November 2008 Clinical Laboratory News: Tissue of Origin Test for Metastatic Cancer

CLN Banner Logo

November 2008: Volume 34, Number 11

Tissue of Origin Test for Metastatic Cancer
Is This the Beginning of Genomic-based Cancer Analysis?
By Gina Rollins 

Metastatic occult primary tumors or cancers of unknown primary (CUP) pose real diagnostic challenges, especially when they are poorly differentiated or undifferentiated. Despite advances in laboratory and imaging studies, identification of a patient’s primary tumor site remains elusive in more than 70% of cases. Yet pinpointing the tissue of origin can have important implications for therapeutic options, and even patient survival. Now FDA’s recent clearance of a first of its kind test may improve treatment options. The new Pathwork Diagnostics Tissue of Origin (TOO) Test, which measures gene expression to determine the original tumor site, is part of the first wave of what is anticipated to be a surge in genomic-based cancer analytics.

“There has been interest for many years in the genetics of cancer, and we’re building capacity and capability so that at some point in the future we will be able to incorporate genetic analysis of cancers into the diagnosis and treatment of the disease,” said Len Lichtenfeld, MD, deputy chief medical officer of the American Cancer Society.

The TOO test combines microarray technology with proprietary normalization and classification algorithms to discern gene expression in 15 of the most common metastatic tumor types, including bladder, breast, colorectal, gastric, germ cell, hepatocellular, kidney, non-Hodgkins lymphoma, non-small cell lung, melanoma, ovarian, pancreatic, prostate, soft tissue sarcoma, and thyroid. Pathwork began offering the test through its CLIA-certified lab in April 2008, and in July FDA cleared the test for marketing, giving labs the option of performing the test in house.


FDA cleared the TOO as an in vitro diagnostic multivariate index assay (IVDMIA), the second one so designated after Agendia’s MammaPrint test, which was given the green light in February 2007 (CLN Nov. 2007). In a statement, Daniel Schultz, MD, director of the FDA Center for Devices and Radiological Health, acknowledged the clinical significance of a second IVDMIA coming on the market. “The clearance of the Pathwork test is another step in the continued integration of molecular-based medicine into standard practice. In the past, scientists have classified different types of cancers based on the organs in which the tumors develop. With the help of microarray technology, they will be able to classify these types of cancers in a standardized non-reader dependent manner based on the patterns of gene activity in the tumor cells.”

Performed on Affymetrix’s GeneChip platform, the new test measures the expression of 1,668 gene probe sets or markers on a microarray to determine the similarity of tumors of uncertain origin to patterns of expression representing 60 morphologies overall in the 15 reference cancers. Pathwork created the test’s algorithm from 2,039 well-characterized tumor specimens from 14 laboratories. The algorithm processes the myriad of expression data from the microarray into 15 individual similarity scores on a 0 to 100 scale, one score for each of the 15 reference tumor types. A score of 0 means there is little similarity between the specimen and the reference tumor type; 100 suggests a high similarity. A score of 30 or higher indicates that there is a greater than 95% probability that the reference tissue in question is present in the specimen. A score less than 5 suggests that there is a 95% probability that a tumor type is not present in the sample. The scores are presented to pathologists and oncologists in a two-page report—including a graph of the similarity scores—so the clinician can make a clinical determination of whether or not any specific type of tumor is in the specimen (See Figure, below).

Tissue of Origin Test

Pathwork Diagnostics’ Tissue of Origin Test uses a microarray to classify different types of cancers based on the organs in which the tumors develop. The report uses an algorithm to assign a similarity score as a measure of the RNA expression pattern of the specimen to the RNA expression pattern of the indicated tissue. The scores range from 0 (very low similarity) to 100 (very high similarity) and the sum of the scores for the 15 tissues equals 100. In this example, the sample had a very high similarity score with colorectal cancer.

IVD Large Chart

“The score is similar to sensitivity and specificity, and it works both ways. It helps identify where a tumor is coming from based on how well it fits the [gene expression] pattern of 1 to 15 sites. And if it fails to identify a match, it will indicate where the tissue is not from,” explained Federico Monzon, MD, medical director of molecular diagnostics at The Methodist Hospital in Houston. Monson was principal investigator for the TOO clinical validation study that tested its analytical performance.

