Personalized dynamic risk-stratification model for childhood cancer survivors

项目摘要/摘要 本申请是对特别利益通知(NOSI)的回应，该通知被确认为 不是-CA-20-038。 心血管并发症已成为儿童长期死亡的主要原因 癌症幸存者。已经有充分的研究表明，这些幸存者患癌症的风险增加了15倍和7倍。 与普通人群相比，发生充血性心力衰竭和心脏事件导致的过早死亡 人口。由于长潜伏期通常发生在临床明显的疾病之前，预测的能力 对扩大以幸存者为基础的研究的影响至关重要，为未来的设计和治疗提供信息 政权。要为幸存者提供适当的随访护理，必须以患者为中心进行风险管理 以个性化和动态的方式结合患者异质性的预测。当有一个 关于风险分层方法的文献越来越多，大多数模型只使用基线因素 与疾病转归有关。心血管负担在儿童癌症中有很好的记录 心脏风险因素随着时间的推移而增加的幸存者。此外，一系列社区和社会经济问题 因素可能会影响特定结果的风险，例如肥胖。利用圣犹大终身修道院的资源 队列研究(SJLIFE)，我们的目标是开发一种新的个性化动态癌症风险统计模型 预测，以改进目前的癌症风险分类范例。假设纵向信息是 对于SJLIFE幸存者的常规测量和记录，最好充分利用可用的纵向 数据完善风险分类，指导及时开展基于风险的预防干预措施，减轻 心脏病。所提出的方法具有概念上的简单性和有效性。 进行包含综合纵向信息的动态预测，其中风险预测/ 分层可以随着收集新的观察结果而更新，以反映患者的最新健康和 行为状态。建议的解决方案将通过我们将开发的开源R软件进行访问 公开提供，并将应用于参加SJLIFE的儿童癌症幸存者的数据。

PROJECT SUMMARY/ABSTRACT This application is being submitted in response to the Notice of Special Interest (NOSI) identified as NOT-CA-20-038. Cardiovascular complications have emerged as the leading cause of death among long-term childhood cancer survivors. It has been well-studied that these survivors have 15- and 7-fold of increased risk of developing congestive heart failure and premature death due to cardiac events compared to the general population. Since a long latency period often occurs before the clinically evident disease, the ability to predict is critical to expand the impact of survivor-based research to inform the future design and treatment regimes. To provide proper follow-up care for survivors, it is imperative to conduct patient-centered risk predictions that incorporate patient heterogeneity in a personalized and dynamic manner. While there is a growing body of literature on risk-stratification methods, most models only utilize baseline factors associated with the disease outcome. Cardiovascular burden is well-documented among childhood cancer survivor with cardiac risk factors increasing over time. In addition, a range of neighborhood and socioeconomic factors can affect risk for specific outcomes such as obesity. Leveraging the resources of the St. Jude Lifetime Cohort study (SJLIFE), we aim to develop a novel statistical model for personalized dynamic cancer risk prediction to improve the current cancer risk classification paradigm. Given that longitudinal information is routinely measured and documented for SJLIFE survivors, it is optimal to make full use of available longitudinal data to improve risk classification and guide timely risk-based prevention interventions to reduce the burden of cardiac diseases. The proposed method is appealing for its conceptual simplicity and efficiency for conducting dynamic prediction incorporating comprehensive longitudinal information, where risk prediction/ stratification can be updated as new observations are gathered to reflect the patient's latest health and behavior status. The proposed solutions will be accessible through open-source R software that we will make publicly available, and will be applied to data from childhood cancer survivors participating in SJLIFE.