To the Editor:
We read with interest the recent letter by Cuneo et al,1 which raises important issues about the diagnosis and classification of acute myeloid leukemia (AML)-M0 and its relation with AML-M1. The authors point out the very high incidence of AML-M0 (8.9%) and AML-M1 (19.5%) in our series2 compared with two large multicentric studies3,4 which yielded an incidence of 0.1% and 4% for AML-M0 and 16.4% and 10% for AML-M1, respectively. Moreover, they argue about the high incidence of chromosome changes among our AML-M1 cases. Accordingly, the authors emphasize the need for uniform criteria for patient selection and centralized review to make a correct diagnosis.
Our apparently higher incidence of AML-M0 is basically due to the inclusion of cases that carried lymphoid antigens. In fact, apart from those cases which fit classical French-American-British (FAB) criteria for AML-M0 diagnosis (<3% of the blasts positive for myeloperoxidase [MPO] and/or Sudan Black B [SBB], positivity for myeloid markers and lack of B-T-lineage associated antigens),5 we regarded as AML-M0 even cases bearing lymphoid antigens, based on the positivity of anti-MPO and negativity for cCD3 or cCD22.2,6,7 We believe that the expression of membrane lymphoid antigens does not necessarily exclude a diagnosis of AML-M0, once documented by immunologic/electron microscopy (EM) assay the presence of MPO. The expression of lymphoid markers might represent only an aspect of a broad heterogeneity of M0 leukemias whose common feature is a poor clinical outcome.7 Conversely, the coordinate expression of anti-MPO, cCD3, or cCD22 defines “genuine biphenotipic” leukemia,8 which is supposed to represent a distinct entity: two of our cases with these features were excluded from the analysis. A routinely immunologic approach for acute leukemia diagnosis should also use monoclonal antibodies (MoAbs) for cytoplasmic MPO, CD3, and CD229 (or as more recently suggested CD79a) to avoid underestimation of AML-M0 and an incorrect diagnosis of acute lymphoid leukemia. In this view, neither C-kit nor antilysozyme MoAb can help to definitively ascertain the diagnosis. In fact, in ours and the experience of others,7,10 CD117 expression was found not to be peculiar of immature AML; in addition, immunologic detection of lysozyme would miss those immature monocytic/monoblastic leukemia that have not yet become competent for lysozyme synthesis.11
Regarding our 19.5% incidence of AML-M1 cases, we do not believe this figure represents a remarkable difference when compared with a 16.4% incidence observed in the large GIMEMA/EORTC trial; Ball et al12 reported an overall incidence of 20% of AML-M1 cases among 339 de novo acute myeloid leukemia. Twenty-eight of 50 observed cases of AML-M1 have been used for cytogenetic comparison as only for them chromosome pattern was available: this was the sole selection criteria and it may account for the higher rate of chromosome abnormalities as compared to the study by Cuneo et al.1 However, most of our cases were recruited to multicentric trials (EORTC/GIMEMA AML8a/b, AML10, AML12, and Amgen protocol G-CSF 91134) and centrally reviewed from a morphological and cytochemical point of view. Finally, the discrepancy in Table 1 between the number of observed cases in the different cytogenetic groups and the total number of patients is obviously due to a typographical error and we apologize for that. In fact, the category “Others” is = 5/19 instead of 6/19.
In conclusion, we certainly agree with the crucial value of a centralized reviewing for the recognition of different AML subsets and the homogeneous application of FAB criteria. For the same reasons centralized reviewing appears desirable even for phenotypic and kariotypic studies.