Analytics in Action News

Predictive Analytics Identifies Patients at Risk of Pancreatic Cancer

A predictive analytics model combined genetic and clinical risk factors with biomarkers to detect patients at high risk of pancreatic cancer.

Predictive analytics identifies patients at high risk of pancreatic cancer

Source: Getty Images

By Jessica Kent

- A predictive analytics model was able to accurately identify patients at higher than normal risk for pancreatic cancer, according to a study published in Cancer Epidemiology, Biomarkers & Prevention.

Pancreatic cancer is the third leading cause of cancer death in the US, researchers noted, with the majority of patients presenting with severe symptoms.

“Pancreatic cancer is a particularly deadly cancer, with about 80 percent of patients diagnosed with advanced, incurable disease,” said the study’s lead author, Peter Kraft, PhD, professor of epidemiology at the Harvard T.H. Chan School of Public Health in Boston. “Catching it at an earlier stage makes it more likely that surgery will be an option, increasing the chances of survival.”

Existing screening techniques like magnetic resonance imaging (MRI) are not recommended for the general public because they may generate an excessive number of false positives. MRIs are most appropriate for people at higher risk of pancreatic cancer, and improving identification of the high-risk population could enhance prevention and screening efforts.

Risk factors for pancreatic cancer include family history, chronic conditions such as diabetes and pancreatitis, and smoking. Prospective studies have also shown that certain circulating biomarkers tied to insulin resistance could increase risk.

“These factors have been investigated individually, and in this study, we wanted to examine the combined effect of clinical factors, common genetic predisposition variants, and circulating biomarkers,” said Kraft.

For the study, researchers examined data from four large prospective cohort studies: the Health Professionals Follow-up Study; the Nurses’ Health Study; the Physicians’ Health Study; and the Women’s Health Initiative. The team analyzed data from 500 patients diagnosed with primary pancreatic adenocarcinoma between 1984 and 2010, as well as 1,091 matched controls.

The study enrolled only US non-Hispanic white patients, because genomic risk variants have not been confirmed in other groups.

Through patient questionnaires, researchers collected information on lifestyle and clinical characteristics from patients, as well as blood samples and genomic DNA from peripheral blood leukocytes of the participants. The team then calculated a weighted genetic risk score based on data from two genome-wide association studies.

Researchers developed three relative predictive analytics models for men and women separately. One included only clinical factors, one included the weighted genetic risk score along with the clinical factors, and the third added biomarkers proinsulin, adiponectin, IL-6, and total branched-chain amino acids.

The models identified subsets of participants who were at three-fold or higher increased risk for pancreatic cancer than the general population. The model featuring only clinical factors identified 0.2 percent of men and 1.5 percent of women who were at three-fold or higher increased risk.

The model that combined clinical and genetic factors identified 0.3 percent of men and 2.3 percent of women at three-fold or greater risk. Finally, the model that added the weighted genetic risk score and circulating biomarkers identified 1.8 percent of men and 0.7 percent of women who were at three-fold or higher increased risk.

The final integrated model identified 2.0 percent of men and 2.3 percent of women who had at least three times greater than average risk in ten years of follow-up. Individuals in the top one percent had a four percent lifetime risk of pancreatic cancer.

Although researchers noted that the model would need to be tested and studied in other populations, the study suggests that combining biomarkers with clinical and genetic factors can help identify patients who could benefit from early screening of pancreatic cancer.

“Like most cancers, pancreatic cancer is multifactorial,” Kraft said. “The more we are able to combine information from multiple domains, the better we will become at identifying those who could benefit from screening.”