Efficacy and tolerability of gemtuzumab ozogamicin (anti-CD33 monoclonal antibody, CMA-676, Mylotarg®) in children with relapsed/refractory myeloid leukemia

Background Gemtuzumab ozogamicin (GO) is a cytotoxic anti-CD33 monoclonal antibody that has given promising preliminary results in adult myeloid CD33+ AML. We conducted a retrospective multicenter study of 12 children treated with GO on a compassionate basis (median age 5.5 y). Three patients (2 MDS/AML, 1 JMML) were refractory to first-line treatment, 8 patients with de novo AML were in refractory first relapse, and one patient with de novo AML was in 2nd relapse after stem cell transplantation (SCT). CD33 expression exceeded 20% in all cases. Methods GO was administered alone, at a unit dose of 3–9 mg/m2, once (3 patients), twice (3 patients), three (5 patients) or five times (1 patient). Mean follow-up was 128 days (8–585 d). Results There were three complete responses (25%) leading to further curative treatment (SCT). Treatment failed in the other nine patients, and only one patient was alive at the end of follow-up. NCI-CTC grade III/IV adverse events comprised hematological toxicity (n = 12), hypertransaminasemia (n = 2), allergy and hyperbilirubinemia (1 case each). There was only one major adverse event (grade IV allergy). No case of sinusoidal obstruction syndrome occurred. Conclusion These results warrant a prospective trial of GO in a larger population of children with AML.

An alternative to chemotherapy (CT) is to target leukemic blasts with monoclonal antibodies. Approximately 80% to 90% of pediatric AML patients have myeloid blast cells that express the CD33 surface antigen [10]. This antigen is present on normal hematopoietic progenitor cells but not on normal hematopoietic stem cells or on non hematopoietic cells [11]. Gemtuzumab ozogamicin (GO) is an immunoconjugate consisting of a humanized anti-CD33 IgG 4κ antibody linked to the cytotoxic compound Nacetyl-γ-calicheamicin dimethylhydrazine, a member of the enediyne antitumor antibiotic family [12,13]. GO selectively targets CD33+ cells and was specifically developed for the treatment of AML. After receptor binding, the complex is rapidly internalized and calicheamicin is released intracellularly. Calicheamicins are known for their extreme cytotoxic potency, attributed to doublestranded DNA cleavage at specific sequences [14,15]. In phase I/II studies, approximately 30% of adults with relapsed AML responded to GO [12,13,16]. Severe myelosuppression is common, however, and platelet recovery can be slow, probably owing to CD33-expressing platelet precursor damage [13]. Toxicity is relatively mild compared with classical multiagent CT, especially with regard to mucositis and infections [17], but GO can cause severe liver toxicity in the form of a sinusoidal obstruction syndrome (SOS) [18,19] Several factors can increase the risk of hepatotoxicity, including previous stem cell transplantation (SCT) [20]. Prior exposure to GO is also known to increase the risk of SOS in patients who subsequently undergo myeloablative SCT [21] GO has been approved in the United States for the treatment of elderly patients with relapsed AML [16].
GO has rarely been used in children. Sievers et al reported preliminary results of a phase I ascending-dose study of GO in 18 children with relapsed or refractory AML [22]. Likewise, Zwaan et al used GO (up to three doses) to treat 15 children with relapsed/refractory CD33+ AML, on a compassionate-use basis [23,24]. More recently Arceci et al reported a dose-escalation study of 29 children with multiple relapsed or primary refractory AML [25]. Here we report our experience with GO monotherapy in 12 children with relapsed or refractory AML.

