Population-level impact of the BMJ Rapid Recommendation for colorectal cancer screening: a microsimulation analysis.

OBJECTIVE: In 2019, a BMJ Rapid Recommendation advised against colorectal cancer (CRC) screening for adults with a predicted 15-year CRC risk below 3%. Using Switzerland as a case study, we estimated the population-level impact of this recommendation. DESIGN: We predicted the CRC risk of all respondents to the population-based Swiss Health Survey. We derived the distribution of risk-based screening start age, assuming predicted risk was calculated every 5 years between ages 25 and 70 and screening started when this risk exceeded 3%. Next, the MISCAN-Colon microsimulation model evaluated biennial faecal immunochemical test (FIT) screening with this risk-based start age. As a comparison, we simulated screening initiation based on age and sex. RESULTS: Starting screening only when predicted risk exceeded 3% meant 82% of women and 90% of men would not start screening before age 65 and 60, respectively. This would require 43%-57% fewer tests, result in 8%-16% fewer CRC deaths prevented and yield 19%-33% fewer lifeyears gained compared with screening from age 50. Screening women from age 65 and men from age 60 had a similar impact as screening only when predicted risk exceeded 3%. CONCLUSION: With the recommended risk prediction tool, the population impact of the BMJ Rapid Recommendation would be similar to screening initiation based on age and sex only. It would delay screening initiation by 10-15 years. Although halving the screening burdens, screening benefits would be reduced substantially compared with screening initiation at age 50. This suggests that the 3% risk threshold to start CRC screening might be too high.

Emulator-based Bayesian calibration of the CISNET colorectal cancer models.

Purpose: To calibrate Cancer Intervention and Surveillance Modeling Network (CISNET) 's SimCRC, MISCAN-Colon, and CRC-SPIN simulation models of the natural history colorectal cancer (CRC) with an emulator-based Bayesian algorithm and internally validate the model-predicted outcomes to calibration targets. Methods: We used Latin hypercube sampling to sample up to 50,000 parameter sets for each CISNET-CRC model and generated the corresponding outputs. We trained multilayer perceptron artificial neural networks (ANN) as emulators using the input and output samples for each CISNET-CRC model. We selected ANN structures with corresponding hyperparameters (i.e., number of hidden layers, nodes, activation functions, epochs, and optimizer) that minimize the predicted mean square error on the validation sample. We implemented the ANN emulators in a probabilistic programming language and calibrated the input parameters with Hamiltonian Monte Carlo-based algorithms to obtain the joint posterior distributions of the CISNET-CRC models' parameters. We internally validated each calibrated emulator by comparing the model-predicted posterior outputs against the calibration targets. Results: The optimal ANN for SimCRC had four hidden layers and 360 hidden nodes, MISCAN-Colon had 4 hidden layers and 114 hidden nodes, and CRC-SPIN had one hidden layer and 140 hidden nodes. The total time for training and calibrating the emulators was 7.3, 4.0, and 0.66 hours for SimCRC, MISCAN-Colon, and CRC-SPIN, respectively. The mean of the model-predicted outputs fell within the 95% confidence intervals of the calibration targets in 98 of 110 for SimCRC, 65 of 93 for MISCAN, and 31 of 41 targets for CRC-SPIN. Conclusions: Using ANN emulators is a practical solution to reduce the computational burden and complexity for Bayesian calibration of individual-level simulation models used for policy analysis, like the CISNET CRC models. In this work, we present a step-by-step guide to constructing emulators for calibrating three realistic CRC individual-level models using a Bayesian approach.

Colorectal Cancer Screening Decision Based on Predicted Risk: Protocol for a Pilot Randomized Controlled Trial.

BACKGROUND: Incidence of and mortality from colorectal cancer (CRC) can be effectively reduced by screening with the fecal immunochemical test (FIT) or colonoscopy. Individual risk to develop CRC within 15 years varies from <1% to >15% among people aged 50 to 75 years. Communicating personalized CRC risk and appropriate screening recommendations could improve the risk-benefit balance of screening test allocations and optimize the use of limited colonoscopy resources. However, significant uncertainty exists regarding the feasibility and efficacy of risk-based screening. OBJECTIVE: We aim to study the effect of communicating individual CRC risk and a risk-based recommendation of the FIT or colonoscopy on participants' choice of screening test. We will also assess the feasibility of a larger clinical trial designed to evaluate the impact of personalized screening on clinical outcomes. METHODS: We will perform a pilot randomized controlled trial among 880 residents aged 50 to 69 years eligible to participate in the organized screening program of the Vaud canton, Switzerland. Participants will be recruited by mail by the Vaud CRC screening program. Primary and secondary outcomes will be self-assessed through questionnaires. The risk score will be calculated using the open-source QCancer calculator that was validated in the United Kingdom. Participants will be stratified into 3 groups-low (<3%), moderate (3% to <6%), and high (≥6%) risk-according to their 15-year CRC risk and randomized within each risk stratum. The intervention group participants will receive a newly designed brochure with their personalized risk and screening recommendations. The control group will receive the usual brochure of the Vaud CRC screening program. Our primary outcome, measured using a self-administered questionnaire, is appropriate screening uptake 6 months after the intervention. Screening will be defined as appropriate if participants at high risk undertake colonoscopy and participants at low risk undertake the FIT. We will also measure the acceptability of the risk score and screening recommendations and the psychological factors influencing screening behavior. We will also assess the feasibility of a full-scale randomized controlled trial. RESULTS: We expect that a total sample of 880 individuals will allow us to detect a difference of 10% (α=5%) between groups. The main outcome will be analyzed using a 2-tailed chi-squared test. We expect that appropriate screening uptake will be higher in the intervention group. No difference in overall screening uptake is expected. CONCLUSIONS: We will test the impact of personalized risk information and screening recommendations on participants' choice of screening test in an organized screening program. This study should advance our understanding of the feasibility of large-scale risk-based CRC screening. Our results may provide insights into the optimization of CRC screening by offering screening options with a better risk-benefit balance and optimizing the use of resources. TRIAL REGISTRATION: ClinicalTrials.gov NCT05357508; https://www.clinicaltrials.gov/study/NCT05357508. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): DERR1-10.2196/46865.

