[Morphology versus prognosis in renal cell carcinoma a statistical approach]

In a group of 102 selected cases of renal cell carcinoma which at various periods after nephrectomy terminated fatally, age, sex, and 23 gross and/or histologic tumor characteristics were examined, subclassified, and evaluated. An attempt was made to correlate these parameters statistically to postoperative survival time. A newly developed and programmed statistical method, i.c. "dichotomic variance analysis", proved to be superior to both multiple regression and cluster analysis. Using infiltrative-destructive tumor growth, polymorphy, and chromatin density of tumor cell nuclei, extension of renal vein invasion, age and sex, as the only 6 required out of 25 distinguished tumor parameters, this statistical multivariate method comprised 7 distinctive groups of different mean geometrical postoperative survival time, obviously corresponding to 7 tumor types of increasing degree of malignancy. By step-wise dichotomic splitting of tumor groups it therefore was possible to delinate schematically some of the complex connections between the criteria or variables of prognostic significance. Conceivable sources of errors possibly influencing reproducibility and practical applicability of the presented dichotomic variance analysis in evaluating prognostic criteria in renal cell carcinoma are discussed.