Bladder cancer is usually found with a camera. Patients with blood in the urine or a history of the disease undergo cystoscopy, a procedure that passes a scope into the bladder, and a urine cytology test that looks for cancer cells under a microscope. The cytology test is easy, but the paper notes its sensitivity is modest, especially for low-grade tumors. A study published in Nature Medicine on Oct. 2 describes a urine test that reads RNA instead, and it found 95% of localized bladder cancers in the team's samples.

The headline holds up better than one of the claims built on it, and the difference is worth explaining.

The Detection Numbers

The test, called uRARE-seq, comes from Stanford Medicine and the VA Palo Alto Health Care System. Researchers led by Maximilian Diehn, Ash Alizadeh and Joseph Liao sequenced fragments of cell-free RNA in urine and trained a model to recognize cancer. They analyzed 683 urine samples from 515 people.

In the training group of 251 samples, 100 controls (including 52 people with blood in the urine or urinary symptoms) and 151 bladder cancers, the model reached an AUC of 0.97, a measure where 1.0 is perfect. It detected 95% of cancers at 90% specificity, meaning 1 in 10 controls was wrongly flagged. Sensitivity was 86% for low-grade tumors confined to the lining, and 95% to 100% for higher-grade and invasive disease, according to the paper.

A locked version of the model was then tested on 142 cancers and 140 controls it had not seen, and performed nearly identically.

Beating Cytology and Urine DNA

The comparison with cytology is stark. In 93 patients with both tests, uRARE-seq detected 89 cases (95.7%) and cytology detected 32 (34.4%). Cytology detected none of the low-grade tumors; uRARE-seq caught 28 of 31.

Against urine DNA tests, run at a matched 87% specificity, the RNA method was more sensitive than a version that does not need a tumor sample (99% versus 85%) and matched one that does (96% for both). The authors also found that field-effect mutations, which can show up in healthy-looking bladder lining and confuse DNA tests, did not affect the RNA test.

Blogerroom has covered similar efforts to replace invasive follow-up with simpler tests, from a breath test for lung nodules to the Mayo model that flags pancreatic cancer risk from routine records. The bladder test is further along than the breath test in one respect: it was validated on a held-out group.

Residual Disease After Surgery and BCG

The paper also followed 36 patients with non-muscle-invasive bladder cancer through surgery and BCG, an immunotherapy delivered into the bladder. Among patients who did not relapse, all had detectable tumor RNA before surgery, 58% before BCG, and none after BCG. Among those with a high-grade recurrence, all were positive before surgery and before BCG, and 92% remained positive after BCG. A positive post-BCG result carried a hazard ratio of 41 for recurrence.

That is a strong association, though from 36 patients at one health system.

The Prediction Claim and Its 19 Patients

The headline treatment claim is that the test can predict whether a patient will respond to BCG or to intravesical chemotherapy, with an AUC of 0.93. That figure needs context. In the Results section, the 0.93 comes from a training group of 19 patients, 11 who did not clear their tumor RNA after BCG and 8 who did, tested with leave-one-out cross-validation. The abstract's "114 patients" is the total across the training group, a 57-patient BCG validation group and a 38-patient chemotherapy group.

The validation results are encouraging. The locked score was linked to recurrence in the BCG group and, in reverse, to recurrence after chemotherapy. But only 5 of the 38 chemotherapy patients relapsed, the groups were not randomized, and the authors note that treatment choice may have been shaped by unmeasured factors. In my view, this is a hypothesis worth a clinical trial, not a basis for choosing treatment.

Conflicts and Limits

The study's design is case-control, and all patients came from two nearby institutions, as the authors state. Specificity held in people with benign urinary conditions, but numbers for each diagnosis were modest. Prospective studies are needed to show clinical usefulness, the paper says.

One more detail matters. Monica Nesselbush, a co-author, is co-founder and CEO of Resero Bio, which issued a press release describing the study as evidence for its cell-free RNA platform. That is a normal route from lab to market, and the paper is open access with code and sequencing data shared, but readers should know a company stands to benefit.

What Comes Next

Three results would change how this is used. The first is a prospective trial in people who present with blood in the urine, which is the real screening population, not a case-control set. The second is testing at other hospitals. The third is whether treatment-selection scores hold up in a randomized comparison of BCG and chemotherapy. The authors point out that BCG supply shortages recur, which is one reason a way to select patients for chemotherapy could matter.

For now this is a research result, not an available test. Anyone who notices blood in the urine should see a doctor promptly. This article is general information, not medical advice.