; 95% CI: 1.08-1.92 per pg/mL). The nomogram achieved an apparent AUC of 0.834 (95% CI: 0.767-0.901) and a bootstrap-corrected AUC of 0.813. At the optimal probability cutoff of 0.264, sensitivity was 81.2% and specificity 75.9%. Calibration showed good agreement (Hosmer-Lemeshow P=0.41), and DCA demonstrated net clinical benefit across a threshold probability range of 5-80%.

This nomogram combining five preoperative predictors shows satisfactory discriminative ability for CLNM in patients with aspect ratio ≥1 PTC. However, given the single-center design, modest sample size, and lack of external validation, these findings should be considered preliminary. External validation in independent multicenter cohorts is required before clinical application.