研究背景

我国人群结直肠癌负担严重，且呈现出逐步升高的趋势，是严重威胁我国居民健康的疾病之一。结直肠癌筛查传统策略因技术准确性不足、资源分配不均及参与率低下等问题，难以满足防控需求。基于新型及部分常用技术构建的结直肠癌分级筛查新策略尚缺乏长期流行病学效果评价，但传统基于人群的评价方法耗时长、花费多且灵活度较差。微观模拟模型通过动态模拟结直肠癌发生发展自然史与筛查干预效果，为策略优化提供了高效评估工具。然而，现有模型多基于欧美人群数据，缺乏符合中国本土特征的参数体系与风险分级路径，限制了精准防控政策的制定。

研究目标

本研究旨在构建中国人群特异性的结直肠癌筛查微观模拟模型，结合人群结直肠癌筛查真实参数，系统评估新型筛查技术及分级筛查策略的长期健康效益与卫生经济学价值，并提出适应我国资源现状与人群异质性的结直肠癌筛查优化策略。

研究方法

本研究的核心方法是构建符合中国人群特征的结直肠癌筛查微观模拟模型。模型设计基于“腺瘤—癌”自然史路径，结合马尔可夫链与蒙特卡罗模拟技术，动态模拟个体从无结直肠病变、非进展期腺瘤、进展期腺瘤、临床前期结直肠癌（I-IV期）及临床期结直肠癌（I-IV期）到死亡的疾病发生发展过程。模型框架包含17种疾病状态和64条状态转移路径，参数来源包括中国结直肠癌筛查策略比较的随机对照试验（TARGET-C）数据库、国家肿瘤登记数据库、全球疾病负担数据库以及公开文献库等。模型关键参数如年度状态转化率等均使用遗传算法进行严格校准，保证模型输出值与中国结直肠癌发病率与死亡率目标值一致。最终通过Bootstrap重抽样、外部验证及初步应用进一步确认模型稳健性。

基于TARGET-C结直肠癌筛查队列的真实世界研究数据，系统评估多靶点粪便FIT-DNA技术的筛查效能及其人群应用价值。首先以结肠镜病理诊断为金标准，在相同样本集中平行比较五种筛查方法的诊断性能，包括基于不同阳性截断值的定性免疫化学法粪便隐血试验FIT-1（8.0 μg Hb/g）、FIT-2（14.4 μg Hb/g）、FIT-3（20.8 μg Hb/g）、采用厂商推荐截断值的定量FIT（20 μg Hb/g）以及多靶点粪便FIT-DNA检测（定量评分165）；随后通过调整定量FIT的阳性截断值，在维持筛查阳性率与FIT-DNA检测相当的条件下，分别以结直肠癌和进展期肿瘤（结直肠癌+进展期腺瘤）的检出情况作为结局指标，对比两种筛查技术的筛查准确性，评价指标包括筛查技术的灵敏度、特异度、约登指数、准确率等。在此基础上，整合改良亚太结直肠癌风险评分构建风险分级筛查策略，并通过基于TARGET-C队列基线人群特征构建的微观模拟模型，对分级筛查方案进行长达20年模拟周期的筛查效果和卫生经济学评价。

基于前期构建的微观模拟模型及收集到的不同筛查策略参数，最终通过构建符合2020年全国45–74岁人群特征的模拟人群，对不筛查、结肠镜、免疫化学法粪便隐血试验、多靶点粪便FIT-DNA以及风险分级筛查等策略的长期效果与成本效果进行评价与比较。此外，提出全国预算分析框架，评估2025–2060年间不同筛查策略对结直肠癌发病率、死亡率、成本及医疗资源使用情况的影响，并从公共卫生服务及卫生经济学的角度对不同资源禀赋和人群特征的地区提出结直肠癌筛查策略建议。

