[Development of a 36-gene signature for predicting recurrence in non-muscle-invasive bladder cancer and validation in a Chinese cohort].

OBJECTIVE

To develop a transcriptomic-based recurrence prediction model for non-muscle-invasive bladder cancer (NMIBC), to comprehensively validate its efficacy in an independent Chinese cohort, and to explore the underlying biological mechanisms driving tumor recurrence.

METHODS

Transcriptome profiling data from the multicenter European Urothelial Cancer Molecular Classification Consortium (UROMOL) cohort (n=535) were used as the training cohort. The least absolute shrinkage and selection operator (LASSO) Cox regression was employed to identify prognostic genes and construct a risk score model. Internal validation was strictly performed using Bootstrap resampling and 10-fold cross-validation. External validation was conducted in an independent Chinese cohort (n=85) from Peking University People ' s Hospital. Gene set enrichment analysis (GSEA) was applied to elucidate the distinct biological differences between high- and low-risk groups.

RESULTS

A 36-gene prognostic signature was identified from 12 170 genes to construct the risk score model. In the training cohort, the model achieved a concordance index (C-index) of 0.702 (95%CI: 0.664-0.740), with 1-, 3-, and 5-year areas under the receiver operating characteristic curve (AUC) of 0.782, 0.811, and 0.799, respectively. The 5-year recurrence-free survival (RFS) rates for the high- and low-risk groups were 35.2% and 72.8%, respectively (Log-rank P < 0.001). Internal validation via Bootstrap (C-index=0.702) and 10-fold cross-validation (C-index=0.708) demonstrated robust stability. In the independent Chinese validation cohort, the model maintained a favorable C-index of 0.685 (95%CI: 0.631-0.739), with 1-, 3-, and 5-year AUC of 0.705, 0.743, and 0.665, respectively. Its predictive performance was comparable to the traditional European Organization for Research and Treatment of Cancer (EORTC) risk score (C-index=0.652). GSEA revealed that the low-risk group was enriched in energy metabolism and protein homeostasis pathways, while the high-risk group was associated with extracellular matrix (ECM) remodeling and cell adhesion pathways.

CONCLUSION

The 36-gene signature developed in this study demonstrates favorable predictive performance and stability across both the training and independent Chinese cohorts. While exhibiting comparable predictive efficacy to the traditional EORTC scoring system, the model serves as an independent prognostic factor that provides crucial supplementary molecular stratification. Furthermore, the model underscores the critical roles of energy metabolism and ECM remodeling in NMIBC recurrence, providing a valuable molecular basis for precise individualized management of NMIBC patients in China.