Detecting racial disparities in cancer survival by integrating multiple high-dimensional observational studies

项目摘要/摘要 尽管总体上有所改善，但族裔或种族差距继续扩大，这表明 在理解差异方面的研究设计存在缺陷。例如，与 2017年美国人口普查，大多数观察性癌症研究被发现过度代表高加索人 并且低估了非裔美国人和亚洲人的代表性。如何利用这些研究来检测和 理解种族差异仍然具有挑战性。 这项建议是由波士顿肺癌生存队列(BLCSC)推动的，该队列是 全球最大的肺癌队列，包括自1992年以来登记的肺癌病例 在Dana-Farber癌症研究所(DFCI)和马萨诸塞州综合医院(MGH)，以及 已经扩展到MD Anderson癌症中心(MDACC)和梅奥诊所。这个丰富的数据库 为研究癌症结局中的种族差异以及 在各种族群体中协变量不平衡的情况下提出了一项挑战。我们还可以访问 国际肺癌和癌症联合会(ILCCO)，一个成立于 2004，其数据结构类似于BLCSC。 利用这些癌症队列，我们开发方法的共同目标是有效地 通过整合高维观测来识别癌症预后中的种族差异 对多个种族群体的研究。像BLCSC和ILLCO这样的丰富数据集非常适合集成、 明确检测癌症结局中的种族差异，并产生统计数据 调查结果可推广到现实和包容的更多人口。

PROJECT SUMMARY/ABSTRACT Despite overall improvements, ethnic or racial disparities continue to increase, suggesting deficiencies in research designs for understanding disparities. For example, compared to the 2017 US Census, most observational cancer studies were found to over represent Caucasians and underrepresent African Americans and Asians. How to utilize these studies to detect and understand racial disparities remains challenging. This proposal is motivated by the Boston Lung Cancer Survival Cohort (BLCSC), one of the largest lung cancer cohorts globally, which consists of lung cancer cases registered since 1992 at the Dana-Farber Cancer Institute (DFCI) and the Massachusetts General Hospital (MGH), and has expanded to the MD Anderson Cancer Center (MDACC) and Mayo Clinic. This rich database provides a unique opportunity for studying racial disparities in cancer outcomes as well as presents a challenge with unbalanced covariates across racial groups. We also have access to the International Lung and Cancer Consortium (ILCCO), an international cohort established in 2004 with a data structure similar to BLCSC. Leveraging these cancer cohorts, we develop methods with a common goal of effectively identifying racial disparities in cancer outcomes by integrating high dimensional observational studies with multiple racial groups. Rich datasets like BLCSC and ILLCO are ideal for integrative, unconfounded detection of racial disparities in cancer outcomes, and for generating statistical findings generalizable to a realistic and inclusive larger population.