A predictive model based on liquid biopsy for non-small cell lung cancer to assess patient’s prognosis: Development and application

Abstract

Background and objectiveImproving ability to predict the prognosis of patients with progressive lung cancer is an important task in the era of precision medicine. Here, a predictive model based on liquid biopsy for non-small cell lung cancer (NSCLC) was established to improve prognosis prediction in patients with progressive NSCLC. MethodsClinical data and blood samples of 500 eligible patients were collected and screened from the electronic case database and blood sample center of Hwa Mei Hospital, University of Chinese Academy of Sciences and Ningbo Institute of Life and Health Industry, University of Chinese Academy of Sciences. Patients were randomly assigned to training set (300 cases) and validation set (200 cases) in a ratio of 3:2 by random number method. Baseline levels of the two datasets were compared. Progression-free survival (PFS) analysis was performed on the training set using Kaplan-Meier method. The independent prognostic factors affecting patients’ PFS were determined by multivariate Cox regression analysis. The prognosis predictive model of patients was constructed by using the nomogram. Calibration curve and C-index were used to evaluate the accuracy of the prognosis predictive model in both internal and external validations. ResultsIn training set, the age distribution of patients was 59.00 (46.00, 71.00) years, including 137 (45.7 %) females and 163 (54.3 %) males, 198 cases (66.0 %) with Eastern Cooperative Oncology Group (ECOG) score 0–1, and 102 cases (34.0 %) with ECOG score 2. In verification set, the age distribution of patients was 60.00 (48.25, 73.00) years, including 92 females (46.0 %) and 108 males (54.0 %), 130 cases (65.0%) with ECOG score 0–1, and 70 cases (35.0 %) with ECOG score 2. Patients in training set showed PFS differences stratified by gene mutation type (p < 0.0001), differentiation degree (p < 0.0001), circulating tumor cell (CTC) content (p = 0.00026), and brain metastasis (p < 0.0001). Besides, multivariate Cox regression analysis indicated that gene mutation type, differentiation degree, CTC content (p = 0.002), and brain metastasis (p = 0.005) are independent prognostic factors for PFS. These factors were included in the nomogram parameters, and both internally validated calibration curve (C-index = 0.672) and externally validated calibration curve (C-index = 0.657), showing good predictive performance of the model. ConclusionThe predictive model has a good predictive ability for prognosis of patients with progressive NSCLC. Notably, the differentiation degree and CTC content are both impact factors for PFS of patients, and the performance of these indicators in predicting the survival of patients with progressive NSCLC needs to be clarified in the future.