Predicting pancreatic cancer in the UK Biobank cohort using polygenic risk scores and diabetes mellitus.

Diabetes mellitus (DM) is known to be associated with Pancreatic ductal adenocarcinoma (PDAC), particularly, new-onset DM (NODM). Others have developed polygenic risk scores (PRS) associated with PDAC risk. We aimed to compare the performance of these PRS in an independent cohort to determine if they can discriminate between NODM and long standing DM (LSDM) patients with PDAC.Cases (1,042) and matched cancer free controls (10,420) were drawn from the UK Biobank. Five PRS models were calculated using single nucleotide polymorphisms (SNPs) from previous studies (Nakatochi, Galeotti, Molina, Jia and Rashkin) and a combination of these. Regression models were used to assess the association between PDAC and PRS adjusted for ancestry, smoking, DM, waist circumference, and a family history of digestive cancer. Receiver operator characteristic (ROC) curves and the area under the curve metrics (AUC) were used to assess the performance of each PRS for classifying PDAC risk.The combined PRS model achieved the highest AUC (0.605), and significantly improved a clinical risk model in this cohort (AUC=0.83, P =0.0002). Individuals within the 5th quintile have a 2.74-fold increased risk of developing PDAC versus those in the 1st quintile (P <0.001), and have a 3.05-fold increased risk of developing PDAC if they have DM versus those without DM (P <0.001). The positive predictive value (PPV) was 11.9% in participants without DM, 23.9% with LSDM and 86.7% with NODM.The PDAC related common genetic variants are more strongly associated with DM. This PRS has the potential for targeting individuals with NODM for PDAC secondary screening measures.