Colorectal Tumor Risk Prediction in the PLCO Trial
Despite slight declines in colorectal cancer (CRC) incidence, it remains the second leading cause of cancer death. Paradoxically, it is among the most preventable and treatable of neoplastic diseases when detected early via screening. In the U.S., endoscopic screening, particularly colonoscopy, is the most commonly used strategy by the population at risk; however, it is costly, invasive, and carries risks, leading to overall low utilization in the population. Current screening recommendations are based only on age, family history of CRC, and previous screening results, whereas incidence of CRC varies substantially in the population and most cases occur in those without a positive family history. Therefore, the goal of this proposal is to develop, calibrate, and validate a comprehensive risk-prediction model incorporating genome-wide genetic data, as well as lifestyle and environmental risk factors, such as obesity, medication, smoking, and diet, which can provide a more accurate risk stratification for those at risk. The model would permit identification and prioritization of individuals at higher CRC risk for targeted screening and intervention, while reducing emphasis for those at low risk. In Aim 1 we will develop a comprehensive CRC risk-prediction model within the large existing Genetics and Epidemiology of Colorectal Cancer Consortium (GECCO), which includes over 14,000 cases and 15,300 controls with existing detailed genetic data from genome-wide genotyping arrays and whole genome sequencing, as well as harmonized clinical and epidemiologic variables (U01-CA137088, U01-CA164930). In Aim 2 we will undertake independent validation of the model in the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial. In PLCO, in addition to using the rich clinical and epidemiologic data, we will genotype and impute genetic variants across the genome in all 1,522 incident CRC cases, all 2,604 advanced adenoma cases with DNA, and 3,000 randomly selected controls. Including advanced adenoma will uniquely allow us to evaluate risk prediction for an important precursor lesion of CRC. Risk-prediction model for left- and right-sided CRC will be compared to account for flexible-sigmoidoscopy screening. In Aim 3 we will compare within PLCO our comprehensive model with current screening guidelines and published risk-prediction models to estimate the improvement of our model in comparison with existing guidelines and other prediction models. Our proposal leverages the extensive efforts to discover genetic and environmental risk factors for CRC, and offers the potential for pragmatic application to the population. Genetic testing is already part of routine care and it is expected that genetic data wil increasingly become part of an individual's medical record. Utilizing genetic and non-genetic risk-factor information in clinical and preventive settings is a critical step towards developing more personalized medicine based on risk stratification. Our model can be used for tailored screening and prevention strategies that are more cost-efficient and may increase adherence.