Daniel A. Arber, MD, 1 Anthony S. Stein, MD, 2 Nora H. Carter, MS, 3 David Ikle, PhD, 3 Stephen J. Forman, MD, 2 and Marilyn L.

Hematopathology / ACUTE MYELOID LEUKEMIA CLASSIFICATION Prognostic Impact of Acute Myeloid Leukemia Classification Importance of Detection of Recurring Cytogenetic Abnormalities and Multilineage Dysplasia on Survival Daniel A. Arber, MD, 1 Anthony S. Stein, MD, 2 Nora H. Carter, MS, 3 David Ikle, PhD, 3 Stephen J. Forman, MD, 2 and Marilyn L. Slovak, PhD 1 Key Words: Acute myeloid leukemia; Classification; World Health Organization; French-American-British classification; FAB DOI: 1309/EM7KCQR4GLMHRCX4 Abstract To evaluate the prognostic impact of acute myeloid leukemia (AML) classifications, specimens from 300 patients with 20% or more bone marrow myeloblast cells were studied. Specimens were classified according to the French-American-British Cooperative Group (FAB), the World Health Organization (WHO), the Realistic Pathologic Classification, and a cytogenetic risk group scheme. Cases with fewer than 30% blast cells did not have a 5-year survival significantly different from cases with 30% or more blast cells, and survival was similar for the low blast cell count group and cases with multilineage dysplasia and 30% or more blasts. Categories of AML with recurrent cytogenetic abnormalities of t(15;17), t(8;21), inv(16)/t(16;16), and 11q23 showed significant differences in 5-year survival. No significant difference was identified between AMLs arising from myelodysplasia and de novo AMLs with multilineage dysplasia, but all cases with multilineage dysplasia had a worse survival than all other AMLs and other AMLs without favorable cytogenetics. FAB types M0, M3, and M4Eo showed differences in survival compared with all other FAB types, with M0 showing a significant association with high-risk cytogenetics and 11q23 abnormalities. Other FAB groups and WHO AML, not otherwise categorized subgroups did not show survival differences. These findings suggest that the detection of recurring cytogenetic abnormalities and multilineage dysplasia are the most significant features of current AML classification. The classification of acute myeloid leukemia (AML) has evolved from the primarily morphologic and cytochemical system of the early French-American-British Cooperative Group (FAB) proposal 1,2 to systems that consider the results of cytogenetic studies. 3,4 Recurring balanced translocations, particularly t(8;21), t(15;17), inv(16)/t(16;16), and 11q23 translocations, represent a substantial percentage of cytogenetic abnormalities in AML. These aberrations constitute 34% to 47% of pediatric 5-7 and 21% to 28% of adult 6,8-10 AML cytogenetic abnormalities. Additional abnormalities, including trisomies, deletions, and complex karyotypes, contribute to another large percentage of adult AMLs 6,10,11 and show overlap with cases of myelodysplasia. The detection of these various cytogenetic abnormalities in AML is known to have prognostic significance, 5-12 but many of them do not correlate well with FAB classification disease groups. Despite the detection of significant cytogenetic abnormalities in many AML cases, a substantial subset of cases show no karyotypic abnormality. 13 Cryptic translocations are present in some of these cases, and other molecular genetic abnormalities presumably are present in the leukemic cells of these patients. Some karyotype abnormalities in AML are associated with characteristic morphologic and/or immunophenotypic findings. These include leukemias with neoplastic promyelocytes showing loss of HLA-DR with t(15;17) and its variants, 14-16 CD19+/CD34+ myeloid cells with abundant and distinctive granules of t(8;21), 17-20 and abnormal eosinophils of inv(16) AML. 21-24 The morphologic detection of multilineage dysplasia often correlates with older age and the presence of complex karyotypes, monosomies, or trisomies, but also may occur in the absence of cytogenetic 672 Am J Clin Pathol 2003;119:672-680 672 DOI: 1309/EM7KCQR4GLMHRCX4

Hematopathology / ORIGINAL ARTICLE abnormalities. 