Statistical graphics in pharmacokinetics and pharmacodynamics: a tutorial.

Abstract

OBJECTIVE: To discuss the use of statistical graphics in the analysis of pharmacokinetics and pharmacodynamics data. METHODS: Information on graphic techniques and their application was retrieved from a MEDLINE search (January 1980-March 1997) of the English-language literature and bibliographic reviews of review articles and books. Data used to generate plots were extracted from some new drug applications submitted to the Food and Drug Administration and by simulation. DATA SYNTHESIS: In carrying out data analysis, we should look at data in several ways, construct a number of plots, and do several analyses, letting the results of each step suggest the next. The information from a plot should be relevant to the goals of the analysis. Thus, in choosing a graphic method, it is necessary to match the capabilities of the method to the need in the context of the application. For example, if linear relationships among variables in a set of multidimensional data are relevant, scatter plots such as the pairs plot with smoothing is likely to be more informative than other graphic methods. It is necessary to recognize what kinds of perceived structure are attributable to the data, and what kinds are artifacts of the display technique itself when using graphs for data analysis. CONCLUSIONS: Graphic techniques enable the data analyst to explore data thoroughly, look for patterns and relationships, confirm or disprove the expected, and discover new phenomena. An important element of statistical graphic techniques is flexibility, both in tailoring the analysis to the structure of the data and in responding to patterns that successive steps of analysis uncover. Statistical graphics can and should be used to enhance numeric statistical analyses.