Developing methods and software for fitting the Cox proportional hazards model to partly interval-censored data

标题： 部分区间删失数据的Cox比例风险模型拟合方法及软件开发 摘要： 部分间隔删失的事件间隔时间数据经常出现在涉及周期性疾病的医学研究中 检查，如癌症、艾滋病毒、传染病和糖尿病。例如，无进展生存 (PFS)，三期癌症临床试验中最常用的主要终点，实际上部分是区间- 被审查为患者的死亡时间将被准确地知道，而疾病进展的时间将被 只知道落在两个成像评估时间点之间。 由于缺乏完善的方法和软件，部分区间删失数据被视为 在临床试验和医学研究中用于分析的右审查数据，这在一开始就引入了偏见 并可能导致无效的统计推断和错误的医学结论。建议的项目包括 四个目标：(1)发展一种贝叶斯半参数方法来拟合COX比例风险模型。 最常用的生存回归模型，用于部分区间截尾数据；(2)发展贝叶斯 半参数方法拟合具空间脆弱性与空间相关性的Cox比例风险模型 部分区间截尾数据；(3)发展了一种贝叶斯半参数方法来拟合多重脆弱性 部分区间删失数据聚类的脆弱选择比例风险模型 经过验证的新方法，以构建高效、可靠和用户友好的医疗R包 调查人员可以使用。R是最常用的统计计算和图形软件，对公众免费。 因此，这样的工具将有助于寻找有效的新药和识别危险因素，从而导致 对许多影响重大的疾病改进患者治疗和疾病预防 公共卫生。

Title: Developing methods and software for fitting the Cox proportional hazards model to partly interval-censored data Abstract: Partly interval-censored time-to-event data arise frequently in medical studies for diseases that involve periodic examinations, such as cancer, HIV, infectious diseases, and diabetes. For example, progression-free survival (PFS), the most commonly used primary endpoint in phase III cancer clinical trials, is actually partly interval- censored as the time of death for a patient will be exactly known while the time of disease progression will be only known to fall between two imaging assessment timepoints. Due to the lack of well-developed methods and software, partly interval-censored data have been treated as right-censored data for analysis in clinical trials and medical studies, which introduces bias at the very beginning and may lead to invalid statistical inferences and erroneous medical conclusions. The proposed project contains four aims: (1) to develop a Bayesian semiparametric method for fitting the Cox proportional hazards model, the most commonly used survival regression model, to partly interval-censored data; (2) to develop a Bayesian semiparametric method for fitting the Cox proportional hazards model with spatial frailty to spatially correlated partly interval-censored data; (3) to develop a Bayesian semiparametric method for fitting a multiple-frailty proportional hazards model with frailty selection to clustered partly interval-censored data; (4) then based on the validated new methods, to construct an R package that is efficient, reliable, and user-friendly for medical investigators to use. R is the most used statistical computing and graphics software and is free to the public. Hence, such a tool will facilitate the search for effective new drugs and identification of risk factors, thereby lead to improvement in patient treatment and disease prevention for many diseases that are significantly affecting public health.