Finding Ovarian Cancer
A Multiplex Test Offers Improved Sensitivity
By Deborah Levenson
Among American women, ovarian cancer is the fifth leading cause of cancer-related death. Lack of an effective screening marker means that most diagnoses occur at the disease’s later stages, when it is more difficult to treat. Monitoring response to treatment and recurrence often depends on CA-125, which lacks sensitivity. But a team of researchers has developed a diagnostic panel that adds other markers to CA-125, one of several such multiplex tests now being developed for detecting early disease. A recent paper about the panel reports sensitivity and specificity that far exceed that of CA-125. This issue of Strategies examines its findings.
Despite being only 10% as common as breast cancer, ovarian cancer is much more lethal, striking about 20,000 American women and killing about 15,000 annually. Because it presents with few specific symptoms, only 20% of women with the disease are diagnosed at stages I and II, when the disease is confined to the ovary and the 10-year survival rate is nearly 90%. At more advanced stages, that figure drops to 20%. The vast majority of these patients with advanced disease will initially respond to chemotherapy, but only about 10% will remain in permanent remission. The ovarian cancer marker currently used to monitor response to therapy and recurrence, CA-125, has not been approved for screening or diagnosis because it is elevated in some normal conditions, like pregnancy, and benign ones, like endometriosis. However, some clinicians include CA-125 within a screening strategy for high-risk women that also includes annual pelvic examinations and transvaginal ultrasound. But CA-125’s sensitivity is less than 60% in the early stages of the disease, and adding ultrasound boosts the positive predictive value of such screening to just 20%.
Adding other biomarkers to CA-125 in a multiplex test may yield levels of sensitivity and specificity that make such a panel a viable diagnostic tool or even a screening test. A team of researchers investigating this approach added five additional proteins—leptin, prolactin, osteopontin (OPN), insulin-like growth factor II (IGF-II), and macrophage inhibitory factor (MIF)—to CA-125. In a recent paper, the team reported that its test detected ovarian cancer with 95.3% sensitivity and 99.4% specificity, perhaps laying the groundwork for a new multimarker test (Clinical Cancer Research 2008; 14: 1065–1072).
Separating Training and Test Sets
The research team studied samples from 362 healthy controls and 156 newly diagnosed women at various stages of ovarian cancer. First, they evaluated all six markers in a training set consisting of 181 samples from the control group and 113 samples from the ovarian cancer group, with the goal of determining if they could replicate results of a previous study of the markers leptin, prolactin, OPN, and IGF-II that used ELISA assays (PN A S 2005; 102: 7677-7682). To increase the assay’s sensitivity and specificity, they added MIF—another protein that is overexpressed in sera of ovarian cancer patients—and CA-125.
All of these proteins, except OPN, CA-125 and MIF, are related to the normal physiology of the ovary. Produced by either the surrounding supportive cells or as a response to signals originating from the ovary, a delicate balance between each of the cellular components of the ovary maintains the proteins’ expression levels. The researchers hypothesize that the presence of abnormal cells may alter intercellular communication and disrupt the balance, resulting in the abnormal expression levels. “Levels of proteins produced by the sroma and surrounding tissue change in response to the presence of abnormal cells, and those proteins allow us to identify ovarian cancer in its early stages,” explained lead researcher Gil Mor, MD, PhD, Associate Professor of Obstetrics and Gynecology at Yale School of Medicine.
After determining that the markers effectively detected cancer in the training set, Mor’s team then blindly evaluated the panel in a test set consisting of different samples from both the control and ovarian cancer groups (181 and 43, respectively) to make sure the assay could differentiate normal from cancerous samples in a group separate from the one used to develop it. AUCs in the test set decreased for each of the biomarkers, which the researchers maintain shows the importance of using different cohorts for the training and test versions of the assay. But each marker’s AUC remained above 0.5, meaning that each biomarker still indicated ovarian cancer, the researchers note.
Mor emphasized the importance of using distinct training and test cohorts. “Many studies just use one cohort to determine what’s normal and cancerous, and then use those same samples for validation. Then you’ll get perfect validation. We used different patients in each test set and tested in a blind manner to see if the test truly can recognize what’s cancer and what’s not using the model built with the first cohort.” He added that the data described in the study was generated by independent statisticians selected by the Early Detection Research Network, which was not involved in the rest of the trial.
The test, which has been licensed to both LabCorp and Teva-Oncotest in Tel Aviv, Israel, is now undergoing further validation in 2,000 samples from the Prostate, Lung, Cervical and Ovarian Cancers Trial run by NCI. At Yale, researchers are also running a longitudinal study of the test in 300 high-risk women. When both of these studies are complete, Mor plans to use the data to seek FDA approval.
Another ovarian cancer researcher cautioned against exuberant optimism about either Mor’s test or any other multiplex ovarian cancer assay. While praising both the use of separate training and test sets and the reported sensitivity and specificity of the assay, Daniel W. Chan, PhD, Professor of Pathology, Oncology, Radiology, and Urology at The Johns Hopkins University, said the data must hold up in further validation. He is Director, Clinical Chemistry and Co-Director of the Pathology Core Lab at Johns Hopkins University Hospital in Baltimore, MD, and is working on another ovarian cancer panel for early detection that he plans to submit to the FDA soon. Detailed in a 2004 paper (Cancer Research 2004; 64: 5882–5890), his panel was the first designed for early detection and has been licensed to Vermillion in Fremont, Calif. “Lab people are skeptical when it comes to research about multiplex tests. If it seems too good to be true, it usually is. Labs are very careful not to jump into a new test unless they see that is works as well as what’s claimed.”
Noting that Mor’s study based its calculations of specificity and sensitivity on samples from women at all four stages of ovarian cancer, Chan commented, “It’s questionable if those figures are applicable to screening situations.” He added that the test’s use of six markers could make it cumbersome to set up. “But if the test were available, labs could do it,” he said, noting the test must hold up in validation in multiple populations that harbor various types of ovarian cancer.
In the not so distant future, labs should expect to be using multiplex tests to diagnose ovarian and other cancers, Mor and Chan agreed. “There’s no magic bullet here. There’s no one protein that’s specific to an organ. We need to put them together to create a profile,” Mor explained. Added Chan, “Labs need to recognize that that multiplex tests represent the future of cancer diagnosis. So everyone needs to be thinking about how to gear up lab operations to accommodate them.”
Was this issue of Strategies of interest to you? Please help us improve by rating this article.