Minimal residual disease in acute myeloid leukemia in children--standardization and evaluation of immunophenotyping in the AML-BFM-98 study

BACKGROUND: Minimal residual disease is a prognostic factor in AML. However, the impact on treatment stratification is not established. The AML-BFM 98 MRD study started in 1/2000 in order to evaluate, standardize and establish immunophenotyping in AML in children. METHODS: In a first phase the participating laboratories in Muenster, Goettingen, Vienna and Prague agreed on identical antibody-panels and standardized procedures of sample processing, analysis and data management. The consensus panel was evaluated and adapted to 3- and 4-color flowcytometry. The complete panel was applied to each follow-up sample in orderto minimize the risk offalse negative results due to the loss or shift of antigens during treatment, a known phenomenon in myeloid blasts. Between 1/2000 and 9/2001 165 of 198 protocol patients were analysed at diagnosis, in 149 children at least two follow-up samples were available. RESULTS: Three kinds of immunophenotypes could be defined. Asynchronous expression of stem cell and myeloid antigens i. e. CD34/CD117 combined with CD13/CD15 had a low specificity because precursors in regenerating or normal bone marrow expressed this pattern in 0.47 % (0.1 to 1.5 %). The aberrant co-expression of stem cell antigens and lymphatic antigens such as CD7 or CD2 showed a median level of specificity (0.07 % (0.04 to 0.19 %). Aberrant expression ofstem cell antigens combined with B-lymphatic (CD19, CD10) or NK-cell antigen (CD56) showed the best specificity. The maximal level in normal bone marrow was 0.05 %. Sensitivity of different immunophenotypes was evaluated by diluting known leukemic blasts in regenerating bone marrow. Minimal level of sensitivity was found to be at 10 (-3) to 5 x 10 (-4). According to these data highiy specific immunophenotypes could be detected in 33 %, median specificity was seen in 71 % and low specificity was seen in 88 % of the protocol patients. Two laboratories analyzed simultaneously 17 samples of children with AML from diagnosis and during therapy. A high correlation of blast quantification could be demonstrated (correlation r (2) = 0.98; blasts < 5 % r (2) = 0.91). In addition, two independent explorers quantified the raw data of 16 samples. All results correlated well (r (2) = 0.97; blasts < 5 % r (2) = 90.94). Conclusion: The prospective study phase, started 1/2002, aims to test the impact of MRD diagnostics as an independent prognostic factor in AML in children. This might facilitate future treatment stratification and consequently optimize outcome.