The prognostic impact of variant histology timing in bladder cancer: a comparative analysis of primary and secondary detection

Variant/subtype histology (V/SH) in bladder cancer is associated with distinct clinical behaviour and adverse oncological outcomes. However, the prognostic significance of the timing of V/SH detection - whether present at initial diagnosis (primary V/SH) or emerging during follow-up (secondary V/SH) - remains unclear. This retrospective, single-centre study aimed to compare clinical characteristics and survival outcomes between patients with primary and secondary V/SH in bladder cancer, with particular attention to the potential influence of immortal time bias. Data from all consecutive 1,120 bladder cancer patients who underwent transurethral resection or radical cystectomy between 2015 and 2024 were retrospectively evaluated. Of these, 171 patients were identified with V/SH and categorised as primary V/SH ( n = 129) or secondary V/SH ( n = 42). Kaplan–Meier analysis with log-rank testing and multivariable Cox proportional hazards regression were performed to assess cancer-specific (CSS) and overall survival (OS). A total of 57 cancer-specific deaths were recorded. Patients with secondary V/SH had significantly longer follow-up (median 54.0 vs. 10.0 months, p = 0.001) and higher rates of recurrence (93% vs. 46%, p = 0.001) and progression (86% vs. 54%, p = 0.001). Unadjusted analyses showed significant differences in CSS ( p = 0.021) and OS ( p = 0.002). In a time-fixed multivariable Cox model, timing was not an independent predictor of CSS (HR: 0.59, 95% CI: 0.29–1.20, p = 0.146). However, in a time-dependent Cox regression model accounting for immortal time bias, secondary V/SH status showed a trend toward increased cancer-specific mortality (HR: 1.84, 95% CI: 0.88–3.84, p = 0.104), a directionally different result from the time-fixed model. The prognostic significance of variant histology timing appeared sensitive to the analytical approach used, with time-fixed and time-dependent models yielding directionally different, non-significant estimates, highlighting the importance of accounting for immortal time bias in similar retrospective study designs. Tumour size and intravesical therapy consistently emerged as independent prognostic factors across all models. These findings suggest an individualised, risk-adapted approach based on tumour characteristics. Larger, prospective, multicentre studies are warranted to clarify the true prognostic role of variant histology timing.