Validation of a Colorectal Cancer Risk Prediction Model in US Black Women

Area of Science:

• Oncology
• Epidemiology
• Biostatistics

Background:

• The National Cancer Institute's Colorectal Cancer Risk Assessment Tool (NCI-CCRAT) was developed using data from US White populations.
• Black Americans have the highest incidence of colorectal cancer (CRC), yet the NCI-CCRAT's performance in this demographic is unknown.
• External validation is crucial to ensure risk assessment tool equity and accuracy across diverse populations.

Purpose of the Study:

• To externally validate the NCI-CCRAT for predicting 5-year CRC risk in Black women.
• To assess the calibration and discriminatory accuracy of the NCI-CCRAT in a Black population.
• To identify potential reasons for discrepancies in risk prediction.

Main Methods:

• Utilized data from 53,324 Black Women's Health Study (BWHS) participants (aged 40-84).
• Predicted 5-year absolute CRC risks for three periods: 2001-2005, 2006-2010, and 2011-2015.
• Calculated calibration (Expected/Observed ratio, E/O) and discriminatory accuracy (Area Under the Curve, AUC).

Main Results:

• The NCI-CCRAT significantly overestimated CRC cases in BWHS participants (E/O=1.34).
• Calibration was good for 2001-2005 (E/O=1.10) but overestimated risk in later periods.
• Age-adjusted AUC was 0.62 (95% CI: 0.57-0.66) for 2001-2005, indicating moderate discriminatory accuracy.

Conclusions:

• The NCI-CCRAT overestimates CRC risk in Black women, likely due to population-specific differences in risk factor associations.
• Current risk factors (physical activity, diet, BMI, aspirin use) showed weaker associations in BWHS compared to the NCI-CCRAT development cohort.
• Further research is needed to identify key CRC predictors in Black Americans and develop a tailored risk prediction model.