• pancreatic cancer risk score

Pancreatic cancer is a rare disease which displays a very high mortality rate, being currently the fifth cause of cancer-related deaths in Europe. The most common form of the disease, which represents about 95% of total cases, is the pancreatic ductal adenocarcinoma (PDAC). Since most cases are asymptomatic in early stages and due to the lack of specific biomarkers for detection, PDAC is usually diagnosed in advanced stages. Because of this situation most patients die within one year of diagnosis. Therefore, the best way to decrease PDAC mortality would be to increase early detection, and a feasible strategy for achieving this goal is to survey subjects with high risk of developing the disease. Though, the complex etiology of pancreatic cancer makes it difficult to identify high risk subjects. Pancreatic cancer shows a multifactorial etiology, which means that the risk of developing the disease is influenced by both epidemiological and genetic risk factors. Known epidemiological risk factors are tobacco smoking, heavy alcohol consumption, type II diabetes mellitus (T2D), obesity, chronic pancreatitis and family history of pancreatic cancer. Many common low-penetrance germline variants increase the risk of developing PDAC. Four genome-wide association studies (GWASs) (PanScan I in 2009, PanScan II in 2010, PanScan III in 2014, PanC4 in 2015) and the following imputation analysis of PanScan in 2016 identified several susceptibility loci in populations of Caucasian origin.
The aim of this study was to generate a multigenic score for PDAC risk prediction, combining the effect of known risk SNPs. This can help risk stratification, i.e. the identification of different risk classes in the population.
The study was conducted on 3238 PDAC cases and 5242 controls from populations of European origin. Twenty-six single nucleotide polymorphisms (SNPs) were selected among all the ones known to be associated with risk and the ones close to the significance threshold, emerged from previous studies. Part of the data on genotypes was already available, as they had been generated in previous projects. Genotyping completion was carried out on all the subjects belonging to the study population using a system based on allele-specific PCR. Two kinds of polygenic scores were built for each subject: unweighted and weighted. For the weighted score, the OR of the association of each SNP with risk was used as a coefficient to weight their relative effect. The scores were used both as a continuous and as a categorical variable, calculating the quintiles and the deciles based on the controls distribution. The association between the scores and PDAC risk was tested by GLM for the continuous variable and by logistic regression for the categorical variable. The multigenic score reached high significance in all models, both considering the variable as continuous and as categorical. The best result was observed for the weighted score (OR=1.96, 95% C.I.=1.66-2.30, p=6.66X10-16 for highest vs. lowest quintile).
As expected, the combination of different SNPs has led to a large increase in the ORs. The majority of the ORs observed for the individual SNPs in this study are below 1.5, while the highest decile of the unweighted score reached OR=3.45 (95% CI = 2.41-4.94).
Although the multigenic score identifies high-risk subgroups, the increase in risk prediction is limited and none of the models help identifying a fraction of the population at extremely high-risk that would benefit from regular screening.
Furthermore, an exploratory analysis for the construction of a multifactorial score was conducted. We began collecting retrospective data on epidemiological risk factors and we tested the association of smoking, diabetes and diabetes occurred before PDAC with risk within this study population. The multifactorial scores were constructed by complementing the weighted genetic score with one variable at a time or with pairs of variables.
This first analysis was very encouraging because it showed a large increase in the ORs, reaching OR>5, that would help to identify a subgroup of the population at very high risk who would benefit from regular screening. For example, the weighted score that includes the factors smoking and diabetes showed OR=9.26 (95% CI=5.14-16.68) and p-value=1.19x10-13. However, we need to use caution in interpreting the results because data of the covariates are largely incomplete. Though, the combination of the genotypes and the data on risk factors seems a suitable way for the construction of a score that leads to the identification of a subgroup of subjects with very high risk.
In perspective, the implementation of the score with new genetic risk variants, which are continuously discovered, and with complete data on epidemiological risk factors can lead to the achievement of a tool for risk stratification of clinical utility.