Real-world outcomes of immunotherapy in lung cancer patients: Prompt evidence generation beyond clinical trials

Abstract

The introduction of immune checkpoint inhibitors (ICIs) has transformed the treatment of advanced non-small cell lung cancer (NSCLC) without targetable driver mutations. Evidence for these treatments mainly originates from randomized controlled trials (RCTs), but outcomes in clinical practice may differ because real-world patients are more heterogeneous and treatment and follow-up schedules are less standardized. Real-world evidence can therefore complement RCT data by evaluating effectiveness in routine care. Therefore, this thesis aimed to complement data on survival outcomes reported in RCTs of patients with advanced NSCLC with data from patients treated with immunotherapy in clinical practice, while also evaluating methodological challenges and advancements for prompt evidence generation. Chapter 2 evaluated outcomes of immunotherapy in clinical practice. In stage III NSCLC, durvalumab after chemoradiotherapy improved progression-free survival (PFS) and overall survival (OS) compared with historical controls receiving chemoradiotherapy alone. Median PFS in clinical practices was longer than reported in the PACIFIC trial, whereas OS was similar, likely reflecting differences in follow-up and outcome assessment between trials and clinical practice. In metastatic NSCLC, first-line pembrolizumab combined with chemotherapy resulted in shorter OS than in the corresponding RCT among patients with non-squamous tumors and PD-L1 expression <1%, while no major differences were observed in other PD-L1 subgroups. In patients with PD-L1 <1%, nivolumab plus ipilimumab combined with chemotherapy showed numerically better response rates and slightly longer PFS than pembrolizumab plus chemotherapy, suggesting potential benefit in clinical practice, although longer follow-up is required for OS evaluation. Another study investigated the impact of concomitant medications on immunotherapy outcomes. Antibiotic and opioid use were associated with worse survival in both immunotherapy-treated patients and chemotherapy-treated historical controls, indicating that these findings are more likely explained by underlying patient characteristics than by direct effects on immunotherapy effectiveness. Chapter 3 focused on methodological challenges and advancements related to prompt evidence generation. A scoping review showed that observational studies frequently report PFS inconsistently, with substantial variation in definitions and measurement methods compared with RCTs. To improve the scalability of outcome assessment, a text-mining algorithm was developed to identify disease progression from electronic health records. The algorithm accurately detected progression across multiple Dutch hospitals, and PFS estimates were comparable to manual chart review, supporting its use for large multicenter real-world studies. In addition, a Bayesian survival model was developed to enable early comparison of accumulating real-world data with fixed RCT data. This approach provided reliable estimates of survival differences before completion of data collection, demonstrating its potential for timely evidence generation. Overall, this thesis demonstrated that survival outcomes of patients with advanced NSCLC treated with immunotherapy in clinical practice can differ from those reported in RCTs, highlighting the added value of real-world evidence. It also showed that measuring PFS in routine care is challenging and susceptible to bias, while methodological innovations such as text mining and Bayesian survival modeling can support more rapid and reliable generation of real-world evidence