13,25-27 Several studies suggest that the presence of multilineage dysplasia is associated with a worse prognosis in AML. 25,27-29 Despite the correlation between acute promyelocytic leukemia (FAB type M3) and t(15;17), and acute myelomonocytic leukemia with eosinophilia (FAB type M4Eo) and inv(16)/t(16;16), the FAB classification does not incorporate cytogenetic results into its system. In addition, the presence of multilineage dysplasia is not considered part of that classification. These parameters, however, are major elements of the recently published World Health Organization (WHO) classification of tumors of hematopoietic and lymphoid tissues, 3 which includes AML, as well as in the recently proposed Realistic Pathologic Classification (RPC) of AML. 4 These classification systems also differ from the FAB classification by lowering the blast cell count to 20% or more for most cases of AML and not requiring a specific blast cell count in the presence of recurring cytogenetic abnormalities. The present study was undertaken to evaluate the prognostic significance of the recently proposed classifications systems, the FAB classification, and a risk-group cytogenetic stratification. Specific elements of interest were the impact of lowering the blast cell count to 20% in AML and the significance of subgroups with recurrent cytogenetic abnormalities and multilineage dysplasia. The significance of morphologic and cytochemical subclassification of AMLs without recurrent cytogenetic abnormalities or multilineage dysplasia also was evaluated. Materials and Methods This retrospective study included 300 cases of AML or refractory anemia with excess blasts in transformation diagnosed based on a blast cell count of 20% or more at the City of Hope National Medical Center, Duarte, CA, between January 1990 and October 2000. Cases were included in the study if overall survival information was available and if original diagnostic pathology material was available for review. Morphologic review on all cases was performed by one of the authors (D.A.A.), and bone marrow aspirate material or peripheral blood smears containing blast cells were required for all cases. Cytochemical results and results of immunophenotyping studies also were reviewed. Based on the slide review, cases were classified using the published criteria of the FAB, 2 WHO, 3 and RPC 4 classifications with correlation with available presentation cytogenetic studies. Cytogenetic results were available for 228 cases. Cytogenetic risk group stratification also was performed by one of the authors (M.L.S.), using criteria similar to those previously described. 10 Cases were considered to be favorable or low risk in the presence of t(8;21), inv(16)/t(16;16), or t(15;17). The presence of other abnormalities with these translocations did not change the risk group. Unfavorable or high-risk abnormalities were complex (more than 3) abnormalities, 7, inv(3q), del(9q) without t(8;21), 11q23, 17p, 20q, or 21q abnormalities, t(9;22), t(6;9), +13, and dmin/hsrs. The intermediate-risk cytogenetic group included those with normal karyotypes; single chromosomal abnormalities such as +8, +11, and Y; and 12p abnormalities. Overall survival data were obtained from the City of Hope Biostatistics database and the City of Hope Cancer Registry database. The study was approved by the City of Hope Institutional Review Board and Cancer Protocol Review and Monitoring Committee (approval No. 95124). Five-year Kaplan-Meier survival estimates were tested using the Mantel-Haenszel (log-rank) test. 30 The 95% confidence intervals (CIs) on the Kaplan-Meier estimates were calculated using a logit transformation of the Greenwood estimate of the variance. The Pearson χ 2 test 31 was used to test for differences in the proportion of cytogenetic abnormalities among different groups within the FAB classification. All P values are 2-sided, and values less than.05 were considered significant (SAS/STAT Software, Changes and Enhancements, Release 8.02, SAS Institute, Cary, NC). Results The cases included 154 males and 146 females with mean and median ages of 4 and 41 years, respectively (range, 1-76 years). Thirteen patients were younger than 18 years. The overall 5-year survival for the group was 35.8% with a 95% CI of 3% to 41.4%. The median follow-up for patients who were alive (n = 112) was 7 months (range, 9.2-127.2 months) and for patients who died (n = 188) was 12.0 months (range, -58.9 months). Classification Groups The breakdowns for the FAB, WHO, and RPC classifications are given in Table 1, Table 2, and Table 3. Numeric differences between the groups are due to differences in diagnostic criteria for each classification. In the FAB classification, 21 (7.0%) were unclassified owing to insufficient cytochemical results. Only 228 cases were classified using the WHO criteria, owing to the requirement for cytogenetic studies for classification. In the therapy-related AML category of the WHO classification, 2 cases were not further classified owing to a combination of previous therapy or radiation therapy. In the RPC classification, there were 34 AML cases, not otherwise specified (NOS), with monocytic or myelomonocytic features by cytochemistry. Cytogenetic risk-group stratification included 67 low risk (29.4%), 100 intermediate risk (43.9%), and 61 high risk (26.8%). Am J Clin Pathol 2003;119:672-680 673 673 DOI: 1309/EM7KCQR4GLMHRCX4 673

