Risk-adjusted colorectal cancer screening using the FIT and routine screening data: development of a risk prediction...

Area of Science:

• Gastroenterology
• Oncology
• Medical Diagnostics

Background:

• The faecal immunochemical test (FIT) is increasingly used for colorectal cancer screening.
• Rising screening uptake and FIT positivity necessitate improved referral accuracy to manage colonoscopy service demands.
• A novel risk prediction model was developed to enhance screening referral precision.

Purpose of the Study:

• To develop and validate a risk prediction model integrating FIT concentration with routine screening data.
• To improve the accuracy of referrals for colonoscopy following positive FIT results.
• To enhance the diagnostic yield of colorectal cancer and advanced adenomas in screening.

Main Methods:

• Logistic regression and feedforward neural networks were employed to build the risk prediction model.
• The model utilized FIT concentration, age, sex, and previous screening history from 1810 positive FIT cases.
• Model performance was evaluated using discrimination, calibration, sensitivity, specificity, and ROC curves.

Main Results:

• The risk-adjusted model improved discrimination from 0.628 to 0.659 (P=0.01).
• Calibration was strong (Hosmer-Lemeshow test = 0.90).
• Clinical sensitivity increased from 30.78% to 33.15% at a FIT threshold of 160 μg/g, with similar specificity. Neural network further enhanced accuracy.

Conclusions:

• Integrating routine risk predictors with FIT significantly enhances screening sensitivity.
• This approach increases the diagnostic yield of high-risk adenomas.
• The developed model offers a more accurate method for colorectal cancer screening referrals.