Accurately Predicting Pancreatic Cancer Risk

An estimated 10 percent of aggressive and highly fatal cases of the disease are caused by inherited genes. "Even if there is a 100 percent chance that an individual carries a pancreas cancer gene, their risk for developing the disease is only 20 to 25 percent over their lifetime," says Alison Klein, Ph.D., assistant professor and director of the National Familial Pancreas Tumor Registry at Johns Hopkins. "So, while it's a rare disease, the need for screening in these persons is important."

The risk "calculator," based on similar tools for breast and colon cancer, calculates a percentage score of probability that a person carries a pancreas cancer gene. Called PancPRO, it also computes an individual's lifetime risk of developing the disease.

Although researchers have not yet identified specific genes that cause the disease, they can estimate high risk based on clusters of family members with a history of pancreas cancer. "We know how genes behave, and coupled with information about a family - who has the disease, their age, family size, and causes of death - our model can provide a good estimate of an individual's risk," says Klein.

Early risk assessment has long been sought for pancreas cancer that runs in families, Klein says because of the difficulty doctors have diagnosing it before it has already spread. Survival rates are extremely low.

To test the model's effectiveness for predicting cancer, Klein and her colleagues fed the software family history information given by more than 6,000 individuals in 961 families when they initially joined the Hopkins pancreas cancer registry several years ago. Klein's team divided registrants into groups representing the number of pancreas cancer patients in each family - from one to three or more members. They compared predictions from the PancPRO model with what actually occurred in these families from one to 11 years later. On average, the model calculated higher risk scores for individuals who developed pancreatic cancer than for those who remained disease free.

Current practices for identifying pancreatic cancer risk without the new model are slightly better than a coin toss at 61 percent, but PancPRO correctly assesses risk 75 percent of the time, taking into account all potential threshold values for defining risk. Overall, 26 individuals developed cancer, and PancPRO predicted slightly more at 31. Their published results appear in the April 10 issue of the Journal of Clinical Oncology.

Studies are under way to determine the effectiveness of screening, based on risk assessment, which includes an ultrasound procedure conducted through an endoscope inserted through the mouth, esophagus and stomach to detect precancerous changes. The test provides sound-wave pictures of the pancreas, which sits next to the stomach.

Klein says that the new model also needs to be assessed in other populations besides the Hopkins registry, and she hopes to determine how insurers would pay for screening procedures. Her group also will be studying whether PancPRO is more effective in selecting candidates for screening than current criteria based only on the number of family members with the disease.

The authors caution that results from PancPRO should only be interpreted by a genetic counselor and a physician. 

Each year in the U.S., pancreas cancer strikes more than 37,000 people and kills about the same number. Most patients succumb to the disease within six months of being diagnosed. The five-year survival rate is five percent.

Funding for the study was provided by the National Cancer Institute's Specialized Programs of Research Excellence and the Michael Rolfe Foundation.

Co-authors include Wenyi Wang, Sining Chen, Kieran A Brune, Ralph H. Hruban, and Giovanni Parmigiani of Johns Hopkins.

Citation: Wang, W. et al. "PancPro: Risk Assessment for Individuals with a Family History of Pancreatic Cancer." JCO, Vol 25, No 11.