Artificial intelligence (AI)-powered immune phenotype (IP) predicts differential benefit from nivolumab plus ipilimumab (NIVO+IPI) versus sunitinib (SUN) in advanced clear cell renal cell carcinoma (ccRCC)

Background While NIVO+IPI places a standard first-line combination treatment for advanced ccRCC, predictive biomarkers to optimize its therapeutic benefit remain limited. This study evaluated whether AI-based immune phenotyping can identify patients who derive benefit from NIVO+IPI versus SUN.

Methods Spatial distribution of tumor-infiltrating lymphocytes were analyzed on H&E-stained slides using Lunit SCOPE IO to define IPs of tumors as inflamed or non-inflamed. Associations between IPs and progression-free survival (PFS), overall survival (OS), and response rate were evaluated in advanced ccRCC patients treated with either NIVO+IPI or SUN. Transcriptional relevance of the IP was assessed in The Cancer Genome Atlas dataset by comparing AI-based classification with established inflamed gene expression signatures.

Results The AI-defined IP showed strong concordance across multiple inflamed transcriptional signatures. Compared to non-inflamed tumors, inflamed tumors demonstrated significantly longer PFS (HR=0.280, 95% CI: 0.158-0.498, P<0.001), OS (HR=0.353, 95% CI 0.156-0.800, P=0.013), and higher response rates (60.5% vs. 23.2%, P<0.001) with NIVO+IPI (N=125), but not with SUN (N=128). Furthermore, while treatment outcomes of NIVO+IPI was superior to those of SUN in inflamed tumors, no significant difference was observed in non-inflamed tumors. Favorable PFS (HR=0.521, 95% CI 0.333-0.813, P=0.004) and OS (HR=0.508, 95% CI=0.288-0.895, P=0.019) were validated in an independent cohort of ccRCC patients treated with NIVO+IPI (N=128).

Conclusions We suggest the AI-defined IP as a promising biomarker to identify patients with advanced ccRCC more likely to benefit from NIVO+IPI versus SUN. This approach may help navigate more effective and personalized first-line treatment strategy in advanced ccRCC.