Development and external validation of a FISH-clinical nomogram for predicting overall survival in bladder cancer patients after radical cystectomy

TL;DR

Researchers developed a model combining FISH test results and clinical data to predict survival in bladder cancer patients after surgery, validated in multiple patient groups.

Contribution

A novel FISH–clinical nomogram was developed and externally validated for predicting bladder cancer survival after radical cystectomy.

Findings

A FISH–clinical model with six independent predictors showed good calibration and discrimination (C-indexes 0.772–0.705).

Chromosome 7 aneuploidy and p16 locus loss were significant predictors incorporated into the model.

Patients were stratified into high-, medium-, and low-risk groups based on the model.

Abstract

Bladder cancer has notable heterogeneity. The urine-based fluorescence in situ hybridization (FISH) test can detect bladder cancer noninvasively. In this study, we aimed to construct a nomogram based on FISH results and clinical features (referred to as the FISH–clinical model) to predict the overall survival (OS) of bladder cancer patients following radical cystectomy (RC).

A total of 261 eligible patients were enrolled for this study. The SYSMH cohort was divided into training (n = 138) and internal validation (n = 70) sets; the SYSUTH cohort was used for external validation (n = 53). Multivariate Cox proportional hazards regression was applied for FISH–clinical model construction, and model performance was evaluated according to analyses of calibration, discrimination ability, and clinical usefulness.

FISH-identified chromosome 7 and 17 aneuploidies correlated significantly with increased pT stage; the former was associated with lymph node metastasis. Six variables, age, tumor size, pT stage, lymphovascular invasion, chromosome 7 aneuploidy, and p16 locus loss, were found to be independent predictors of OS and were incorporated into our FISH–clinical model. The model demonstrated good calibration and discrimination, with C-indexes (95% CIs) of 0.772 (0.693–0.851), 0.712 (0.605–0.819) and 0.705 (0.587–0.822), in the training, internal validation and external validation sets respectively. Decision curve analysis demonstrated the model’s clinical utility. Furthermore, all enrolled patients were successfully categorized into high-, medium- or low-risk groups, and stratified analyses were performed.

Preoperative FISH has predictive value for OS, and we developed a FISH–clinical model for OS prediction in bladder cancer patients who have not received neoadjuvant chemotherapy or immunotherapy. This model showed favorable predictive efficacy with internal and external validation.

The online version contains supplementary material available at 10.1186/s12885-025-14677-w.

Figures