Skip to main content
Premium Trial:

Request an Annual Quote

Evidence for Targeted Treatment Builds in Cancers of Unknown Primary but Randomized Data Pending


NEW YORK – New data presented over the weekend at the American Society of Clinical Oncology's virtual annual meeting, has added to growing evidence that genomic sequencing, in particular liquid biopsy tests, can identify actionable alterations in patients with cancers of unknown primary (CUP).

Adding to this, retrospective evidence presented at the meeting showed that at least some of these tumors may respond to targeted treatments, while other studies illustrated the challenges of implementing precision medicine, including obtaining the drugs indicated by genomic profiling tests.

In particular, experts debated the relative utility of tissue of origin classifiers and tumor profiling using next-generation sequencing in facilitating precision treatment strategies in this difficult-to-treat and heterogeneous patient population.

The emergence of molecular tumor origin or cancer-type classifier assays for CUP in the last decade raised hopes that these tests would improve patient outcomes by enabling treatment with more effective site-specific regimens.

However, recent randomized trials of such assays have yielded somewhat disappointing results, including the Phase III GEFCAPI study presented late last year, which concluded that patients treated on the basis of molecular tumor origin or primary site testing did no better than those treated without this added information.

Some have argued that instead of treating patients by trying to predict where in the body their cancer started, they should focus on targeted treatments that shut down the genomic drivers of patients' disease. Randomized trials have been designed to test this hypothesis. Most notably, the Roche-sponsored CUPISCO trial is randomizing CUP patients to either standard treatment or treatment guided by Foundation Medicine's FoundationOne CDx tumor profiling test.

As the field waits this prospective data, new presentations at ASCO included an updated report on the detection of putatively actionable mutations using blood-based sequencing tests, case reports that speak to the utility of specific biomarkers, and discussions of the challenges the field may face with the implementation of genomic strategies.

In one of the ASCO presentations on Saturday, researchers from Central South University in Changsha, China reported on a small retrospective study of CUP patients with ALK alterations who were treated with ALK inhibitors. The investigators queried their institution's cancer database for cases of ALK-positive mediastinal CUP patients treated with ALK inhibitors in whom both NGS and PD-L1 immunohistochemistry testing had been performed. Overall, 48 patients with ALK-positive cancers were registered in the database.

Among these, researchers identified two mediastinal CUP patients who responded to ALK inhibition, though one more strongly than the other. The first patient harbored an ALK/TP53 co-mutation, had wild type TERT, and low PD-L1 expression. This individual showed a "remarkable response" to alectinib, according to the study authors. The second case, who had stable disease when treated with crizotinib, showed mutations in ALK, TP53, and TERT, as well as high PD-L1 expression.

Karim Fizazi, head of the department of cancer medicine at the Institut Gustave Roussy in France, said during the session that based on the two cases reported — as well as on previous case reports which together showed about a 50 percent response rate to ALK inhibitors in ALK-positive CUP — testing for ALK rearrangements in patients with CUP and mediastinal disease is likely to be informative.

If a patient is ALK positive, Fizazi added, "an ALK inhibitor is a very reasonable option assuming this is feasible according to local regulations." That said, he cautioned that such cases are very rare, and that there shouldn’t be any larger takeaway about molecularly driven treatment for CUP as a whole.

Access issues

Even if the field is able to confirm the actionability of ALK and other genomic targets via the CUPISCO trial and other efforts, challenges will remain, meeting participants stressed. One of these is actually getting patients access to genotype-indicated drugs.

Although it didn't deal specifically with CUP patients, another study addressed this question through an analogous population: patients with rare cancers. Investigators from a group of Australian hospitals looked at how NGS impacted the care of patients with rare cancers with poor prognosis that lack standard treatment options.

Between July 2017 and Nov 2019, the group enrolled 121 patients across four sites, of which 83 percent had their tissue successfully sequenced. In about half the group, researchers identified a potentially actionable alteration, and they matched 27 patients to a clinically validated drug.

However, getting access to these matched drugs was a significant issue, with only 12 patients receiving therapy based on NGS results and the recommendations of a molecular tumor board.

Utility of liquid biopsies

Another pressing question for the field is whether liquid biopsy assays can return results for CUP patients that are as actionable as tissue tests. The ability to test patients noninvasively would be an additional boon in a population where tissue biopsies are often infeasible due to advanced stage of disease and poor health status.

Investigators from the University of California, San Diego first reported on their research using Guardant Health's liquid biopsy assay in this context in 2017. In a publication in Cancer Research, the group wrote that among 442 patients with CUP tested using Guardant360, more than half had at least one targetable alteration — meaning a variant associated with either an US Food and Drug Administration-approved drug or with an investigational drug in a clinical trial.

At ASCO, researchers updated the findings of this study, having identified a total of 2,022 CUP patients who underwent testing with the Guardant360 test between November 2016 and November 2019.

The team classified alterations detected by the assay based on their ranking in the OncoKB database, which categorizes the therapeutic actionability of genomic biomarkers identified in patients' tumors based on the underlying evidence. OncoKB level 1 alterations, for example, are FDA-recognized biomarkers associated with response to an FDA-approved drug. Level 2 is made up of biomarkers recommended by major guidelines or expert panels as predictive of response to an FDA-approved drug. Level R1 markers are those recognized as being predictive of resistance to an FDA-approved drug.

