A nomogram prognostic model for large cell lung cancer: analysis from the Surveillance, Epidemiology and End Results Database

Abstract

Background: Currently, there is no reliable method for predicting the prognosis of patients with large cell lung cancer (LCLC). The aim of this study was to develop and validate a nomogram model for accurately predicting the prognosis of patients with LCLC.

Methods: LCLC patients, diagnosed from 2007 to 2009, were identified from the Surveillance, Epidemiology and End Results (SEER) database and used as the training dataset. Significant clinicopathologic variables (P<0.05) in a multivariate Cox regression were selected to build the nomogram. The performance of the nomogram model was evaluated by the concordance index (C-index), the area under the curve (AUC), and internal calibration. LCLC patients diagnosed from 2010 to 2016 in the SEER database were selected as a testing dataset for external validation. The nomogram model was also compared with the currently used American Joint Committee on Cancer (AJCC) tumor-node-metastasis (TNM) staging system (8^th edition) by using C-index and a decision curve analysis.

Results: Eight variables-age, sex, race, marital status, T stage, N stage, M stage, and treatment strategy-were statistically significant in the multivariate Cox model and were selected to develop the nomogram model. This model exhibited excellent predictive performance. The C-index and AUC value were 0.761 [95% confidence interval (CI), 0.754 to 0.768] and 0.886 for the training dataset and 0.773 (95% CI, 0.765 to 0.781) and 0.876 for the testing dataset, respectively. This model also predicted three-year and five-year lung cancer-specific survival (LCSS) in both datasets with good fidelity. This nomogram model performs significantly better than the 8th edition AJCC TNM staging system, with a higher C-index (P<0.001) and better net benefits in predicting LCSS in LCLC patients.

Conclusions: We developed and validated a prognostic nomogram model for predicting 3- and 5-year LCSS in LCLC patients with good discrimination and calibration abilities. The nomogram may be useful in assisting clinicians to make individualized decisions for appropriate treatment in LCLC.

Keywords: Large cell lung cancer (LCLC); nomogram; prognosis; prognostic model.

Figure 1

Flow-chart illustrating the methodology used for identifying cases of large cell lung cancer in the SEER database during 2007–2016. SEER, Surveillance, Epidemiology and End Results database.

Figure 2

Lung cancer-specific survival (A) and overall survival (B) curves of patients with large cell lung cancer in the training dataset. LCSS, lung cancer-specific survival; OS, overall survival.

Figure 3

Three- and five-year lung cancer-specific survival nomogram for patients with LCLC. LCLC, large cell lung cancer; LCSS, lung cancer-specific survival; M, M descriptor in the 8^th TNM staging system of lung cancer; N, N descriptor in the 8th TNM staging system of lung cancer; Treat-C, chemotherapy; Treat-CR, chemoradiotherapy; Treat-R, radiotherapy; Treat-S, surgery; Treat-SR, surgery combined with radiotherapy; Treat-CRS, surgery combined with chemoradiotherapy; Treat-NO, no treatment; T, T descriptor in the 8th TNM staging system of lung cancer.

Figure 4

Predicted patient survival probability curve corresponding to risk score levels [1–4]. LCSS, lung cancer-specific survival.

Figure 5

Calibration plots of nomogram predictions of three- (A) and five-year (B) lung cancer-specific survival of patients with large cell lung cancer in the training dataset. Redline: equality of the observed and predicted probability. AUC: 0.861 as calculated from the three- (C) and five-year (D) nomogram prognostic model. AUC, the area under the curve; FP, false positive; LCSS, lung cancer-specific survival; TP, true positive.

Figure 6

Calibration plots of nomogram predictions of three- (A) and five-year (B) lung cancer-specific survival of patients with large cell lung cancer in the testing dataset. Redline: equality of the observed and predicted probability. LCSS, lung cancer-specific survival.

Figure 7

Decision curve analysis of the current nomogram model and the 8^th edition AJCC TNM staging system for predicting three- (A) and five-year (B) lung cancer-specific survival. Horizontal line (Threshold probability) referred to the recurrence probability, which was predicted using the nomogram model or the 8^th edition TNM staging system. When the recurrence probability exceeded a certain value, the clinician should consider starting treatment for the patient. After starting the treatment, recurrence patients would be benefited from the treatment (benefit), and those who have not relapsed would be harmed by the treatment (disadvantage). The net benefit = benefit − disadvantage. Decision curve analysis was a tool to evaluate the net benefit of starting treatment. The grey lines represent extreme situations. The horizontal line represents that all patients are untreated and the net benefit is zero. The oblique line indicates that all patients have been treated, and the net benefit is a backslash with a negative slope. Both the nomogram model (blue line) and the 8^th edition AJCC TNM staging system (red line) have obvious gains in net benefits compared to the extreme curve. The decision curve of the nomogram model is better than that of the AJCC TNM staging system.