Comparing prospectively assigned trial and real-world lung cancer patients

We utilized de-identified patient-level structured data sourced from EMRs. We then compared patterns of overall survival (OS) between probable CT patients with those drawn from non-contemporaneous real-world data (RWD) using a two-sided log-rank test, hazard ratios (HRs) using a Cox proportional-hazards model and Kaplan–Meier (KM) survival curves. Each regression estimate was calculated with a corresponding 95% confidence interval. We additionally conducted multiple matching methods to assess their relative performance.

Median (standard deviation) OS was 10.2 (0.7) months for the RWD arm and 11.3 (1.3) for the probable CT arm with a Log rank p-value equal to 0.4771. OS in both cohorts is longer than the reported CT median OS of 9.2 (0.6). The HRs generated under all five assessed matching methods (including without adjustment) were not statistically significant at the 95% confidence level.

Our results suggest, with caveats noted, that survival patterns between real-world and CT cohorts in this NSCLC setting are not statistically significantly different.