Joseph Liao chairs the urology department at Stanford, and when his group’s paper appeared in Nature Medicine on October 2 he summed up the state of his field in one sentence: “There are currently no biomarkers at all that are recommended or approved for surveillance.” For the patients his specialty follows, that absence has a shape. High-risk patients with a history of non-muscle-invasive bladder cancer may need a surveillance cystoscopy every three months, a thin camera threaded up the urethra and into the bladder, and by Stanford’s own account the scope can miss up to 30 percent of cancers anyway. About 85,000 Americans are diagnosed each year, and the disease’s habit of coming back is what makes it so costly to watch.
Liao’s group is offering a cup of urine instead. Their assay, uRARE-seq, reads the cell-free RNA that cells lining the urinary tract shed into urine, on the theory that a tumor leaves a different signature than healthy tissue does. Across 683 urine samples from patients and healthy volunteers, the release says, it identified 95 percent of people with localized bladder cancer and cleared 90 percent of those without it. The cohort-level breakdown shows how that number was built: a 381-gene classifier trained on 151 bladder cancer patients and 100 controls, an area under the curve of 0.97, and a second set of 142 patients and 140 controls on which performance held.
Where the test missed tells you more than where it hit. In the training cohort it caught all 13 carcinomas in situ, all 34 high-grade T1 tumors and all 28 high-grade T2 tumors, the muscle-invasive kind. Its soft spot was low-grade Ta disease, the slow and superficial category, where it found 31 of 36, or 86 percent. If a surveillance test has to be wrong somewhere, that is the place to be wrong.
The comparison a working urologist will care about is against urine cytology, the existing noninvasive standard, in which a pathologist examines shed cells under a microscope. On 93 paired samples, cytology flagged 32 cancers, a sensitivity of 34.4 percent, while uRARE-seq flagged 89, or 95.7 percent. On low-grade Ta tumors cytology caught none at all, and the RNA test found 28 of 31.
| Group | Value (percent) |
|---|---|
| uRARE-seq | 95.7 |
| Urine cytology | 34.4 |
The release also says the RNA test outperformed a DNA-based urine test, and the record behind that line is more interesting than the line. Against a tumor-naive DNA assay, one that goes in knowing nothing about the patient’s cancer, uRARE-seq posted 99 percent sensitivity to the DNA test’s 85 percent across 85 matched cases. Against a tumor-informed DNA assay, one first taught the patient’s own tumor, the two tied at 96 percent on 28 cases. The tumor-naive test lost. The tumor-informed one did not. What the RNA approach has going for it there is that it needs no tissue to learn from, which is a convenience advantage rather than proof of better biology.
The bigger claims come after detection. In 36 patients with intermediate- or high-risk non-muscle-invasive disease, treated with surgery and then BCG immunotherapy, the group tracked whether tumor RNA lingered in the urine. Among patients who stayed free of recurrence, none had detectable cfRNA after BCG; among those who went on to a high-grade recurrence, 92 percent still did. Residual RNA after BCG carried a hazard ratio of 41.17 for high-grade recurrence, against 4.30 when it was measured before BCG. A hazard ratio of 41 drawn from 36 patients is the kind of number that sends a statistician looking for the confidence interval before anyone reaches for the champagne.
The group then reported that a pre-treatment signature predicted molecular response to BCG with an AUC of 0.93 in leave-one-out cross-validation, alongside independent cohorts of 57 BCG-treated patients and 38 treated with intravesical chemotherapy. The stakes are concrete. BCG is given weekly for six weeks and can be repeated as maintenance for up to three years, it comes from a single global supplier, and shortages are common enough that urologists sometimes have to ration it. A test that told them in advance who would respond would change how a scarce drug gets spent.
All of it rests on a case-control design, with cancers and controls assembled and run through the classifier after the fact, which is how a diagnostic gets built but says little about how it performs in a clinic where patients with blood in their urine arrive unsorted. The study’s stated limitations concede the rest: participants came from two affiliated, geographically adjacent institutions, several analyses leaned on small cohorts, and the BCG and chemotherapy cohorts were not randomized. That last point leaves the treatment-prediction result, the most seductive finding in the paper, exposed to exactly the selection confounding a randomized trial exists to remove. The authors’ own conclusion is that larger prospective studies are needed, which puts the trials that would justify skipping a scope still ahead of the test, not behind it.
Then there is the company. uRARE-seq is being commercialized by Resero Bio, a San Carlos startup spun out of Stanford for that purpose, according to the Stanford release. Kevin Liu, the graduate student who is the paper’s lead author, and several coauthors hold ownership interests in Resero, and two coauthors are its employees. The three senior authors, Maximilian Diehn, Ash Alizadeh and Liao, have filed patents on urine cell-free RNA analysis and report consulting or advisory roles with multiple diagnostics and pharmaceutical companies. Every bit of this was disclosed, as the rules require. It also means the 95 percent figure was produced, framed and announced by people with a financial interest in its becoming a product, on work the public helped pay for through four NIH grants, while the upside, if the test survives the trials it has not yet faced, belongs to a private company.
Liao was right that no urine biomarker is approved for bladder cancer surveillance. He is also among the people who have filed patents on the likeliest candidate to change that, which makes his sentence an accurate description of a clinical gap and, read from a slightly different angle, of an open market.