Earlier studies have shown that morphometric analysis of nuclear size and shape features and stereologic analysis of gland architecture in endometrial hyperplasia are useful to predict the risk of progression to cancer. A multivariate discriminatory D-score using these features has a higher sensitivity than qualitative microscopic characteristics (nuclear atypicality and glandular complexity) when an appropriate threshold is chosen. In the present study, the (additional) prognostic value of morphometric features related to the arrangement of nuclei in glands (distance of nuclear centroids to basal membrane, angle of longest nuclear axis with basal membrane) and DNA-ploidy (determined by flow cytometry) has been analyzed in 39 cases of endometrial hyperplasia, seven (18%) of which have progressed to cancer. Variation in stratification of nuclei has prognostic value, in contrast to DNA-ploidy. Multivariate analysis selects the outer surface density of the glands, mean distance, volume percentage of lumen and the coefficient of variation of nuclear axes as having separate and additional value. A multivariate rule called HYPER-score results from these features. With this score a high-risk (score value >/=0.20, nine cases, 78% progressed) and a low-risk group (score value <0.20, 30 cases, none progressed) can be discerned. The prognostic value, sensitivity and specificity of the hyper-score greatly exceeds that of any of the previously described prognostic factors in endometrial hyperplasias. An additional clinical advantage is that with the HYPER-score two groups (instead of four groups as with Kurman's method(1)) with a high sensitivity and specificity can be discerned. The morphometric and stereologic features agree well with the usual qualitative predictors but have much stronger prognostic value.