Prognostic factors for non-cleaved follicular center-cell lymphomas and immunoblastic sarcoma. A Bayesian approach

The Bayesian multivariate statistical method was applied to determine the relative strength and optimal combination of 17 variables in predicting survival of 151 patients with non-Hodgkin's lymphomas assigned as non-cleaved follicular center-cell and immunoblastic sarcoma types according to the classification of Lukes & Collins. Considering all the factors simultaneously, the analysis showed that the combination of stage, Hb level and location of the lymphoma was included in the best predictive model at each survival time studied. Additional factors were erythrocyte sedimentation rate, thrombocyte count and leucocyte count. Of the histological variables, only growth pattern and mitotic ratio in the biopsy specimen remained significant. At manually controlled computer simulation with these best indicators, this model would have given a correct classification for 69-78% of the patients at the 4 survival times studied. One can thus expect about 70% correct prognoses using this model.