Methods
This retrospective study involved 12 children treated with GO between March 1999 and April 2004 on a compassionate-use basis in four pediatric centers. GO therapy was approved by the french agency for health product safety (AFSSAPS), and the guardians' and/or patients' informed consent was obtained. The cutoff date for this analysis was 30 September 2004. The children had myelodysplasia (MDS)/AML refractory to standard induction therapy (n = 2), juvenile myelomonocytic leukemia (JMML) refractory to several cytotoxic drugs and retinoic acid (n = 1), first relapse of AML refractory to reinduction therapy (n = 8), or AML in second relapse after SCT (n = 1). The patients' characteristics at initial diagnosis are shown in Table 1 First-line CT was based on four different intensive regimens consisting of repeated courses of cytarabine plus intercalating agents. The patient with refractory JMML ( Table 2, patient #1) received a combination of 6-mercaptopurine, etoposide, cytarabine, hydroxyurea and 6-thioguanine, plus 13 cis-retinoid acid, without responding. The two patients with primary refractory MDS/AML (patients #5 and #7) had received standard induction CT (cytarabine 200 mg/m 2 /d × 7 days plus mitoxantrone 12 mg/m 2 /d × 5 days), without responding. They were further treated with high-dose cytarabine and amsidine, but again no response was obtained. One of the two patients then received several courses of low-dose cytarabine plus etoposide before undergoing matched unrelated donor (MUD)-SCT while in partial remission. He relapsed 14 months later and was again refractory to low-dose cytarabine plus etoposide at the time of GO therapy.
The nine relapsing patients had been treated with FLAG (a combination of fludarabine, high-dose cytarabine and granulocyte colony-stimulating factor), [26] with or without anthracyclines. Eight of these nine patients (patients #2, #4, #6, #8, #9, #10, #11, #12) were in refractory reinduction CT before receiving GO. The ninth patient (UPN #3) entered CR2 after reinduction CT and underwent cord-blood SCT. This patient had no sign of active hepatic graft-versus-host disease or SOS and had normal transaminase and bilirubin values when GO was administered for very early post-SCT relapse.
The median interval between initial diagnosis and GO administration was 11.5 months (4.8-45.5 months). All patients had CD33+ myeloid leukemia at the time of GO therapy, with a median proportion of bone marrow blasts of 47% (20 -98%). The median WBC count was 3.2 × 10 9 /L (0.2-22.0 × 10 9 /L). GO was given to 8 patients at a total dose of 9 mg/m 2 , administered in a single dose (n = 3) or divided into 3 doses given on days 1, 4 and 7 (n = 5). Patient #9 received 5 doses of 3 mg/m 2 (days 1, 4, 7, 28 and 31). Patient #10 received two doses of 9 mg/m 2 each (days 1 and 14). Patient #11 received two doses of 7.5 mg/m 2 each (days 1 and 16). Patient #12 received one dose of 6 mg/m 2 (day 1) and another dose of 9 mg/m 2 (day 16) (see Table 2). The unit doses and dosing intervals were based on those used in adults and in the pediatric studies of Sievers et al [22] and Zwaan et al [23], and were decided on by the physicians in charge of each patient. Fractionated doses were adopted secondarily, as they were reported to be associated with less SOS in adults (Raffoux E et al, annual meeting of the french hematologic society SFH, 2000, abstract). This schedule was chosen in the ongoing french MyloFrance protocol in adults.
Complete responses to GO were defined on the basis of the following consensus criteria: a bone marrow blast proportion of 5% or less, in the absence of leukemic cells in peripheral blood or elsewhere. The definition of CR included adequate recovery of peripheral blood cell counts (granulocytes > 1 × 10 9 /L and platelets > 100 × 10 9 /L with at least one week of transfusion independence). CRp was defined as a response with incomplete platelet recovery but with platelet transfusion independence [12].

Results
Responses and major toxicities are summarized in Table  2.

