Background: The long-term real-world effect of immunotherapy (IO) is uncertain in metastatic nonsmall cell lung cancer (mNSCLC). This retrospective observational study aimed to describe treatment patterns following the introduction of IO, estimate real-world treatment effects of IO compared to standard of care, and evaluate the impact of introduction of IO on a real-world population, based on a large dataset of over 10 000 patients with several years of follow-up.
Methods: Data from routine care of lung cancer patients were extracted from Flatiron Health including those who received either IO or platinum-based doublet chemotherapy (PBDC) in the first line (1L), or either IO or chemotherapy (CT) in the second line (2L). Real-world overall survival (rwOS) and real-world time to next therapy (rwTTNT) were estimated using Cox regression. Flexible parametric models, relaxing proportional hazard assumptions, were used to evaluate long-term IO effects.
Results: After 1:1 nearest neighbor matching among 16 754 1L and 6548 2L patients, the hazard ratio (HR) was 0.942 (95% CI, 0.902-0.984) in 1L and 0.853 (95% CI, 0.795-0.915) in 2L. Adjusting for crossover effects, HR was 0.887 in 1L and 0.775 in 2L. Over the 7-year follow-up, the mean rwOS benefit was 3.2 months for 1L and 2.7 months for 2L. IO significantly delayed rwTTNT in both 1L and 2L. The IO effects increased and persisted over time, with uncertainty in the time-varying HR estimate.
Conclusion: IO improves survival in patients with mNSCLC, though the effect size is smaller than in trials and long-term survival estimates are uncertain.
Keywords: cancer immunotherapy; crossover effects; metastatic non‐small cell lung cancer; propensity score matching; real‐world evidence; survival analysis.
© 2025 The Author(s). Thoracic Cancer published by John Wiley & Sons Australia, Ltd.