Taking Steps to Identify Carcinoma of Unknown Primary

At the 15th Annual Meeting of the National Comprehensive Cancer Network (NCCN), Richard Klausner, MD, gave a spirited presentation on “New Advances in Molecular Diagnostics and Cancer.” Oncologists may recognize Klausner as director of the National Cancer Institute (NCI) between 1995 and 2001. He also served as executive director for Global Health of the Bill and Melinda Gates Foundation. Today, he is a managing partner with the Column Group, which seeks to fund startup biotechnology companies researching promising therapeutic agents.

Following the first draft of the human genome in 1996, Klausner’s vision as NCI director was to shift the agency’s focus to molecular biology and genetics. That initial map of the human genome ushered in the era of personalized medicine, as investigators teased out a growing number of molecular markers that paved the way to new therapies in cancer. But molecular markers, said Klausner, are also an important tool for making a definitive diagnosis of tumors. “We’d all love to use this information for predictive outcomes and ultimately for determining choice of therapies, and we will be there,” he said, “but…getting diagnosis correct has been important in medicine for the last 100-odd years and it is still important in oncology.”

This perspective prompted Klausner to create the Director’s Challenge program at NCI in 1999, which provided grants for molecular diagnostics. “We were going to reclassify subsets of tumors,” he said. Progress has been made in this area, Klausner said, singling out OncoType DX and MammaPrint gene profile testing as positive steps. What Klausner really hoped for, however, was a way to use gene expression to solve what he called the “low hanging fruit”—tissue of origin. “As you know, this is not a minor problem,” said Klausner, adding that an analysis of the Medicare database identified 53,000 diagnoses of carcinoma of unknown primary (CUP).

Klausner said it is “right and ethical” to align treatment with recommendations like those established by NCCN, but if you are using techniques that offer a “best guess” at tissue origin, then you run the risk of giving the patient a useless—or even harmful—treatment. “The most frequent diagnoses that we get either in switching diagnoses or the unknown [are] colorectal, breast, ovarian, pancreatic, non—small lung cancer, or hepatocellular carcinoma,” Klausner said, “all of which have recommendations in therapy that are distinct from a default occult primary therapy.”

Determining Cell Lineage

No single gene defines a tissue type. “Cell lineage is defined by a generally inalterable pattern of gene expression,” Klausner said. Once you have validated the methods for measuring gene expression, the rest is science, he noted. Using gene expression to predict response to therapy requires years of clinical trials, but cell lineage is what it is, explained Klausner. Once you decipher the gene expression pattern, you have an objective indicator of tumor origin. He contrasted this with immunohistochemistry, a test he described as having “less and less value,” in part because it measures only one protein at a time and is open to subjective interpretation.

Pathwork Diagnostics is a molecular diagnostics company that Klausner singled out as making substantial progress in developing tests to help oncologists identify cell lineage of tumor tissue. Pathwork used a variety of measures to discover the gene pattern characteristic of each cancer, including modern informatics (cloud computing), whole genome microarrays, large training sets of diverse tumor samples, and direct measurement of gene expression (mRNA). The current test analyzes small samples (one microtome slice) from formalin-fixed tissues, core biopsies, or fine-needle aspirates against 15 tissue samples and “provides an objective measure of the probability that it is a tumor from 1 tissue and not from 14 other tissues,” said Klausner.

The test is more effective at facilitating rule-outs, with a negative agreement rate of 99.1%. Positive agreement is 88.5%, which Klausner said allows a positive match “in 9 out of 19 cases.” One of the best parts, he explained, is that the signal has no statistical degradation. Once you identify a tumor as colorectal cancer, there is no chance of finding out later that it is liver or lung cancer. “There is no degradation of the differentiation signature from highly, well-differentiated primaries to undifferentiated met[astases],” Klausner said.

Researchers at the Mayo Clinic used the test on several cases of CUP and received a definitive answer for 76%, most of which were found to be common tumors. Klausner said, “We don’t know how many of the 24% was the test not performing or something that was not in the 15 cancer patterns.” NCCN guidelines required a major change in therapy for 74% of patients, a minor change for 20% and no change for 6% of patients. In a University of Pittsburgh study of 284 sequential cases of malignancy, the ICD codes identified 64% as CUP. The Pathwork Diagnostics test confirmed 42%, found that 57% had been identified incorrectly, and could not make a rule-in on 1%. According to Klausner, in the incorrectly identified group, 43% of cases warranted a major change in therapy based on NCCN guidelines, 40% required a minor change, and only 3% needed no change.

“Clearly, there was uncertainty here...a lot more uncertainty than just what is called CUP or occult primary,” Klausner said. “It raises the question of ethics not to do tissue of origin testing,” he added.

Next Steps

Now that it is possible to definitively answer the tissue of origin question, Klausner said he believes “it’s time for a definitive and dramatic change.” He argued that it is not right to give patients identified as having CUP “very toxic therapies” when there are tools available to ascertain the correct diagnosis 80% to 90% of the time. “It needs to be looked at—and I think this is important—by those that are contemplating and describing practice guidelines,” said Klausner, who added that getting a diagnosis correct before treating also reduces costs.

In the future, Klausner believes the FDA will need to take on oversight of molecular diagnostics testing and set the standards for adoption, which he believes should require evidence that the test results affect therapeutic decisions. To critics who consider lineage “old fashioned” compared with the “new molecular stuff,” Klausner said lineage is molecular. “I strongly disagree with a statement that there is not a role for molecular analysis to define tissue of origin. As a biologist, I will argue that this is the only way to establish tissue of origin,” he said.