Analysis of Unbiased Histopathology Data from Rodent Toxicity Studies (or, Are These Groups Different Enough to Ascribe It to Treatment?)

It is a common problem to distinguish a minor treatment change from background variation, especially when establishing no observed effect levels. Toxicological histopathologists use a wide range of methods at the microscope for comparing groups to help them form their opinions. Although the data produced by these methods can be subjective, all of these methods produce data that can be formally analyzed to give an objective, probabilistic result, provided the observations are unbiased. Other important experimental disciplines make extensive use of completely subjective data to produce objective results, for example, clinical trials using patients' symptoms. It is argued here that pathological experimental data too should be analyzed before an expert opinion (along with the objective evidence for that opinion) is formally reported. The Ordering Method, based on ranking the severity of a putative toxic change, is the most sensitive, robust, and analytically flexible method currently available.