ABSTRACT
Many people in business and medicine regard statistics as at best a nuisance, and at worst a hinderance to science. For example, Einstein liked to say that ‘God doesn’t gamble.’ [54]. That is likely true in the long run (in statistical terms, in the limit), but in the short term while making drugs, we cannot operate with complete certainty and have to depend upon statistics to guide us in making safe, effective, quality products. In clinical drug development, statistics are used to quantify the uncertainty associated with human use of drugs - not to eliminate uncertainty. In a perfect world, it would be perfectly clear whether to use a drug or not (the drug is either safe, effective, and made well or it is not), but in practice, statistics are used to measure the outcome of studies used as tools to assess these properties. If they are not used well, the trend toward increased length and cost in drug development [137] are very likely to continue. Clinical pharmacology and many aspects of drug development are evolv-
ing, and will continue to do so. These changes are good as they would be expected to improve the drugs that are produced for the people who need them. Statistically, changes such as these represent new challenges, but the raw materials to meet the needs of the science are available. Change is not so bad once you get used to it. The future of Statistics in Clinical Pharmacology lies in learning (not
confirming). Design of studies and bioequivalence analyses can be automated by sponsoring companies themselves or by commercial software companies. Indeed, some software companies already claim to do so, and it is to be expected that more will appear in the future. This frees up statisticians to spend more time on the parts of drug development that are in need of attention. Confirmatory work like bioequivalence testing should soon no longer be an activity directly involving statisticians but will be handed off to clinical scientists and pharmacokineticists with statisticians only being consulted as needed. To reiterate, the future of statistics in clinical pharmacology lies in
other areas - in particular, better understanding of quantitative aspects of safety and efficacy assessment in Phases I and IIa. Better control of these should lead to less Phase III portfolio attrition. Particularly, safety in the use of drug products could use some quantitative enhancements.
