The International Workshop on the relationship between prior therapy and balanced chromosome aberrations in therapy-related myelodysplastic syndromes (t-MDS) and therapy-related acute leukemia (t-AL) identified 79 of 511 (15.5%) patients with balanced 21q22 translocations. Patients were treated for their primary disease, including solid tumors (56%), hematologic malignancy (43%), and juvenile rheumatoid arthritis (single case), by radiation therapy (5 patients), chemotherapy (36 patients), or combined-modality therapy (38 patients). 21q translocations involved common partner chromosomes in 81% of cases: t(8;21) (n = 44; 56%), t(3;21) (n = 16; 20%), and t(16;21) (n = 4; 5%). Translocations involving 15 other partner chromosomes were also documented with involvement of AML1(CBFA2/RUNX1), identifying a total of 23 different 21q22/AML1 translocations. The data analysis was carried out on the basis of five subsets of 21q22 cases, that is, t(8;21) with and without additional aberrations, t(3;21), t(16;21), and other 21q22 translocations. Dysplastic features were present in all 21q22 cases. Therapy-related acute myeloid leukemia (t-AML) at presentation was highest in t(8;21) (82%) and lowest in t(3;21) (37.5%) patients. Cumulative drug dose exposure scores for alkylating agents (AAs) and topoisomerase II inhibitors indicated that t(3;21) patients received the most intensive therapy among the five 21q22 subsets, and the median AA score for patients with secondary chromosome 7 aberrations was double the AA score for the entire 21q22 group. All five patients who received only radiation therapy had t(8;21) t-AML. The median latency and overall survival (OS) for 21q22 patients were 39 and 14 months (mo), compared to 26 and 8 mo for 11q23 patients, 22 and 28 mo for inv(16), 69 and 7 mo for Rare recurring aberrations, and 59 and 7 mo for Unique (nonrecurring) balanced aberration (latency P < or = 0.016 for all pairwise comparisons; OS, P < or = 0.018 for all pairwise comparisons). The percentages of 21q22 patients surviving 1 year, 2 years, and 5 years were 58%, 33%, and 18%, respectively. Noticeable differences were observed in median OS between 21q22 patients (n = 7) receiving transplant (BMT) (31 mo) compared to 21q22 patients who received intensive non-BMT therapy (n = 46) (17 mo); however, this was nonsignificant because of the small sample size (log-rank, P = 0.33). t-MDS/t-AML with balanced 21q22 aberrations was associated with prior exposure to radiation, epipodophyllotoxins, and anthracyclines, dysplastic morphologic features, multiple partner chromosomes, and longer latency periods when compared to 11q23 and inv(16) t-MDS/AML Workshop subgroups. In general, patients could be divided into two prognostic risk groups, those with t(8;21) (median OS, 19 mo) and those without t(8;21) (median OS, 7 mo) leukemia (log-rank, P = 0.0007).