Data analyses, interim reviews and design issues of longitudinal study in biomedical research

许多生物医学研究关注个人或群体随时间的发展。本研究的主要目的是发展新的理论和方法，用于分析纵向研究的数据以及研究设计和监测。由于现有的治疗方法、生物学知识和护理标准正在迅速发展，因此对预期治疗效果的性质和程度以及结局指标中某些类型差异的临床意义了解相对较少，因此在许多健康研究中需要采用适度模型假设的方法。第一个具体目标是进一步发展强有力的统计方法，从收集的纵向数据中量化时间趋势和变量之间的关联。临床试验设计中的伦理和经济考虑，以评估新疗法的疗效和可能的毒性作用，以及纵向试验期间信息流的固有顺序性质，是对监测研究和提供适应性设计研究的组序贯方法越来越感兴趣的原因。我们的第二个具体目标是填写临床试验中正式审查的重复置信带方法的开发，并提供有效适应现有信息的研究设计。第三个具体目标是在纵向响应、可能不完整和大量解释变量的情况下开发回归分析方法。这是出于科学兴趣，即建立疾病进展和随时间推移的临床结果与患者的生物标志物或基因表达数据之间的关联。研究结果，说明其应用到最近的医学研究，将通过出版物，会议和软件分发传播给医学调查人员和生物统计从业人员。此外，参与的学生将做好充分的准备，以满足在实践中，特别是在医疗环境中的统计支持的各种需求。

Many biomedical studies are concerned with the development of persons or groups over time. The broad objective of this research is to develop new theory and methods for analyzing data from longitudinal study and for study design and monitoring. Since available therapies, biological knowledge, and standard of care are rapidly evolving and thus relatively little may be known about the nature and magnitude of the expected treatment effect and about the clinical significance of certain types of differences in the outcome measure, approaches with modest model assumption are desirable in many health studies. The first specific objective is further development of robust statistical approaches to quantifying the time trend and association between variables over time from the collected longitudinal data. The ethical and economic considerations in the design of clinical trials to evaluate the efficacy and possible toxicity effects of new therapies, together with the inherent sequential nature of the information flow during the longitudinal trials, are responsible for the increasing interest in group sequential methods for monitoring the studies and for providing studies with adaptive designs. Our second specific objective is to fill out the development of the repeated confidence bands approach for formal reviews in clinical trials, and to provide study designs with efficient adaptation of the available information. The third specific objective is to develop regression analysis methods in the situations with longitudinal response, possibly incomplete, and a large number of explanatory variables. This is motivated by the scientific interest in establishing the association between disease progression and clinical outcome over time and patient's biomarker or gene expression data. The research results, illustrated with their applications to recent medical studies, will be disseminated to medical investigators and biostatistical practitioners via publications, conferences and software distribution. In addition, the students involved will be well-prepared to meet the various demands of statistical support in practice, particularly in medical settings.