Urine-Based Biopsy Shows 97.8% Sensitivity in Prostate Cancer Detection

A urine-based biopsy displayed a 97.8% sensitivity rate across all Gleason grades among patients scheduled for prostate biopsy based on clinical indication, potentially expanding prostate cancer detection while reducing unnecessary invasive procedures, according to findings from a US-based prospective validation study published in Diagnostics.¹

Findings from the study showed among 283 patients across 26 urology practices, the liquid assay identified 180 of 184 cancer cases. Moreover, the specificity rate was 53.3%, with an F1 score of 90.2% identified for the group and an area under the curve (AUC) of 0.91.

For high-grade cancers identified as those with a Gleason score from 8 to 10 (n = 36), the assay conferred a 97.3% specificity rate and 55.6% sensitivity rate with an AUC of 0.87. Furthermore, with the addition of intermediate-grade cancers, defined as Gleason 3 + 4 = 7, the biopsy displayed a 92.0% specificity rate and a 79.1% sensitivity rate with an AUC of 0.93. Finally, for low-to-intermediate grade disease, which included Gleason grade 6 and 3 + 4 = 7, the sensitivity rate was 94.0% and the specificity rate was 68.8%, with an AUC of 0.89.

Model reproducibility was confirmed with cross-validation across independent data partitions, and F1 scores regularly exceeded 68% across grade comparisons. Furthermore, the investigators suggested that the biopsy reliably distinguished biochemical patterns associated with the disease from controls while displaying distinct performance characteristics across the disease continuum.

“This study reflects a different way of approaching cancer detection,” Obdulio Piloto, PhD, chief scientific officer and co-founder of PanGIA Biotech, the developers of the urine-based assay, stated in a news release on the findings.² “Rather than targeting a single biomarker, this urine-based liquid biopsy uses machine learning to interpret complex biochemical patterns. That allows us to capture signals that may be missed by more reductionist approaches and supports detection across the full spectrum of disease.”

Of the 184 patients with biopsy-confirmed prostate cancer, 123 were between the ages of 60 and 70 years of age, with a mean age of 61.2 years (range, 30-100), consistent with the known prostate cancer epidemiology. Additionally, 84 controls who showed no clinical indication of prostate cancer provided voided urine specimens and were comparatively younger in age.

Those scheduled for needle-guided prostate biopsy received the urine collection kits at least 1 week before the procedure, with patients self-collecting first morning-voided urine in pre-labeled containers. Twenty-five mL of the specimen was transferred into a conical container and combined with a 25 mL methanol fixative. These samples were then shipped to a central laboratory for processing.

Patients who were identified as positive for prostate cancer were stratified by Gleason score, prostate-specific antigen (PSA) levels, and clinical staging criteria. A total of 199 patients were initially identified as cancer-positive, with 15 removed via principal component analysis. Of the 84 controls, 9 were removed as outliers, yielding 75 samples for analysis.

Additionally, a machine learning framework using random forest classifiers was employed. The model was trained on the training partition and evaluated on the held-out test for each fold, with repetition occurring 5 times across different, random partitions. Performance metrics in the study included specificity, F1 score, and AUC, and confusion matrices were generated for each analysis.

Serum PSA did not show a strong correlation with positive diagnosis, with the study investigators pointing to evidence suggesting superior early detection of disease of urinary PSA vs serum PSA. They further posited that an integration of the urine-based assay with serum PSA, urinary PSA, and clinical examination could significantly improve diagnostic precision and risk stratification.

“This marks an important external validation of our platform and reinforces the clinical and scientific foundation underlying our diagnostic pipeline,” Holly Magliochetti, chief executive officer and co-founder of PanGIA Biotech, stated in the release.² “It reflects both the rigor of the work and the potential for broader application across multiple cancer types.”

References

1. Hausman MS, Ambert K, Nagesetti B, et al. Urine-based machine learning assay detects prostate cancer. Diagnostics. 2026;16(7):993. doi:10.3390/diagnostics16070993
2. PanGIA Biotech announces peer-reviewed study in diagnostics showing 97.8% sensitivity in detecting prostate cancer using a urine-based liquid biopsy with machine learning. News release. April 2, 2026. Accessed April 2, 2026. https://tinyurl.com/bde9jhr8