Comparison of peripheral blood markers in predicting the occurrence of immune-related adverse events in patients with metastatic genitourinary cancer treated with cabozantinib + nivolumab +/- ipilimumab.

Abstract

e16506 Background: Peripheral blood markers (PBM) have been demonstrated to be prognostic and/or predictive for cancer patients (pts) treated with immune checkpoint inhibitors (ICI). Initial evidence has shown their potential role in predicting the occurrence of immune-related adverse events (IrAE) in pts treated with single-agent ICI. We explored the potential role of different PBM in predicting IrAE in metastatic genitourinary cancer (mGU) pts treated with an ICI-based combination. Methods: We retrospectively evaluated 60 mGU pts treated with Cabozantinib + Nivolumab +/- Ipilimumab (CaboNivo +/- Ipi) in a phase I study at the National Cancer Institute. We retrieved their demographics, the occurrence of IrAE, and the baseline values of different PBM (neutrophil/lymphocyte ratio [NLR], neutrophil/eosinophil ratio [NER], lymphocyte/monocyte ratio [LMR], monocyte/lymphocyte ratio [MLR], platelet/lymphocyte ratio [PLR], systemic immune-inflammation index [SII] [N/L*P], and HALP score [albumin*hemoglobin*L/P]). We evaluated PBM based on the presence or absence of IrAE using the Wilcoxon rank sum test followed by multivariable logistic regression analysis. Receiver operating characteristic (ROC) curve analyses were used to investigate an optimal cut-off for PBM suitable for prediction. Results: The median age of pts was 58.5 years (interquartile range [IQR] 46 - 67), and 13 (21.7%) were female. Twenty (33.3%) pts developed at least one IrAE. Baseline median values of PMB are summarized in the Table. A multivariable logistic regression model showed after backward elimination that only NER had the potential to predict IrAE (p = 0.033; Odds Ratio = 0.989 per unit [95% Confidence Interval 0.979 - 0.999]) and thus, we further investigated the predictive potential of NER for the development of IrAE following CaboNivo +/- Ipi. The ROC curve suggested that an optimal cut-off value for NER was 30.647 (area under the ROC = 0.6351). Accordingly, we classified pts with high NER ( > 30.647) as not having IrAE vs low NER (≤ 30.647) as having IrAE. Using this classification, 24/37 (64.9%) of those without (w/o) IrAE and 14/20 (80%) of those with an IrAE could simultaneously be correctly predicted by NER alone. Conclusions: A baseline low NER value was an independent factor associated with a higher probability of developing IrAE in mGU pts treated with CaboNivo +/- Ipi. Prospective validation on a larger cohort of pts is needed. [Table: see text]