Arber et al / ACUTE MYELOID LEUKEMIA CLASSIFICATION Table 1 Case Distribution Using the French-American-British Cooperative Group Classification Type Description No. of Cases M0 Minimally differentiated acute myeloid leukemia 18 M1 Myeloblastic leukemia without maturation 49 M2 Myeloblastic leukemia with maturation 73 M3 Promyelocytic leukemia 22 M4 Myelomonocytic leukemia 35 M4Eo Myelomonocytic leukemia with eosinophils 33 M5 Monoblastic and monocytic leukemia 12 M6 Erythroleukemia 7 M7 Megakaryoblastic leukemia 6 RAEBT Refractory anemia with excess blasts in transformation 24 Unclassified 21 Total 300 Table 2 Case Distribution Using the World Health Organization Classification Type and Description No. of Cases Acute myeloid leukemia (AML) with recurrent 78 cytogenetic abnormalities AML with t(8;21)(q22;q22), (AML1/ETO) 16 AML with inv(16)(p13q22) or t(16;16)(p13;q22), 31 (CBFβ/MYH11) Acute promyelocytic leukemia (AML with t(15;17)(q22;q12), 20 (PML/RARα and variants) AML with 11q23 (MLL) abnormalities 11 AML with multilineage dysplasia 74 With previous myelodysplastic syndrome 26 Without previous myelodysplastic syndrome 48 AML and myelodysplastic syndrome, therapy-related 7 Alkylating agent related 3 Topoisomerase II inhibitor related 2 Other 2 AML not otherwise categorized 67 AML, minimally differentiated 3 AML, without maturation 19 AML, with maturation 21 Acute myelomonocytic leukemia 16 Acute monoblastic and monocytic leukemia 2 Acute erythroid leukemia 3 Acute megakaryoblastic leukemia 2 Acute basophilic leukemia 0 Acute panmyelosis with myelofibrosis 1 Unclassified 2 Total 228 Impact of Lowering the Blast Cell Count for AML to 20% No significant difference in overall survival was identified when cases with 20% or more blasts but fewer than 30% blasts were compared with all AML cases with 30% or more blasts; with AML cases, NOS [those without recurring cytogenetic abnormalities t(8;21), inv(16), or t(15;17) or multilineage dysplasia]; or with cases of AML with multilineage dysplasia or therapy-related AML. Sixteen of the cases with fewer than 30% blast cells were classified as AML with multilineage dysplasia, 4 as therapy-related AML, and 1 as AML with 11q23 abnormalities. Three cases did not have cytogenetic studies but had morphologic features of AML with abnormal eosinophils suggestive of inv(16) (1 case) or myelodysplasia-associated AML (2 cases) in the RPC classification. The overall survival of the AML group with low blast cell count was essentially identical to the group AML with multilineage dysplasia. The 5-year survival for AMLs with 20% or more to less than 30% blast cells (n = 24) was 18% (95% CI, 2%-34%). Comparison with the survival for other groups was as follows: 37.3% (95% CI, 31.5%-43.3%; P =.1603) for all AML cases with more than 30% blast cells (n = 276); 39.8% (95% CI, 3%-49.3%; P =.1347) for Table 3 Case Distribution Using the Realistic Pathologic Classification Type and Description No. of Cases Acute myeloid leukemia, de novo 183 Acute myeloid leukemia (with or without monocytic features), not otherwise specified 109 Acute myeloid leukemia with changes suggestive of t(8;21) 18 Acute promyelocytic leukemia 22 Acute myeloid leukemia with abnormal eosinophils suggestive of inv(16)(p13q22) or t(16;16)(p13;q22) 31 Acute megakaryoblastic leukemia 3 Acute myeloid leukemia, myelodysplasia-associated 114 Acute myeloid leukemia, treatment-related 8 Acute myeloid leukemia arising from myelodysplasia 31 Acute myeloid leukemia with associated myelodysplasia 75 Acute myeloid leukemia, unclassified 1 Myelodysplasia 2 Total 300 674 Am J Clin Pathol 2003;119:672-680 674 DOI: 1309/EM7KCQR4GLMHRCX4

Hematopathology / ORIGINAL ARTICLE AML cases, not otherwise categorized with more than 30% blast cells (n = 109); and 19.3% (95% CI, 11.1%-27.6%; P =.4904) for AML cases with multilineage dysplasia or AML cases that were treatment related (n = 114). Figure 1 illustrates the similarity in survival between the low blast cell count group and the AML cases with multilineage dysplasia or therapy-related AML. FAB Classification Groups FAB subgroups with 10 or more cases were compared with all other cases for 5-year survival differences Table 4. FAB types M0, M3, and M4Eo differed significantly from the other groups. The M0 cases were reclassified in the WHO classification into AML, not otherwise categorized (minimally differentiated type; 3 cases); AML with multilineage dysplasia (5 cases); and AML with 11q23 abnormalities (3 cases). The remaining cases did not have cytogenetic abnormalities but were classified as AML, NOS (4 cases); AML with multilineage dysplasia (1 case); and therapyrelated AML (2 cases) in the RPC classification. For FAB type M0 vs non-m0 