A LASSO-based model to predict central lymph node metastasis in preoperative patients with cN0 papillary thyroid cancer

Abstract

Introduction: Central lymph node metastasis (CLNM) is common in papillary thyroid carcinoma (PTC). Prophylactic central lymph node dissection (PCLND) in clinically negative central compartment lymph node (cN0) PTC patients is still controversial. How to predict CLNM before the operation is very important for surgical decision making.

Methods: In this article, we retrospectively enrolled 243 cN0 PTC patients and gathered data including clinical characteristics, ultrasound (US) characteristics, pathological results of fine-needle aspiration (FNA), thyroid function, eight gene mutations, and immunoenzymatic results. Least absolute shrinkage and selection operator (LASSO) analysis was used for data dimensionality reduction and feature analysis.

Results: According to the results, the important predictors of CLNM were identified. Multivariable logistic regression analysis was used to establish a new nomogram prediction model. The receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA) curve were used to evaluate the performance of the new prediction model.

Discussion: The new nomogram prediction model was a reasonable and reliable model for predicting CLNM in cN0 PTC patients, but further validation is warranted.

Keywords: central lymph node metastasis (CLNM); clinically negative central compartment lymph nodes (cN0); least absolute shrinkage and selection operator (LASSO); nomogram prediction model; papillary thyroid carcinoma (PTC); prophylactic central lymph node dissection (PCLND).

Figure 1

Identification of the influencing factors by LASSO regression. LASSO regression identified the following 5 most powerful predictors. (A) LASSO coefficient profiles of the 49 characteristics. A coefficient profile plot was produced against the log lambda (λ) sequence. (B) The relationship curve between the partial likelihood deviation (binomial deviation) and log(λ) was plotted.

Figure 2

Heatmap of the top five significant variables identified based on LASSO regression analysis. Heatmap of the top five significant variables identified via LASSO regression, including BRAF ^V600E mutation, age, history of nodular goiter, history of Hashimoto’s thyroiditis and TI-RADS classification.

Figure 3

Evaluation of the performance of the new prediction model. (A) Nomogram for predicting CLNM in PTC patients based on five risk factors. (B) The ROC curve and AUC of the nomogram. ROC, receiver operating characteristic. (C) Calibration plots of the nomogram for predicting CLNM. (D) The DCA method evaluated the performance of the model.