A pre-cystectomy decision model to predict pathologic upstaging and oncologic outcomes in clinical stage T2 bladder cancer.

Abstract

4578 Background: Neoadjuvant chemotherapy is the standard for advanced bladder cancer (BC), although its benefit in lower risk clinically organ-confined muscle-invasive (cT2N0M0) BC is uncertain. This study aimed to define pre-cystectomy factors in cT2N0M0 BC that could predict pathological upstaging at cystectomy and oncological outcomes. Methods: 1,964 institutional BC patients were reviewed. Neoadjuvant chemotherapy-naïve cT2N0M0 patients who underwent cystectomy were included. A multivariate classification and regression decision tree was used to assess hierarchical interactions of univariately significant variables, thereby identifying patients at greatest risk for upstaging and poor outcome. Results: 948 cT2N0M0 patients were identified; 512 (54%) were upstaged at cystectomy. Pathological upstaging was associated with poor RFS and overall survival (OS) (both p < 0.001). Age, lymphovascular invasion (both P = 0.01), multifocality (p = 0.004), deep muscle invasion, tumor growth pattern and hydronephrosis (all p < 0.001) were associated with upstaging. When these factors were included in a multivariate decision tree model, 71% of patients with hydronephrosis were upstaged and had the worst outcome (p < 0.001). In patients without hydronephrosis, tumor growth pattern was a second-tier discriminator; only 37% of patients with papillary tumors were upstaged. In patients with non-papillary tumors, deep muscle invasion was a third-tier discriminator; 70% of patients with invasion were upstaged. Age was a third-tier discriminator in patients with papillary + non-papillary features; 33% of patients ≤65 years were upstaged compared to 47% of patients >65 years. The decision tree was cross-validated and resulted in 3 risk groups – 5 year RFS/OS for low risk patients was 66%/48% versus 46%/27% for high risk patients (p < 0.001). Conclusions: This study identified pre-cystectomy variables that predict upstaging and poor outcomes in cT2N0M0 BC. The decision tree approach highlighted the most crucial prognostic variables, and can aid in identifying low risk patients in a stepwise fashion who have lower probability of upstaging and better outcomes.