Rosetta Genomics released an abstract for this years' annual meeting of the American Society of Clinical Oncology reiterating the concordance and sensitivity of its diagnostic test for cancers of unknown primary origin, and showing that the test frequently influences clinical decision-making.
The abstract, available online, describes how the Rosetta Cancer Origin Test (previously called MiRview Mets) performed when analyzing 258 specimens referred to the company. The data show that the test results were more than 80 percent concordant with doctors' best guesses based on clinical and pathological features of patients' cancer, and that the test impacted doctors' treatment decisions in a majority of cases.
"We already had a lot of good validation data and a lot of good retrospective studies with academic centers, looking at the concordance between the best final diagnosis and the test results. So, with this study, we really wanted to show from a real-world setting with real-world samples, how the test correlates [with clinical and pathological measures], and how doctors actually use the results," Robert Wassman, Rosetta's chief medical officer, told PGx Reporter this week.
"It was a way to try to mine our own data … from work done over the past year and a half … [and] see how our results are impacting care for the doctors that are ordering the test commercially," he said.
The Rosetta Cancer Origin Test is a 64-microRNA expression array, which categorizes metastatic cancer samples as originating from one of 42 specific tumor types. Determining the site of origin for a patient's cancer has implications for how a doctor treats the disease.
In the abstract, Rosetta researchers discuss comparing the predicted cancer type, based on results of the origin test, to other clinical information, pathology findings, supplemental IHC testing results, treatment response, and clinical course, as well as physician perception of the relevance and utility of the test results.
Of 258 samples submitted, the company was able to obtain sufficient tumor tissue from 217, and successfully process 192 using the cancer origin test.
Overall, according to the researchers, the concordance of the test results with the clinical or pathology-based final diagnosis was 86 percent, and when the reporting algorithm yielded a single diagnosis — in about half of the cases — the concordance rose to 89 percent.
Wassman said that the concordance results were "remarkably close" to what the company has seen in its earlier validation studies.
"Doing this in the real world is a little less rigorous than in controlled academic studies," he said, "but when we totaled [it] up and looked at it – 86 percent of the time it matched with clinical or pathological best diagnosis."
"I wouldn’t have been disappointed if it had been 10 points lower since it was a real world [analysis]," Wassman added.
The data also revealed that about 70 percent of the time, the results of the test either prompted physicians to consider a different course of therapy or to be more confident in the therapy they would have chosen without the test information.
According to Wassman, in one case, a physician was treating her patient based on one of two origin options she got from pathology analysis and wasn’t getting a good response. "But getting the results [from our test] that confirmed that diagnosis gave her the confidence to stay on that [treatment] path," he said.
The most common cancer origin predicted by the test was colorectal cancer, in 12 percent of samples, followed by breast cancer in about 10 percent. More than half of the cases represented one of 30 less-common tumor types
According to Wassman, the confirmation of a high proportion of breast and colon cancer cases also has interesting clinical implications. For one, he said, the company found in the recent study that a small percentage — about three percent of the breast cancer cases — were males.
In such rare cases, the test information may be particularly impactful because "those doctors aren’t thinking about breast cancer at all," Wassman said.
While a former competitor, Pathwork Diagnostics, recently went out of business (PGx Reporter 4/10/2013), Rosetta is still joined in the CUP space by BioTheranostics' CancerType ID test, an assay that measures expression of 96 genes to diagnose a metastatic cancer's origin.
BioTheranostics last year published a prospective clinical utility study of its test in 252 patients, showing that patients with cancer of unknown primary who were tested using the CancerType ID assay, and who then received chemotherapy specific to their predicted tumor of origin, survived longer on average than expected for standard empirical CUP treatment (PGx Reporter 10/17/2012).
The company has also announced it is presenting several posters and releasing a number of abstracts for the upcoming ASCO meeting.
Included in this set is a study re-evaluating BioTheranostics' previous outcomes study after adjusting for pre-treatment prognostic factors, which concluded that the "assay results remained a significant independent variable affecting survival outcome in these patients after adjustment for prognostic heterogeneity."
Another abstract describes a modeled cost analysis study concluding that the company's CancerType ID assay was cost effective 94 percent to 99 percent of the time based on a societal willingness-to-pay threshold of $100,000.
In another study, similar to Rosetta's, BioTheranostics tracked how its test impacted physicians' therapeutic decision making by surveying 103 doctors. The survey revealed that assay prediction was concordant with physicians' final clinical diagnosis in 84 percent of cases, and "guided initial treatment choice in 52 percent of cases, changed a considered treatment in 10 percent, excluded a potential treatment in 6 percent, and confirmed a previously determined treatment in 29 percent of cases."
According to the company, three posters at the meeting will present results of BioTheranostics' test specifically for periampullary adenocarcinomas, poorly differentiated neoplasms, and metastatic neuroendocrine tumors.
According to Wassman, Rosetta is now planning prospective studies to establish that its test — when it identifies the correct origin of a cancer and patients are treated according to that diagnosis — actually improves patient outcomes.
Rosetta is also working to organize and underwrite a study in the UK, currently under grant review by the NIH, with the University College London Hospitals' John Bridgewater, which would be a "truly randomized trial or therapeutic change [study] based on [Rosetta's] test or on empiric therapy," Wassman said.
This will be a somewhat difficult undertaking, he said, primarily because it will be necessary to recruit a large enough cohort of patients to show that the test helps guide treatment decisions that significantly improve patient outcomes.
For example, he said, breast cancer is one of the two most common findings the company sees — about 10 percent — "so in a 400 person study you are only going to have about 40 breast patients, and that's a relatively small cohort."
Breast cancer is also one of the limited cancer types for which Rosetta expects treating patients according to the test results versus empirically would make a significant difference in outcomes.
"With lung cancer, even with the best specific treatment it will only be a little difference, and with other cancers, like biliary tract cancer, our ability to treat those cancers hasn’t advanced enough to affect that," he said.
Researchers conducting BioTheranostics' outcomes study also highlighted this as a main difficulty — there are still small numbers of cancer types currently for which clinicians would choose a totally different treatment for a specific, molecularly defined cancer versus unspecified CUP. But, they said that as more and more molecularly targeted cancer therapies are developed, the utility of tests like its CancerType ID — and Rosetta's Cancer Origin Test — will only increase.
In the meantime, Wassman said, Rosetta is looking for other surrogates of clinical utility that it can demonstrate, "like this mined data we have [in the abstract."
He said the company is also working with groups who do mathematical modeling to potentially publish a modeled analysis of how its test should impact clinical outcomes.