Inroads in treating many cancers have largely bypassed pancreatic cancer, which remains one of the most lethal: in the United States about 46,000 people are diagnosed each year and about 40,000 die from it. The challenge in changing this discouraging track record has been that nearly all patients are diagnosed well into the disease process, now thought to progress for up to 2 decades. Despite rigorous research efforts, so far no validated biomarker with sufficient sensitivity or specificity has surfaced to make it a serious screening or diagnostic contender.

Carbohydrate antigen (CA) 19-9—the only biomarker cleared by the Food and Drug Administration to monitor pancreatic cancer progress is a problematic tool for screening or diagnosis because it also is elevated in other diseases such as diabetes and chronic pancreatitis, and up to 15% of people don’t produce this glycoprotein. Given all these factors, the pancreatic cancer scientific community took note of research presented earlier this year at an American Association for Cancer Research conference on pancreatic cancer.

Ayumu Taguchi, PhD, MD, and his colleagues at The University of Texas MD Anderson Cancer Center in Houston reported that a panel of four biomarkers proved better than CA 19-9 alone in distinguishing samples from subjects with pancreatic cancer from those of healthy individuals, those with chronic pancreatitis, and those who had pancreatic cysts, correctly identifying 92%, 85%, and 92%, respectively. In a logistic regression model, the area under the curve of the four biomarker panel—comprised of CA 19-9 and three other not publicly named markers—was 0.80, compared with 0.76 for CA 19-9 alone.

The four-analyte panel also correctly identified as negative for pancreatic cancer 94%, 90%, and 91% of samples from healthy individuals, those with chronic pancreatitis, and those with pancreatic cysts, respectively.

“In our training set, CA 19-9 showed high specificity; however it still failed to identify 40 percent of pancreatic cancer samples. We then developed a decision tree model which included CA 19-9. The performance in the final training set was significantly improved over CA 19-9 alone, identifying 22 out of 23 pancreatic cancer patients,” recalled Taguchi, an assistant professor of translational molecular pathology at MD Anderson. “Even so, I think our biomarker panel is still preliminary.”

Beyond Robust
Though heartened by Taguchi’s findings, other pancreatic cancer specialists agreed that this line of research needs to be fleshed out more. “Any sort of blood marker that is identified for pancreatic cancer really has to have very strong operating characteristics, with very high positive- and negative-predictive values. We’re talking percentages in the high 90s, even 100, to really make this both clinically effective and cost-effective,” said James Farrell, MD, director of the Yale University Center for Pancreatic Diseases in New Haven, Connecticut.

“The challenge really rotates around the fact that this is not a high-prevalence cancer. So if you start getting false-positives, you’re basically going to put a lot of patients through unnecessary anxiety and testing. On the other hand, you don’t want to miss any patients who actually have the disease,” he added. A population-based screening test with even an impeccably high specificity like 99% still could lead to an abundance of false-positives, with all the attendant follow-up testing to rule-out disease. The other issue is how early in the disease process any biomarker would accurately distinguish pancreatic cancer, when it might be surgically resectable. Taguchi emphasized that a strength of his findings is that samples were from patients with early-stage disease, whereas most studies looking at CA 19-9 have been from individuals in later stages.

Whom to Screen?
Some researchers place narrowing the potential screening population on equal footing with finding robust diagnostics. “Unless we can identify appropriate groups for screening, the predictive value of any test is not likely to be adequate for widespread population-based screening,” said Steven Leach, MD, director of the David M. Rubenstein Center for Pancreatic Cancer Research at Memorial Sloan-Kettering Cancer Center in New York City.

Others argue that merely detecting early stage disease among high-risk patients is not the only hurdle any clinically useful marker will need to clear. “Identifying patients who have favorable tumor biology would be fairly important so we would know who would benefit most from aggressive treatment,” explained Jashodeep Datta, MD, Harrison post-doctoral research fellow at the University of Pennsylvania Perelman School of Medicine in Philadelphia.

On the Horizon
With these stringently high standards against the backdrop of a relentless disease, is there reason to expect that a biomarker or biomarker panel will be useful anytime soon in detecting pancreatic cancer when it is treatable? Despite the seemingly long odds, researchers still have high hopes. “There has been a tremendous amount of work over the last 10 years that has inspired a lot of smart people to get into the field. It’s one of the most intensely researched cancers now. That’s where the optimism comes from,” says Farrell.

He was not alone in suggesting that it will be key to keep an open mind not only about candidate proteins but also markers reflecting other aspects of pancreatic cancer pathophysiology. “Protein-based markers potentially are easier to translate into the clinical setting. But we also know that there are about four to five canonical genetic mutations in the progression from normal pancreas to cancer, so exploiting that knowledge and the associated pathways, whether at the DNA or RNA levels, certainly would be helpful,” explained Anil Rustgi, MD, chief of gastroenterology at the Perelman School of Medicine. He also cited as potentially useful any biomarkers that could capture the robust fibrotic response that takes place once pancreatic tumors begin spreading into surrounding tissue, as well as those that might catch tumor cells as they disseminate rapidly throughout the body.

Leach added that while proteins might have an edge right now, “this may be more a function of cost and technology rather than biology, and cost and technology can obviously change rapidly.” Indeed, Taguchi’s team already is collaborating with other researchers at MD Anderson, exploring the diagnostic power of microRNAs and autoantibodies.

Any sort of blood marker that is identified for pancreatic cancer really has to have very strong operating characteristics, with very high positive and negative predictive values.