Abstract
Background: Substantial variability exists in outcomes of patients with metastatic urothelial carcinoma (mUC) receiving PD1/PD-L1 inhibitors. Unfortunately, there is an absence of optimal predictive biomarkers of refractory disease. We hypothesized that with a combination of baseline and early-on-treatment (EOT) parameters we can distinguish patients who will respond from those who will have refractory disease. We investigated readily available baseline and EOT variables from the phase 3 clinical trial IMvigor211, which compared atezolizumab, a PD-L1 inhibitor, with chemotherapy (taxane or vinflunine), in patients with mUC who have progressed on platinum-based chemotherapy.
Methods: Of 931 patients in IMvigor211, 902 were evaluable for our analysis. We built logistic regression models for both atezolizumab and chemotherapy to predict the clinical endpoint of risk of treatment failure defined as progressive disease as best response. We used LASSO to select covariates from an array of clinical (e.g. metastatic site, performance status, prior therapy) and laboratory variables (e.g. albumin, hemoglobin, PD-L1 expression) measured before treatment, and EOT 3-6 weeks after starting therapy. Moreover, we used our models to identify clinical features that are differentially associated with outcomes in response to either therapy.
Results: Based on pre-treatment information, our baseline model achieves a predictive accuracy for treatment failure of AUC = 0.69 for both therapies. The strongest baseline predictors are PD-L1 expression, prior treatment with both cisplatin and carboplatin, and the presence of liver metastases. Moreover, higher TNM-stage of the primary tumor is strongly predictive of worse treatment outcomes for atezolizumab but not for chemotherapy. After the first treatment cycle (3 weeks), the predictive accuracy of the model increases to AUC = 0.83 with differences in importance of variables between the atezolizumab- and chemotherapy-treated patients. No variable stood out as a dominant predictor in the EOT model, with 14 variables significant at p=0.05. We found that 66% of patients have a lower predicted risk of treatment failure for atezolizumab than for chemotherapy at baseline. Our EOT model suggests that 36% of patients treated with atezolizumab would benefit from switching to chemotherapy after the first cycle.
Conclusions: Our prediction models employing readily available and affordable EOT clinical and laboratory variables robustly (AUC 0.83) identified patients with early resistance and may inform patient-specific therapeutic decisions. Further validation is required.
Citation Format: Christopher Graser, Thomas O. McDonald, Guru Sonpavde, Franziska Michor. Use of early dynamics of clinical and laboratory parameters to predict resistance in patients with metastatic urothelial carcinoma receiving post-platinum atezolizumab [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2392.