Pharmacology and statistics: recommendations to strengthen a productive partnership.

Critical to the discovery, development and rational use of drugs and vaccines are the foundational principles and proper application of statistics. However, in too many cases, there has been misuse of statistics and/or overemphasis on statistical significance (p < 0.05), as though this criterion possessed truth-guaranteeing properties. To clarify confusion about the proper use of statistics in pharmacology, we summarize briefly the foundational principles of probability; the role of statistics in assessment of causality; the three basic uses of statistical methods, especially those employed in hypothesis testing; and current statistical issues in pharmacological research. We then review and provide examples of the meaning of statistical significance, the consequences of lack of randomization in epidemiology/observation studies, the criteria for measurement instrument validation, the problems with subgroup analyses, the need for multiple comparison statistical methods, and how to handle dropouts and missing data. Finally, based on sound experimental and statistical principles, we make a series of recommendations to both experimentalists and journal editors to improve published pharmacological experiments. These include widespread use of blinding and randomization and/or random selection of subjects in both basic and clinical pharmacology, mandatory use of rigorous evidentiary criteria in epidemiology/observation studies claiming causal associations, proper interpretation of statistical versus clinical/pharmacological significance, appropriate interpretation of meta-analyses, meaningful validation of methods, and a more rational statistical approach to subgroup analyses and genetic association studies.