研究结果

构建的结直肠癌筛查微观模拟模型（MIMIC-CRC）与中国人群结直肠癌自然史拟合情况较好，模型输出年龄别发病率与死亡率与观察值高度一致，校准准确性达88.6%；模型输出的性别与年龄别结直肠癌发病率及死亡率与全球疾病负担数据比较的风险比均满足20%等效性阈值（P>0.05）；外部验证显示，模型模拟英国乙状结肠镜筛查试验情景，预测的发病率风险比与死亡率风险比与试验实际报告值均保持一致（P>0.05），体现出MIMIC-CRC模型的应用潜能及稳健性。

以进展期肿瘤为研究结局时，多靶点粪便FIT-DNA的灵敏度（26.4%，95%CI：19.5%–34.8%）显著高于阳性截断值较高的定性FIT-3（19.0%，95%CI：12.5%–26.3%；P<0.001）及定量FIT（13.9%，95%CI：7.7%–20.0%；P<0.001），特异度方面则显著较低（P<0.001）。当调整到相同阳性率时的阳性截断值时，多靶点粪便FIT-DNA与定量FIT的筛查准确性差异无统计学意义。而基于多靶点粪便FIT-DNA的分级筛查策略与结肠镜筛查具有可比的疾病负担降低效果，但由于较高的使用成本（1,998元/例），其成本效果劣于传统的结肠镜筛查与基于FIT的分级筛查策略。

基于微观模拟模型的我国结直肠癌分级筛查策略优化研究结果显示，逐步推广FIT筛查与风险分级策略在资源有限地区具有显著优势。在保证较优的防控效果的同时，逐步推广FIT（初始为每两年一次）大幅减少所需结肠镜检查数（减少85.2%），总体成本也大幅降低（减少27.7%）。基于多靶点粪便FIT-DNA分级筛查策略获得的质量调整生命年最高，比结肠镜高1,900万，降低了75.6%的所需结肠镜检查数，但总体成本是结肠镜筛查策略的3.8倍。全国预算分析表明，对于医疗资源丰富的地区，可采用基于多靶点粪便FIT-DNA分级筛查策略；对于医疗成本投入不足但结肠镜资源充足的地区，可采用结肠镜筛查或两年一次FIT筛查；对于医疗成本投入及结肠镜资源均不足的地区，可采用逐步推广FIT（初始为每两年一次）筛查策略。

研究结论

MIMIC-CRC填补了中国人群结直肠癌微观模拟模型的空白，通过参数本土化、多路径疾病转归及校准策略，为结直肠癌精准防控提供了量化分析工具。新型结直肠癌筛查技术多靶点粪便FIT-DNA因其较高的筛查准确性及人群接受度具有潜在的人群筛查应用价值，但未来应重点注意使用成本的降低。逐步推广FIT与风险分级筛查策略协同可实现资源动态适配，在保证卫生经济学价值的同时提升筛查覆盖率，尤其适用于资源有限地区。综上所述，本研究为差异化筛查路径设计提供了直接证据。研究成果不仅服务于我国结直肠癌筛查政策优化，还为全球中低收入国家在资源有限情况下优化癌症筛查策略提供了可复制的方法论。未来需融合多组学数据、区域资源参数及长期随访证据，持续提升模型预测精度与政策指导价值。

Background

The persistently high incidence and mortality of colorectal cancer (CRC) in China have established it as a major public health threat. Conventional CRC screening strategies face challenges in meeting prevention demands due to limitations such as insufficient detection accuracy, uneven resource allocation, and low participation rates. Emerging risk-adapted screening strategies based on novel and existing technologies lack long-term evaluations, while traditional population-based evaluation methods are time-consuming, costly, and inflexible. Microsimulation models dynamically simulate CRC natural history and screening effects, offering an efficient tool for strategy optimization. However, existing microsimulation models primarily rely on Western population data, lacking China-specific parameter support and risk-adapted pathways, which hinders the development of precise prevention policies.

Objective

This study aimed to develop a CRC screening microsimulation model tailored to the Chinese population, systematically evaluate the long-term effectiveness and cost-effectiveness of novel screening technologies and risk-adapted strategies, and propose optimized CRC screening approaches adapted to China’s resource constraints and population heterogeneity.

