2019 HSR&D/QUERI National Conference

4091 — Validation of the National Cancer Institute Colorectal Cancer Risk Assessment Tool in the CSP #380 Veterans Cohort

Lead/Presenter: Laura Musselwhite,  Cooperative Studies Program Epidemiology Center-Durham
All Authors: Musselwhite LW (Cooperative Studies Program Epidemiology Center-Durham, Duke University ), Redding TS (Cooperative Studies Program Epidemiology Center-Durham), Sims K (Cooperative Studies Program Epidemiology Center-Durham) O'Leary M (Cooperative Studies Program Epidemiology Center-Durham) Hauser ER (Cooperative Studies Program Epidemiology Center-Durham, Duke University) Hyslop T (Cooperative Studies Program Epidemiology Center-Durham, Duke University) Gellad ZF (Cooperative Studies Program Epidemiology Center-Durham, Duke University) Lieberman (VA Portland Health Care System, Oregon Health and Science University) Provenzale D(Cooperative Studies Program Epidemiology Center-Durham, Duke University)

Objectives:
Most patients who undergo screening colonoscopy do not derive individual benefit and are exposed to procedural harms and costs. Tailoring screening strategy to colorectal cancer (CRC) risk may improve efficiency for all stakeholders. We applied the National Cancer Institute (NCI) CRC Risk Assessment Tool, which calculates 5- 10-year, and 20-year absolute risk of colorectal cancer to determine whether it could be used to predict baseline risk of colorectal cancer precursors in a Veterans cohort undergoing first screening colonoscopy.

Methods:
We studied Veterans enrolled in CSP #380, a cohort of Veterans aged 50-75 who underwent screening colonoscopy, to validate the NCI CRC Risk Assessment Tool. We applied the NCI CRC Risk Assessment Tool to determine whether it predicted baseline risk of colorectal cancer precursors in a Veterans cohort undergoing first screening colonoscopy. Family, medical, dietary and physical activity histories were collected at study enrollment and used to calculate absolute 5-, 10-, and 20-year CRC risk, and to compare tertiles of expected-to-observed CRC risk by risk factors incorporated into the model and by risk score tertile. Discriminatory accuracy was also measured.

Results:
Of 3,121, 94% had complete data available to calculate risk (N = 2,934, median age 63 years, 100% men, and 15% minorities). 11% (N = 313) were diagnosed with AN on baseline screening colonoscopy. The area under the curve for predicting AN was 0.60 (95% CI; 0.57-0.63, p < 0.0001) at 5 years, 0.60 (95% CI, 0.57-0.63, p < 0.0001) at 10 years and 0.58 (95% CI, 0.54-0.61, p < 0.0001) at 20 years. At 5 years, we calculated the sensitivity (0.18, 95% CI; 0.14-0.22), specificity (0.91, 95% CI; 0.90-0.92) positive predictive value (0.19, 95% CI; 0.15-0.24) and negative predictive value (0.90, 95% CI; 0.89-0.91) considering the top 10th percentile of risk tool scores as a positive result.

Implications:
The NCI CRC Risk Assessment Tool had modest discriminatory function for predicting AN risk at 5, 10 and 20 years. The Tool's specificity and negative predictive value were quite good, highlighting its usefulness in risk prediction.

Impacts:
This tool may be used to inform the benefit-risk assessment of screening colonoscopy vs. other CRC screening modalities for patients with competing comorbidities.