Unified group sequential designs for clustered data (eye level) in randomized eye trials

摘要 中期数据监测对于成功设计、实施和报告长期 眼科疾病的随机临床试验。组序贯方法是最常用的 采用中期监测计划的研究方法，控制I类错误率，保护统计学 权力，并避免假阳性/阴性结果。然而，当患者的双眼（群集） 被列入研究的，组序贯方法占眼间相关性最好 根据我们的知识，基本上不存在所有的设计选项（不同的随机化方式， 眼睛或患者的一只眼睛）。在本申请中，通过 乔治华盛顿大学和威尔斯眼科医院，结合理论和应用统计 专业知识，我们建议开发和评估一个统一的艾德组顺序设计的视觉研究 该方法适应所有可能的设计选项，同时考虑到眼间相关性， 患者的眼睛包括在研究中。这些设计选项包括：1）只有一只眼睛符合资格， 每人随机化; 2）双眼设计; 3）双眼设计+单眼病例 仅; 4）同一研究组的两只眼睛（聚类）加上仅贡献一只眼睛的病例;以及5）两只眼睛 相同或不同组的患眼加上仅贡献单眼的病例-统一艾德设计， 包括前四个设计作为特殊情况。建议的项目旨在达致以下目的 具体目标：1）提出，调查和评估组序贯试验的设计属性， 聚类（视线水平）连续终点; 2）提出、调查和评估设计属性 对于具有群集（眼水平）二元终点的组序贯试验;以及3）提出，调查 并评估具有群集（眼水平）生存终点的组序贯试验的设计特性。 这些目标的实现将为研究人员提供一套新的统计工具， 改进长期试验的设计和监测，尽量减少临时试验的可能性， 可能导致不必要的研究继续或提前停止的决定。每一个具体目标都将是 通过以下步骤实现：1）严格建立所提出的理论性质 方法; 2）通过广泛的模拟研究检查有限样本性能;以及 3)将所提出的方法应用于来自真实的示例眼试验的中间数据（眼相关性试验）， 疾病研究，眼部HOME监测研究，早产儿视网膜病变的冷冻治疗 试验和早产儿视网膜病变早期治疗研究），以证明其有效性 提出的方法。将编制实施新方法的计算机程序 通过R Shiny App提供，具有Web界面，便于从业人员轻松访问和方便 使用.将提供详细的使用指南和文档。

ABSTRACT Interim data monitoring is essential to the successful design, conduct, and reporting of long-term randomized clinical trials for eye diseases. Group sequential methods are the most commonly used methods for studies with interim monitoring plans, which control type I error rate, protect statistical power, and avoid false positive/negative findings. However, when both eyes (clustered) from a patient are included in a study, group sequential methods accounting for the inter-eye correlation to the best of our knowledge, essentially do not exist for all design options (different ways of randomizing both eyes or one eye of a patient) in the literature. In this application, through a collaborative effort between George Washington University and Wills Eye Hospital, that combines theoretical and applied statistical expertise, we propose to develop and evaluate a unified group sequential design for vision research that accommodates all possible design options while accounting for inter-eye correlations when both eyes of a patient are included in the study. These design options include: 1) only one eye eligible and randomized per person; 2) paired eye design; 3) paired eye design plus cases contributing one eye only; 4) both eyes on the same study arm (clustered) plus cases contributing one eye only; and 5) two eyes either on the same or different arms plus cases contributing one eye only - unified design, which includes the first four designs as special cases. The proposed project aims to achieve the following specific aims: 1) to propose, investigate and evaluate design properties for group sequential trials with clustered (eye level) continuous endpoints; 2) to propose, investigate and evaluate design properties for group sequential trials with clustered (eye level) binary endpoints; and 3) to propose, investigate and evaluate design properties for group sequential trials with clustered (eye level) survival endpoints. Completion of these proposed aims will provide investigators with a new suite of statistical tools to improve the design and monitoring of long-term trials and minimize the likelihood of making interim decisions that might lead to unnecessary study continuation or early stopping. Each specific aim will be achieved through the following steps: 1) rigorously establish the theoretical properties of the proposed methodology; 2) examine the finite-sample performance through extensive simulation studies; and 3) apply the proposed methods to interim data from real example eye trials (the Age-Related Eye Disease Study, the HOme Monitoring of the Eye Study, the Cryotherapy for Retinopathy of Prematurity Trial, and the Early Treatment for Retinopathy of Prematurity Study) to demonstrate the usefulness of the proposed methods. Computer programs implementing the new methodology will be made available through an R Shiny App with a web interface for practitioners' easy access and convenient use. Detailed usage guides and documentation will be provided.