Abstract

Clinical patient management is dynamic. It is not based on a single decision but on a sequence of decisions, with adjustments of therapy overtime, where adjustment are personalized to individual patients. However, strategies allowing for such adjustments are infrequently studied. In the treatment of serious bacterial infections, there are two therapeutic decision points: empirical therapy when the patient is first seen by the clinician, and definitive therapy when drug susceptibility results are available to help guide therapeutic choices. COMparing Personalized Antibiotic StrategieS (COMPASS) is a trial design that compares strategies consistent with clinical practice rather than specific treatments. A strategy is a decision-rule guiding treatments of bacterial infections both at the empirical and at the definitive stages. Sequential multiple assignment randomized (SMART) COMPASS allows evaluation of strategies when there are multiple definitive therapy choices. SMART COMPASS is a pragmatic design, mirroring antibiotic treatment decision-making as they unfold in clinical practice and addressing the most relevant question for treating patients: identification of the patient-management strategy that optimizes ultimate patient outcomes. Complex statistical challenges are encountered during the design of SMART COMPASS including how to: appropriately estimate effects and associated standard errors within the context of sequential randomization, identify the best of several strategies while controlling trial-wise error, and calculate sample size and power when trying to identify the optimal strategy. Design considerations for SMART COMPASS are discussed.

摘要

临床患者管理是动态的。它不是基于单个决策，而是基于一系列决策，并随着治疗时间的推移而进行调整，其中调整是针对个别患者的。但是，很少研究允许这种调整的策略。在严重细菌感染的治疗中，有两个治疗决策点：临床医生首次见到患者时的经验性治疗，以及可获得药物敏感性结果以帮助指导治疗选择的确定性治疗。比较个性化抗生素策略（COMPASS）是一项试验设计，比较了符合临床实践而非特定疗法的策略。一种策略是在经验阶段和确定阶段都指导细菌感染治疗的决策规则。顺序多重分配随机（SMART）COMPASS可在有多种确定的治疗选择时评估策略。SMART COMPASS是一个务实的设计，反映了抗生素治疗决策在临床实践中的进展，并解决了与患者治疗最相关的问题：确定可优化患者最终治疗效果的患者管理策略。在SMART COMPASS的设计过程中遇到了复杂的统计挑战，包括如何：在顺序随机化的背景下适当估计效果和相关的标准误差，在控制试验误差的同时确定几种策略中的最佳方法，并在尝试时计算样本量和功效确定最佳策略。讨论了SMART COMPASS的设计注意事项。