BACKGROUND: Intermediate-risk (IR) non-muscle-invasive bladder cancer (NMIBC) represents a heterogeneous disease group with variable oncologic outcomes. To address the need for consistent risk stratification, the International Bladder Cancer Group (IBCG) proposed a simplified model incorporating five clinical risk factors. OBJECTIVE: We sought to externally independently validate the IBCG risk stratification model among patients with IR NMIBC. DESIGN, SETTING, PARTICIPANTS: We performed a large multi-center retrospective analysis of patients diagnosed with IR NMIBC and treated with transurethral resection of the bladder between 2005 and 2025. As defined by IBCG guidelines, all high-grade tumors were excluded. Based on the number of IBCG-defined clinical risk factors, patients were stratified into IR-low (0), IR-intermediate (1-2), and IR-high (≥3). Competing-risk analyses assessed recurrence and progression. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: The primary outcomes were recurrence and progression. Competing-risk analyses were used to evaluate associations between risk groups. RESULTS AND LIMITATIONS: A total of 2822 patients with IR NMIBC were included: 1143 (41%) IR-low, 1535 (54%) IR-intermediate, and 144 (5%) IR-high. At 3 years, we observed stepwise increases in recurrence (12% vs 38% vs 77%) and stage progression (1% vs 3% vs 7%) across the IR-low, IR-intermediate, and IR-high groups. Fine-Gray models confirmed significantly increased hazards of recurrence and progression with higher risk categories (all p<0.001). Limitations include the retrospective design and variability in treatment administration. CONCLUSIONS: We present the largest validation to date of the IBCG stratification model for IR NMIBC. Our findings confirm that patients with IR NMIBC can be reliably classified into 3 distinct prognostic groups with significantly different outcomes. These data provide robust external validation of the IBCG scoring system and support its integration into clinical practice and future guideline development.
Siva et al. (Mon,) studied this question.