Nomogram Predicting Overall Survival Benefit of Stereotactic Ablative Radiotherapy for Early-Stage Non-Small Cell Lung Cancer.

OBJECTIVES: To develop and validate a nomogram that predicts overall survival (OS) for patients with early-stage non-small cell lung cancer (NSCLC) treated with stereotactic ablative radiotherapy (SABR) vs. observation. MATERIALS AND METHODS: Adults with biopsy-proven T1-T2N0 NCSLC treated with SABR (30-70 Gy in 1-10 fractions with biologically effective dose ≥100 Gy10) or observation between 2004 and 2015 in the National Cancer Database (NCDB) were identified. Propensity score was used to match SABR and observation cohorts on prognostic demographic and clinicopathologic factors identified by logistic regression. Using backward selection, a multivariable Cox proportional hazard was identified predicting 2- and 5-year OS via a nomogram. Model prediction accuracy was assessed by time-dependent receiver operating characteristic (ROC) curves and integrated area under the ROC curve (AUC) analysis. RESULTS: A total of 22,073 adults met inclusion criteria and 4418 matched pairs (total n = 8836) were identified for nomogram development. The factors most strongly associated with improved OS on multivariable analysis included younger age (HR 0.82 by decade, P < .001), female sex (HR 0.81, P < .001), lower comorbidity index (HR 0.65 for 0 vs. ≥3, P < .001), smaller tumor size (HR 0.60 for ≤3 cm vs. 5.1-7 cm, P < .001), adenocarcinoma histology (P < .001), and receipt of SABR (P < .001). Interaction between SABR and histology was significantly associated with OS (P = .017). Relative to adenocarcinoma, patients with squamous cell carcinoma who were observed (HR 1.44, 95% CI 1.33-1.56) or treated with SABR (HR 1.24, 95% CI 1.14-1.35) had significantly worse OS. The nomogram demonstrated fair accuracy for predicting OS, with an integrated time-dependent AUC of 0.694 over the entire follow-up period. CONCLUSION: This nomogram estimates OS at 2 and 5 years based on whether medically inoperable early-stage NSCLC patients receive SABR or elect for observation. Incorporation of other variables not captured within the NCDB may improve the model accuracy.

Duke Scholars

Author Xiaofei Wang Biostatistics & Bioinformatics, Division of Biostatistics

Author Joseph Kamel Salama Radiation Oncology

Author Christopher Ryan Kelsey Radiation Oncology