Purpose: To explore the ccRCC clinical and immune characteristics correlated with IL-23 expression level and build pre-operative prediction models based on contrast CT scans.
Materials and methods: The study included the cancer genome atlas kidney renal clear cell carcinoma cases to build a bioinformatics cohort. The cases with qualified contrast CT images were selected as radiographic and radiomics cohort. The IL-23 expression level groups were defined by median-based thresholding. The clinical characteristics were compared between groups. The impacts of IL-23 on immune microenvironment composition were measured via the CIBERSORT. Two radiologists evaluated the pre-operative contrast CT images. The radiomics features were automatically extracted. IL-23 group-specific radiographic and radiomics features were collected and used for prediction model establishment via Orange Data Mining Toolbox. P < 0.05 was set as statistically significant.
Results: For total, 530 ccRCC cases were included. The IL-23 group was significantly associated with survival, histologic grade, AJCC tumor stage, AJCC cancer stage, and plasma calcium level. Except for Treg and other T cells, IL-23 showed correlation with NK cell, mast cell, monocyte infiltration. Axial length was the only significant radiographic measurement between IL-23 groups. The radiomics features established an IL-23 group prediction model with the highest 10-fold cross-verification AUC of 0.842.
Conclusion: The clear cell renal cell carcinoma IL-23 expression level had prognosis and immune microenvironment correlation and could be predicted by pre-operative radiomics features.