types, the frequency of high-risk cytogenetics (50% vs 18.4%, respectively; P =.0036) and the frequency of 11q23 abnormalities (17% vs 2.8%, respectively; P =.0222) were significantly higher in the M0 group. No differences in the frequency of previous therapy or multilineage dysplasia were found between the AML-M0 and the non-m0 cases. Significance of Recurrent Cytogenetic Abnormality Subgroups The 5-year overall survival rates for the groups with recurrent cytogenetic abnormalities were as follows: t(8;21), 40% (n = 16; 95% CI, 15%-66%); t(15;17), 60% (n = 20; 95% CI, 38%-81%); inv(16)/t(16;16), 61% (n = 31; 95% CI, 43%-78%); and 11q23 abnormalities, 12% (n = 11; 95% CI, 0%-34%) (overall P =.0266). The survival differences of these categories are illustrated in Figure 2. Significance of Morphologic Detection of Multilineage Dysplasia Of the cases studied, 106 were AML with morphologic features of multilineage dysplasia, including 31 with preexisting myelodysplasia. Eight cases of therapy-related AML also were identified. No significant differences in overall survival were identified between cases of therapy-related disease, AML arising from myelodysplasia, and AML with associated myelodysplastic changes and no history of myelodysplasia (P =.9823) Figure 3. This combined group (n = 114), however, had a lower 5-year overall survival (18.4%; 95% CI, 11.1%-25.7%) compared with cases of AML without associated myelodysplastic changes (n = 186; 5-year survival, 46.7%; 95% CI, 39.3%-54.1%; P <.0001). This difference remained significant when compared with AML, NOS, which excludes the favorable prognostic groups of t(8;21), inv(16), and t(15;17) AML (n = 109; 5-year survival, 39.8%; 95% CI, 3%-49.3%; P =.0014). Lack of Significance of Subclassification of AML, Not Otherwise Categorized/NOS The 109 cases in the RPC AML, NOS group and the 70 cases in the WHO AML, not otherwise categorized group were evaluated to attempt to assess the significance of subcategories of each. Cytochemical evidence of monocytic or myelomonocytic differentiation (n = 34; 5-year survival, 41%; 95% CI, 24%-58%) did not identify a significant diagnostic group when compared with cases without such differentiation (n = 74; 5-year survival 38%; 95% CI, 27%-50%; P =.7945). Comparisons of overall survival for the WHO, not otherwise categorized groups were made for AML without maturation (n = 19; 5-year survival, 25%; 95% CI, 5%-45%), AML with maturation (n = 21; 5-year survival, 41%; 95% CI, 20%-63%), and acute myelomonocytic leukemia (n = 16; 5-year survival, 25%; 95% CI, 2%-49%) and the other AML, not otherwise categorized cases. There were insufficient cases in the remaining groups for statistical comparison. No significant differences in survival were identified by separating these 3 subcategories of AML, not otherwise categorized (P =.7876). Comparison of Diagnostic Groups by Treatment In an attempt to evaluate some treatment differences and the role of different treatments in the differences in outcome of the major classification groups, cases were separated in the RPC classification into AML, NOS (n = 109; 5-year AML with 30% blasts and multilineage dysplasia AML with 20%- <30% blasts Figure 1 Similarity in overall survival between acute myeloid leukemia (AML) with <30% blasts and AML with multilineage dysplasia (including therapy-related AML). P =.4904. Am J Clin Pathol 2003;119:672-680 675 675 DOI: 1309/EM7KCQR4GLMHRCX4 675

Arber et al / ACUTE MYELOID LEUKEMIA CLASSIFICATION Table 4 Five-Year Survival by French-American-British (FAB) Type FAB Type * No. of Cases Median Age (Range) (y) 5-Year Survival, % (95% CI) P M0 18 43.5 (6-76) 2 (1.3-4).0345 M1 49 45 (6-73) 32.2 (19.0-45.4).4360 M2 73 36 (4-75) 38.9 (27.5-5).6169 M3 22 29.5 (5-65) 61.5 (4-82.5).0144 M4 35 41 (20-73) 26.4 (11.2-41.6).0980 M4Eo 33 42 (18-68) 56.0 (38.5-73.5).0120 M5 12 4 (19-63) 22.2 (-47.4).3283 RAEBT 24 47 (11-73) 18.3 (2.3-34.4).1603 Unclassified 21 42 (16-73) 28.6 (9.2-47.9).3899 CI, confidence interval; RAEBT, refractory anemia with excess blasts in transformation. * FAB types M6 (n = 7) and M7 (n = 6) were not analyzed owing to the small number of cases in each group. Log-rank test. The P value is based on a comparison of the FAB type with all other cases. inv(16)/t(16;16) AML t(15;17) AML t(8;21) AML 11q23 AML Figure 2 Overall survival of patients with acute myeloid leukemia (AML) with recurrent cytogenetic abnormalities. P =.0266. AML, therapy-related AML arising from myelodysplasia AML with associated dysplasia Figure 3 Comparison of overall survival of patients with therapy-related acute myeloid leukemia (AML), AML arising from myelodysplasia, and AML with dysplasia and no history of myelodysplasia. P =.9823. survival, 39.8%; 