Bladder cancer has resource-intensive surveillance requirements, creating a need for noninvasive biomarkers that can support cancer detection and molecular characterization. Urine is particularly suited for liquid biopsy approaches because it is in direct contact with bladder tumors, but urine cell-free RNA (cfRNA) has not been extensively explored, with prior studies mainly focusing on a small number of genes.
In a study published in Nature Medicine on October 2, 2026, researchers developed urine random priming and affinity capture of cell-free RNA fragments for enrichment analysis by sequencing (uRARE-seq) to investigate whether broader urine cfRNA profiling could enable bladder cancer detection and molecular characterization. The study, titled “Urine cell-free RNA for bladder cancer detection and treatment response prediction,” also evaluated the approach for minimal residual disease detection and prediction of response to intravesical therapy.
Authors: Kevin J. Liu, William Y. Shi, Monica C. Nesselbush, Nick A. Phillips, Kathleen E. Mach, Bogdan Luca, Ravi K. Bajpai, Isabel Jabara, Jordan C. Cheng, Takeshi Sugio, Navika Shukla, Emily C. Shen, Ivy Lee Koh, Douglas J. Wu, Chih-Long Liu, Diego Almanza, Soyeong Jun, Feng Tian, Emily G. Hamilton, Christian R. Hoerner, Rui Wang, Mohammad S. Esfahani, Angela Hui, Vinh La, Grace Holton, Gabriela Rodriguez, Timothy J. Lee, Xi Ling, Nikitha Kalahasti, Spoorthi Vallamkonda, Patrick Mullane, Simon B. Chen, Eugene Shkolyar, Kris Prado, Ali Raza Khaki, Emily Chan, Mark K. Buyyounouski, Alice C. Fan, Eila C. Skinner, Ash A. Alizadeh, Joseph C. Liao, and Maximilian Diehn.
Detecting Bladder Cancer With Urine cfRNA
The researchers developed a tumor-naive bladder cancer detection model using urine cfRNA from 151 patients with bladder cancer and 100 noncancer controls. The model selected 381 genes and achieved an area under the curve (AUC) of 0.97, with 95% sensitivity at 90% specificity. Sensitivity remained high across different tumor grades and stages. The model detected:
- 86% of low-grade Ta tumors (31 of 36)
- 95% of high-grade Ta tumors (38 of 40)
- 100% of carcinoma in situ cases (13 of 13)
- 100% of high-grade T1 tumors (34 of 34)
- 100% of high-grade T2 tumors (28 of 28)
The locked detection model was then applied to an independent validation cohort comprising 142 patients with bladder cancer and 140 noncancer controls, where performance was similar to that observed in the training cohort. The investigators also evaluated whether field-effect mutations, which can affect the specificity of urine tumor DNAapproaches, influenced cfRNA-based detection. In the evaluated subset, uRARE-seq performance was similar regardless of the presence of field-effect mutations.
Comparison With Cytology and Urine Tumor DNA
In 93 paired samples, uRARE-seq detected 89 bladder cancer cases (95.7%), compared with 32 cases (34.4%) detected by urine cytology. The difference was particularly notable in low-grade Ta disease. Urine cytology detected none of the low-grade Ta tumors, while uRARE-seq detected 28 of 31 cases (90%).
The investigators also compared uRARE-seq with urine tumor DNA analysis. Among 85 matched cases, sensitivity was 99% with uRARE-seq versus 85% with a tumor-naive urine tumor DNA assay. In another 28 matched cases, uRARE-seq and a tumor-informed, field-effect-informed urine tumor DNA approach both demonstrated 96% sensitivity.
Assessing Tumor Grade and Muscle Invasion
Urine cfRNA profiles also contained information related to tumor grade and invasiveness. A machine-learning model designed to distinguish low-grade from high-grade bladder cancer achieved an AUC of 0.84with 10-fold cross-validation in the training cohort and an AUC of 0.84 in the independent validation cohort.
Among tumors classified as low grade by histopathology, higher cfRNA grade-model scores were associated with a greater risk of recurrence. The researchers separately developed a model to distinguish non-muscle-invasive bladder cancer (NMIBC) from muscle-invasive bladder cancer (MIBC). This model achieved an AUC of 0.82 in the training cohort and 0.75 in the independent validation cohort.
