An Externally Validated Nomogram to Predict Distant Metastases after Stereotactic Body Radiation Therapy in Early Stage Non-Small Cell Lung Cancer Patients

Abstract

Stereotactic body radiation therapy (SBRT) is the standard of care for medically inoperable early stage non-small cell lung cancer (NSCLC) patients and has shown excellent local control rates. However, the development of distant metastases (DM) is the most common form of failure. While there have been multiple international efforts to incorporate systemic therapy to SBRT, there is little guidance on determining which patients are at high risk of DM and would most benefit from systemic therapy. As such, we developed a nomogram to predict risk of DM in early stage NSCLC patient after SBRT. We conducted a retrospective cohort study utilizing a multi-institutional community practice database to identify consecutive patients with biopsy proven T1-3N0M0 NSCLC treated with definitive SBRT from 2006-2015. Univariate cause-specific Cox proportional hazards models were used to identify variables associated with DM, which were incorporated into a multivariate model if they had P<0.1 or if there was clinical rationale for their inclusion. The final model was internally validated with bootstrapping with 100 replications. A nomogram was generated from the model and discriminatory performance was evaluated with C-index. A separate academic institutional database was used to externally validate the model. Of 1,328 patients in the initial database, 821 met the study criteria for analysis. Patients were excluded if they did not have biopsy proven NSCLC, received chemotherapy, had synchronous primaries, and did not receive a BED10 of at least 100 Gy. The median time to the development of DM was 18 months, and the median follow-up time was 19 months. The raw rate of DM at 2 years was 9.6% (79/821). On multivariate analysis, higher PET SUV (<3 vs ≥3: HR: 2.52, P=0.022), larger tumor size (HR: 1.39, P=0.030), and adenocarcinoma as opposed to squamous cell carcinoma (HR: 1.62, P=0.036) were correlated with a higher rate of distant metastases. The final nomogram included PET SUV, size, age, performance status, histology, T-stage, gender, and prior lung cancer. All variables in the final model had a univariate P < 0.1 except for prior lung cancer, which was included based on clinical significance. On internal validation, the model had a C-index of 0.67 for DM. The model was validated on an institutional database of 185 patients with biopsy proven T1-3N0 NSCLC who did not receive chemotherapy, did not have synchronous primaries, and received a BED10 of at least 100 Gy and achieved an externally validated C-index of 0.63. We developed a nomogram from a large, heterogeneous patient population that uses clinical information to predict the risk of distant metastases in early stage NSCLC patients after SBRT, and externally validated the model on an independent dataset. The nomogram may be helpful in guiding clinical decision making regarding adjuvant systemic therapy and clinical trial development. Further research is ongoing to increase the predictive performance of the model.