A biomarker-based prediction model for risk of locoregional recurrence in pathologic stage IIIA-N2 non-small cell lung cancer.

objectiveTo investigate risk factors for locoregional recurrence (LRR) of pathologic stage IIIA-N2 non-small cell lung cancer (pIIIA-N2 NSCLC) and construct a prediction model for risk score to determine a patient's risk for LRR and guide the selection of postoperative radiotherapy (PORT).

methodsThe clinical, pathologic, and biological data of 107 patients with pIIIA-N2 NSCLC treated at Fujian Provincial Hospital between May 2012 and December 2018 were analyzed retrospectively. None of the patients had positive surgical margins, and none received preoperative treatment or PORT. The Kaplan-Meier method was used for a univariate analysis of possible factors for locoregional recurrence-free survival (LRFS). The Cox regression model was used in a multivariate analysis to identify independent risk factors for LRFS, which were used to construct a prediction model for risk score. The concordance index was calculated to evaluate discrimination.

resultsThe median follow-up time was 31.2 months. During the follow-up, 69 (64.5%) patients had LRR and/or distant metastasis (DM). Among them, 46 (43%) patients had LRR (with or without DM), and 56 (52.3%) patients had DM (with or without LRR). The 1-year LRFS, distant metastasis-free survival, disease-free survival, and overall survival rates were 78.2%, 78%, 69.8%, and 90.2%, respectively; the 3-year rates were 50.6%, 41.2%, 31.2%, and 66.3%, respectively. Multivariate analysis showed that surgical approach (hazard ratio [HR], 0.348; 95% confidence interval [CI], 0.175-0.693; P = 0.003), metastatic N2 lymph node ratio (HR, 3.597; 95% CI, 1.832-7.062; P = 0.000), epidermal growth factor receptor status (HR, 3.666; 95% CI, 1.724-7.797; P = 0.001), and lymphocyte-to-monocyte ratio (HR, 2.364; 95% CI, 1.221-4.574; P = 0.011) were independent risk factors for LRFS. These independent risk factors were used to construct a prediction model for risk score and stratify patients into the low-risk group (risk score: 0-2), medium-risk group (risk score: 3-5), and high-risk group (risk score: 6-13). The 1-year LRFS rates of these groups were 91.9%, 85.3%, and 54.6%, respectively; the 3-year LRFS rates were 71.4%, 57.3%, and 13.6%, respectively. These between-group differences were significant (P = 0.000). The prediction model showed good discrimination (concordance index = 0.747, 95% CI, 0.678-0.816).

conclusionOur prediction model for risk score based on characteristics of pIIIA-N2 NSCLC patients may help clinicians predict a patient's risk for LRR. Further investigations of PORT with patients in different risk groups are warranted.