Detecting Molecular Residual Disease After Treatment
The study next assessed whether urine cfRNA could identify residual disease after treatment in 36 patients with intermediate- or high-risk NMIBC undergoing surgery followed by BCG induction. Urine samples were analyzed before surgery, before BCG treatment, and at the first follow-up visit after BCG. Among patients who did not develop recurrence, bladder cancer cfRNA was detectable in all patients before surgery, in 58% before BCG, and in none after BCG.
Among patients who subsequently developed high-grade recurrence, bladder cancer cfRNA remained detectable in all patients before surgery and before BCG and in 92% after BCG. CfRNA positivity was associated with worse high-grade recurrence-free survival both before BCG (HR 4.30; P=0.02) and after BCG (HR 41.17; P<0.0001).
The investigators also used serial cfRNA results to distinguish molecular responses after surgery from responses occurring after BCG. Ten patients had undetectable bladder cancer cfRNA after surgery and were classified as having a molecular complete response after surgery. Of 26 patients with residual cfRNA following surgery, 15 subsequently had undetectable cfRNA after BCG, while 11 continued to have detectable disease. Patients with molecular complete responses after either surgery or BCG had substantially lower recurrence risk than thosewith persistent molecular residual disease after treatment.
Molecular Features Associated With BCG Response
Pretreatment urine cfRNA showed distinct biological patterns according to subsequent BCG response. Patients who achieved molecular complete response after BCG had increased expression of immune-related pathways, including interferon response and T-cell-associated signatures. Urine cfRNA from BCG responders was also enriched for previously described signatures associated with tumor-reactive CD4 and CD8 T cells, tissue-resident memory T cells, tertiary lymphoid structures, major histocompatibility complex class I, and antigen processing and presentation.
In contrast, patients with persistent molecular residual disease after BCG showed increased expression of proliferation-related pathways, including mitotic spindle and G2M checkpoint genes.
T-cell receptor analysis also showed that patients responding to BCG had more T-cell clones and greater T-cell receptor diversity than patients who did not respond. The authors interpreted these findings as suggesting that the presence of a preexisting antitumor immune response may be important for BCG activity.
Predicting Response to Intravesical Therapy
Based on the immune- and proliferation-related differences observed between BCG responders and nonresponders, the researchers developed an intravesical therapy response prediction model using presurgery urine cfRNA.
Using leave-one-out cross-validation, the model achieved an AUC of 0.93 for predicting molecular response to BCG. Higher scores were associated with molecular complete response after BCG but were not associated with complete molecular response following surgery. The locked model was subsequently assessed in an independent cohort of 57 patients with NMIBC treated with surgery followed by BCG induction, where response-prediction scores were strongly associated with recurrence.
Because patients with persistent disease after BCG showed greater expression of proliferation-related transcripts, the model was also evaluated in an independent cohort of 38 patients with NMIBC treated with intravesical chemotherapy. In this cohort, patients who did not develop recurrence had lower response-prediction scores, while those who developed recurrence had higher scores.
Together, these findings suggest that pretreatment urine cfRNA may contain information relevant to the likelihood of benefit from BCG versus intravesical chemotherapy. However, this treatment-selection approach remains investigational.
Takeaway
The study showed that urine cfRNA profiling with uRARE-seq could detect localized bladder cancer with 95% sensitivity at 90% specificity, show higher sensitivity than urine cytology and tumor-naive urine tumor DNA analysis in the evaluated comparisons, and provide information about tumor grade and muscle invasion.
Serial urine cfRNA analysis also identified molecular residual disease after surgery and BCG, while pretreatment transcriptional profiles were associated with subsequent outcomes following BCG and intravesical chemotherapy.
The study had a case-control design, participants were enrolled at two affiliated and geographically adjacent institutions, and some analyses involved relatively small cohorts. Importantly, the BCG and intravesical chemotherapy validation cohorts were not randomized, leaving the possibility of unmeasured treatment-selection confounding.
The authors concluded that urine cfRNA analysis represents a promising biomarker approach for bladder cancer, but larger prospective studies are required to establish its clinical utility, including its potential role in treatment selection.
The full article is available in Nature Medicine.
Find more videos on OncoDaily YouTube TV.
