Comparative performance of EUA and AUA risk stratification models and development of an integrated model for predicting ≥pT2 upper tract urothelial carcinoma: A ROBUUST analysis

Document Type

Conference Proceeding

Publication Date

3-1-2026

Publication Title

Eur Urol

Keywords

Urology & Nephrology

Abstract

Introduction & Objectives: As upper tract urothelial carcinoma (UTUC) management is shifting toward nephron-sparing when oncologically safe, the EAU and AUA each propose risk stratification systems to guide treatment. However, their comparative accuracy is unclear. We compared their ability to predict non-organ-confined (≥pT2) disease and developed an integrated model using the most informative variables from both. Materials & Methods: We analyzed patients from the multicenter ROBUUST database who underwent nephroureterectomy for M0 UTUC. Each patient was classified according to both EAU and AUA models. Variables from each score underwent backward stepwise multivariable logistic regression (MLRA) to identify independent predictors of ≥pT2 and build an integrated model. A weighted score (based on rounded Odds Ratios [OR]) defined 3 risk groups: low (≤3), intermediate (4-6), and high (≥7). Receiver operating characteristic/area under curve (ROC/AUC) evaluated performance in predicting ≥pT2 for EAU, AUA, and integrated models. Results: Among 1132 patients, 522 (46%) had ≥pT2 disease, 71 % had high-grade (HG) biopsy, 63% had local invasion on preop imaging, 48% hydronephrosis, and 22% multifocality. Suspicious lymph nodes (cN1) were in 8%, previous lower tract involvement in 29%, and HG cytology in 38%. Median tumor size was 2.6 cm. At MLRA, HG biopsy (OR 4.17, 95%CI 3.23-5.42, p<0.05), invasive aspect (OR 2.71, 95%CI 2.15-3.44; p<0.05), hydronephrosis (OR 1.60, 95%CI 1.29-2.00, p<0.05), multifocality (OR 1.45, 95%CI 1.03-2.04, p=0.035), cN1 (OR 2.77, 95%CI 1.81-4.33; p<0.05), tumor size (OR 1.04 for each cm, 95%CI 1.02-1.07; p<0.05) were independently associated with ≥pT2 disease and included in the integrated model. With the novel model, 164, 586 and 382 patients were classified as Low-, Intermediate-, and High-risk, respectively. At ROC, EAU and AUA systems showed modest discrimination (AUC 0.56 and 0.66, respectively). The integrated model showed superior predictive performance (AUC 0.74; p<0.001 vs. both). Conclusions: An integrated model incorporating high-yield variables from AUA and EAU systems improves discriminatory accuracy for invasive disease and may provide a more robust tool to guide shared decision making for preoperative counseling and treatment selection.

Volume

89

First Page

1

Find It @ Sladen

Share

COinS