The researchers reported that more than 90 percent of the CUP patients had an alteration detected in their circulating cell-free DNA, and nearly half had alterations in one of the top three OncoKB categories: 20.7 percent had a level 1 alteration, 9.5 percent a level 2 alteration, and 23.9 percent an R1 alteration.

"While many of these alterations are associated with approvals in specific cancer types, the identification of these alterations suggests many CUP patients may have targeted therapy options," the authors wrote.

Presenting the data in a virtual address, UC San Diego oncologist Shumei Kato also highlighted a clinical case study, of a 62-year-old man with poorly differentiated CUP in whom the Guardant assay detected two alterations — one in KRAS the other in ARID1A — suggesting potential sensitivity to trametinib (Novartis' Mekinist) and olaparib (AstraZeneca's Lynparza), respectively.

Using the matching protocols advanced in the previously published I-PREDICT study, the patient was treated with a combination of those two drugs and experienced tumor shrinkage. As of this month, that patient has been on treatment for 13 months, Kato said.

Uncertainties with immunotherapy

Although most of the discussion at the ASCO session centered around the promise of molecularly guided treatment, not all efforts in CUP reflect a push toward individualization. In a fourth presentation, researchers also described efforts that they hope might be useful across the CUP population, using the immune checkpoint inhibitor nivolumab (Bristol Myers Squibb's Opdivo).

In a multicenter Phase II study, a collaborative group from several Japanese institutions recruited 56 CUP patients, most of whom were previously treated with more than one line of systemic therapy. Anyone who was determined to be in a prognostically favorable group based on IHC testing, clinical features, or other criteria was excluded.

Of the 45 previously treated patients, 11 had either a complete response or partial response, with a median progression-free survival of 5.4 month and overall survival of 15.1 months across the cohort.

Among 11 previously untreated patients, only one had a partial response. There were no treatment-related deaths, but immune-related adverse effects occurred in more than half of patients.

The authors concluded that the treatment showed a clinical benefit and suggested that the data support a potential for nivolumab to become a new standard of care therapy for CUP, but Kizazi disagreed, saying that he had concerns over the study's short progression-free survival curve and the failure of non-progressors to reach a plateau in the Kaplan-Meier analysis.

Tennessee Oncology physician Anthony Greco argued that the nivolumab data illustrate even further that CUP is not one cancer type but represents a range of cancers, some of which would be expected to respond favorably to immune checkpoint inhibition and others not.

Investigators did analyze patients' PD-L1 status in the trial and reported that higher PD-L1 expression was associated with a greater likelihood of response. However, Kizazi said, this trend was not very strong, reflecting perhaps the biomarker and personalization challenges the immunotherapy field faces across cancers, not just in CUP.

Prioritizing testing

Greco and Kizazi grappled over the question of how oncologists should prioritize NGS testing alongside the available cancer-type and tissue-of-origin classifiers, for which recent randomized trial results have been so disappointing.

Kizazi, who was the principal investigator of the GEFCAPI trial, reiterated that the weight of evidence seems to have come down hard on tissue origin classifiers used on their own. Such tests, and the types of therapies they inform, "have not made a clinically meaningful difference," he said.

Although the definitive, prospective data isn’t in yet on the value of NGS, Kizazi said he would probably look to NGS first for CUP patients if tissue was limited and he had to choose one type of test over the other.

But Greco argued that if genomic medicine is going to be effective for CUP, it will have to be coupled with tumor origin prediction. While some pan-cancer biomarkers like microsatellite instability status have emerged in recent years, he pointed out that the genomic targets in widespread use in oncology are largely histology specific.

Because CUP is not itself a cancer type but rather a shared manifestation of what can be virtually any type of cancer, genomic treatments stand to benefit CUP patients just as much, or as little, as they do the oncology population as a whole, Greco said.

As the field pushes for integration of genomic sequencing into CUP care, he added, oncologists should be wary that using genomic information to guide initial care without information about a tumor's origin could lead to inappropriate first-line therapy choices.

"It's illogical to treat most CUP patients … initially with NGS data alone rather than the standard effective [and] accepted therapy for their cancer type," he said, using as an example a CUP breast cancer with an EGFR mutation being treated with the EGFR inhibitor erlotinib, which has dramatic efficacy in lung cancer but has not shown itself effective in breast cancers.

Encouragingly, as molecular testing technologies continue to advance, oncologists may not have to choose between origin prediction and genotyping when treating CUP patients.

For example, as firms like Grail have worked to validate blood-based tests for early cancer detection, they have created new methods to determine a cancer's tissue of origin based on the circulating DNA signals these new assays rely upon. Potentially, such methods might be incorporated into liquid biopsy genomic profiling in the future, so that a single assay both identifies a tumor origin or cancer type and reads out any targetable molecular drivers.

Tissue-based technologies are also evolving to allow both cancer typing and genotyping. Caris Life Sciences, for example has developed what it calls the Caris Molecular Disease Classifier, an algorithmic approach that determines the source of unknown cancers using data from the company's 592-gene NGS panel.

In data presented at last year's ASCO meeting, the firm reported that it was able to make classifications of tumor samples with 85 percent to 95 percent accuracy, and for cases of CUP, the classifier generated a site prediction in the vast majority.