OPTIMAL COLONOSCOPY SCREENING POLICIES FOR COLORECTAL CANCER PREVENTION AND SURVEILANCE

Purpose: Our purpose is to determine the optimal colorectal cancer (CRC) screening and surveillance policies that maximize the quality adjusted life years (QALYs) of a patient based on age, gender, personal history of polyp and CRC, and risk of having colorectal lesion.      

Method: We develop a partially observable Markov decision process (POMDP) model that optimizes the colonoscopy screening decisions for patients in three risk levels: asymptomatic, higher-risk (having history of adenomatous polyp), and post-CRC (having history of CRC). We use the clinical data from Mayo Clinic, SEER database, and literature to obtain the inputs for the model. We estimate the unobservable annual probabilities of polyp-to-CRC progression and mortality from undetected CRC via calibration.  

Result: The following figure shows the optimal annual colonoscopy decisions of higher risk patients for different ages and risks of having adenomatous polyp and CRC. The risk combinations for “Do nothing” decision increase with age implying less aggressive colonoscopy screening for older men. For example, the optimal intervals for the follow-up colonoscopy after adenomatous polyp removal are three, four, and six years for 60, 70, and 80 year-old men with average risk, respectively. Moreover, there are gender-based differences in optimal colonoscopy policies. For instance, a 75-year-old woman and man should undergo a follow-up colonoscopy five and four years after a polypectomy, respectively. The optimal policies outperform the current guidelines and no screening option in QALYs improvement and CRC risk reduction. For example, optimal policy for the higher-risk patients improves the QALYs by 1.77% and reduces CRC risk by 83% compared to the no screening option. The improvements are 0.15% and 39% when the optimal policy is compared to the current guidelines.

Conclusion: Our model promotes personalized screening policies by determining the colonoscopy decisions based on factors, including the risk of colorectal lesions, which are not considered by the current guidelines. Optimal colonoscopy screening intervals increase as age decreases and risk level increases. Our optimal policies recommend more frequent colonoscopy screening and surveillance for asymptomatic and higher-risk men, and post-CRC women. In general, our model recommends more frequent colonoscopy sessions than the current guidelines, therefore provides justification for the clinicians who practice shorter screening intervals.