Predictable early onset high-dose-glucocorticoid-associated-irAE and its predictive role in anti-PD-1 monotherapy treated advanced melanoma patients.

Abstract

9544 Background: Though uncommon, a subgroup of patients with melanoma develop early-onset severe immune-related adverse effects (irAEs) that require immunosuppressive treatment with high-dose glucocorticoids (HD-GCCs). We aimed to examine the impact of early onset HD-GCC-associated-irAE (EO-HGA-irAE) in the setting of anti-PD-1 monotherapy initiation on prognosis and to develop a predictive scoring system. Methods: Clinical data was collected retrospectively from advanced melanoma patients treated with anti-PD-1 monotherapy at Massachusetts General Hospital from Sept 2009 to Dec 2017. The relationship between EO-HGA-irAE and PFS (defined as time from anti-PD-1 monotherapy initiation to PD) was assessed using 8-week conditional landmark analysis (Kaplan-Meier curves, log-rank test) and time-dependent COX regression model (multivariate analysis). Demographic characteristics and baseline routine laboratory variables were collected and correlated with occurrence of EO-HGA-irAE using logistic regression modeling. Best cutoff values were identified using ROC curve (Youden index) to dichotomize continuous variables. Results: Among 146 patients, 13 (8.9%) developed EO-HGA-irAE. In 8-week landmark analysis, median PFS was 2.9 (95% CI, 2.8-3.0) and 17.5 (95% CI, 10.8-24.2) months (P = .001) for patients with and without EO-HGA-irAE, respectively. Multivariate analysis revealed that EO-HGA-irAE was independently correlated with significantly higher risk of disease progression with hazard ratio of 2.4 (95% CI 1.4-4.0) (P = .001). Potential predictive variables (P < .25 for continuous variables, P < .1 for dichotomous variables) in favor of the occurrence of EO-HGA-irAE included age, baseline glucose, and neutrophil, eosinophil, and WBC count. After dichotomization and further validation using a logistic regression model, a scoring system (score range 0-3) composed of 3 dichotomized predictors, including age, baseline glucose, and baseline neutrophil count was developed with odds ratio of 3.4 (95% CI, 1.6-8.8) (P = .001). The probabilities of patients scoring 0,1,2,3 of developing EO-HGA-irAE were 1.1%, 3.5%, 11.1%, and 29.9%, respectively. Conclusions: Development of EO-HGA-irAE is correlated with shorter PFS in advanced melanoma patients treated with anti-PD-1 monotherapy. A scoring system based on age and simple, easily accessible, routinely tested biomarkers can be used to help predict the risk of EO-HGA-irAE occurrence. Validation with a larger sample size is required.