OBJECTIVES: Predicting survival for patients with non-small cell lung cancer (NSCLC) using the traditional Kaplan Meier (KM) model or Cox proportional hazard (CPH) model leads to over-simplification as competing events are ignored. The Fine-Gray (FG) competing risk model accounts for these competing events. This study compared the methodologies for analyzing the impact of various prognostic factors on survival.

METHODS: This cross-sectional study identified patients with NSCLC in the Surveillance, Epidemiology, and End Results (SEER) registry from 2016-2019. The prognostic factors assessed were age, race, gender, year of diagnosis, tumor size, primary site, stage, laterality, histology, income, marital status, surgery, chemotherapy, and radiation. The cumulative incidence (CI) and hazard ratio (HR) for NSCLC-specific deaths were estimated at 12, 24, and 36 months by the KM and the CPH model respectively, and by the FG model, where death from other causes was considered a competing event.

RESULTS: In the SEER registry, 110,416 NSCLC patients were identified, of which 41% had NSCLC-specific deaths, and 9% died from other causes. CI of deaths due to NSCLC at 12/24/36 months were approximately 37%/49%/55% using the KM-model and 35%/46%/52% using the FG-model. At 36-month, CI of NSCLC-specific death in males was 60% in the KM-model [CPH-HR: 1.205, p< 0.0001] and 56% in the FG-model [FG-HR: 1.166, p< 0.0001]; while for patients with ‘distant’ stage of cancer, it was 80% using KM-model [CPH-HR: 6.303, p< 0.0001] and 75% using FG-model [FG-HR: 5.589; p< 0.0001]. For patients aged ≥ 75 years, CI at 36-month was 61% using KM-model [CPH-HR:1.167, p< 0.0001] and 55% using FG-model [FG-HR:1.119; p< 0.0001].

CONCLUSIONS: The CI estimates for the KM-model were greater than the FG-model for every factor, and these differences increased with prolonged follow-up. However, the differences between the two models were minor potentially due to smaller proportions of patients with competing events.