Quantitative decision-making in clinical drug development incorporating biomarkers

这一由MRC资助的博士培训伙伴关系(DTP)将尖端分子和分析科学与数据分析中的创新计算方法结合在一起，使学生能够在与行业保持一致的优先领域解决重要的应用生物医学研究问题。这是一个为期4年的课程，第一年包括一系列的教学模块和两个基于实验室的研究项目，这些项目导致了跨学科生物医学研究的硕士学位。前两个学期包括精选的教学模块，使学生能够在多学科的科学中获得坚实的基础。学生们还参加了由学术和行业专家主持的一系列大师课程，这些专家涉及分子、细胞和组织动力学、微生物学和传染病、应用生物医学技术以及人工智能和数据科学等领域。在第三学期和暑期，学生们在他们选择的实验室里进行两个为期十一周的研究项目。项目：新药和疗法的开发必须通过三个临床里程碑，即第一阶段(安全性评估)、第二阶段(初步疗效评估)和确认性第三阶段试验。这条管道的每个阶段都提供了有用的决策点，以便对积累的数据进行批判性评估，并使用合理的量化方法做出重要决定。考虑到高流失率，特别是在早期阶段，根据与特定开发阶段相关的指标做出正确的通过或不通过决策是至关重要的。传统上，这些依赖于假设检验框架，涉及使用统计检验来实现基于p值的相关重要性阈值。然而，考虑到很高的流失率，特别是在肿瘤学，只关注p值可能是违反直觉的。最近，已经提出了基于模型的替代方法，基于使用治疗效果的上下限可信区间，例如目标值(TV)是期望的效果，而下参考值(LRV)表示最小的临床有意义的效果。这些方法将观察到的结果的证据强度更有力地分类为3种结果：通过、不通过决定和考虑区域。拟议的项目将旨在通过进一步纳入将患者群体分为两个不同亚组的生物标记物来制定决策标准的概括性。这将通过模拟第二阶段生物标记物分层肿瘤学试验来实现，该试验的操作特性将被详细研究。这应该会改善使用各种临床终点和整个药物开发流水线的统计设计的量化决策。

This MRC-funded doctoral training partnership (DTP) brings together cutting-edge molecular and analytical sciences with innovative computational approaches in data analysis to enable students to address important applied biomedical research questions in priority areas aligned with industry. This is a 4-year programme whose first year involves a series of taught modules and two laboratory-based research projects that lead to an MSc in Interdisciplinary Biomedical Research. The first two terms consist of a selection of taught modules that allow students to gain a solid grounding in multidisciplinary science. Students also attend a series of masterclasses led by academic and industry experts in areas of molecular, cellular and tissue dynamics, microbiology and infection, applied biomedical technologies and artificial intelligence and data science. During the third and summer terms students conduct two eleven-week research projects in labs of their choice. Project:Development of new drugs and therapies must pass three clinical milestones, Phase 1 (evaluation of safety), Phase 2 (initial evaluation of efficacy), and confirmatory Phase 3 trials. Each stage of this pipeline provides useful decision points to critically evaluate the accumulated data and make important decisions using sound quantitative methods. Given the high attrition rate, especially in early phases, it is paramount to make correct Go or No-Go decisions based on metrics relevant to that particular stage of development. Traditionally, these have relied on a hypothesis testing framework involving the use of statistical tests to achieve the relevant thresholds of significance based on p-values. However, given the high attrition rate especially in oncology, focusing on p-values alone can be counter-intuitive. Recently, alternative model-based approaches have been proposed based on using the upper and lower confidence intervals of the treatment effect such as the Target Value (TV) being the desired effect whereas the Lower Reference Value (LRV) representing the smallest clinically meaningful effect. Such methods provide a more robust categorisation of the strength of evidence of observed results into 3 outcomes: Go, No-Go decision and Consider zone. The proposed project will aim to formulate a generalisation of the decision-making criteria by further incorporating a biomarker that splits the patient population into two distinct subgroups. This will be achieved by simulation of a Phase II biomarker stratified oncology trial, whose operating characteristics will be studied in detail. This should improve quantitative decision-making using a variety of clinical endpoints and statistical designs across the drug development pipeline.