Although the prognostic value of cytogenetic analysis has previously been demonstrated in myelodysplastic syndromes (MDS), karyotype had not been included in previously published scoring systems, such as Bournemouth and Sanz's scores. We studied karyotype at diagnosis in 408 cases of de novo MDS (after excluding therapy-related MDS). Karyotypes were classified in 10 groups: normal; isolated del 5q; del 5q and other rearrangements; isolated +8; isolated -7 or del 7q; del 20q; isolated -Y; miscellaneous single rearrangements; -7 or del 7q and other rearrangements; miscellaneous complex rearrangements. Karyotypes were considered complex when at least three chromosomes were rearranged. Complex karyotypes included all patients with del 5q and other rearrangements, -7 or del 7q and other rearrangements, and miscellaneous complex rearrangements (i.e. three of the 10 groups). Median actuarial survival of the 408 patients was 28 months, and 90 patients (22%) progressed to acute myeloid leukemia (AML). For survival, bone marrow (BM) blasts, circulating blasts, white blood cell (WBC), neutrophil count, platelet count, hemoglobin, age, sex, FAB classification, and Bournemouth and Sanz's scores had strong prognostic value. Cytogenetics also had strong prognostic value. An unfavorable cytogenetic group (patients with complex karyotypes) was identified. On the other hand, although patients with isolated del 20q and del 5q had a somewhat better prognosis than other patients with non-complex cytogenetic findings, they could not be statistically individualized as a favorable group, possibly owing to their relatively limited number. By multivariate regression analysis, a three-variable new scoring system could be designed based on karyotype (1 point for complex karyotype; 0 for other groups), platelets (0 for > 75 x 10(9)/l; 1 for < 75 x 10(9)/l), and BM blasts (0 for < 5%, 1 for 5-10%, 2 for > 10%). The total score (addition of points for the three variables) was able to separate patients in three groups with low (score 0) intermediate (score 1 or 2), and high risk (score 3 or 4) which included 34%, 47%, and 19% of the patients, and had a median survival of 55, 24, and 6 months, respectively (chi 2 = 110, p < 10(-4)). This new score (Lille score) was able to subdivide risk groups according to Bournemouth and Sanz's scores into further subgroups of different prognoses. For progression to AML, BM blasts, circulating blasts, WBC count, hemoglobin, FAB type, and karyotype had prognostic value by univariate analysis.(ABSTRACT TRUNCATED AT 400 WORDS)