95% CI, 3%-49.3%); RPC with features of favorable cytogenetic groups t(8;21), inv(16), and t(15;17) (n = 71; 5-year survival, 56%; 95% CI, 44%-68%); and AML with multilineage dysplasia or therapy-related AML (n = 114; 5-year survival, 18.4%; 95% CI, 11.1%-25.7%). Significant survival differences (P <.0001) were recognized using this separation. These groups then were compared by therapy type, with broad categories of allogeneic transplantation (n = 144), autologous transplantation (n = 71), and no transplantation (n = 79). Details of the overall 5-year survival for each group are given in Table 5. The disease groups, however, remained significantly distinct with each treatment type (P =.0175 for autologous transplantation; P =.0233 for allogeneic transplantation; P =.006 for no transplantation). Clinically Significant Disease Groups Table 6 shows a modification of the RPC classification, based on the significant disease groups of this study. These include AMLs with evidence of t(8;21), t(15;17), or inv(16)/t(16;16); AML with multilineage dysplasia or therapy-related AML; and AML, NOS. Figure 4 and Figure 5 illustrate the overall survival of the significant disease groups in the RPC and WHO classifications. Discussion The FAB classification of AML has been the major system used by hematopathologists for more than 20 years. This system was useful in that it provided structured criteria for the diagnosis of a variety of morphologic and cytochemical subtypes of AML. The present study, however, indicates that the majority of categories in this system do not delineate significant disease groups, based on overall survival. The only significant FAB groups identified were M0, M3, and M4Eo. The significance of M3 and M4Eo are not surprising because of the high correlation of these subgroups with 676 Am J Clin Pathol 2003;119:672-680 676 DOI: 1309/EM7KCQR4GLMHRCX4

Hematopathology / ORIGINAL ARTICLE Table 5 Five-Year Overall Survival for Disease Groups by Transplant Type Disease Group * No. of Cases Mean Age (y) 5-Year Survival, % (95% CI) P Autologous transplant AML, favorable cytogenetics 21 39 79.7 (61.9-97.5).0175 AML, NOS 29 40 61.3 (43.3-79.2) AML/MDS 21 45 36.7 (15.6-57.8) Allogeneic transplant AML, favorable cytogenetics 37 27 46.6 (29.9-63.3).0233 AML, NOS 52 34 37.7 (23.7-51.7) AML/MDS 55 42 24.4 (12.8-36.0) No transplant AML, favorable cytogenetics 13 47 46.2 (19.1-73.3).0060 AML, NOS 28 51.5 2 (5.5-36.1) AML/MDS 38 56 0 AML, acute myeloid leukemia; CI, confidence interval; MDS, myelodysplasia-associated or therapy-related; NOS, not otherwise specified. * Favorable cytogenetics indicates cases with features of t(8;21), inv(16)/t(16;16), or t(15;17). Log-rank test. t(15;17) and inv(16)/t(16;16), respectively. These cytogenetic abnormalities are known to be associated with significantly improved survival with current therapies compared with other AML types. 12,32,33 The significantly worse prognosis of AML-M0 compared with other AMLs in the present study seems to be related to its high association with the poor-prognosis cytogenetic groups of AML, including AML with 11q23 abnormalities. The recently proposed AML classification systems include categories that encompass the significant FAB disease groups, and the remaining FAB groups seem to be of little, if any, significance. Several studies have demonstrated the impact of specific cytogenetic abnormalities on survival in AML, with t(15;17) and its variants, t(8;21), inv(16)/t(16;16), and abnormalities involving 11q23 being the most common. 6,10,12 The present study confirms the significance of these abnormalities, with t(15;17) and inv(16)/t(16;16) AMLs having a more favorable prognosis, t(8;21) conferring an intermediate prognosis, and 11q23 abnormalities associated with a generally poor prognosis. Because AMLs with 11q23 abnormalities do not have sufficiently specific morphologic and immunophenotypic findings to suggest this abnormality and because of the reported variability in prognosis based on the 11q23-translocation partner, these leukemias were not designated as a subtype in the Realistic Pathologic Classification. 