Impact of oncological biomarkers and clinical factors on survival outcomes in renal cell carcinoma with spinal metastasis: a retrospective analysis.

OBJECTIVE: Renal cell carcinoma (RCC) is a common malignancy that metastasizes to the spine, leading to complex treatment challenges and reduced survival. The aim of this study was to evaluate the prevalence of oncological biomarkers in patients with RCC spinal metastasis and analyze how these biomarkers, along with clinical factors, impact survival outcomes. METHODS: This retrospective cohort study included patients with RCC spinal metastasis who were treated surgically at a single academic center between 2013 and 2024. Clinical, surgical, and treatment data were collected. Immunohistochemical analysis of the 10 most prevalent biomarkers was performed on spinal tumor specimens. Predictive modeling of overall mortality was conducted using multivariate logistic regression, decision tree, and random forest algorithms. Model performance was assessed using the area under the curve, and key interactions were identified through interaction depth analysis. Unsupervised clustering was used to stratify patients into biomarker-defined risk groups. RESULTS: Thirty-six patients (mean age 60 years) were included in the analysis. The overall mortality rate was 61.1%, with a mean follow-up duration after diagnosis of spinal metastasis of 16.1 months. CK7 and AE1/AE3 were the most significant predictors of mortality, with CK7-positive and AE1/AE3-negative patients having an 80% mortality rate versus 25% in AE1/AE3-positive patients (p < 0.01). CAM5.2 and KRAS expression were associated with 100% and 80% mortality, respectively. Patients with EGFR positivity and CK7 negativity had 0% mortality (p = 0.01). Decision tree analysis identified CAM5.2, CK7, and AE1/AE3 as key hierarchical classifiers. In the random forest analysis, CK7 and AE1/AE3 (mean minimal depths of 2.04 and 2.34, respectively) had the highest variable importance scores (Gini p < 0.01) (area under the curve: logistic regression = 0.74; decision tree = 0.72; random forest = 0.81). Unsupervised clustering stratified patients into three molecular subgroups with distinct mortality risks: cluster 1 (61.5%), cluster 2 (50.0%), and cluster 3 (66.7%). CONCLUSIONS: Immunohistochemical biomarkers, particularly CK7, AE1/AE3, CAM5.2, and EGFR, hold significant prognostic value in RCC spinal metastasis and can help stratify patients into high- and low-risk groups. The integration of these biomarkers into tree-based machine learning models provides interpretable data-driven decision tools to support personalized surgical and systemic treatment planning. These findings warrant validation in larger multi-institutional prospective cohorts and further analysis of biomarker interactions.