Validation Results

The TOO clinical validation study submitted to the FDA used a set of 545 metastatic, poorly differentiated or undifferentiated tumor specimens that other methods had previously confirmed to be one of the 15 types the test screens for. Pathwork reported an overall 89% positive agreement with the already confirmed samples and a 99% negative agreement. In preliminary data from the validation study conducted by Monzon involving 477 specimens, the percent positive agreement ranged from 75% in the case of germ cell tumors to 100% for thyroid tumors, while the percent negative agreement ranged from 98.3% for breast cancer to 100% for several other tumor types, including bladder, germ cell, and prostate. The percent positive agreement was higher than 90% for six tumors.

Similarity scores reflecting the conditional probability for a positive result ranged from 82.4 for gastric tumors to 100 for bladder, germ cell, hepatocellular, non-Hodgkins lymphoma, ovarian and prostate tumors; similarity scores indicating the conditional probability for a negative result ranged from 91.1 for non-Hodgkins lymphoma to 100 for pancreatic tumors.

Researchers also tested the validity of the TOO test with 10 CUP specimens in which the pathologist had identified possible sites of origin but had been unable to make a definitive determination. In six of the 10, the TOO test produced a high similarity score for a site already suspected by the pathologist; in two it suggested sites not already suspected but which were plausible based on clinical history and the location of the metastatic tumor. For two additional specimens, the TOO test produced indeterminate results, but still would have assisted with diagnosis because in both instances it produced significant negative similarity scores for nine tumor types, making it possible to confidently exclude those as the primary site, according to Monzon. Results for both the final validation data submitted to FDA and the published study were similar, according to Shawn Becker, MD, vice president of marketing and reimbursement at Pathwork.

A separate study tested the analytic performance of the TOO test and looked at a key issue in adopting microarray technology as a routine diagnostic practice—interlaboratory variability (J Mol Diagn 2008;10:67–77). Four labs with different personnel, reagents, instrumentation, and protocols tested 60 archived tumor specimens from poorly differentiated and undifferentiated tumors. Researchers found an 89.4% overall concordance between the labs in terms of the final report to physicians. When the analysis included only specimens that met Pathwork parameters for tumor content—more than 60% tumor and less than 20% necrosis—the average percent agreement between the test locations rose to 93.8%. At the same time, 18 specimens with RNA degradation failed Pathwork’s quality control parameters, which “strongly accentuates the need for an RNA extraction quality control program for successful implementation of this test in the clinical setting,” researchers noted.

Since both RNA degradation and adequate specimen quality are factors in the performance of the TOO test, “integrity of the tissue after it’s out of the patient is critical,” said Monzon. Tissue should be frozen as rapidly as possible, and Monzon also recommends performing a morphological evaluation of the tissue either before or after it is frozen but prior to analysis to verify that there is sufficient tumor content.

Closing the Diagnosis Gap

Improvements in immunohistochemistry (IHC) analysis have narrowed the diagnosis gap for CUP, and now with the availability of the TOO test that gap may shrink. National Comprehensive Cancer Network practice guidelines now call for IHC in conjunction with imaging studies to select the best possible treatment option for patients with CUP. However, even though the panel of available IHC markers is expanding, it covers a more modest range of malignancies than the TOO test, and, at least for certain tumors, is not particularly sensitive. “There are still gaps in [IHC] being able to identify the tumor of origin, but it’s much less than before the advent of IHC. The [TOO] test has a much higher resolution of discrimination than IHC. We’re talking 15 to 20 IHC stains versus more than 1,500 genes,” noted Thomas M. Wheeler, MD, Harlan J. Spjut professor and chair of pathology at Baylor College of Medicine in Houston. Wheeler also serves as chair of the council of scientific affairs for the College of American Pathologists.

