Multiple myeloma (MM) is a plasma-cell malignancy characterized by marked epidemiological, biological, and clinical heterogeneity. The goal of this study was to find a genetic basis for this heterogeneity. Using fluorescence in situ hybridization, we analyzed a prospective cohort of 901 patients with various plasma-cell disorders--monoclonal gammopathies of undetermined significance, smoldering MM, MM, and primary plasma-cell leukemia--for genetic abnormalities involving the 13q14 and 14q32 chromosomal regions; the patients were consecutively enrolled in the Intergroupe Francophone du Myélome clinical trials, We performed statistical analyses comparing these chromosomal abnormalities in terms of immunological (ie, immunoglobulin types and light-chain subtypes) and clinical status and, to some extent, prognostic features. It was found that 14q32 translocations and del(13) are the most frequent chromosomal abnormalities, observed in 75% and 45% of the patients, respectively, and are not randomly distributed, but interconnected. Second, correlations between them allowed us to define 4 major genetic categories of patients: (1) patients lacking any 14q32 abnormality (25%) and generally also lacking del(13); (2) patients presenting either t(4;14) or t(14;16), almost always associated with a del(13) (15% of patients); (3) patients with other 14q32 abnormalities and presenting del(13) (25%); and (4) patients with other 14q32 abnormalities but not presenting del(13) (35%). Third, we show that this genetic stratification is highly correlated with immunological status and clinical presentation and with some major prognostic factors. For the first time, this study gives genetic support to the heterogeneity observed in patients with MM and demonstrates that the 14q32 and 13q chromosomal abnormalities are not randomly distributed. The strong correlations we found might be the basis for a novel genetic classification of MM, as has been previously demonstrated for leukemias and lymphomas. Furthermore, our study supports different models for MM oncogenesis.