Prognostic factors in children and adolescents with acute myeloid leukemia(excluding children with Down syndrome and acute promyelocytic leukemia): univariate and recursive partitioning analysis of patients treated on Pediatric Oncology Group (POC) Study 8821

The purpose of the paper was to define clinical or biological features asso ciated with the risk for treatment failure for children with acute myeloid leukemia. Data from 560 children and adolescents with newly diagnosed acute myeloid leukemia who entered the Pediatric Oncology Group Study 8821 from June 1988 to March 1993 were analyzed by univariate and recursive partition ing methods. Children with Down syndrome or acute promyelocytic leukemia we re excluded from the study. Factors examined included age, number of leukoc ytes, sex, FAB morphologic subtype, cytogenetic findings, and extramedullar y disease at the time of diagnosis. The overall event-free survival (EFS) r ate at 4 years was 32.7% (s.e. = 2.2%). Age greater than or equal to 2 year s, fewer than 50 x 10(9)/l leukocytes, and t(8;21) or inv(16), and normal c hromosomes were associated with higher rates of EFS (P value = 0.003, 0.049 , 0.0003, 0.031, respectively), whereas the M5 subtype of AML (P value = 0. 0003) and chromosome abnormalities other than t(8;21) and inv(16) were asso ciated with lower rates of EFS (P value = 0.0001). Recursive partitioning a nalysis defined three groups of patients with widely varied prognoses: fema le patients with t(8;21), inv(16), or a normal karyotype (n = 89) had the b est prognosis (4-year EFS = 55.1%, s.e. = 5.7%); male patients with t(8;21) , inv(16) or normal chromosomes (n = 106) had an intermediate prognosis (4- year EFS = 38.1%, s.e. = 5.3%); patients with chromosome abnormalities othe r than t(8;21) and inv(16) (n = 233) had the worst prognosis (4-year EFS = 27.0%, s.e. = 3.2%). One hundred and thirty-two patients (24%) could not be grouped because of missing cytogenetic data, mainly due to inadequate marr ow samples. The results suggest that pediatric patients with acute myeloid leukemia can be categorized into three potential risk groups for prognosis and that differences in sex and chromosomal abnormalities are associated wi th differences in estimates of EFS. These results are tentative and must be confirmed by a large prospective clinical trial.