Development and external validation of a novel nomogram model for predicting postoperative recurrence-free survival in non-muscle-invasive bladder cancer

Abstract

Background: Transurethral resection of the bladder tumor with or without adjuvant intravesical instillation (IVI) has been the standard treatment for non-muscle-invasive bladder cancer (NMIBC), whereas a high percentage of patients still experience local tumor recurrence and disease progression after receiving the standard treatment modalities. Unfortunately, current relevant prediction models for determining the recurrent and progression risk of NMIBC patients are far from impeccable.

Methods: Clinicopathological characteristics and follow-up information were retrospectively collected from two tertiary medical centers between October 2018 and June 2021. The least absolute shrinkage and selection operator (LASSO) and Cox regression analysis were used to screen potential risk factors affecting recurrence-free survival (RFS) of patients. A nomogram model was established, and the patients were risk-stratified based on the model scores. Both internal and external validation were performed by sampling the model with 1,000 bootstrap resamples.

Results: The study included 299 patient data obtained from the Affiliated Hospital of Xuzhou Medical University and 117 patient data obtained from the First Affiliated Hospital of Guangxi Medical University. Univariate regression analysis suggested that urine red blood cell count and different tumor invasion locations might be potential predictors of RFS. LASSO-Cox regression confirmed that prior recurrence status, times of IVI, and systemic immune-inflammation index (SII) were independent factors for predicting RFS. The area under the curve for predicting 1-, 2-, and 3-year RFS was 0.835, 0.833, and 0.871, respectively. Based on the risk stratification, patients at high risk of recurrence and progression could be accurately identified. A user-friendly risk calculator based on the model is deposited at https://dl0710.shinyapps.io/nmibc_rfs/.

Conclusion: Internal and external validation analyses showed that our model had excellent predictive discriminatory ability and stability. The risk calculator can be used for individualized assessment of survival risk in NMIBC patients and can assist in guiding clinical decision-making.

Keywords: NMIBC; bladder cancer; nomogram; predictive indicator; risk calculator; risk factor; tumor recurrence.

Figure 1

The flow chart for study inclusion and exclusion in the training cohort and external validation cohort. NMIBC, non-muscle-invasive bladder cancer; IVI, intravesical instillation; TURBT, transurethral resection of the bladder tumor.

Figure 2

(A) An UpSet plot showing the intersections between tumor invasion locations. (B) Plot showing the ten-fold cross-validation via minimum criteria for the selection of the optimal value of tuning parameter (λ). Dotted vertical lines were drawn at the value with the minimum criteria and one standard error of the minimum criteria. (C) The least absolute shrinkage and selection operator coefficient profiles of the 21 clinicopathologic features associated with recurrence-free survival. A dotted vertical line was drawn at the optimal λ value identified through ten-fold cross-validation. The resulting 10 predictors with non-zero coefficients were identified based on the log (λ1se) value. (D) Heat map showing the correlation between the patients’ clinicopathologic features based on Spearman’s rank correlation coefficient. BMI, body mass index; IVI, intravesical instillation; SII, systemic immune-inflammation index = platelet* neutrophil/lymphocyte.

Figure 3

Multivariate Cox regression analysis of the patients’ clinicopathologic features. N, number of patients; IVI, intravesical instillation; SII, systemic immune-inflammation index = platelet* neutrophil/lymphocyte.

Figure 4

(A) The constructed nomogram for predicting recurrence-free survival of non-muscle-invasive bladder cancer patients after transurethral resection of the bladder tumor. (B) Kaplan-Meier curves of low-risk, intermediate-risk, and high-risk groups based on the prediction of the nomogram. (C) Harrell’s concordance index for 1-, 2-, and 3-year recurrence-free survival of two models. (D) Decision-curve analyses demonstrate the net benefit of using the models. Model 1, the nomogram; model 2, model based on age, gender, T category, prior recurrence status, pathology grade, tumor number, and maximum tumor diameter. IVI, intravesical instillation; SII, systemic immune-inflammation index = platelet* neutrophil/lymphocyte.

Figure 5

(A) Time-dependent ROC curves of the nomogram in the training cohort. (B) Time-dependent ROC curves of the nomogram in the external validation cohort. (C) Time-dependent ROC curves of model 2 in the training cohort. (D) Time-dependent ROC curves of model 2 in the external validation cohort. Model 2, model based on age, gender, T category, prior recurrence status, pathology grade, tumor number, and maximum tumor diameter. ROC, receiver operating characteristic; AUC, area under the curve.

Figure 6

(A) Calibration plot of the nomogram done by bootstrapping with 1,000 resamples for predicting recurrence-free survival(RFS) in the training cohort. (B) Calibration plot of the nomogram done by bootstrapping with 1,000 resamples for predicting recurrence-free survival in the external validation cohort.