The Natural Disease Course of Pancreatic Cyst-Associated Neoplasia, Dysplasia, and Ductal Adenocarcinoma: Results of a Microsimulation Model.

BACKGROUND & AIMS: Estimates on the progression of precursor lesions to pancreatic cancer (PC) are scarce. We used microsimulation modeling to gain insight into the natural disease course of PC and its precursors. This information is pivotal to explore the efficacy of PC screening. METHODS: A Microsimulation Screening Analysis model was developed in which pancreatic intraepithelial neoplasms and cysts can evolve from low-grade dysplasia (LGD) to high-grade dysplasia (HGD) to PC. The model was calibrated to Dutch PC incidence data and Japanese precursor prevalence data (autopsy cases without PC) and provides estimates of PC progression (precursor lesion onset and stage duration). RESULTS: Mean LGD state durations of cysts and pancreatic intraepithelial neoplasms were 15.8 years and 17.1 years, respectively. Mean HGD state duration was 5.8 years. For lesions that progress to PC, the mean duration was 4.8-4.9 years for LGD lesions and 4.0-4.1 years for HGD lesions. In 13.7% of individuals who developed PC, the HGD state lasted less than 1 year. The probability that an individual at age 50 years developed PC in the next 20 years was estimated to be 1.8% in the presence of any cyst and 6.1% in case of an LGD mucinous cyst. This 20-year PC risk was estimated to be 5.1% for individuals with an LGD pancreatic intraepithelial neoplasm. CONCLUSIONS: Mean duration of HGD lesions before development of PC was estimated to be 4.0 years. This implies a window of opportunity for screening, presuming the availability of a reliable diagnostic test. The probability that an LGD cyst will progress to cancer was predicted to be low.

Risk prediction for advanced neoplasia using longitudinal adherence measures to fecal immunochemical test-based colorectal cancer screening programs.

BACKGROUND: Patterns of longitudinal adherence may predict advanced neoplasia (AN) detection in subsequent rounds of colorectal cancer (CRC) screening. However, after more than five rounds, it is important to obtain a simplified measure. The aim was to determine the best simplified measure of longitudinal adherence to predict AN detection in CRC screening. METHODS: Individuals with four invitations from a Dutch Fecal immunochemical testing (FIT-)based pilot study and two Italian FIT-based CRC screening programs were included. We calculated AN detection in the fourth round, stratified by prior adherence. Five simplified measures were compared to full information (permutations) using chi-squared goodness-of-fit: adherence previous invitation, consistency, frequency, frequency + adherence previous invitation, and proportion of invitations covered. RESULTS: AN detection in the fourth round was highly dependent on prior adherence behavior. For inconsistent adherence, detection in the fourth round was strongly dependent on frequency and time since last participation. The performance of the simplified measures to capture this variation differed considerably. 'Adherence previous invitation' scored worst in predicting AN detection. 'Frequency+adherence previous invitation' had lowest chi-squared goodness-of-fit. DISCUSSION: The simplified measure 'frequency+adherence previous invitation' is the best measure to reflect patterns of longitudinal adherence and could be used to emphasize to individuals the importance of CRC screening.

An Evolutionary Algorithm to Personalize Stool-Based Colorectal Cancer Screening.

BACKGROUND: Fecal immunochemical testing (FIT) is an established method for colorectal cancer (CRC) screening. Measured FIT-concentrations are associated with both present and future risk of CRC, and may be used for personalized screening. However, evaluation of personalized screening is computationally challenging. In this study, a broadly applicable algorithm is presented to efficiently optimize personalized screening policies that prescribe screening intervals and FIT-cutoffs, based on age and FIT-history. METHODS: We present a mathematical framework for personalized screening policies and a bi-objective evolutionary algorithm that identifies policies with minimal costs and maximal health benefits. The algorithm is combined with an established microsimulation model (MISCAN-Colon), to accurately estimate the costs and benefits of generated policies, without restrictive Markov assumptions. The performance of the algorithm is demonstrated in three experiments. RESULTS: In Experiment 1, a relatively small benchmark problem, the optimal policies were known. The algorithm approached the maximum feasible benefits with a relative difference of 0.007%. Experiment 2 optimized both intervals and cutoffs, Experiment 3 optimized cutoffs only. Optimal policies in both experiments are unknown. Compared to policies recently evaluated for the USPSTF, personalized screening increased health benefits up to 14 and 4.3%, for Experiments 2 and 3, respectively, without adding costs. Generated policies have several features concordant with current screening recommendations. DISCUSSION: The method presented in this paper is flexible and capable of optimizing personalized screening policies evaluated with computationally-intensive but established simulation models. It can be used to inform screening policies for CRC or other diseases. For CRC, more debate is needed on what features a policy needs to exhibit to make it suitable for implementation in practice.