NEW YORK – Results from a study published last week in JCO Precision Oncology have added to the growing evidence for the value of paired tumor origin testing and molecular profiling to guide the treatment of patients with cancers of unknown primary (CUP).
Cancer of unknown primary presents as metastatic disease without a clinically identified primary site. CUP has been associated with especially poor outcomes, at least historically, because patients were limited to physicians' best guess at treatment with empiric chemotherapy.
Early studies of tumor origin testing failed to show a benefit, but those trials were conducted without the ability to direct patients to targeted therapy, before the advent of immunotherapy, and with other limitations, such as an overrepresentation of tumor types with no good treatments, like pancreatic cancer.
With the advent of more modern site-specific chemo regimens — and more recently, molecularly targeted treatments and immunotherapies — there are now possibilities for improvement, as long as both the molecular and biological nature of a patient's cancer can be untangled.
In the newly published study, investigators led by senior author Anthony Greco, an oncologist with Tennessee Oncology and cofounder of the Sarah Cannon Research Institute, analyzed data collected by Hologic's Biotheranostics on patients who received testing that included a combination of the company's 93-gene expression assay, CancerType ID, and molecular profiling performed by NeoGenomics.
Greco, who is also an adviser to Biotheranostics, said the two companies currently comarket their tests, with NeoGenomics targeted sequencing-based NeoType assays serving as a reflex for patients who have a tumor origin predicted by CancerType ID. Patients who receive certain CancerType results might have site-specific NeoType tests, while others might get pan-cancer sequencing.
Investigators mined Biotheranostics' MOSAIC database cataloging the rate of cancer type identifications and the frequency of actionable biomarkers in a select set of genes, including KRAS, IDH1/2, BRCA1/2, and BRAF.
The team reported that CancerType ID identified a specific tumor type in nearly 93 percent (2,929 of 3,168) of CUP cases in the MOSAIC database. The most commonly identified histological type was adenocarcinoma, at 75 percent, with pancreaticobiliary being the most common molecularly diagnosed cancer, at 25 percent. Alterations in KRAS, IDH1/2, BRCA, and BRAF genes were identified in 19 percent of biopsies. A cancer-specific US Food and Drug Administration-approved or investigational targeted therapy was potentially available for 25 percent of these patients.
Greco said that while the current report concentrated on a subset of oncogenes, the group has studied many others that could aid in personalized treatment decision-making, despite not being FDA-approved for the specific tumor types CancerType ID predicts.
Being a retrospective database, without patients' clinical treatment details, MOSAIC didn't offer the authors the opportunity to analyze how many patients eligible for a targeted therapy actually got one or track their outcomes.
But evidence from the most recent studies of site-specific molecularly targeted or immunotherapy treatment in CUP patients has made clear that it improves patient outcomes compared to those treated without this information available, Greco said this week.
One such study, the CUPISCO trial, found in 2023 that genomic profiling and molecularly guided treatment improved outcomes for patients with CUP. In that trial, newly diagnosed patients were randomized to receive either routine platinum-based chemotherapy or targeted therapy and cancer immunotherapy guided by comprehensive genomic profiling. They were not tested for their cancer type or tissue of origin.
Greco highlighted another study, published earlier this year in Lancet Oncology, in which investigators found that patients with CUP whose treatment was guided by a Canhelp-Origin — a gene expression test similar to CancerType ID — with subsequent biomarker analysis had improved survival compared to those who received empirical chemotherapy. Median overall survival was 28.2 months in the site-specific therapy group and 19.0 months in the empirical chemotherapy group.
Canhelp Genomics, the test's developer, is currently seeking strategic partnerships to help secure regulatory approval in the US and other countries.
Greco said that with the weight of evidence demonstrating a benefit, additional randomized trials that include an untested empiric chemotherapy arm would be unethical at this point, but he believes other study designs could still help build clinical utility evidence.
Additionally, he and his colleagues' discovery of pooled groups of patients with certain types of CUP findings also speaks to the value of gene-expression-based cancer type testing.
For example, Greco highlighted work he and his team published on occult kidney cancers presenting as CUP, which they were able to substantiate using immunohistochemistry. That study included 24 cases with malignant cancer and no evidence of a primary tumor, with CancerType ID results indicating kidney cancer, but Greco said he has seen many more since then.
"If you treat them as renal cell carcinoma, for instance, with immunotherapy, they do remarkably well. That's what they have. It's just that you have to diagnose them properly in order to know how to treat them," Greco said.
He and his colleagues have also seen cases amassing of CancerType ID-predicted lung adenocarcinoma presenting without a primary site, again confirmed by immunohistochemistry.
"We've used this test in over 1,600 patients, so we have a wealth of experience. The fact that all CUP comes from specific cancers is becoming clear now, and most experts who see what's happening in the evolving data, they don't challenge it anymore," Greco said.
Several companies offering comprehensive genomic sequencing have shared in recent years that they are working with algorithmic or AI tools that can glean tumor type, or origin, from their genomic data. These include Caris Life Sciences, which has been offering an AI-driven tumor origin predictor, MI GPSai, since December 2019. The company has published results showing 93 percent success in determining an origin across cancers, but only about a 72 percent rate for CUP cases.
Memorial Sloan Kettering has also recently made similar forays using its MSK-IMPACT genomic profiling assay.
Greco cautioned that these other tools haven't been specifically validated in a CUP population as performance in a larger group of cancers doesn't necessarily translate to success in the CUP population.
"I'm an adviser to Biotheranostics, but I don't believe I am biased. If they can validate that their tests are accurate in unknown primary cancer, then I'm going to applaud them, but they haven't done that yet," Greco said. "It's easy if you see several mutations that are usually seen in lung cancer, and you say, 'Yeah that's probably lung cancer.' But [in CUP] there are many overlapping mutations and many cancer types."