Methods for Post Marketing Surveillance and Comparative Effectiveness Research

食品和药物管理局授权的哨兵倡议将导致大量的 正在计划的上市后研究，需要分析和组合来自几个不同的 问题研究拟议的项目将通过开发新的灵活方法来应对这一挑战， 使用各种模型的荟萃分析，包括二进制和离散数据的模型， 纵向数据和事件发生时间数据模型。还将解决的一个相关问题是设计， 样本量和使用这些类型的荟萃分析模型的功效考虑。这些模型和数据 收集的上市后数据在设计未来的临床研究（如非劣效性， 等效性和优效性癌症临床试验。拟议的项目还将开发元分析方法， 诊断测试的研究，以促进循证医学。我们还将创造灵活和 用于准确比较不同药物的癌症罕见不良事件发生率的可靠方法， 确定这些比率如何受到重要预后因素的影响。拟议项目还将 探索分析大型癌症数据集的统计方法，以校准治疗剂量， 存在潜在的冲突因素，例如寿命和生活质量以及经济成本。我们将 严格探索这些权衡，使用传统上用于分析的基于效用的方法， 人口层面的健康政策。拟议的统计方法将广泛适用于 III期临床试验和上市后研究中出现的复杂、大规模数据集。

The Sentinel Initiative mandated by the Food and Drug Administration will lead to an enormous number of studies being planned post-market that will require analyzing and combining data from several different studies. The proposed project will address this challenge through developing new and flexible methods for meta-analysis using a variety of models, including models for binary and discrete data, models for longitudinal data, and models for time-to-event data. A related issue that will also be addressed is design, sample size, and power considerations using these types of meta-analytic models. Such models and data collected post-market can be quite useful in designing future clinical studies such as non-inferiority, equivalence, and superiority cancer clinical trials. The proposed project will also develop methods for metaanalytic studies of diagnostic tests to facilitate evidence-based medicine. We will also create flexible and robust methodology for accurately comparing rare adverse event rates in cancer for different drugs and for determining how those rates are affected by important prognostic factors. The proposed project will also explore statistical methods for the analysis of large cancer data sets for calibrating treatment dose in the presence of potentially conflicting factors, such as length and quality of life and economic costs. We will explore these tradeoffs rigorously, using a utility based approach traditionally employed in the analysis of health policy at the population level. The proposed statistical methodology will be broadly applicable to complex, large scale, data sets arising in phase III clinical trials and post-marketing studies.