Endocrine immune-related adverse event is a prognostic biomarker independent of lead-time bias.

ABSTRACT

OBJECTIVES: Immune-related adverse events (irAEs) are known to be associated with clinical efficacy and better prognoses in patients receiving immune checkpoint inhibitors. In particular, endocrine irAE (e-irAE) is related to better prognoses. Since the incidence of irAEs increase as treatment duration becomes longer, we should consider lead-time bias not to overvalue the result. We evaluated the impact of e-irAE on the outcome before and after 6-, 9-, and 12-week landmark analyses. MATERIALS AND METHODS: We evaluated 222 patients with advanced or recurrent non-small cell lung cancer who received anti-PD-1 antibodies such as nivolumab or pembrolizumab from January 2016 to April 2021. Treatment efficacy and outcomes of patients with or without e-irAE (e-irAE group or no e-irAE group) were retrospectively evaluated. In addition, we performed 6-, 9-, and 12-week landmark analyses to exclude the effect of lead-time bias. RESULTS: Median progression free survival (PFS) was significantly longer in the e-irAE group than in the no e-irAE group (overall 15.3 vs 3.9 months, p < 0.0001; 6-week 15.3 vs 4.9 months, p < 0.0002; 9-week 19.8 vs 6.1 months, p = 0.0012, 12-week 19.8 vs 8.4 months, p = 0.017). Overall survival (OS) was significantly longer in the e-irAE group (overall not reached (NR) vs 15.4 months, p = 0.0003; 6-week NR vs 19.1 months, p = 0.0049, 9-week NR vs 22.2 months, p = 0.006; 12-week NR vs 23.3 months, p = 0.04). We used the multivariate cox proportional hazard model to adjust for confounding factors and found that e-irAE had better impact on both PFS and OS (PFS overall hazard ratio 0.37 [95% confidence interval 0.23-0.56], 6-week 0.41 [0.26-0.63], 9-week 0.43 [0.24-0.63], 12-week 0.52 [0.31-0.84]; OS overall 0.40 [0.22-0.68], 6-week 0.46 [0.25-0.79], 9-week 0.47 [0.24-0.84], 12-week 0.58 [0.29-1.08]). CONCLUSION: The occurrence of endocrine irAE was associated with better efficacy and prognoses regardless of the lead-time bias.