Urinary peptidomic signatures predict overall and progression-free survival in patients with bladder cancer

Clinicopathologic calculators for bladder cancer moderately predict survival and fail to depict the underlying molecular phenotype. We applied urinary capillary electrophoresis-mass spectrometry (CE-MS) to identify prognostic signatures linked to bladder cancer outcome. In a discovery cohort (n=131; mean follow-up 623 days), 114 survival-associated peptides, predominantly collagen fragments, were significant prognostic factors of survival and were integrated into a classifier (BC110), resulting in an accuracy of 0.80 (p-value < 0.001). Validation of the classifier in an independent cohort (n=102; mean follow-up 1605 days) confirmed correlation with survival (AUC: 0.78; p-value=0.03). Survival analysis using the BC110 scores resulted in significant prediction of both overall (p-value<0.0001) and progression-free survival outcome (p-value < 0.0001). To test biological plausibility, a previously reported collagen-focused model (COL210) was subsequently investigated and demonstrated concordant prognostic separation, reinforcing extracellular matrix remodeling as the underlying signal. These urine-based classifiers enable non-invasive risk stratification and may complement guideline calculators by identifying high-risk patients for adjuvant therapy and low-risk groups for reduced surveillance, potentially lowering reliance on repeated cystoscopy. Significance StatementAccurate non-invasive risk stratification in bladder cancer remains a major unmet clinical need, given the diseases high recurrence rates and the need for cystoscopic surveillance. We applied a standardized urinary peptidomics workflow using capillary electrophoresis-mass spectrometry (CE-MS) to develop a machine learning based classifier (BC110) enriched in collagen and other extracellular matrix (ECM) components, in a discovery cohort (n=131) of mainly NMIBC patients. In an independent validation cohort (n=102), BC110 achieved robust quartile-based separation of overall and progression-free survival, with a PFS hazard ratio of 7.23 (highest vs. lowest quartile; p<0.0001). Prognostic performance was further confirmed with a previously reported fibrillar collagen model (COL210), identifying ECM remodeling as the biological driver of the peptide alterations depicted in urine in association with BC progression. These classifiers provide a reproducible, urine-based approach to non-invasive survival prediction, enabling risk-adapted surveillance and earlier therapeutic intervention. While urinary peptidomics requires prospective validation, it is poised to reduce reliance on invasive cystoscopy and to advance personalized care through clinical implementation.