Development of a nomogram for predicting lymph node metastasis in patients with urothelial carcinoma of the bladder

Abstract

Background and Objectives:

Multiple nomograms have previously been described to predict recurrence, overall, and cancer-specific survival following radical cystectomy (RC) for urothelial carcinoma of the bladder (UCB).Our aim was to develop a nomogram for the preoperative prediction of lymph node (LN) metastasis in UCB.

Methods:

We prospectively collected data from 483 patients which were used in construction of the statistical model. The variables considered as predictors in the model were demographic, histopathological and radiological factors.

Results:

The full model containing all 12 covariates produced an AIC (Akaike information criterion) value of 448.9. After model selection T-stage, grade, CIS (carcinoma in situ), pathology, LV (lymphovascular) invasion and CT (computed tomography) were included as the most parsimonious model while retaining predictive accuracy. Ta in 82 (17%), T1 in 214 (445) and T2 in 187 (38%) patients. This model had an AIC of 436.4, indicating a significant improvement in model fit after the removal of unimportant predictors. The C-indices were 0.821 and 0.808 for the reduced model and the full model, respectively, indicating greater discrimination ability for the reduced model.The nomogram further emphasises the effect of CT and LV invasion on the risk of LN positivity. Specifically, regardless of all other variables, a patient with a CT will have 100 points more than a patient without a CT, corresponding to a difference in risk of approximately 40%. The odds of LN positivity for patients with a CT are 7.45 times that of patients without a CT, regardless of all other covariates. LV invasion, pathology, CIS and T-stage are also statistically significant (p = 0.05).

Conclusion:

This nomogram is a preoperative prediction tool that uses different preoperative variables with acceptable predictive accuracy for LN metastasis in patients with BC.