Methods

1. The MIcrosimulation Model for prevention and Intervention of Colorectal Cancer in China (MIMIC-CRC) was developed based on the adenoma-carcinoma natural history pathway. The model integrated Markov chain and Monte Carlo simulations to dynamically simulate disease progression from healthy status to non-advanced adenoma, advanced adenoma, preclinical CRC (stages I–IV), clinical CRC (stages I–IV), and death. The framework included 17 disease states and 64 state transition pathways. Parameters were derived from the TARGET-C randomized controlled trial database, national cancer registry data, the Global Burden of Disease database, and published literature. Key parameters such as annual transition rates were calibrated using genetic algorithms to align model outputs with observed Chinese CRC incidence and mortality. Model robustness was validated through bootstrap resampling, external validation, and preliminary applications.

2. Real-world data from the TARGET-C cohort were used to evaluate the performance of multi-target stool FIT-DNA testing. Diagnostic accuracy of five screening methods was compared using colonoscopy as the gold standard: three qualitative FIT tests (FIT-1: 8.0 μg Hb/g; FIT-2: 14.4 μg Hb/g; FIT-3: 20.8 μg Hb/g), quantitative FIT (20 μg Hb/g), and FIT-DNA (quantitative score ≥165). At matched positivity rates, FIT-DNA and quantitative FIT showed no significant differences in accuracy for detecting CRC or advanced neoplasia (CRC and advanced adenomas). Risk-stratified screening strategies integrating FIT-DNA and the modified Asia-Pacific CRC risk score were evaluated for 20-year outcomes using MIMIC-CRC.

3. Long-term effectiveness and cost-effectiveness of five strategies—no screening, colonoscopy, FIT, FIT-DNA, and risk-stratified screening—were simulated for China’s 2020 population aged 45–74. A national budget impact analysis (2025–2060) assessed CRC incidence, mortality, costs, and resource utilization. Region-specific recommendations were proposed based on healthcare resources and economic capacity.

Results

1. MIMIC-CRC demonstrated high calibration accuracy (88.6%), with age-specific CRC incidence and mortality outputs closely matching observed data. Sex- and age-stratified risk ratios met 20% equivalence thresholds (P>0.05) against global benchmarks. External validation using UK flexible sigmoidoscopy trial data confirmed model robustness (P>0.05 for predicted vs. observed risk ratios).

2. FIT-DNA showed superior sensitivity for advanced neoplasia (26.4%, 95% CI: 19.5%–34.8%) compared to higher-cutoff FIT-3 (19.0%, 95% CI: 12.5%–26.3%; P < 0.001) and quantitative FIT (13.9%, 95% CI: 7.7%–20.0%; P < 0.001) but lower specificity. Risk-stratified FIT-DNA strategies achieved comparable burden reduction to colonoscopy but were less cost-effective (CNY 1,998 per test) than FIT or colonoscopy.

3. The roll-out FIT-based strategy reduced colonoscopy demand by 85.2% and costs by 27.7% while maintaining effectiveness. The multitarget stool FIT-DNA screening strategy yielded the highest quality-adjusted life years (QALYs), exceeding colonoscopy by 19 million QALYs, but its total cost was 3.8 times higher than colonoscopy-based screening. National budget analyses recommended FIT-DNA for resource-rich regions, colonoscopy or biennial FIT for areas with sufficient endoscopic capacity, and gradual FIT adoption for resource-limited settings.

Conclusions

MIMIC-CRC fills a critical gap in China-specific CRC microsimulation, enabling precision screening through localized parameters and validated calibration. FIT-DNA demonstrates potential for population screening but requires cost reduction. Gradual FIT rollout with risk stratification optimizes resource allocation and cost-effectiveness, particularly in resource-constrained settings. This study provides evidence for tailored screening strategies, informing CRC policy in China and offering replicable methodologies for low- and middle-income countries. Future integration of multi-omics data and regional resource parameters will enhance model precision and policy relevance.