Morphological characteristics of tumor cells have been employed in the prognosis of lymphomas and solid tumors. This report documents an attempt to predict survival from the known cytologic heterogeneity in multiple myeloma. Myeloma cells in bone marrow smears from patients at diagnosis were evaluated by assigning them to morphologically defined categories. Cox's multivariate regression model for censored survival data was used to generate optimal weights, which served as coefficients in two regression equations to estimate death risk from cellular morphology. Step-wise procedures excluded redundant parameters. “Myeloma morphology score” (MMS) discriminates significantly (p less than 0.0001) among 3 stages, with median survival times of 42.5, 30.7, and 9.1 mo. For clinical routine application, “myeloma progression scorex201D; (MPS), the weight sum of the proportion of plasmablasts and the extent of bone marrow plasma cell infiltration, is suggested as a simple prognostic tool. Its discriminative power is very high [p less than 10(-9)]. Median survival times of greater than 71.5, 23.4, and 6.1 mo were found for good, moderate, and poor risk groups, respectively. However, staging is not confined to three subgroups, grouping is flexible, and pairs of data can be matched. This fact may prove to be valuable in designing prognosis-controlled clinical trials or theoretical studies on cellular differentiation. Preliminary results suggest changes in morphology due to disease progression and/or the effect of therapy on tumor kinetics. Most importantly, staging according to MPS or MMS may facilitate the adaption of therapy to the current state of the disease in patients with multiple myeloma.