• effectiveness
• immunotherapy
• lung
• nsclc
• pembrolizumab
• real-world

This single-center, retrospective observational study assessed the efficacy-effectiveness (EE) gap in a cohort of 518 patients with stage IIIC-IV non-small cell lung cancer (NSCLC) treated with pembrolizumab, with or without chemotherapy, between 2017 and July 2024. Patients were stratified according to PD-L1 tumor proportion score (TPS) and histology to explore variations in the EE factor across subgroups.
Eligible patients had histologically confirmed NSCLC, received first-line pembrolizumab, and had available tumor and survival data. Patients were categorized as "trial-eligible" or "trial-ineligible" based on the major inclusion criteria of randomized clinical trials (RCTs). The primary aim was to calculate the EE factor, which compares the real-world overall survival (OS) to the pooled OS reported in RCTs for pembrolizumab. A median EE factor of 1.0 would indicate no efficacy-effectiveness gap, while values less than 1.0 would suggest shorter survival in the real-world population.
The results demonstrated a significant EE gap, with a median EE factor of 0.60 (p<0.001), indicating that real-world patients had 40% shorter survival compared to clinical trial patients. Stratified analysis showed the lowest EE factor in patients with PD-L1 TPS ≥50% (median EE factor 0.55, p<0.001), while patients with squamous histology and PD-L1 TPS <1% had the highest EE factor (median 1.31, p=0.221), suggesting closer alignment with RCT outcomes. Multivariate analysis identified poor eastern cooperative oncology group (ECOG) performance status (PS), bone metastases, chronic steroid therapy, and the presence of immune-related adverse events (irAEs) as significant factors affecting the EE factor.
This study highlights the substantial EE gap in real-world NSCLC patients treated with pembrolizumab, emphasizing that real-world outcomes are often slightly worse than those reported in RCTs. The EE factor provides a valuable metric for understanding the discrepancies between trial and real-world efficacy, driven by patient-specific factors and clinical conditions often underrepresented in RCTs. However, our median OS was generally consistent with data from pivotal trials KEYNOTE-024, KEYNOTE-189, and KEYNOTE-407, and when real-world patients are selected using criteria similar to those in RCTs this gap nearly disappears, indicating that patient selection plays a crucial role in survival outcomes.
In conclusion, this study confirms pembrolizumab's long-term effectiveness and safety, emphasizing the importance of real-world data to guide clinical decision-making. Although a modest EE gap exists, primarily driven by patient selection (especially ECOG PS and steroid use), the real-world results closely align with those from clinical trials, reinforcing pembrolizumab’s role in treating advanced NSCLC.