Personalized, Dynamic Risk-based Lung Cancer Screening

项目总结/摘要 肺癌是美国癌症相关死亡的主要原因。大多数患者 被诊断为晚期疾病，可用的治疗干预措施提供的生存贝内极小。 尽管最近在筛查和治疗方法方面取得了进展，但早期发现对于实现治愈和 加强疾病管理。低剂量计算机断层扫描已成为标准的筛查方式， 国家肺筛查试验（National Lung Screening Trial）报告20%的肺癌特异性 降低死亡率。然而，对于筛查合格人群，最佳的 筛查间隔，以及肺癌筛查的开始和停止年龄，导致 现有建议。此外，低剂量计算机断层扫描与潜在危害相关，包括， 假阳性结果辐射暴露和过度诊断现有的肺癌筛查指南分层 基于年龄和吸烟史的个体，忽略了与肺癌相关的其他重要风险因素 发展 这项研究旨在通过开发个性化的、动态的、基于风险的 通过随机动态决策模型筛选策略。这个项目利用了已发表的肺癌 自然史模型，用于在没有任何干预的情况下模拟疾病进展，沿着肺 癌症特异性风险预测模型，以个性化水平估计发生肺癌的风险。我们 将制定肺癌筛查问题作为一个有限的地平线，离散时间部分可观测马尔可夫 决策过程（POMDP），以优化随机条件下的肺癌筛查检查序列 健康进展和不完善的状态信息。POMDP模型的目标是最大化预期 通过筛查有患肺癌风险的无症状个体获得的寿命。该模型 将把筛查史沿着个人患肺癌的风险纳入决策 提供最先进的个性化筛选策略的过程。预期的最优筛选策略 将在成本效益分析中进行测试，以检查与肺癌筛查相关的成本 是合理的，因为它带来了健康贝内。 本项目为肺癌筛查排程研究提供了一个新的方向。重要危险因素 包括年龄、性别、种族/民族、筛查史和癌症家族史等， 肺癌筛查的有效性。拟议的研究承认其重要性和地址 将这些因素的动态演变纳入决策的筛选调度问题 过程该项目的结果将成为制定具有成本效益的准则的基础， 基于风险的个性化肺癌筛查。拟议的分析模型有可能成为 扩展到解决其他影响肺癌筛查的棘手问题。

Project Summary/Abstract Lung cancer is the leading cause of cancer related deaths in the United States. The majority of patients are diagnosed with advanced stage disease for which available treatment interventions offer minimal survival benefit. Despite recent advancements in screening and treatment methods, early detection is vital to achieve cure and enhance disease management. Low-dose computed tomography has become the standard screening modality for lung cancer after the conclusion of the National Lung Screening Trial which reported 20% lung cancer-specific mortality reduction. However, there is considerable debate over the screen eligible population, the optimal screening interval, and the starting and stopping ages of lung cancer screening, causing discrepancies in the existing recommendations. Moreover, low-dose computed tomography is associated with potential harms including, false-positive results, radiation exposure, and overdiagnosis. Existing guidelines for lung cancer screening stratify individuals based on age and smoking history, ignoring other important risk-factors associated with lung cancer development. The proposed research aims to improve lung cancer screening by developing individualized, dynamic risk-based screening strategies through stochastic, dynamic decision models. This project leverages a published lung cancer natural history model to simulate the disease progression in the absence of any intervention, along with a lung cancer-specific risk prediction model to estimate the risk of developing lung cancer on a personalized level. We will formulate the lung cancer screening problem as a finite horizon, discrete time partially observable Markov decision process (POMDP) to optimize the sequence of lung cancer screening examinations under stochastic health progression and imperfect state information. The objective of the POMDP model is to maximize the expected lifetime gained from screening asymptomatic individuals at risk of developing lung cancer. The proposed model will incorporate screening history along with the personal risk of developing lung cancer into the decision making process providing state-of-the-art individualized screening strategies. The anticipated optimal screening policies will be tested in a cost-effectiveness analysis to examine whether the cost associated with lung cancer screening is justifiable by the health benefits gained. This project presents a new direction in lung cancer screening scheduling research. Important risk factors including age, gender, race/ethnicity, screening history, and family history of cancer, among others, influence the effectiveness of lung cancer screening. The proposed research acknowledges their significance and addresses the screening scheduling problem incorporating the dynamic evolution of these factors into the decision making process. The findings of this project will form the basis for the development of cost-effective guidelines for personalized, risk-based lung cancer screening. The proposed analytical models would have the potential to be extended to address other vexing problems affecting lung cancer screening.