Grading of prostatic cancer: III. Multivariate analysis of prognostic parameters.

This work is based on the two previous publications which are concerned with grading prostatic carcinoma. In the present communication the technique of multivariate analysis is applied to quantify the prognostic importance of parameters which were earlier identified as contributing in a significant manner to grading. The results of this analysis are used to construct a scoring system which differentiates five significantly different prognostic groups. When compared to the results of grading, the new system is shown to be superior. The large group of 228 patients with G2 tumors can be split into three groups, each with a significantly different prognosis. The correlation of the scoring system to local tumor extension (pT category) is described. Recommendations for grading of prostatic carcinoma resulting from this work are the following: Only three independent parameters are of significance for grading prostatic cancer: tumor architecture, nuclear anaplasia, and the presence or absence of mitoses. The combination of these parameters in a scoring system identifies five separate prognostically different groups. The separation achieved by this new system is superior to conventional grading and is suggested for grading prostatic cancer.