4 However, the prognostic importance of 11q23 abnormalities indicates that the detection of such abnormalities should be noted in an amended pathology report for a complete diagnosis. Continued studies to evaluate the prognostic significance of other balanced translocations in AML are needed, and additional subgroups of AML with recurrent cytogenetic abnormalities should be considered as additions to any future classification systems. The impact of multilineage dysplasia in AML also has been studied previously. The frequency of trilineage dysplasia in de novo AML was reported as 11.6% in a series Table 6 Modified Realistic Pathologic Classification Acute myeloid leukemia, de novo Acute myeloid leukemia with changes suggestive of t(8;21)(q22;q22) Acute promyelocytic leukemia Acute myeloid leukemia with abnormal eosinophils suggestive of inv(16)(p13q22)/t(16;16)(p13;q22) Acute megakaryoblastic leukemia * Acute myeloid leukemia, not otherwise specified Acute myeloid leukemia with multilineage dysplasia Acute myeloid leukemia, therapy-related * Larger studies are needed to confirm the significance of de novo acute megakaryoblastic leukemia. Reports for cases in this category that are found to have 11q23 abnormalities should be amended to indicate the significance of this cytogenetic abnormality. Further studies are needed to determine whether this category should be separated from the category acute myeloid leukemia with multilineage dysplasia. of 336 cases by Goasguen and colleagues 25 and 12% in a series of 162 cases by Tamura and associates. 29 The current definition of AML with multilineage dysplasia requires dysplastic changes in only 2 cell lines, accounting for the higher percentage (38%) of cases in that category in the present study. 3,4 Goasguen et al 25 found that the presence of dysgranulopoiesis was associated with a significant decrease in the ability to achieve a complete remission, while Tamura et al 29 found the presence of trilineage dysplasia to be associated with lower disease-free survival. Dysplastic changes of granulocytes and megakaryocytes were associated with a worse event-free survival in a series of 102 patients with AML by Gahn et al. 28 The present study suggests that the less rigid criteria of multilineage dysplasia (2 cell lines involved rather than 3) retain clinical significance, with a significantly worse overall survival observed in these patients. In addition, the present study suggests that AMLs with multilineage dysplasia behave in a similar manner whether they arise from myelodysplasia or are de novo. Both the WHO and RPC classifications include a lowering of the blast cell count for acute leukemia to 20%, Am J Clin Pathol 2003;119:672-680 677 677 DOI: 1309/EM7KCQR4GLMHRCX4 677

Arber et al / ACUTE MYELOID LEUKEMIA CLASSIFICATION AML, with changes of recurrent favorable cytogenetic abnormalities AML, not otherwise specified AML with multilineage dysplasia or therapy-related AML with recurrent favorable cytogenetic abnormalities AML, not otherwise categorized AML with multilineage dysplasia or therapy-related 11q23 AML Figure 4 Comparison of the overall survival for the major disease groups in the Realistic Pathologic Classification. P <.0001. Figure 5 Comparison of overall survival of significant disease groups in the World Health Organization classification. P <.001. compared with 30% in the FAB classification. A study by Estey et al 34 showed a similar outcome for patients treated with AML-type therapy with lower blast cell counts compared with patients with 30% or more blasts. The present study supports the lowering of the blast cell count for a diagnosis of AML. The majority of cases with 20% to 30% blast cells were classified as AML with multilineage dysplasia or therapy-related AML, and patients had a virtually identical overall survival to the AML groups with 30% or more blasts. After removal of cases with multilineage dysplasia or recurring cytogenetic abnormalities, few cases remained in many of the disease subgroups of the WHO classification of AML, not otherwise categorized. Of those with sufficient numbers for analysis, no significance could be determined for such subdivisions of this category. However, taken as a whole, the AML, not otherwise categorized of the WHO classification and AML, NOS of the RPC classification were of intermediate risk compared with AMLs with favorable recurrent cytogenetic abnormalities and AML with multilineage dysplasia or therapy-related AML (Figure 4). Therefore, these general categories seem to be significant, but there seems to be little relevance to subdividing them based on cytochemical findings. Future subdivision of this group based on molecular genetic differences may be of value as more recurring abnormalities are defined. Although the same therapy regimen was not used in all cases in the present study, the major disease groups remained significantly different when subsets of cases in general treatment groups were analyzed. These groups included different types of transplantation (allogeneic and autologous) and cases without transplantation. A similar study of a large number of patients treated in an identical manner would be needed to definitely determine the significance of each of the proposed disease groups. However, such a study would be difficult to undertake owing to the recognized advantages of certain treatment strategies that are common for specific leukemia types, such as the core binding factor leukemias and acute promyelocytic leukemia. 