# Optimizing Ki-67 and E-Cadherin Thresholds for Improved Grade and Stage Classification in Urothelial Bladder Cancer

## TL;DR

This study identifies optimal thresholds for Ki-67 and E-cadherin to better classify bladder cancer stages and grades, improving clinical decision-making.

## Contribution

The study proposes new cut-off values for Ki-67 and E-cadherin and demonstrates that their combination improves bladder cancer classification.

## Key findings

- Ki-67 with thresholds ≥40% for MIBC and ≥30% for HG showed the strongest single-marker performance.
- Combining Ki-67 and E-cadherin improved classification accuracy with AUCs of 0.851 for MIBC and 0.838 for HG.
- E-cadherin, evaluated on an inverted scale, provided complementary value to Ki-67 in tumor classification.

## Abstract

Background: Bladder cancer exhibits substantial heterogeneity, and accurate discrimination between non-muscle-invasive (NMIBC) and muscle-invasive disease (MIBC), as well as between low-grade (LG) and high-grade (HG) tumors, remains essential for appropriate clinical management. Established immunohistochemical (IHC) markers, such as p53, Ki-67, and E-cadherin, could be used in a new setting, but standardized cut-off values and their combined predictive value remain unclear. This study aimed to identify optimal cut-offs for these markers and to evaluate whether biomarker combinations enhance the discrimination of tumor grade and stage. Methods: A retrospective dataset of 568 cases of bladder cancer was analyzed. For each case, the expression of p53, Ki-67, and E-cadherin was quantified, and tumors were classified as NMIBC or MIBC, and as LG or HG. ROC-based cut-off selection was performed using Youden’s J criterion with 10-fold stratified cross-validation. E-cadherin was modelled using an inverted scale to reflect biological loss. Logistic regression models were used to evaluate the discriminatory performance of single markers, two-marker combinations, and a three-marker model. Cross-validated AUC values and optimal thresholds were reported. Results: Ki-67 showed the strongest single-marker performance for predicting both MIBC (AUC 0.842) and HG disease (AUC 0.813), with optimal cut-offs of ≥40% and ≥30%, respectively. p53 demonstrated moderate discrimination (AUC 0.778 for MIBC and 0.776 for HG), while E-cadherin, evaluated on an inverted scale, showed acceptable performance (AUC 0.746 for MIBC; 0.780 for HG). Combining markers yielded modest improvements, with the best performance observed for Ki-67 + E-cadherin (AUCs of 0.851 for MIBC and 0.838 for HG). Conclusions: Ki-67 is the most effective single biomarker for distinguishing invasive and HG bladder cancer, while E-cadherin provides complementary value. A two-marker panel combining Ki-67 and E-cadherin, using appropriate cut-offs, offers the highest overall performance and may serve as a practical tool for enhanced pathological stratification.