Heirarchical Clustering of Immunophenotypic Cell Surface Antigen Expression Identifies Clinically Meaningful Cohorts in Childhood AML: A Report from the Children's Oncology Group Protocol AAML0531

Background: Acute myeloid leukemia (AML) is a heterogeneous disease with diverse cell surface antigen expression. Multidimensional flow cytometry (MDF) quantifies the expression of such surface antigens on neoplastic cells, permitting an objective means for sub-classification of this disorder.

Objective: To investigate the relationship between the presenting immunophenotype and response to therapy in a large, controlled study of pediatric AML patients.

Methods: Of 1022 newly diagnosed pediatric patients with de novo AML enrolled on protocol AAML0531, 769 satisfied three criteria for this study: (1) submission of a blood or bone marrow sample for MDF at diagnosis, (2) proper consent for specimen testing and (3) leukemia comprising >10% of non-erythroid cells by MDF. The diagnostic AML tumor population was identified by gating on CD45 vs log-SSC. A fifteen-dimensional immunophenotypic expression profile (IEP) was defined by computing mean fluorescent intensities, antigen intensity coefficient of variations, and light scatter characteristics of each leukemia without analyst imposed cut-offs. 

Unsupervised hierarchical clustering analysis was performed to mathematically define groups of patients with similar diagnostic IEPs. Selection of the appropriate number of clusters was accomplished by minimizing within-cluster variation while maximizing between-cluster variation. The rate of measurable residual disease (MRD) by MDF at the end of initial induction therapy (EOI1) and event free survival (EFS) was determined for each group and compared to the mean MRD rate and EFS for the entire cohort. Of the 769 total patients, 643 (84%) had bone marrow specimens submitted for MRD by MDF at EOI1.

Results:

Of the 643 patients with bone marrows submitted for MRD by MDF at EOI1, 223 (35%) were positive. The 5 year EFS rate was 49% for the study (N=769). Hierarchical clustering analysis of the IEPs defined 11 groups of patients (A-K) with similar quantitative diagnostic immunophenotypes.

Five immunophenotypic groups of patients were associated with comparable EOI1 MRD rates (15%-36%) and EFS (50%-69%) to the study average. These 5 groups (A-E) consisted of immunophenotypes close to the normal counterpart of myeloid progenitor cells (Groups A, B, C; N=266, 77 and 106) or normal monocytes (Groups D and E; N=41 and 68).

Two immunophenotypic groups (Group F and G) of patients were associated with significantly higher rates of MRD (Group F: 54%, p<0.001; Group G: 88%, p<0.001) and lower EFS (Group F: 28%, p<0.001; Group G: 19%, p=0.006) than average for the study. The defining different-from-normal immunophenotypic features of Group F (N=81) were expression of CD34 with a lack of HLA-DR. The defining immunophenotypic features of Group G (N=16) were high intensity CD56 expression, a lack of HLA-DR, dim-to-negative CD38 expression and dim-to-negative CD45 expression. The majority of patients in both groups were stratified as low or standard risk (86% and 100%).

Three immunophenotypic groups (H-J) had low rates of MRD (20%-24%) but poor EFS (27%-39%). The defining different-from-normal immunophenotypic features of Group H (N=52) were expression of CD56 and CD13 with a lack of CD34. The defining different-from-normal immunophenotypic features of Group I (N=11) were CD34 positive expression with a lack of CD117. The defining different-from-normal features of Group J (N=28) were bright CD11b expression with a lack of CD34 and CD117. The majority of patients in these groups were stratified as standard risk (94%, 100%, and 75% respectively).

One immunophenotypic group (K) had a high rate of MRD (50%) and comparable EFS (52%) to the study average. The defining different-from-normal features of Group K (N=23) were expression of CD56 with a lack of CD34 and CD13. 

Conclusions: IEPs can provide a powerful tool to discriminate patient cohorts that have a high probability of relapse and subsequently elucidate when MRD by MDF at EOI1 is predictive of patient outcome. Notably 2 IEP groups (97 patients, 13% of total) had significantly higher MRD rates and were associated with a high probability of disease relapse. IEPs can identify these 2 groups of patients otherwise stratified as good or standard risk, which can provide clinicians with an additional analytical method to more appropriately stratify patients.