Letter to the Editor

Dear Editors,

The purpose of this letter is to point out 3 problems I noted in a manuscript published by Andrews and colleagues in the Sept 20, 2020 issue of JAALAS.¹

First, the authors report that 3 experienced individuals scored rabbits for pain using a Grimace Scale.⁸ Although the scorers were blind to the treatment that each rabbit had received, the authors did not provide any information regarding inter-rater consistency. Without this assessment, it cannot be determined whether rater bias or inter-rater error has skewed the results, as discussed in relevant statistical literature.^2,7

Second, the authors state in the statistical methods that a trend was declared if the P value was less than 0.10 but greater than 0.05. The term trend has several definitions that are all rooted in identifying temporal or spatial changes in some value. As applied in this paper, the term trend and the use of a P value of less than 0.10 to identify a trend is not valid.^4,5 To determine if a trend exists in a data set, a trend analysis must be conducted. Numerous statistically valid methods can be applied for this purpose.^6,9,11 The only conclusion that can be drawn from a P > 0.05 in the analysis that was used is that the probability is greater than 1 in 20 that a difference between groups this size or larger might occur by chance alone.

Third, figures 2 through 5 have error bars that are not defined as far as I can tell. These bars could represent standard deviation, standard error, or even range. If these data are intended as a basis for dosing recommendations, then reporting confidence intervals (CI) would be most appropriate and helpful.^3,10 Providing CI would allow readers to determine dosing concentrations and intervals that would provide appropriate analgesic blood levels of drug over time for the most animals.