ABSTRACT
It is important to recognize drug development as a decision-based science. Current evidence suggests that this recognition is not yet in place as inadequate consideration is placed on the consequences of wrong decisions in any part of the process, and that any such wrong decision will contribute to the failure rate. Evidence for this oversight is illustrated by examining the resource distribution in any pharmaceutical company, academic laboratory, or research division. Data gathering, as a generic term for the design, execution and collection, as well as storing of data in a given experiment, is the
focus of most of the scientific research. Furthermore, planning, funding, personnel, and even visibility focus on these activities. Then, what proportion of time is spent conducting these data gathering activities versus the time spent conducting the analysis and performing a careful synthesis of the information? How carefully are decisions made from the data scrutinized and discussed among scientists before a final decision is taken? In almost all the research settings, it is obvious that this is the last, most hurried, and least collaborative activity. In many research settings, the individual best trained to perform the analyses, the statistician, is not even involved. This is especially, and perhaps unitarily, true in the early stage of research. For some reason, error in these early stages is tolerated, even expected. Then, when the paradigm shifts and the experiments are carried out for registration purposes (phase III clinical studies bound for registration/submission), the team would be ready to put more serious effort into analysis, to ensure correct decision making. In some sense, they are forced by regulatory agencies to ensure that they make correct decisions, and statisticians suddenly find themselves consulted on the study design, monitoring, and final analysis as predefined by an approved analysis plan. This gives rise to the following questions: Are the experiments carried out early in the R&D process that impact compound selection, dose selection, safety assessment, or preclinical efficacy any less important? Is error in the decision making more tolerable in these early stages? Is the magnitude of such error smaller in these early stages? Do most of today’s research organizations quantify the level of error or manage it to be the customary 5% Type I and 20% Type II levels typical in the late stage research? Is there any consideration given to the presence of such errors in these early stages? Are optimal scientific designs being implemented in these early phases? Unfortunately, for the majority of pharmaceutical companies as well as academic laboratories, the answers to all the above questions are negative. It is our personal belief that this is a significant contributor to the high failure rate we observe under the current processes.
