In the pharmaceutical industry, the process of developing a new medicinal entity (NME) is associated with considerable risk and uncertainty. The present paper investigates how variability in the decision making process in pharmaceutical industry research and development (R&D) projects influences the expert group size needed to take informed go/nodecisions. In other words, we wanted to investigate how go/no go decision process is influenced by the degree of inherent project uncertainty. Since the information gap during the early as well as late phases of the R&D process of an NME is commonly prominent, we made a statistical forecasting of the impact of individual variations in go/no-go judgments and decision making. In respect to inherent project uncertainty, throughout the discovery/development process of 8 – 10 formal go/no-go steps, we then simulated the size of the expert group size needed to take a meaningful and coherent group decision based on individual real expert judgments.
In the study, we used data from 52 experts in the pharmaceutical industry and allied sectors, making a series of go/no-go judgements in 4 different drug R&D case scenarios derived from real R&D cases from the pharmaceutical industry. The 4 case-scenarios related to go/no-go judgment decisions over phases of drug discovery (early) as well as development (late) ranging from target selection/pharmacology, toxicology, biopharmacy/galenics to clinical development/market introduction.
Based on the intrinsic variability found in the real go/no-go (or go/no-go and recycle) judgments made by the experts, as each of the 4 cases gradually evolved, we modelled the impact of the real intrinsic judgment variability found, on a fictive R&D decision chain based on 10000 bootstrap samples involving between 5 -20 different major go/no-go decisions. We found, that when serial mean judgements were expressed as clear go or stop, some 10 – 15 experts were needed in order to arrive at a coherent go or stop group decision. However, when both go and stop (recycle) mean decisions were encountered at any step in the process development, the number of experts needed in a group to arrive at a decisive go or no-go judgment tended to be very large. In spite of the fact that large groups were involved, there was a substantial inherent uncertainty that remained in the decision chain.
We conclude that in pharmaceutical industry R&D, rational decision making from initial drug discovery to late development and marketing, can commonly be managed by a group of 10 – 15 experts when mean group judgements over a series of decision points are clear go decisions. However, when mean group judgements from one decision point to another varies from go to stop in a specific case, i.e. involves a recycle component, there will be a need to expand R&D expert input substantially. In such cases the drug development process more or less takes on the form of an open innovation process. The present research findings may be used to construct a new model on how to plan and model the size of expert input in structured decision processes similar to those practised in the pharmaceutical industry.