Construction of predictive models for recurrence and progression in >1000 patients with non-muscle-invasive bladder cancer (NMIBC) from a single centre.

OBJECTIVE: To construct predictive models based on the objectively calculated risks of progression and recurrence of non-muscle-invasive bladder cancer (NMIBC) in a large cohort of patients from a single centre. PATIENTS AND METHODS: Between October 1984 and March 2009 a cohort of 1019 patients (877 males; 142 females; median age 44 years) with histologically confirmed NMIBC was included in this study. Among these patients, 74% received bacillus Calmette-Guérin (BCG)-based therapy. Complete transurethral resection of bladder tumour of all visible tumours was carried out in all patients, and the stage and grade were determined. Univariate analysis and multivariate Cox regression were used to identify predictors of recurrence and progression. The studied predictors included age, sex, stage, grade, associated carcinoma in situ, tumour size, multiplicity, macroscopic appearance of the tumour, history of recurrence and type of adjuvant intravesical therapy. Multivariate logistic regression models were used to develop the 12- and 60-month recurrence and progression predictive models. The predictive accuracy of the models was assessed for discrimination as well as calibration. RESULTS: The median (range) follow-up was 44 (6-254) months. On multivariate analysis, stage, multiplicity, history of recurrence and adjuvant intravesical therapy were significantly associated with recurrence, whereas for progression only tumour grade and size were significant independent predictors. The constructed nomograms had a 64.9% and 69.4% chance of correctly distinguishing between two patients, one destined to have a recurrence and one not at 12 and 60 months, respectively. The constructed nomograms had a 70.2% and 73.5% chance of correctly distinguishing between two patients, one destined to progress and one not at 12 and 60 months, respectively. All predictive models were well calibrated. CONCLUSIONS: Based on multivariate analysis of the studied prognostic factors nomograms for predicting recurrence and progression in NMIBC were constructed. Most of the studied patients had received BCG-based therapy, making these models more closely applicable to contemporary practice than others. These predictive models have reasonable discriminative ability and are well calibrated, but require external validation before they can be applied to other populations.