On reclaiming data from the literature: Literature data "R and R"(recovery and reanalysis)

It is common practice to perform a literature review on a selected topic in order to accomplish one or more goals (e.g., preparation of a seminar, lecture, grant application, review paper, etc.). A critical analysis of the literature often requires recovering data from a publication. The latter can be a challenge when data are presented graphically, especially when non-uniform or non-linear (e.g., logarithmic) scales are used. The recovered data then provides a basis for reanalysis and reinterpretation. The purpose of this communication is to discuss methods available for recovering graphical data from the literature with emphasis on scanning and digitization procedures. Data from graphs in several publications have been recovered after scanning the graph into a file and using a software program which permits on-screen digitization. Only publications which had a tabular listing of the plotted values were examined in order to provide some estimate of the accuracy of the digitization process. Recovered values are accurate and rarely differ by more than about 2-3 percent from the correct (i.e., literature) value. Furthermore, the between-investigator range in the recovered values is small (based upon digitization by 11 investigators). The utility of data recovery is shown with a literature example. Data in one publication afforded the opportunity to extend the authors' original analysis to create a pharmacokinetic/pharmacodynamic model to describe cocaethylene (ethylcocaine) in the dog. We conclude that scanning of literature graphs followed by digitization, provides a means of accurately reclaiming data from the literature and offers a unique opportunity to reanalyze the data and, perhaps, to generate a different perspective.