12,32,35,36 The present study suggests that lowering the blast cell count for AML to 20% is justified, with most cases added by this criterion having an overall survival that is similar to AML with multilineage dysplasia. The detection of recurrent cytogenetic abnormalities and multilineage dysplasia provides useful prognostic information in the classification of AML. Six significant disease types were confirmed in the present study: (1) AML with t(15;17), (2) AML with inv(16), (3) AML with t(8;21), (4) AML with 11q23, (5) AML with multilineage dysplasia or therapy-related AML, and (6) AML, NOS. Diseases such as acute megakaryoblastic leukemia without multilineage dysplasia, which occurs more often in infants, also may be of significance. However, de novo acute megakaryoblastic leukemia was not sufficiently represented in the present study for any conclusion to be made. Proposed subclassifications of the broad category of AMLs without multilineage dysplasia or recurrent cytogenetic abnormalities do not seem appropriate at present, particularly based on cytochemical evidence of monocytic or myelomonocytic differentiation. The identification of additional cytogenetic or molecular genetic abnormalities in AML should permit the future identification of other clinically significant disease subgroups in the AML with multilineage dysplasia and the AML, NOS disease groups. A revision of the RPC that reflects the present findings is provided in Table 6. This revised approach continues 678 Am J Clin Pathol 2003;119:672-680 678 DOI: 1309/EM7KCQR4GLMHRCX4

Hematopathology / ORIGINAL ARTICLE to recognize the significance of recurrent cytogenetic abnormalities, and a group of AML with 11q23 abnormalities could be taken from the AML, NOS group after cytogenetic studies are complete. From the Divisions of 1 Pathology, 2 Hematology/Bone Marrow Transplantation, and 3 Biostatistics, City of Hope National Medical Center, Duarte, CA. Supported in part by grants CA-33572 and CA-30206 from the National Institutes of Health, Bethesda, MD. Address correspondence to Dr Arber: Clinical Laboratories, Stanford University Medical Center, 300 Pasteur Dr, H1507, Stanford, CA 94305-5627. References 1. Bennett JM, Catovsky D, Daniel MT, et al. Proposals for the classification of the acute leukemias. Br J Haematol. 1976;33:451-458. 2. Bennett JM, Catovsky D, Daniel MT, et al. Proposed revised criteria for the classification of acute myeloid leukemia: a report of the French-American-British Cooperative Group. Ann Intern Med. 1985;103:626-629. 3. Jaffe ES, Harris NL, Stein H, et al, eds. Pathology and Genetics of Tumours of Haematopoietic and Lymphoid Tissues. Lyon, France: IARC Press; 2001. World Health Organization Classification of Tumours. 4. Arber DA. Realistic pathologic classification of acute myeloid leukemias. Am J Clin Pathol. 2001;115:552-560. 5. Martinez-Climent JA, Lane NJ, Rubin CM, et al. Clinical and prognostic significance of chromosomal abnormalities in childhood acute myeloid leukemia de novo. Leukemia. 1995;9:95-101. 6. Grimwade D, Walker H, Oliver F, et al. The importance of diagnostic cytogenetics on outcome in AML: analysis of 1,612 patients entered into the MRC AML 10 trial. Blood. 1998;92:2322-2333. 7. Raimondi SC, Chang MN, Ravindranath Y, et al. Chromosomal abnormalities in 478 children with acute myeloid leukemia: clinical characteristics and treatment outcome in a cooperative Pediatric Oncology Group study: POG 8821. Blood. 1999;94:3707-3716. 8. Fenaux P, Preudhomme C, Lai JL, et al. Cytogenetics and their prognostic value in de novo acute myeloid leukaemia: report on 283 cases. Br J Haematol. 1989;73:61-67. 9. Dastugue N, Payen C, Lafage-Pochitaloff M, et al. Prognostic significance of karyotype in de novo adult acute myeloid leukemia. Leukemia. 1995;9:1491-1498. 10. Slovak ML, Kopecky KJ, Cassileth PA, et al. Karyotypic analysis predicts outcome of preremission and postremission therapy in adult acute myeloid leukemia: a Southwest Oncology Group/Eastern Cooperative Oncology Group Study. Blood. 2000;96:4075-4083. 11. Grimwade D, Walker H, Harrison G, et al. The predictive value of hierarchical cytogenetic classification in older adults with acute myeloid leukemia (AML): analysis of 1065 patients entered into the United Kingdom Medical Research Council AML11 trial. Blood. 2001;98:1312-1320. 12. Bloomfield CD, Lawrence D, Byrd JC, et al. Frequency of prolonged remission duration after high-dose cytarabine intensification in acute myeloid leukemia varies by cytogenetic subtype. Cancer Res. 1998;58:4173-4179. 