Reading Tumor Activity from Urine: Bladder Cancer RNA Test Also Explores Treatment Response

For patients with bladder cancer, tumor removal is often just the beginning of a long period of follow-up. The cancer is prone to recurrence, and repeated cystoscopies also impose a burden. Now, a Stanford team is trying to answer more detailed questions from a urine sample: whether a tumor is still present, and whether its biological activity can offer clues to the next step in treatment.

A study published on October 2 in *Nature Medicine* evaluated a urine cell-free RNA sequencing method called uRARE-seq, analyzing 683 urine samples from 515 participants. The research was conducted collaboratively by teams from Stanford, the US Department of Veterans Affairs Palo Alto Health Care System, and other institutions, with Resero Bio also participating. RNA reflects ongoing gene activity, so beyond detecting cancer signals, it may also describe the state of a tumor and the surrounding immune response.

However, RNA in urine is both scarce and highly fragmented. According to accounts from Stanford and LabMedica, the team adapted a sequencing technique originally used for plasma RNA, designing a dedicated capture panel to focus on transcripts that are uncommon in healthy urine but more abundant in genitourinary tumors, then using a machine learning model to identify bladder cancer signals.

In a training set of 251 samples, the localized bladder cancer detection model achieved 95% sensitivity at 90% specificity; this also means that, under the study conditions, approximately one in ten samples without cancer could still be classified as positive. After the model was fixed, independent validation using 142 bladder cancer samples and 140 noncancer controls showed similar performance. In a separate comparison of 93 paired cancer samples, uRARE-seq detected 89, while urine cytology detected 32; this comparison supports its detection potential but does not yet establish that it can replace cystoscopy.

Another direction for the test is tracking disease signals that remain after treatment. The study analyzed urine before and after surgery and intravesical bacillus Calmette–Guérin (BCG) immunotherapy, finding that persistent tumor RNA was associated with a higher risk of recurrence. BCG stimulates an immune response to eliminate cancer cells; identifying which patients still have residual disease at the molecular level after surgery could provide a basis for subsequent follow-up and treatment research.

Pretreatment urine also showed differences: patients who later responded to BCG had more abundant T-cell and immune signals, while nonresponders had stronger cell proliferation signals. Using pretreatment samples from 114 patients, the team developed biomarkers to predict responses to BCG and chemotherapy, achieving an area under the curve (AUC) of 0.93. This measures the model’s ability to distinguish between outcomes; it does not mean that treatment can be correctly selected for 93% of patients, nor does it demonstrate that changing treatment based on the model would improve prognosis.

Moving from a research tool into clinical practice still hinges on validation that more closely reflects actual medical care. This study used a case-control design, with samples from two interconnected institutions located near each other; treatment groups were also not randomly assigned and could have been influenced by unmeasured factors. The team plans larger, multicenter prospective studies to confirm the test’s performance and clinical benefit. Stanford also disclosed that some authors hold equity interests in Resero Bio, which is working to commercialize the technology, and are involved in related patents. Further independent validation would also help assess whether the findings are broadly applicable.