Biostatistical evaluation of focal hepatic preneoplasia.

Abstract

Qualitative analyses of focal hepatic preneoplasia are relatively easy and fast but hypothesis tests based on these analyses often lack statistical power. Evaluating focal hepatic preneoplasia quantitatively, on the other hand, requires more effort but is rewarded by an increased ability to detect differences between treatment groups and by the possibility to investigate the mechanism of a treatment under study. Due to the stereological problems inherent in the data a statistical analysis that concentrates on the evaluation of area fraction will provide clear results whereas the analysis of focal transection density and size distribution can produce misleading results. In addition, the area fraction is a valid variable even in the presence of confluent foci. The number and size distribution of focal transections in liver sections cannot be directly translated to the number and sizes of foci in the liver. As no general statements about the relationship between focal transection density and foci density as well as between focal transection size and foci size distribution can be made, there is need for a parametric mechanistic model to link the number and size distribution of focal transections to those of the underlying foci. The stereological problem therefore can be avoided by introducing a model for foci appearance and change of volume that then can be used to conclude whether the treatment induces foci and whether it changes their volume.