A novel computed tomography-based multi-parameter decision tree algorithm model for preoperatively predicting the risk of lymph node metastasis in surgically resectable synchronous multiple primary lung cancer.

BACKGROUND

Chest thin-section computed tomography (TS-CT) has the potential to provide evidence for the prediction of lymph node metastasis (LNM) in synchronous multiple primary lung cancer (SMPLC). The present study aims to develop and validate a new CT-based multi-parametric decision tree algorithm (CT-DTA) model capable of accurate risk evaluation for LNM in SMPLC preoperatively.

METHODS

A total of 235 patients with surgically resected SMPLC from Sun Yat-Sen University Cancer Center (SYSUCC), the First Affiliated Hospital of Sun Yat-Sen University (FAH-SYSU) and Sichuan Provincial People's Hospital (SPPH) were finally included. We initially retrieved all the CT-derived quantitative signs in the training cohort (139 cases from SYSUCC) and selected those with statistical significance to build a DTA model. The discriminative power of CT-DTA model for the occurrence of LNM was further externally validated among the validation cohort (96 patients from FAH-SYSU and SPPH). In addition, the performance of CT-DTA model was also assessed across different subgroups of the entire cohort.

RESULTS

Five key quantitative covariables measured on chest TS-CT constituted a CT-DTA model with seven leaf nodes, and long-axis diameter of the solid portion was the most dominant risk contributor of LNM. This CT-DTA model gained a satisfactory predictive accuracy, revealed by an area under the curve >0.80 in both the training cohort (0.905; P<0.001) and the validation cohort (0.812; P<0.001). Moreover, our CT-DTA model was also exhaustively demonstrated to perform as an independent predictor for risk stratification of LNM in both the training cohort (odds ratio: 12.01; P=0.003) and the validation cohort (odds ratio: 8.11; P=0.033). Its potent performance for risk prediction still remained stable across nearly all of the subgroups stratified by clinicopathological characteristics.

CONCLUSIONS

This CT-DTA model could serve as a noninvasive, user-friendly and practicable risk prediction tool to aid treatment decision-making in surgically resectable SMPLC.