Integrated Planning of Drug Development Programs

在临床研究中，前期研究和后续验证性研究的规划是密切相关的。这就产生了同时规划的想法，目的是确保预先确定的成功概率。具体来说，问题出现在为试点研究选择样本量的最佳策略，以及是否以最经济的方式利用总资源继续进行验证性研究（“进行/不进行的决定”）的决策规则。这可以通过最大化预先指定的效用函数的期望值来实现，其中包括项目成本（试点/验证研究的固定和可变每位患者成本），未来收益和某种开发风险（成功项目的期望值），超过试点研究的样本量和进行/不进行决策规则的阈值。在正在进行的DFG研究项目“临床研究和临床研究的综合规划”中，得出了一些方法，以便能够充分地映射临床研究中的几种情况。例如，可以计划具有事件时间、二元或正态分布终点、多组、多个共同主要终点或偏倚校正治疗效果估计的研究。为了便于实际实施，相应的生物统计方法在编程语言R中实现。到目前为止，提供了用户友好的R Shiny应用程序（“Apps”）的原型版本和R包（drugdevelopR）以进行最优规划。应用程序和软件包可通过https://web.imbi.uni-heidelberg.de/drugdevelopR/.It访问，希望提供一个充分验证和专业质量保证的药物开发软件版本，用于临床研究。因此，申请该项目的目的是延长DFG研究项目的成功工作，以增加研究软件drugdevelopR的功能、可用性和最重要的质量保证（“适合用途”），以保证可持续使用并转移到研究项目之外的新研究环境（“适合重用”）。药品开发人员的质量保证（例如，确认、验证、测试和审核）和增强其用户友好性是通过遵循“质量保证概念”来实现的。这包括归档、版本控制、bug报告/跟踪和源代码文档的措施。软件描述的概念规定将原型软件转换为专业的CRAN版本（包括例如详细的用户手册和/或教程），并在同行评审的期刊上至少发表一篇文章。GitHub.com上的开源出版物和适当的drurgdevelopr软件许可模型（MIT许可）旨在确保学术使用和开发是免费的。

In clinical research, the planning of pilot and subsequent confirmatory study are strongly linked. This leads to the idea of simultaneous planning with the aim of ensuring a predetermined probability of success. Specifically, the question arises as to an optimal strategy for choosing the sample size for the pilot study and the decision rule as to whether or not to proceed with conducting the confirmatory study (“go/no-go decision”) with the most economical use of the total resources. This can be done by maximizing the expected value of a pre‐specified utility function – which includes costs of the program (fixed and variable per-patient costs for pilot/confirmatory study), future benefit and some kind of development risk (expected value of a successful program) – over the sample size of the pilot study and the threshold value for the go/no-go decision rule. In the ongoing DFG research project “Integrierte Planung von Pilotstudien und konfirmatorischen Studien in der klinischen Forschung” methods were derived so that several situations in clinical research could adequately be mapped. For example, studies with time-to-event, binary or normally distributed endpoint, multiple arms, multiple co-primary endpoints or bias adjusted treatment effect estimates can be planned. In order to facilitate the practical implementation, the corresponding biostatistical methods were implemented in the programming language R. Up to now, prototype versions of user friendly R Shiny applications (“Apps”) and an R package (drugdevelopR) are provided for optimal planning. The Apps as well as the package are accessible via https://web.imbi.uni-heidelberg.de/drugdevelopR/.It is desirable to provide an adequately validated and professionally quality assured version of the drugdevelopR software for use in clinical research. Therefore, the aim of the project to be applied for is to prolong the successful work of the DFG research project in order to increase the functionality, usability, and – most importantly – the quality assurance (“Fit for Purpose”) of the research software drugdevelopR to guarantee sustainable use and transfer to new research contexts beyond the research project (“Fit for Re-Use”). Quality assurance (i.e. e.g. validation, verification, testing and audit) of drugdevelopR and enhancement of its user friendliness is to be achieved by following a “quality assurance concept”. This includes measures for archiving, versioning, bug reporting/tracking and source code documentation. The concept for the description of the software stipulates the transformation of the prototype software to a professional CRAN version (including e.g. detailed user manuals and/ or tutorials) and provision of at least one publication in a peer-reviewed journal. The open-source publication on GitHub.com and an appropriate licensing model (MIT License) of the durgdevelopR software are intended to ensure that the academic use and development is free of charge.