MP66-12: Urinary Metabolic Profiling Allows Rapid Bladder Cancer Diagnosis

Monday, September 13, 2021 8:00 PM to 10:00 PM
Abstract

Information

Authors: Caleb Seufert, Xiaowei Song, Kathleen Mach, Timothy Lee, Richard Zare, Joseph Liao

Introduction: Non-invasive strategies for bladder cancer (BC) detection are recognized as a major unmet need. Metabolomics, the profiling of small molecule substrates, intermediates, and products of cell metabolism, can be used to identify malignancies using cells or biological fluids. Urine represents a non-invasive carrier of choice for BC diagnostics.  Conventional mass spectrometry is limited by complex sample processing creating a barrier to clinical translation.  In this pilot study, we investigated conductive polymer ionization mass spectrometry (CPSI-MS) of urine samples and analyzed the results with machine learning (ML) to identify the metabolomic signature of BC.

Methods: With IRB approval and participant consent, urine samples were collected from individuals with and without tissue-proven BC. Histopathological grade distribution included 27.5% (11) low grade and 72.5% (29) high grade.Initial metabolite identification and ML model development was conducted with 47 control and 40 BC urine specimens. For CPSI-MS processing, 3 ml of urine was loaded directly onto the tip of the conductive polymer.  Total assay time from sample loading to data acquisition averaged 15 seconds. Mann-Whitney U test was used to discover the potential metabolite markers in the univariate analysis. Using the optimal ML model, support vector machine (SVM) sets of metabolic profiles were identified to distinguish between BC and control samples.  A 5-fold cross validation was introduced to assess CPSI-MS/ML generalization ability to identify BC.

Results: A set of 49 candidate urine metabolites was selected based on fold difference, statistical significance, and metabolic pathway origin. These metabolites are involved in arginine, proline, carnitine, and histidine metabolic pathways. The SVM model distinguished BC from benign control urine specimens with an accuracy of 87.4%, and sensitivity and specificity were both 87%.  Preliminary analysis suggests metabolomic profiling could be used to discern pathological grade.

Conclusions: The CPSI-MS/ML assay present here is a feasible tool for rapid, non-invasive, and sensitive urine metabolite analysis that warrants validation in a larger patient cohort. CPSI-MS allows for direct urine analysis and is amenable to automated detection of bladder cancer in clinical practice.

Source of Funding: None

Therapeutic Area
Oncology: Bladder/Urothelium/Urethra