Statistical Methods for Clinical Studies

项目摘要/摘要：对癌症的遗传和生化机制的了解的增加，导致了癌症诊断和分类的新技术，现在正在开发一系列可能对具有特定分子属性的癌症有效的治疗方法。这些新的治疗方法为临床试验提供了改进设计和数据适应性分析方法的机会和必要性。具体而言，本研究将考虑以下方面：1)新的靶向治疗的II期和III期研究。一些新的抗癌药物提供的临床益处因疾病的靶表达而异；因此，需要更好的设计来避免错过有前途的药物。策略将包括子组的联合测试和收缩方法。2)探索患者预后的自适应回归方法。涉及靶向治疗的新研究结果的复杂性要求更好地理解遗传属性与治疗效果之间的关系。为不同预后和治疗效果的患者亚组构建规则的计算方法将被评估。3)纵向标记过程数据。还需要改进的方法来了解顺序测量的生物标志物及其对治疗的影响和相互作用的关联。我们将考虑因果模型构建，以估计存在潜在时间依赖性混淆的生物标志物对患者预后的影响。还将实施软件，以方便使用作为本提案一部分制定的方法。评估降低癌症死亡率和发病率的新干预措施具有重大的公共利益。在过去的几年中，分子靶向治疗的发展和潜在生物标志物的鉴定取得了快速进展。对这些新的治疗方法和生物标志物进行严格和有效的评估，以更好地为患者服务，并扩大对这些复杂疾病的认识，这一点至关重要。公共卫生相关性：本提案的主要重点是开发适合单独使用或与其他当前癌症治疗联合使用的靶向药物的设计和分析方法。我们将开发和评估灵活的临床试验设计的操作特点，包括基于分子属性的生物异质性。我们还将研究自适应统计算法，以模拟患者的结果，并确定可能从这些新治疗中获益最多的患者群体。

DESCRIPTION (provided by applicant): PROJECT SUMMARY/ABSTRACT Increased understanding of the genetic and biochemical mechanisms of cancer has led to new technologies for diagnosis, classification of cancers and now to the development of an array of treatments that may have efficacy for cancers with specific molecular attributes. These new treatments provide both the opportunity and necessity to develop improved designs and data adaptive analysis methods for clinical trials. Specifically, this research will consider the following: 1) Phase II and Phase III studies for new targeted treatments. Some new anticancer agents offer clinical benefits that vary with respect to target expression of the disease; therefore, better designs are needed to avoid missing promising agents. Strategies will include joint testing of subgroups and shrinkage methods. 2) Adaptive regression methods for exploring patient outcome. The complexity of results from new studies involving targeted therapy demands a better understanding of the relationships between genetic attributes and treatment efficacy. Computational methods that construct rules for patient subgroups with differing prognoses and treatment efficacy will be evaluated. 3) Longitudinal marker process data. Improved methods are also needed to understand the association of sequentially measured biomarkers and their impact and interactions with respect to treatment. We will consider causal modeling constructs to estimate effects of biomarkers in the presence of potentially time-dependant confounding on patient outcome. Software will also be implemented to facilitate the use of methods developed as part of this proposal. The evaluation of new interventions to reduce mortality and incidence of cancers is of significant public interest. Over the last few years there has been rapid progress in the development of molecular targeted therapies and in the identification of potential biomarkers. It is crucial that these new treatments and biomarkers be evaluated in a rigorous and efficient manner to best serve patients and to expand knowledge of these complex diseases. PUBLIC HEALTH RELEVANCE: The major focus of this proposal is the development of design and analysis methods appropriate for targeted agents used alone or in combination with other current cancer therapies. We will develop and evaluate the operating characteristics of flexible clinical trial designs which incorporate biologic heterogeneity based on molecular attributes. We will also study adaptive statistical algorithms for modeling patient outcome and for identifying of groups of patients who may benefit most from these new treatments.