Profiling May Help in Cancer of Unknown Origin

Profiling May Help in Cancer of Unknown Origin


Molecular profiling identified treatable targets in 80% of 1,400 patients with cancer of unknown primary (CUP) site, leading to improved treatment and evidence of clinical benefit, according to a study reported at ECCO 2013 in Amsterdam.

Biomarker profiling often pointed investigators and clinicians toward guideline-recommended therapies. In a substantial number of cases, however, biomarkers suggested drugs not traditionally considered for CUP, Zoran Gatalica, MD, DSc, of Caris Life Sciences in Phoenix, reported at the European Cancer Congress.

"Only about 18% of cancers, after extensive profiling, still remained without any identifiable target," said Gatalica. "What was more important is that we were able to find both conventional chemotherapy targets as well as the biological targets, providing an option for drugs that may not be as expensive as what we think."

"Secondly, these targeted therapeutics usually were not those that you would consider in CUP. They are really associated mechanistically with other solid tumors. Because this is an unknown primary, we believe that's a solid basis to recommend those drugs."

CUP accounts for 3% to 5% of all solid tumors. Considered in isolation, CUP does not rank high on a list of major causes of cancer-related morbidity and mortality. However, when considered within the worldwide prevalence of solid tumors, the number of patients affected by CUP ranges into tens or even hundreds of thousands, Gatalica noted.

By definition a diagnostic challenge, CUP poses even greater problems with respect to therapeutic decision making. As a result, CUP offers a prime target for application of recent gains in knowledge related to molecular diagnostics and treatment.

The emergence of targeted therapies and companion diagnostic tests has fueled a transformation in the approach to treating solid tumors, Gatalica continued. Earlier this year, investigators in another study of CUP concluded that "tissue of origin and histological subtype are insufficient" to guide therapy. Examination for the presence of so-called druggable genetic alterations is "mandated," the authors stated.

Following the line of thought from investigators in that earlier study, Gatalica and colleagues searched the Caris database comprising profiling results for more than 50,000 cancer patients from throughout the world. The query yielded more than 1,400 cases of CUP, about 3% of the total number of cases in the database, consistent with previous estimates for the general population.

Females accounted for a majority of the CUP cases (55%). Mean age was 61 and few patients were younger than 20, even so, patient age ranged from 1 to 92.

The most common biopsy site was the liver (24% of cases), followed by lymph nodes (16%), and skin (14%). The investigators could not determine whether a patient had more than one involved site. CUP histology was adenocarcinoma in 45.3% of cases and other types of carcinoma in 37.5%.

Gatalica and his team employed a multiplatform approach to molecular profiling to maximize the odds of identifying a marker for which one or more therapeutic agents exist. The profile consists of three types of mutational analysis, assessment of gene copy number, MGMT methylation analysis, and protein expression.

Biomarker prevalence varied by CUP histology. For example, epidermal growth factor receptor expression was identified in 66% of carcinomas versus 16% of neuroendocrine tumors, androgen receptor in 7% of carcinomas versus 3.3% of undifferentiated and/or neuroendocrine tumor types, and KRAS mutations in 19% of carcinomas versus 2% of neuroendocrine tumors.

"When we identified a predictive biomarker, we performed a literature search to identify clinical evidence-based drug associations," said Gatalica.

As an example, a 35-year-old women with multiple bone metastases of adenocarcinoma origin. Molecular profiling revealed EGFR activating mutations, leading to treatment with erlotinib (Tarceva). The patient had a "remarkable" near-complete response.

Another case involved a 64-year-old women with a thoracic spinal extramedullary tumor, which a referring pathologist characterized as metastatic melanoma. Molecular profiling revealed a GNA11 mutation, which excluded melanoma. The interpretation was primary leptomeningeal melanoma or melanocytoma. Follow-up assessment led to identification of a primary leptomeningeal melanoma.

The most commonly identified mutations involved TP53 (32.8%), followed by KRAS (17.5%), PIK3CA (7.4%), APC (3.6%), BRAF (2.9%), and EGFR (1.1%). "Surprisingly" low and absent rates were seen for PDGFRA (0%), ERBB2 (0%), KIT (0.3), and PTEN (1.1%).

In 2% of cases, biomarkers predicted only lack of benefit, and no drug association could be found in 18% of cases. For the remaining 80% of tumors, associations were found with a wide range of therapies, including tyrosine kinase inhibitors, monoclonal antibodies, mTOR inhibitors, conventional alkylating agents, anthracyclines, pyrimidine antagonists, antimetabolites, and taxanes.

Conventional drugs were indicated in 41% of cases, conventional and targeted agents in 38%, and targeted agents only in 1%.

The study produced a "treasure trove" of data that nonetheless raised many questions, Lajos Pusztai, MD, DPhil, of Yale University, said during a discussion of the results. Unanswered questions included the distribution of anomalies, the definition of "targetable" biomarkers, differences in target distribution between CUP and non-CUP cancers, and frequency of co-occurrence of molecular aberrations.

Proving the value of novel diagnostics for CUP will be challenging, Pusztai continued. A randomized trial comparing conventional versus assay-driven therapy is unlikely to be feasible. The heterogeneous nature of CUP would likely rule out retrospective analyses.

A national registry might be useful, provided that it included information about treatment and patient outcomes, Pusztai said in conclusion.


Source: MedPage Today:

Login to your account

Did you forget your password?