We have retrospectively analyzed the impact of prognostic factors on the outcome of serologically HLA-matched unrelated donor (UD) BMT for CML. For this purpose, we have studied a cohort of 366 patients transplanted in Europe between January 1985 and December 1994. The median age of the 211 males and 155 females was 34 years; 238 patients were transplanted in first chronic phase and 116 in advanced phases. The median interval from diagnosis to BMT was 827 days. GVHD prophylaxis consisted of CsA and MTX in 202 patients or of ex vivo or in vivo T cell depletion (TCD) in 129. Recently, DNA-based methods of HLA-class II typing have been used to improve donor selection. We obtained complete data on 300 donor/recipient (D/R) pairs. Among them, we have identified three groups of patients, according to specific HLA-DRB1 D/R compatibility. Two hundred and ten patients received marrow from donors identical for HLA-DRB1 (group 1). Thirty-one patients received BMT from a donor who was HLA-DRB1 mismatched (group 2) and 59 from a donor in whom specific HLA-DRB1 typing was not performed (group 3). The overall survival was 37 +/- 3% at 2 years and leukemia-free survival (LFS) was 31 +/- 3%. In univariate analysis, five variables had a favorable effect on LFS: transplant in first chronic phase (P = 0.0001), time interval from diagnosis to BMT shorter than the median (P = 0.01), prophylaxis of GVHD without TCD (P + 0.001), acute GVHD < grade III (P = 0.0009) and HLA-DRB1 D/R matching (P = 0.0001). Transplant-related mortality (TRM) was 49 +/- 4% in group 1, 79 +/- 8% in group 2 and 80 +/- 6% in group 3 (P = 0.0001). Multivariate analysis confirmed that HLA-DRB1 matching was the most significant factor influencing survival (P = 0.04), LFS (P = 0.013) and TRM (P = 0.0049). From these results, we have defined a 'good risk' group, ie patients transplanted in first chronic phase, from an HLA-DRB1 matched donor, without TCD as prophylaxis against GVHD. The 2 year LFS, TRM and relapse incidence for this group were 51 +/- 5%, 47 +/- 5% and 2 +/- 2%, respectively. This suggests that the long-term outcome of patients with favorable prognostic features can approach that of patients transplanted from geno-identical siblings. In contrast, the TRM for patients transplanted for advanced disease from non HLA-DRB1-identical donors was 94%. Such a high TRM clearly indicates that UD BMT is not justifiable for these individuals.