Transforming clinical research and development through the exploitation of patient level data sources

目前，90%以上正在进行临床试验的新疗法未能进入市场，而开发每一种成功新药的成本估计约为30亿美元，生命科学行业迫切需要降低研究失败率，以实现更好的投资回报，并降低医疗干预的成本。虽然有些失败是由于新药本身不安全或无效，但有些失败是由于用于测试新药候选效果的临床试验设计不佳。通过有效的前瞻性临床试验计划，可以完全避免不良的研究设计。我们的目标是通过改进研究设计和加速将新疗法推向市场的过程来减少临床试验失败。为了实现这一目标，该项目将确定患者级别的数据源，并使用它们为关键治疗领域（如癌症、呼吸系统疾病和罕见疾病）创建一个数字图书馆，用于智能软件信息生态系统，扩展我们现有的临床研究计划工具KERUS Cloud的功能。KERUS Cloud通过允许研究人员生成研究模拟来了解不同因素如何影响研究结果，从而优化临床研究设计，从而选择最佳设计。这个软件已经被证明，当使用这个工具设计时，它可以将实际研究的成功率提高40%。基于真实患者数据的信息库将进一步改进这一过程，因为它允许将不同疾病类型的数据也输入到KERUS Cloud生成的模拟中，使它们成为最真实的研究模拟。在此基础上，后期的开发阶段将涉及在生态系统中添加机器学习元素，该元素可以将模拟结果与实际结果进行比较，以确定导致实际试验中可变性的关键因素。这些信息将被我们的KERUS云软件用于设计下一个研究，提高其成功的机会。然而，这里概述的项目将确定合适的数据源，并生成准备和清理数据所需的算法，用于创建KERUS Cloud使用的数字数据库，以通知研究模拟。这种利用真实的患者级数据和机器学习为试验设计提供信息的方法将彻底改变这一过程，为真正的试验取得成功提供动力。

With over 90% of new medical treatments undergoing clinical trials failing to reach the market and the cost of developing each new successful drug now estimated to be around US$3 billion, there is an urgent need for the life science industry to reduce their study failure rates to achieve a better return on investment and bring down the cost of medical interventions. Whilst some failure is down to the novel drugs themselves being unsafe or ineffective, some failure is a result of poor design of the clinical trials used to test the effects of new drug candidates. Poor study design is completely avoidable with effective prospective clinical trial planning.Our aim is to reduce clinical trial failure by improving study design and accelerating the process of bringing new treatments to market for patients. To achieve this, this project will identify patient-level data sources and use them to create a digital library for key therapeutic areas such as cancer, respiratory and rare diseases for use within an intelligent software information ecosystem, extending the capability of our existing clinical study planning tool, KERUS Cloud. KERUS Cloud already optimizes clinical study design by allowing researchers to generate study simulations to understand how different factors can affect a study's outcome, so that the best design is chosen. This software has already been shown to improve a real study's success when designed with this tool by 40%. A library of information that is based on real patient data will improve this process further by allowing data for different disease types to also be fed into the simulations generated by KERUS Cloud, making them the most realistic study simulations yet possible. Following on from this, a later development stage will involve adding a machine learning element to the ecosystem that can compare simulations with real outcomes to identify the key factors that account for the variability seen in real trials. This information will then be used by our KERUS Cloud software in designing the next study, improving its chances of success. However, the project outlined here will identify suitable data sources and generate the algorithms required to prepare and clean the data for use in creation of the digital data libraries to be used by KERUS Cloud to inform study simulations. This approach of harnessing real patient-level data and machine learning to inform trial design will revolutionize the process, powering real trials for success.