Analysis of the association between prospectively collected immune-related adverse events and survival in patients with solid tumor treated with immune-checkpoint blockers, taking into account immortal-time bias

Background: Numerous retrospective studies and reviews have reported a positive association between immune-related adverse events (irAEs) and survival in non-small cell lung cancer (NSCLC) and melanoma patients treated with immune checkpoint blockers (ICBs). However, some results are controversial and the studies, whose results converge, should be interpreted cautiously because most of them do not deal appropriately with the immortal-time bias. Here, we report an observational real-life study of the association between prospectively collected irAEs and survival of patients treated with ICBs while dealing with the immortal-time bias. Methods: Data from patients treated at Gustave Roussy from June 2014 to October 2017 with anti-PD-(L)1 antibodies for a melanoma or NSCLC have been prospectively collected in the REISAMIC database, a pharmacovigilance registry dedicated to irAEs. Adverse events of grade 2 and higher were collected prospectively. To study the association between the occurrence of irAEs and survival, we used both a landmark analysis and a Cox regression model with time-dependent covariate. Results: 577 patients were treated with anti-PD-(L)1 antibodies for melanoma (60.3 %) or NSCLC (39.7 %). The occurrence of an irAE was significantly associated with improved overall survival (OS): HR 0.56, 95 % CI [0.41; 0.75], p = 0.0001 and progression-free survival (PFS): HR 0.63, 95 % CI [0.47; 0.83], p = 0.001 using a Cox regression model with time-dependent covariate. In a 12-week landmark analysis, median OS was 21.2 months (95 % CI, 12.2 to 35.7) and 16.4 months (95 % CI, 12.4 to 21.3) p = 0.26 and median PFS was 14.3 months (95 % CI, 9.5 to 24.6) and 13.4 months (95 % CI, 10.2 to 18.3) p = 0.66, for patients with and without irAEs, respectively. Conclusions: In our real-life study of patients with melanoma and NSCLC treated with anti-PD-(L)1 antibodies, we confirm that irAEs are associated with improved survival using a time-varying Cox regression model. Analysis with a landmark method showed no difference in OS or PFS between patients who experienced irAE during the first 12 weeks of treatment and those who did not. Retrospective analysis and reviews including studies that do not deal with the immortal-time bias and studies insufficiently powered for a landmark analysis should be interpreted with caution.