The efficacy and safety of four different PD-1 checkpoint inhibitors in advanced non-small cell lung cancer: a real-world retrospective study

Background While various PD-1 inhibitors are approved for advanced non-small cell lung cancer (NSCLC), direct comparisons of their real-world efficacy and safety are limited. This retrospective study evaluates four PD-1 inhibitors (pembrolizumab, sintilimab, tislelizumab, and camrelizumab) in real-world clinical practice to supplement registered clinical trial (RCT) data. Methods We retrospectively screened 563 patients with stage IIIB-IV NSCLC who received pembrolizumab, sintilimab, tislelizumab, or camrelizumab (as monotherapy or combined with chemotherapy) between February 1, 2019, and December 31, 2022. The follow-up cutoff date was September 15, 2023. Results A total of 409 eligible patients were included (pembrolizumab: 136; sintilimab: 115; tislelizumab: 123; camrelizumab: 35). For squamous NSCLC, no significant differences in effectiveness were observed among the four inhibitors, regardless of the treatment line. For nonsquamous NSCLC, second or later-line camrelizumab was associated with a higher risk of disease progression (progression-free survival [PFS] hazard ratio [HR] 2.29; 95% CI 1.01-5.19) and mortality (overall survival [OS] HR 3.14; 95% CI 1.28-7.74) compared to pembrolizumab, despite similar objective response rates. Regarding safety, immune-related adverse events (irAEs) were consistent and manageable. Notably, sintilimab demonstrated a significantly lower incidence of irAEs than the other three inhibitors across any grade ( P = 0.004) and grades 3-5 ( P = 0.005). Conclusions In this real-world, single-center retrospective study, sintilimab and tislelizumab showed efficacy estimates broadly similar to pembrolizumab in advanced NSCLC within prespecified strata. Camrelizumab was associated with shorter PFS/OS in previously treated nonsquamous NSCLC; however, this finding is hypothesis-generating given small subgroup sizes, treatment heterogeneity, and potential residual confounding. Overall irAEs were manageable; differences in recorded incidence should be interpreted cautiously.