Study Offers Promise for Detecting Pancreatic Cancer
Using novel protein microarray technology, researchers have identified a panel of 10 protein biomarkers found in blood samples that proved highly accurate at detecting the existence of pancreatic cancer and identifying truly negative samples.
Several biomarker experts called the findings preliminary but promising. But there is still much work to be done before a commercially available test could reach the clinic, acknowledged the study's leader, Dr. Anna E. Lokshin of the University of Pittsburgh School of Medicine.
"We are still looking for more potential biomarkers in order to increase the sensitivity of this screening method to 100 percent," she said. "So far, these preliminary results are very encouraging, but we need to reach 100 percent accuracy before this test can be widely used."
Dr. Lokshin presented the study results on Monday in Boston at the American Association for Cancer Research's (AACR) Frontiers in Cancer Prevention Research meeting. The study was conducted in conjunction with researchers from Harvard and Northwestern Universities.
The research team began with a panel of 44 biomarkers, which, as Dr. Lokshin explained, "represent the dialogue between the tumor and the body" - that is, proteins that are secreted by tumor cells, the vasculature that feeds the tumor, immune system cells, and cells in the tumor microenvironment generated as a result of a tumor's presence.
They then used a protein array system to analyze blood samples from 100 patients with both resectable and nonresectable pancreatic cancer, including some patients with early-stage disease, and a control group of 400 healthy participants. They found marked differences in the expression of many of these proteins in the patient samples compared with those from controls.
Using "our own, quite powerful algorithm," Dr. Lokshin said, they discovered a 10-biomarker panel that correctly identified 97 percent of patient samples. In addition, the panel's sensitivity and specificity - its ability to correctly identify a malignancy and rule out cancer - was 95 percent and 98 percent, respectively.
"Since the completion of this study, we have been able to reproduce these results in a blinded validation set, with strong diagnostic power, and 87 percent sensitivity and 98 percent specificity," Dr. Lokshin said.
Dr. Sudhir Srivastava, head of NCI's Early Detection Research Network (EDRN), which funded the study, lauded the team's work, but cautioned against reading too much into the results.
"We must narrow down the panel so it becomes more organ specific," he explained. While some of the biomarkers used in the study are expressed in pancreatic cancer, he noted, they are also overly expressed in other cancers, so there is a significant risk of "cross reactivity."
The results presented in Boston did show that the 10-biomarker panel specifically recognized patients with pancreatic cancer, but not patients with other cancers, including lung, esophageal, head and neck, ovarian, breast, endometrial, and melanoma.
The diagnostic platform used to test the samples also will need to be standardized, Dr. Srivastava added, and the refined biomarker panel tested in different patient cohorts to ensure that the results are reproducible from one site to the next.
Perhaps most important, he said, "it has to be tested on prospectively collected samples, so we can determine what kind of benefit you get in terms of lead time." In other words, can the test detect the disease early enough in its course that treatment will actually improve survival? The statistical algorithms used to analyze the data must also be validated, he added.
"The validation steps will be crucial to determine whether these initial exciting results hold up," said Dr. Teri Brentnall, a researcher at the University of Washington Medical Center, who also is doing work in this area.
By Carmen Phillips, with additional reporting by Heather Maisey