As a newly introduced test, Pathwork’s test faces some hurdles. The American Cancer Society estimates that in 2008 there will be an estimated 31,490 new cases of cancer with an unspecified primary site. However, Pathwork believes the TOO test has a wider potential use, for situations where the primary site is uncertain after the initial diagnostic workup. Feedback from its market research with pathologists and oncologists suggested that for every CUP case there are two-to-three of uncertain origin, according to Pathwork’s Becker. An estimated 2% to 10% of all newly diagnosed cancers—projected to be 1.4 million in 2008—are identified as CUP. By including tumors for which the site of origin is uncertain after initial testing, Pathwork estimates that the potential market for TOO test might reach 200,000 cases annually.

Even if that number ultimately proves to be valid, one of the drawbacks of the TOO test is that at the present time it can only be performed on frozen tissue biopsy specimens, as opposed to other methodologies such as formalin-fixed, paraffin-embedded (FFPE) samples and needle biopsy specimens. “It will do better when it can be used with tissues in formalin block. That will be a blockbuster, but for now it’s useful but probably won’t be widely embraced,” predicted Lawrence Weiss, MD, chair of pathology at City of Hope in Duarte, Calif. Right now, it is not standard practice to reserve a portion of biopsied tissue for frozen sections. Wheeler estimates that perhaps half of tumor biopsies include a frozen section. However, that could change if the utility of the TOO test becomes widely recognized. “If the technology is accepted in the medical community, frozen sections could be done automatically, probably up to 90% of tumor specimens,” he indicated. Pathwork has just such a scenario in mind, and is performing after-market research to document its efficacy further. “As with any new diagnostic test, it’s important to demonstrate its clinical utility and how it fits in with existing clinical workup practices,” said Becker. “A key question is whether it makes a difference in patient outcomes. This is a step in the journey, but it needs more validation in the real world of diagnosis and treatment,” Lichtenfeld noted.

The company also is working on versions of the test that can be performed with FFPE and needle biopsy specimens. It hopes to be able to offer the FFPE version before the end of the year, according to Becker. “We started with a test that worked with frozen sections because there was experience with the microarray in working with frozen sections. When we began, clinical diagnostic use of microarrays that would analyze paraffin-based specimens hadn’t been developed. The frozen section version gives us a beachhead while we develop a test that works with paraffin-based specimens,” he explained.

Therapeutic Insights

Determining the tissue of origin has taken on new importance in light of the development of cancer-specific treatments. Targeted chemotherapies can be effective even when there is metastasis, and have become more nuanced over time. For instance, 5-fluorouracil had been the standard treatment for virtually all gastrointestinal adenocarcinoma tumors, including colon, pancreatic, or gastroesophageal. But now these cancers are more likely to be treated with different chemotherapeutic regimens. For example, 5-FU, epirubicin, and cisplatin might be used for gastroesophageal tumors, while gemcitabine might be employed against pancreatic cancer. Aside from being more effective, targeted chemotherapeutic regimens typically are better tolerated than broad-spectrum agents.

Confirming the site of origin also can make a difference in the range of available treatment options, according to Lichtenfeld. “If the physician thinks the primary site is the pancreas, the choice of therapy is quite limited. But if a test shows the primary site to be lung cancer instead, now he can be more aggressive and perhaps a bit more optimistic.”

Pathwork plans to continue offering the TOO test in its own lab. The fee is $3,750, and it takes 4 to 5 days, according to Deborah Neff, president and CEO. Pathwork’s test kit, which will enable other clinical labs to perform the TOO test, will be available soon at a cost of $1,875; however, labs must transmit the analytic data file to Pathwork for analysis and report generation. Pathwork plans to offer training and technical support to customers as needed. At present, the test is not covered by third-party payers, but Becker is working actively to obtain reimbursement for the TOO test.

While it will take time to judge the diagnostic impact of the new test, what is clear is that the TOO test is an important milestone in cancer diagnostics. As Monzon puts it, “it signals a new wave of diagnostics based on multi-analyte testing, which incorporate microarrays and bioinformatics.”

Gina Rollins is a freelance writer in Silver Spring, Md.  

Page Access: