Implementation and pratical aspects of quantitative decision-making in clinical drug development

TL;DR

This paper discusses the practical implementation of quantitative decision-making (QDM) in clinical drug development, covering its principles, applications at various levels, and insights from a collaborative industry-academia group.

Contribution

It provides a comprehensive overview of QDM principles, frameworks, and best practices for implementation in clinical drug development based on expert discussions.

Findings

QDM can be applied at study, project, and portfolio levels.

Common principles for QDM implementation include framework construction and communication.

Discussions highlighted variability in methodologies and terminology.

Abstract

Quantitative decision-making (QDM) principles address the issues related to the mapping of results to decisions, the synthesis of information and the quantification of uncertainty. Since the clinical drug development involves a succession of decisions to be made, QDM methods can be applied at various levels. At the study level, it can be used to properly design a study, and improve the decisions that are made either during the trial or at its end. Establishing decision criteria ahead of the study is essential here to address the need for speedy decisions, potentially in real time. At the project level, QDM can be used to inform decisions to continue, adapt or stop a drug development programme based on results from previous studies. At the portfolio level, QDM can be used to choose, prioritise and optimise the development portfolio, e.g. using the probability to reach market access or target sales within a predefined timeline. The increasing interest in QDM and its statistical nature led in 2017 to the development a cross-industry and academia Special Interest Group on QDM within the Society and the European Federation of Statisticians in the Pharmaceutical Industry (PSI and EFSPI). The activities of the group included discussing QDM examples, some published in the literature and some real anonymised ones covering several settings. While the methodologies, and to some extent the terminology, employed also varied depending on the context, discussions within the group distilled common principles to be considered when implementing QDM, particularly around the construction of QDM frameworks, assessment of operating characteristics and communication with the clinical team. The present manuscript presents those points to consider, hoping they can be helpful to statisticians interested in implementing QDM.