13. Schoch C, Kern W, Krawitz P, et al. Dependence of agespecific incidence of acute myeloid leukemia on karyotype [letter]. Blood. 2001;98:3500. 14. Paietta E, Andersen J, Gallagher R, et al. The immunophenotype of acute promyelocytic leukemia (APL): an ECOG study. Leukemia. 1994;8:1108-1112. 15. Guglielmi C, Martelli MP, Diverio D, et al. Immunophenotype of adult and childhood acute promyelocytic leukaemia: correlation with morphology, type of PML gene breakpoint and clinical outcome: a cooperative Italian study on 196 cases. Br J Haematol. 1998;102:1035-1041. 16. Khalidi HS, Medeiros LJ, Chang KL, et al. The immunophenotype of adult acute myeloid leukemia: high frequency of lymphoid antigen expression and comparison of immunophenotype, French-American-British classification, and karyotypic abnormalities. Am J Clin Pathol. 1998;109:211-220. 17. Kita K, Nakase K, Miwa H, et al. Phenotypical characteristics of acute myelocytic leukemia associated with the t(8;21)(q22;q22) chromosomal abnormality: frequent expression of immature B-cell antigen CD19 together with stem cell antigen CD34. Blood. 1992;80:470-477. 18. Nakamura H, Kuriyama K, Sadamori N, et al. Morphological subtyping of acute myeloid leukemia with maturation (AML- M2): homogeneous pink-colored cytoplasm of mature neutrophils is most characteristic of AML-M2 with t(8;21). Leukemia. 1997;11:651-655. 19. Hurwitz CA, Raimondi SC, Head D, et al. Distinctive immunophenotypic features of t(8;21)(q22;q22) acute myeloblastic leukemia in children. Blood. 1992;80:3182-3188. 20. Arber DA, Glackin C, Lowe G, et al. Presence of t(8;21)(q22;q22) in myeloperoxidase-positive, myeloid surface antigen negative acute myeloid leukemia. Am J Clin Pathol. 1997;107:68-73. 21. Larson RA, Williams SF, Le Beau MM, et al. Acute myelomonocytic leukemia with abnormal eosinophils and inv(16) or t(16;16) has a favorable prognosis. Blood. 1986;68:1242-1249. 22. Le Beau MM, Larson RA, Bitter MA, et al. Association of an inversion of chromosome 16 and abnormal marrow eosinophils in acute myelomonocytic leukemia. N Engl J Med. 1983;309:630-636. 23. Adriaansen HJ, te Boekhorst PAW, Hagemeijer AM, et al. Acute myeloid leukemia M4 with bone marrow eosinophilia (M4Eo) and inv(16)(p13q22) exhibits a specific immunophenotype with CD2 expression. Blood. 1993;81:3043-3051. 24. Paietta E, Wiernik PH, Andersen J, et al. Acute myeloid leukemia M4 with inv(16) (p13q22) exhibits a specific immunophenotype with CD2 expression [letter]. Blood. 1993;82:2595. 25. Goasguen JE, Matsuo T, Cox C, et al. Evaluation of the dysmyelopoiesis in 336 patients with de novo acute myeloid leukemia: major importance of dysgranulopoiesis for remission and survival. Leukemia. 1992;6:520-525. 26. Head DR. Revised classification of acute myeloid leukemia. Leukemia. 1996;10:1826-1831. Am J Clin Pathol 2003;119:672-680 679 679 DOI: 1309/EM7KCQR4GLMHRCX4 679

Arber et al / ACUTE MYELOID LEUKEMIA CLASSIFICATION 27. Lemez P, Galikova J, Haas T. Erythroblastic and/or megakaryocytic dysplasia in de novo acute myeloid leukemias M0-M5 show relation to myelodysplastic syndromes and delimit two main categories. Leuk Res. 2000;24:207-215. 28. Gahn B, Haase D, Unterhalt M, et al. De novo AML with dysplastic hematopoiesis: cytogenetic and prognostic significance. Leukemia. 1996;10:946-951. 29. Tamura S, Takemoto Y, Wada H, et al. Significance of trilineage myelodysplasia in de novo acute myeloid leukaemia during remission rather than at diagnosis. Br J Haematol. 1998;101:743-748. 30. Kaplan EL, Meier P. Non-parametric estimation from incomplete observations. J Am Stat Assoc. 1958;53:457-481. 31. Pearson K. In the chi-square test of goodness of fit. Biometrika. 1922;9:22-27. 32. Tallman MS, Andersen JW, Schiffer CA, et al. All-transretinoic acid in acute promyelocytic leukemia. N Engl J Med. 1997;337:1021-1028. 33. Lo Coco F, Nervi C, Avvisati G, et al. Review and meeting report: acute promyelocytic leukemia: a curable disease. Leukemia. 1998;12:1866-1880. 34. Estey E, Thall P, Beran M, et al. Effect of diagnosis (refractory anemia with excess blasts, refractory anemia with excess blasts in transformation, or acute myeloid leukemia [AML]) on outcome of AML-type chemotherapy. Blood. 1997;90:2969-2977. 35. Ghaddar HM, Plunkett W, Kantarjian HM, et al. Long-term results following treatment of newly-diagnosed acute myelogenous leukemia with continuous-infusion high-dose cytosine arabinoside. Leukemia. 1994;8:1269-1274. 36. Byrd JC, Dodge RK, Carroll A, et al. Patients with t(8;21)(q22;q22) and acute myeloid leukemia have superior failure-free and overall survival when repetitive cycles of high-dose cytarabine are administered. J Clin Oncol. 1999;17:3767-3775. 680 Am J Clin Pathol 2003;119:672-680 680 DOI: 1309/EM7KCQR4GLMHRCX4