Validation of Gene Expression Signatures to Identify Low-risk Clear-cell Renal Cell Carcinoma Patients at Higher Risk for Disease-related Death

Background Approximately 5–10% of patients with “low-risk” clear cell renal cell carcinoma (ccRCC), as stratified by externally validated clinicopathologic prognostic algorithms, eventually have disease relapse and die. Improving prognostic algorithms for these low-risk patients could help to provide improved individualized surveillance recommendations. Objective To identify genes that are differentially expressed in patients with low-risk ccRCC who did and did not die of their disease. Design, setting, and participants Using the Mayo Clinic Renal Registry, we identified formalin-fixed paraffin-embedded samples from patients with low-risk ccRCC, as defined by Mayo Clinic stage, size, grade, and necrosis score of 0–3. We conducted a nested case-control study between patients who did (cases) and did not (controls) have ccRCC relapse and death, using two independent sets (discovery and validation). We performed RNA sequencing of all samples in the discovery set to identify differentially expressed genes. In the independent validation set, we assessed the top 50 expressed genes using the nCounter Analysis System (NanoString Technologies, Seattle, WA, USA). Results and limitations In the discovery set of 24 cases and 24 controls, 92 genes were differentially expressed with p < 0.001. The top 50 genes were validated in an independent set of 22 cases and 22 controls using linear mixed models. In the validation set, 10 genes remained differentially expressed between the groups. Conclusions RNA signatures from formalin-fixed paraffin-embedded blocks can identify patients with low-risk ccRCC who die of their disease. This finding provides an opportunity to help guide improved surveillance in patients with low-risk ccRCC. Patient summary In the current study we identified RNA signatures from low-risk clear cell renal cell carcinoma patients who died from this disease. Improving prognostic algorithms for these low-risk patients could help to provide improved individualized surveillance recommendations.