Effect of digital Ki67-based risk modeling on prognostic precision in non–muscle-invasive bladder cancer and high-risk subsets in mixed-grade tumors.

642 Background: Grade is a major determinant of progression in non–muscle-invasive bladder cancer (NMIBC), but tumor heterogeneity and subjective histologic grading complicate risk stratification. Ki67, a proliferation marker, provides an objective, reproducible measure of tumor biology. We evaluated whether digitally quantified Ki67 predicts recurrence, grade progression, and T1+ stage progression across low-grade (LG), mixed-grade (MG), and high-grade (HG) NMIBC. Methods: We retrospectively reviewed 472 patients with non-invasive papillary urothelial carcinoma (2012–2023). Exclusions included carcinoma in situ, BCG-unresponsive disease, lymphovascular invasion, prostatic urethral involvement, and re-staging specimens. Histology was reclassified per 2022 WHO criteria. Ki67 IHC was digitally quantified (Visiopharm) recording total and strong nuclear staining. Associations with recurrence, grade progression, and T1+ stage progression were evaluated using Kaplan–Meier, Fine–Gray competing risk, and multivariable Cox regression adjusting for tumor size, multifocality, prior bladder cancer, and intravesical therapy. Results: The cohort included 472 patients (LG 125, MG 192, HG 155) with a mean follow-up of 41.7 months. Ki67 expression increased significantly with tumor grade, both in total (LG 16.1%, MG 21.1%, HG 38.9%) and strong nuclear staining (LG 7.9%, MG 11.3%, HG 25.0%) (p < 0.001). In multivariable Cox models, strong nuclear Ki67 independently predicted recurrence (HR 1.01 per 1% increase, 95% CI 1.003–1.02, p = 0.007) and T1+ stage progression (HR 1.06, 95% CI 1.03–1.08, p < 0.001). Total Ki67 predicted T1+ progression (HR 1.04, 95% CI 1.02–1.06, p < 0.001) but not recurrence after adjustment. Threshold analyses: Ki67 ≥25.6% for recurrence (HR 1.43, p = 0.039) and ≥35% for T1+ progression (adjusted HR 3.1, p = 0.027). High Ki67 identified MG tumors with aggressive, high-grade–like behavior not apparent morphologically. Tumor size (HR 1.2, p = 0.03) and multifocality (HR 1.5, p = 0.02) also independently predicted progression. Conclusions: Digital Ki67 quantification provides an objective, reproducible proliferation measure that improves prognostic precision in NMIBC. High Ki67 (> 25–35%) identifies patients at increased risk for recurrence and T1+ progression, particularly within MG tumors. Incorporating digital Ki67 into risk models may enhance individualized management and surveillance.