A pilot study on the frequency structure of histological neoplasm diagnoses in rats

There are large differences in the frequencies of different kinds of neoplasms in a human or animal population. The question arises, whether the set of these frequencies shows some characteristic feature. Our fitting results on neoplasm frequency data relating to laboratory rats show that the frequencies are approximately lognormally distributed. At the same time, fitting results with the logarithmic series distribution, also frequently used in similar studies, are poor. A good fit of the Zipf-Mandelbrot distribution, fitted to the descendingly ordered dominant and subdominant frequencies can be achieved, sometimes after omitting some diagnoses. We point out the possibility that the omitted frequencies may be considered "unnatural" ones. The good fit of a particular frequency distribution to the diagnosis frequency set suggests a corresponding chance mechanism in forming the occurrence probabilities of different kinds of neoplasms. A frequently used scalar feature of frequency distributions of categorical data is the concentration or diversity. It was found that in the female rat population the diagnoses are more concentrated among the diagnosis categories. A possible explanation may involve the fact that females mature faster than males. The study of the distribution properties of the diagnosis frequencies promises the observation of new epidemiological phenomena.