PREDICTION OF IMMUNE CHECKPOINT INHIBITOR THERAPY IN UROTHELIAL CARCINOMA: INTEGRATED ANALYSIS OF MOLECULAR PATTERNS AND CLINICAL FACTORS
Hermann-Váradi Melinda
Surgical Medicine Division
Dr. Szijártó Attila
SE Belgyógyászati és Onkológiai Klinika, Simonyi terem
2025-06-23 15:30:00
Urology
Dr. Nyirády Péter
Dr. Szarvas Tibor
Dr. Bödör Csaba
Dr. Beöthe Tamás
Dr. Dank Magdolna
Dr. Kocsmár Ildikó
Dr. Tóvári József
In recent years, ICIs have become available for the systemic treatment of advanced UC. PD-1 and PD-L1 inhibitor therapies have resulted in durable therapeutic effect in a subset of patients. The molecular background of the large individual heterogeneity regarding the response to ICIs is still poorly understood.
The aim of my research was to identify factors affecting ICI therapy effectiveness in a real-world UC cohort. We assessed the clinical characteristics of more than 200 UC patients receiving ICI treatment in routine clinical settings and compared the effectiveness of pembrolizumab and atezolizumab against results from respective clinical trials. Our findings revealed that the ORRs and OS times were comparable to those reported in the clinical trials. Moreover, we identified worse ECOG PS, metastases, low hemoglobin, and Bellmunt risk factors as independent predictors of shorter OS.
The molecular part of this work aimed to identify potential ICI-predictive genes through gene expression analysis of the tumor using the NanoString technology. Our gene expression analysis identified 23 genes with prognostic value for OS. For ORR and DCR, 20 and 43 genes, respectively, showed significantly different expressions. Although seven genes were associated with both endpoints (OS and DCR), PSMB10 was the only one validated in the IMvigor210 transcriptome data for both outcomes.
Additionally, we examined how different molecular subtypes of UC relate to therapy response and survival outcomes. Using various classifications, we found the highest ORR and longest OS in the neuronal, TCGA-luminal infiltrated, and MDA-p53-like subtypes. To address the limitation of small sample sizes, we analyzed neuronal signature scores and found high scores associated with improved OS and PFS. This indicates that even patients with non-neuronal subtypes but elevated neuronal scores may benefit from ICI therapy.
We applied the Random Survival Forest model to identify the best combination model incorporating clinicopathological and molecular factors. Our prognostic model effectively distinguished patients with longer survival on ICI therapy.
Finally, we evaluated the prognostic value of soluble PD-L1 for ICI-treated UC patients. Higher pre-treatment serum PD-L1 concentrations were associated with poor OS. In addition, we observed a significant sPD-L1 flare in on-treatment serum samples from atezolizumab-treated but not pembrolizumab-treated patients.