Responses
Median follow-up after the beginning of GO treatment was 128 days (8-585 d). As shown in Table 2  Alive in PR ♣ BM: bone marrow; SCT, stem cell transplantation; MUD, matched unrelated donor; CR, complete remission; CRp, complete remission without total platelet recovery; PR, partial remission; DP, disease progression. *Bone marrow blast percentage. CD33 expression was greater than 20% before GO in all cases. ¥All patients experienced NCI-CTC grade III to IV hematological toxicity, which is not mentioned here. ‡This patient was treated with 3 × 3 mg/m 2 of GO leading to a reduction in BM blasts from 86 to 20% at d28. Because of good tolerance, two more doses of 3 mg/m 2 were administered and CR was obtained at d39. §This patient relapsed 69 days after MUD-SCT but a drastic reduction of BM blasts was obtained with 5 weekly doses of 3 mg/m 2 GO without any significant adverse effects. He received GO maintenance therapy but died 12 months after MUD-SCT. V Death ♣ This patient was scheduled to receive MUD-SCT.
related to the underlying disease. He died in blast crisis 238 days after GO initiation. Patient #12 did not respond by day 31, after two doses of GO (day 1 = 6 mg/m 2 , day 16 = 9 mg/m 2 ). He then received alternative treatments and was alive in partial remission 120 days after GO administration. MUD-SCT was scheduled one month later.
Only one of the 12 patients was alive at the cutoff date for this analysis.

Toxicity
In the three responding patients (#5, #9, #11), GO was well tolerated, with the exception of NCI-CTC grade III-IV hematologic toxicity. Patient #1 developed severe infusion-related hypotension (NCI-CTC grade IV) and fever, necessitating fluid support.

Discussion
We treated 12 children with relapsed or refractory myeloid leukemia with GO (3-18 mg/m 2 , up to 5 infusions) on a compassionate-use basis. The prognosis of such children is very poor: three of our patients with newly diagnosed JMML or MDS/AML were refractory to several different induction regimens; the remaining nine patients were either refractory to reinduction therapy or in second relapse after SCT. Further use of intensive CT is limited by its toxicity [28,29].
GO monotherapy thus yielded a response rate of 25% (3/ 12 patients). This rate is similar to that obtained in adults with relapsed AML [12,13] Considering the toxicity of GO, NCI-CTC grade IV hematologic toxicity occurred in all 12 children of our series. In responders, it was probably related to expression of the CD33 target by normal bone marrow progenitors. In non responders, it was difficult to determine the respective roles of GO and the underlying leukemia. With regard to non hematologic adverse events, GO was relatively well tolerated compared to intensive CT, with no cases of mucositis and only one noteworthy infection. This is in keeping with previous pediatric results [23,25]. GO therapy is an established risk factor for SOS [18][19][20]. No cases of SOS occurred in our series, even after SCT in the three responding patients. Zwaan et al reported only one case of SOS among 15 GO-treated children, occurring after SCT [23]. Arceci et al described an 24% overall incidence of SOS [25]. SOS developed in 6 of 7 patients after they underwent SCT. They had been exposed to GO within 3.5 to 4 months of transplantation. Two other patients who underwent SCT more than 4 months from GO exposure did not developed SOS. These findings are consistent with similar adult series [21]. Taken together, these data suggest that the time from treatment with GO to SCT for refractory/relapsed AML appears to be important with respect to the incidence of SOS after SCT. The incidence of liver dysfunction in GO-treated children is similar to that observed in the largest adult series [13], in which liver dysfunction was generally transient and never life-threatening.
GO was well tolerated, even during repeated infusions. One of our patients (#9) received a cumulative dose of approximately 45 mg/m 2 over a 14-month period. One patient in Zwaan's series was treated repeatedly with GO, with relatively long intervals between the infusions, and responded each time without showing signs of additional toxicity [23]. Palliative treatment thus appears to be feasible with repeated GO infusions.

Conclusion
This report confirms that GO is clinically active in children with relapsed/refractory CD33+ AML. Despite its good tolerability, GO monotherapy only induces a response rate of 25 to 33%, similar to that previously reported in adults. In a recent trial, GO was combined with intensive CT as first-line treatment for AML in 72 patients aged 17 from 59 years [31]. Certain schedules of CT-GO combination induction therapy yielded CR in 91% of patients, 78% of whom were in continuous CR at 8 months. GO should be tested prospectively in a larger population of children with AML, with more stringent eligibility criteria and treatment schedules, in association with CT.