Construction and validation of a nomogram to predict the overall survival of small cell lung cancer: a multicenter retrospective study in Shandong province, China

Abstract

Abstract Background Patients diagnosed with small cell lung cancer (SCLC) typically experience a poor prognosis, and it is essential to predict overall survival (OS) and stratify patients based on distinct prognostic risks. Methods Totally 2309 SCLC patients from the hospitals in 15 cities of Shandong from 2010 − 2014 were included in this multicenter, population-based retrospective study. The data of SCLC patients during 2010–2013 and in 2014 SCLC were used for model development and validation, respectively. OS served as the primary outcome. Univariate and multivariate Cox regression were applied to identify the independent prognostic factors of SCLC, and a prognostic model was developed based on these factors. The discrimination and calibration of this model were assessed by the time-dependent C-index, time-dependent receiver operator characteristic curves (ROC), and calibration curves. Additionally, Decision Curve Analysis (DCA) curves, Net Reclassification Improvement (NRI), and Integrated Discriminant Improvement (IDI) were used to assess the enhanced clinical utility and predictive accuracy of the model compared to TNM staging systems. Results Multivariate analysis showed that region (Southern/Eastern, hazard ratio [HR] = 1.305 [1.046 − 1.629]; Western/Eastern, HR = 0.727 [0.617 − 0.856]; Northern/Eastern, HR = 0.927 [0.800 − 1.074]), sex (female/male, HR = 0.838 [0.737 − 0.952]), age (46–60/≤45, HR = 1.401 [1.104 − 1.778]; 61–75/≤45, HR = 1.500 [1.182 − 1.902]; >75/≤45, HR = 1.869 [1.382 − 2.523]), TNM stage (II/I, HR = 1.119[0.800 − 1.565]; III/I, HR = 1.478 [1.100 − 1.985]; IV/I, HR = 1.986 [1.477 − 2.670], surgery (yes/no, HR = 0.677 [0.521 − 0.881]), chemotherapy (yes/no, HR = 0.708 [0.616 − 0.813]), and radiotherapy (yes/no, HR = 0.802 [0.702 − 0.917]) were independent prognostic factors of SCLC patients and were included in the nomogram. The time-dependent AUCs of this model in the training set were 0.699, 0.683, and 0.683 for predicting 1-, 3-, and 5-year OS, and 0.698, 0.698, and 0.639 in the validation set, respectively. The predicted calibration curves aligned with the ideal curves, and the DCA curves, the IDI, and the NRI collectively demonstrated that the prognostic model had a superior net benefit than the TNM staging system. Conclusion The nomogram using SCLC patients in Shandong surpassed the TNM staging system in survival prediction accuracy and enabled the stratification of patients with distinct prognostic risks based on nomogram scores.

Keywords

• Lung cancer
• Clinical characteristics
• Prognostic factors
• Survival rate
• Multi-center