Negative Prognostic Impact Of High CD33 Expression Is Negated With The Use Of Gemtuzumab Ozogamicin: A Report From The Children’s Oncology Group

CD33 is expressed on leukemic blasts of most patients with acute myeloid leukemia (AML) and is the target for gemtuzumab ozogamicin (GO), a toxin-conjugated anti-CD33 monoclonal antibody. CD33 expression of leukemic blasts was prospectively quantified within the context of COG AAML0531, a phase III randomized study for de novo AML in which patients were randomized to receive conventional chemotherapy (Arm A) vs. GO + conventional chemotherapy (Arm B) to determine the impact of CD33 expression on outcome within the context of this GO randomization.

CD33 mean fluorescent intensity (MFI) of leukemic blasts was prospectively quantified in 825 diagnostic specimens. Patients were divided into quartiles (Q1-Q4) based on CD33 expression values and these levels were correlated with disease characteristics and outcome by treatment arm for the total study cohort and by cytogenetic/molecular disease risk-group.

Analysis of 3 year outcome by treatment arm (N= 412 for Arm A vs. N=414 for Arm B) demonstrated that patients with high CD33 expression (Q4) in Arm A (no GO) had an overall survival (OS) from diagnosis of 55% vs. 70% for those with lower CD33 expression (Q1-3, P=.014) with a corresponding disease-free survival (DFS) from complete remission (CR) of 41% and 57%, respectively (P=.010). In contrast, for the patients in Arm B (receiving GO therapy) those with and without high CD33 expression had a similar OS from diagnosis (67% vs. 72%, P=.290) with a corresponding DFS from CR of 57% vs. 64%, respectively (P=.255). Comparison of the patients with the highest CD33 expression (Q4) who were treated with (N=105) and without (N=101) GO demonstrated that those who received GO had an OS from diagnosis of 67% versus 55% (P=.196) with a corresponding DFS from CR of 57% vs. 41% (P=.052).

Analysis by cytogenetic/molecular disease risk group also showed that the effect of CD33 expression levels on outcome differed by treatment arm. Among intermediate risk (IR) patients on Arm A (N=200), those with high CD33 expression (Q4) had an OS from diagnosis of 52% vs. 62% for those with lower CD33 expression (P=0.194) with a corresponding DFS from CR of 28% vs. 53% respectively (P=.012). Conversely, for IR patients treated with GO (N=197), outcomes were similar for patients with high (Q4) and low (Q1-3) CD33 expression (OS from diagnosis of 65% vs. 64%, P=.923, DFS from CR of 50% vs. 53%, P=.687). The loss of prognostic impact of high CD33 expression for patients in Arm B may be due to improved response to GO in those with high CD33 expression (OS of IR patients in Q4 from study entry: Arm A (N=65) 52% vs. Arm B (N=70) 65%, P= .234, DFS from CR of IR patients in Q4: Arm A 28% vs. Arm B 50%, P=.033). Accurate sub analysis of the high-risk (HR) group was not feasible due to the very small number of HR patients with high CD33 expression (Q4) in Arm A (N=9) and Arm B (N=16). Similar trends were, however, observed in the low-risk (LR) group. LR patients with high (Q4) CD33 expression treated on Arm A (no GO) had an OS from diagnosis of 69% vs. 84% for those with lower CD33 expression (P=.092) with a corresponding DFS from CR of 68% vs. 64% respectively (P=0.803). For patients in Arm B (GO) those with and without high CD33 expression had an OS from diagnosis of 94% vs. 86%, respectively (P=.316) with a corresponding DFS from CR of 85% vs. 76% (P=.344). Like IR patients, those LR patients with high CD33 expression (Q4) who received GO trended towards improved outcome compared to Q4 patients treated without GO (LR OS from diagnosis: Arm A 69% vs. Arm B 94%, P= .069, LR DFS from CR: Arm A 68% vs. Arm B 85%, P=.195). However, given the small number of LR patients in Q4 (N= 27 Arm A, N=17 Arm B) we cannot state the significance of this finding with certainty.

Taken together our results suggest that, for patients enrolled on AAML0531, high CD33 expression was associated with adverse outcome for those who received standard therapy and GO treatment negated the negative effect of high CD33 expression on clinical outcome for the entire study cohort and in IR and LR patients. This finding may reflect GO’s CD33 dependent mechanism of targeting and the potential for more efficient targeting in the setting of high antigen expression.It is also plausible that repeated exposure to GO, as seen within the context of treatment for all LR and some IR patients (e.g. those that did not undergo hematopoietic stem cell transplant), may also contribute therapeutic benefit within the context of high CD33 expression.