Molecular Test Aids in Identifying Cancers of Unknown Origin

Prospective Trials Needed, Say Experts

Nick Mulcahy

July 22, 2011

July 22, 2011 — A molecular profiling assay developed to aid in tumor diagnosis is "highly accurate," according to a report of 45 cancer cases published in the July 15 issue of Clinica Chimica Acta.

The small retrospective study of the Tissue of Origin (TOO) test (Pathwork Diagnostics) confirms the utility of the test and comes from a laboratory at the University of California, San Francisco (UCSF), which was not involved with a previous larger controlled validation study, and, as a result, represents a different practice setting.

The identity of most tumors can be determined "with confidence" by clinicians with histologic, clinical, and radiographic findings, say the study authors, led by James P. Grenert, MD, from the Department of Pathology at UCSF. "However, when tumors are poorly differentiated or metastatic with no clear primary, identifying the tissue of origin is difficult," they write.

Immunohistochemical stains may or may not be helpful in identifying the tumor origin. Molecular testing of tumors is an additional means of characterizing these tumors, the authors say.

TOO is the only molecular diagnostic test for cancer tissue of origin cleared by the US Food and Drug Association, the manufacturer notes. It is limited to identifying 15 types of tumor tissues: bladder, breast, colorectal, gastric, testicular germ cell, kidney, hepatocellular, nonsmall-cell lung, non-Hodgkin's lymphoma, melanoma, ovarian, pancreas, prostate, thyroid, and sarcoma. These 15 tumor types represent 90% of all cancers, the company points out.

With 1 exception, the investigators in this study used tissue samples belonging to the 15 tissue types that TOO is able to identify.

The study samples were fixed in formalin, embedded in paraffin, and stored for 3 to 10 years in the UCSF archives. The diagnosis of each tissue sample was established before being processed but was blinded. Eight of the samples were excluded for technical reasons, leaving 37 for the study. Investigators accurately identified 35 of these samples with TOO (95%). In other words, a very high percentage of the evaluations agreed with the archived "reference diagnosis." The sensitivity was 95% and the overall specificity was 99.6%, report the authors.

These results are better than those from the larger validation study that examined 462 samples and showed that overall agreement with the reference diagnosis was 88.5% (J Mol Diagn. 2011;13:48-56).

Still Finding Its Niche?

However, the test has not been wholly endorsed by experts.

For example, the National Comprehensive Cancer Network (NCCN)'s panel on occult cancer does not recommend the use of any gene-expression profiling product, including TOO, as "part of routine evaluation."

"While gene-expression profiling looks promising, prospective clinical trials are necessary," reads the 2011 NCCN occult cancer guideline. Trials will confirm whether patient prognosis could improve with testing and refine the related treatment choices, the guideline explains.

However, the NCCN acknowledges TOO, and says that gene-expression profiling assays in general are "an emerging diagnostic tool to help identify tissue of origin."

At the 2010 annual meeting of the NCCN, an expert suggested that the high-tech assays will eventually make their way into guidelines.

Molecular-profile testing "will find its niche," said Charles Handorf, MD, PhD, from the University of Tennessee Cancer Institute in Chattanooga, who is a member of the NCCN panel on occult cancer.


The TOO test employs a computer algorithm, which is a 2000-gene model that quantifies the similarity between RNA expression patterns of a specimen and the 15 tissues on the test panel.

A TOO test report is generated for each case/specimen, and includes a similarity score that reflects the closeness of the RNA expression pattern of the specimen to the RNA expression pattern of the indicated tissue, Dr. Grenert and coauthors explain.

Similarity scores range from 0 (very low similarity) to 100 (very high similarity), and sum to 100 across all 15 tissues on the panel. The highest similarity score indicates the likely tissue of origin. For instance, a 58-year-old woman with biopsied tissue from her right lung had a similarity score of 86.6 for nonsmall-cell lung cancer, but scores of less than 5 for the other cancers, such as breast (3.3) and bladder (2.4).

Of the 45 cases that were screened, 37 met the entry criteria. One case was off-panel (squamous cell carcinoma of the larynx), 6 had inadequate tumor purity to meet the inclusion criteria, and RNA from 1 specimen was lost in transport.

Of the 37, 6 were high-grade/poorly differentiated tumors and 5 were metastatic tumors. Patients were 22 to 74 years of age (median, 55 years). There were roughly equal numbers of male and female patients (17 vs 20).

With inclusion of the off-panel specimen, 35 of 38 (92.1%) were in agreement. Of the 2 nonagreements, one was a seminoma (called ovarian) and the other was an ovarian mucinous cystadenocarcinoma (called colon).

The practical value of the test was described by the study authors. "Ruling out colon cancer in a patient with a history of colon cancer and a new lung tumor would help confirm a new primary tumor, rather than high-stage, metastatic disease. The TOO test can be very helpful in this type of role," they write.

The study was funded in part by a grant from Pathwork Diagnostics.

Clin Chim Acta. 2011;412:1462–1464. Abstract


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.