Association of CD33 Expression Level with Disease Risk-Group Classification and Induction Response In Pediatric AML: A Report From the Children's Oncology Group

Abstract 2732

Our previous analysis of diagnostic AML bone marrow (BM) samples from a subset of patients enrolled on Children's Oncology Group (COG) AAML03P1, a pilot study in which conventional chemotherapy was used in combination with the CD33 targeted therapeutic gemtuzumab ozogamicin (GO), demonstrated that CD33 expression is highly variable in pediatric AML and that low or absent CD33 expression was not associated with inferior clinical response. Moreover, patients with the highest CD33 expression did not have superior outcomes. Low CD33 expression was associated with low risk disease [core binding factor (CBF) AML e.g. t(8;21), inv(16) or t(16;16), CEBPA mutated AML, NPM1 mutated AML] whereas the highest CD33 expression levels were seen in patients with high-risk disease [FLT3/ITD+ disease with allelic ratio >0.4, high risk cytogenetics e.g. -7, -5, -5q]. These findings refute previous adult AML data in which CD33 expression is directly correlated with response to single agent GO and suggest, within AAML03P1, that clinical response is linked to underlying disease biology. In this larger analysis, we prospectively evaluated CD33 expression levels of AML blasts isolated from 676 pediatric diagnostic BM samples (238/340 and 438/968 patients enrolled on COG AAML03P1 and COG AAML0531 respectively) to determine whether this association persists in a larger cohort. CD33 expression, as defined by mean fluorescent intensity (MFI) of the blast population, varied over 2-log fold, and a median MFI of 128 was observed (range 3–1550.07). The study population was divided into quartiles (Q) based on CD33 expression (n= 169 patients per quartile). Median MFI was 37.49 (range 3–62) for Q1, 90 (range 62.21–128) for Q2, 171 (range 128–245) for Q3 and 349 (range 245.52–1550.07) for Q4. Samples were also screened for disease-relevant molecular mutations: 89/650 (14%) were FLT3/ITD positive; 69/587 (12%) were positive for NPM1 mutations; and 35/585 (6%) positive for CEBPA mutations. FLT3/ITD prevalence significantly increased with increasing CD33 quartile (Q1 7%, Q2 10%, Q3 17%, Q4 20%; p<0.001), whereas no definitive trend in prevalence was observed for NPM1 (Q1 7%, Q2 14%, Q3 15%, Q4 12%; p=0.158) or CEBPA mutations (Q1 6%, Q2 8%, Q3 6%, Q4 4%; p=0.307). Cytogenetic data was available for 613 (91%) samples; 177 (29%) CBF AML samples were identified and their prevalence declined with increasing quartile (Q1 51%, Q2 40%, Q3 20%, Q4 6%; p<0.001). There was no apparent association between CD33 expression and high-risk cytogenetics; however, analysis was limited by the small number of patients (9/613, 1.4%) with such mutations. For risk-group classification, complete cytogenetic and molecular data were available for 535 (79%) samples: 204/535 (38%) were classified as low-risk and 64/535 (12%) were defined as high-risk. There was an inverse association between CD33 expression and prevalence of low-risk AML (Q1 59%, Q2 50%, Q3 27%, Q4 17%; p<0.001). In contrast, the prevalence of high-risk disease increased with each quartile (Q1 5%, Q2 8%, Q3 17%, Q4 18%; p<0.001). We observed a higher median CD33 MFI with high-risk disease (median MFI 195.315; range 12–720) than with low-risk (median MFI 80.5; range 5–1550.07) or intermediate-risk (i.e., neither low- nor high-risk) disease (median MFI 163.06; range 7–1351) (p<0.001). Response from end of induction I (CR) was also determined for our patient cohort. Rates of CR were similar across CD33-expression quartiles (Q1 78%, Q2 75%, Q3 75%, Q4 72%; p=0.581). Moreover, CR rates for each risk group did not vary across quartiles (low-risk: Q1 84%, Q2 85%, Q3 88%, Q4 82%; p=0.917; intermediate-risk: Q1 77%, Q2 65%, Q3 72%, Q4 71%; p=0.594; high-risk: Q1 57%, Q2 67%, Q3 60%, Q4 72%; p=0.801). This large scale analysis supports our earlier finding that CD33 expression in pediatric AML is heterogeneous and associated with conventional risk-group criteria. As outcome data matures from COG AAML0531, the phase III counterpart of AAML03P1 in which patients are randomized to GO treatment, we will determine whether the addition of GO improves survival of patients with low versus high CD33 expression and whether this finding translates into enhanced outcomes for associated disease-risk groups.