An Absolute Risk Model to Identify Individuals at Elevated Risk for Pancreatic Cancer in the General Population

PURPOSE: We developed an absolute risk model to identify individuals in the general population at elevated risk of pancreatic cancer. PATIENTS AND METHODS: Using data on 3,349 cases and 3,654 controls from the PanScan Consortium, we developed a relative risk model for men and women of European ancestry based on non-genetic and genetic risk factors for pancreatic cancer. We estimated absolute risks based on these relative risks and population incidence rates. RESULTS: Our risk model included current smoking (multivariable adjusted odds ratio (OR) and 95% confidence interval: 2.20 1.84-2.62), heavy alcohol use (>3 drinks/day) (OR: 1.45 1.19-1.76), obesity (body mass index >30 kg/m(2)) (OR: 1.26 1.09-1.45), diabetes >3 years (nested case-control OR: 1.57 1.13-2.18, case-control OR: 1.80 1.40-2.32), family history of pancreatic cancer (OR: 1.60 1.20-2.12), non-O ABO genotype (AO vs. OO genotype) (OR: 1.23 1.10-1.37) to (BB vs. OO genotype) (OR 1.58 0.97-2.59), rs3790844(chr1q32.1) (OR: 1.29 1.19-1.40), rs401681(5p15.33) (OR: 1.18 1.10-1.26) and rs9543325(13q22.1) (OR: 1.27 1.18-1.36). The areas under the ROC curve for risk models including only non-genetic factors, only genetic factors, and both non-genetic and genetic factors were 58%, 57% and 61%, respectively. We estimate that fewer than 3/1,000 U.S. non-Hispanic whites have more than a 5% predicted lifetime absolute risk. CONCLUSION: Although absolute risk modeling using established risk factors may help to identify a group of individuals at higher than average risk of pancreatic cancer, the immediate clinical utility of our model is limited. However, a risk model can increase awareness of the various risk factors for pancreatic cancer, including modifiable behaviors.