NEW YORK – A team at the University of Michigan has laid the groundwork for what they hope can be a valuable tool in the personalized care of patients with kidney cancer, publishing initial results on their development and early validation of a gene-expression score that can distinguish higher versus lower-risk tumors.
Gene expression signatures have found purchase in several tumor types, most notably breast and prostate cancer, but earlier efforts to apply the same approach in kidney cancer had fallen short, said Simpa Salami, a urologist and professor at the University of Michigan school of medicine.
He and his colleagues' recent publication in JCO Precision Oncology described their discovery work that resulted in a 15-gene expression signature, which the team then validated in multiple independent sample sets, demonstrating that the test's prognostic readout was independent of various other existing risk factors.
The group is now working on a prospective validation study and is in conversations with potential commercial partners after sharing early data at a recent meeting of the American Urological Association.
"About 80,000 new cases [of kidney cancer] are diagnosed every year, and about 15,000 die from it every year, including patients with cancers that are localized … and currently we don't have a clinically useful biomarker in the kidney cancer space," Salami said.
In their effort to remedy this, the Michigan team assembled a training set of 110 patients that underwent kidney removal surgery, half of whom later recurred with systemic disease and half of whom remained disease-free. Using whole-transcriptome sequencing, investigators looked for genes that were differentially expressed in patients who recurred versus those who did not.
Machine-learning algorithms then allowed them to narrow down to the resulting 15-gene score, which they showed was associated with both risk of recurrence and risk of death.
Salami emphasized that this held true even when the team did a multivariable analysis, adjusting for clinical and pathological risk factors, as well as multi-factor risk scores developed by other institutions, and a prior molecular score that was at one time pursued by Myriad Genetics as the MyPlan Renal test. Authors reported a hazard ratio of over 11 for disease-free survival and nearly 10 for disease-specific survival for their new signature.
In their JCO paper, the Michigan team also reported on the validation of their score in two independent cohorts — the first, a group of 380 kidney cancer cases from The Cancer Genome Atlas, and the second, another 380 individuals assembled from several other smaller datasets.
The signature's independent association with both disease-free survival and death from kidney cancer held strong in these independent validations, though with lower but still significant hazard ratios, Salami said.
In the TCGA samples, the group was able to use their locked cutoff from the initial assay development work. For the other cohort, they had to use what is known as a Z-score approach, because sequencing techniques differed amongst the source datasets.
The next step, which Salami and his colleagues have already begun, is to follow this up with prospective validations, he said.
Because of the research value in having data from the whole transcriptome, the team is trying to maintain that methodology in their ongoing efforts, but Salami said that he and his colleagues are also working on converting their signature to an RT-PCR platform.
"It's only 15 genes, so we have the option to develop a more locally applicable and robust assay," he said.
After establishing their analytical validity, prior gene-expression tests for breast and prostate cancer faced a long road in proving their utility — their value in informing or altering clinical decision-making. The same will likely be true for Salami and his colleagues.
"I do biomarker research, but in the clinic, if I do a test, I ask myself, 'How is this test going to change what I do?'" Salami said.
One thing he and his team hope for the assay is that it could help doctors better tailor treatment for localized disease.
"One of my collaborators did a study that looked at data where they did biopsies of small renal masses, tumors that were less than 4 centimeters, [looking] at what factors determined who got treated versus who didn't. And the main driver of treatment was not whether cancer was found in the biopsy or not. It was the size. Anything more than 2 centimeters tended to get treated, and anything less was put on surveillance," Salami said.
A molecular test could potentially identify patients with small tumors who may be at higher risk, as well as patients with larger tumors who might not actually need treatment.
A second application could be in patients who receive surgery to identify those at an elevated risk who would be referred to adjuvant systemic treatment.
"Right now, the guidelines recommend that for patients with stage III cancer, and some patients with stage II who have adverse pathologic features, a doctor should consider additional treatment. But only about 30 percent of patients who get additional treatment actually benefit from it. So rather than give it to everybody, what if we can use this biomarker to identify those who are truly high-risk," Salami said.
He added that in early work with clinical collaborators, the signature's prognostic power has been holding up, but demonstrating predictiveness — that the test can determine who will respond to a particular treatment — has been more variable. Hence, the importance of continuing to use RNA sequencing, which could allow the team to tweak their signature to work better in different clinical settings.
Salami also mentioned a third potential application for his team's score, which would be in helping to personalize post-treatment surveillance strategies. For patients with a low molecular risk, clinicians could consider longer time periods for follow-up imaging or vice versa for those with high-risk test results.
In terms of future commercialization, Salami said that that the group has been focused on sharing results at scientific meetings and is considering its options in terms of potential out-licensing or spinning out its own company.
"We want to do whatever we think is going to be the best in bringing this to our patients," he said.