The majority of patients with acute myeloid leukemia (AML) achieve complete remission (CR) after induction therapy. Despite intensive consolidation regimens, less than one-third of adult patients with AML are cured, mainly because of a high incidence of relapse.1 In a recently published randomized trial, treatment with azacitidine increased the overall survival (OS) relative to supportive care in patients with myelodysplastic syndrome (MDS).2 Therapy with decitabine (DAC) also improves the outcomes in MDS, including time to AML transformation or death.3 DAC is also active in AML and produces CRs in a number of patients.4

We and others have reported that DNA methylation increases at relapse in patients with AML and may predict the occurrence of relapse, suggesting its involvement in disease progression and providing a rationale for the use of epigenetic agents in maintenance.5, 6 In this study, we compared DAC with conventional care (CC), including low-dose subcutaneous cytarabine, prolonged intensive chemotherapy (IC) or observation, in patients with AML in their first or subsequent CR.

Inclusion criteria included age ⩾18 years; verified CR or CRp (CR without platelet recovery to >100 × 109/l); adequate organ function (creatinine ⩽2.0 mg/dl, bilirubin ⩽3.5 mg/dl, alanine transaminase (ALT) and aspartate transaminase (AST) ⩽3 times upper limit of normal); and Eastern Cooperative Oncology Group (ECOG) performance status of 0–2. Patients eligible for an allogeneic stem cell transplantation (SCT) were allowed to participate. Exclusion criteria were active and uncontrolled disease or infection, HIV, pregnancy and lactation. The study was conducted between August 2006 and October 2009.

A total of 50 patients (median age 57 years, range 24–79) were enrolled, of which 45 (19 male and 26 female) patients (including 35 in first CR and 10 in subsequent CR) were evaluable. Among them, 27 patients were ⩽60 years and 18 patients were >60 years of age. Seven patients had treatment-related AML and one had secondary AML. Median follow-up was 44.9 months. Among the 45 evaluable patients, 20 received DAC 20 mg/m2 i.v. daily × 5 every 4–8 weeks. DAC was then continued for a total of 12 cycles, until toxicity ⩾grade 3 developed, which was considered unacceptable, or until disease relapsed. Twenty-five patients were randomized to CC, including 6 (24%), 9 (36%) and 10 (40%) patients who were assigned to observation, low-dose subcutaneous ara-C (LDAC) and IC , respectively. The length of CC treatment was at the discretion of the treating physicians. The primary end point was incidence of relapse at 1 year. We monitored minimal residual disease (MRD) by multi-parameter flow cytometry (MFC) and studied DNA methylation as a predictor of outcome.

Baseline characteristics were relatively balanced between patients treated with DAC and those treated with CC (Table 1a). Patients receiving DAC were older (median age, 62 years (range, 24–79)) than those in the CC arm (median age, 53 years (range, 31–79)) (P=0.11). In all, 55% of patients >60 years in the DAC group versus 28% patients in the CC group. All patients in CC group who received IC were ⩽60 years. Overall,10 (22%) patients had an unfavorable karyotype (including −5/−7/complex, 11q/complex, isolated −5 or −7), including 25% in the DAC and 20% in the CC arm. Thirty-four (76%) patients had an intermediate karyotype (DAC, 75%; CC, 76%) and one relapsed patient with favorable cytogenetics (inv(16)) received CC.

Table 1a Patients’ characteristics
Full size table

The median number of cycles of DAC administered was 4.5 (range, 1–12). For the CC group, the median number of LDAC cycles was 5 (range, 1–14) and the median number of IC cycles was 1 (range, 1–2). Two patients underwent SCT in the CC arm, of which one is alive to date. Three patients underwent SCT in the DAC arm, of which one is alive. Toxicity in the DAC arm was limited; grade 3 and 4 events are shown in Table 1b.

Table 1b Toxicities in the DAC arm
Full size table

The study was terminated early by the institutional data safety monitoring board (DSMB) because of the higher incidence of relapse at 1 year in the DAC arm (Figures 1a–c). At a median follow-up time of 44.9 months (range, 6.3–62.3), 10 patients on the DAC arm and 10 patients on the CC arm remained in remission, and 10 of 20 (50%) on DAC versus 15 of 25 (60%) on the CC arm relapsed (P=0.7) (Figure 1c). No significant difference in OS or event-free survival (EFS) was observed, with OS rate 45% in the DAC versus 36% in the CC group (P=0.9) and the EFS rate 35% in the DAC versus 32% in the CC group (P=0.9) (Figures 1a and b). None of the seven patients who received maximum number of cycles (11–12 cycles of DAC) have relapsed, and six of them are alive.

Figure 1
figure 1

(a–c) Kaplan–Meier estimates for 45 patients, including 25 patients on the CC arm and 20 patients on the DAC arm, of (a) OS, (b) EFS and (c) incidence of relapse are shown. There was no significant difference noted between the groups (P-values of 0.9, 0.9 and 0.7, respectively). (d–f) Kaplan–Meier estimates for MFC for 38 patients, including 30 patients with negative and 8 patients with positive MFC, at the time of study entry. (d) OS, (e) EFS and (f) cumulative incidence of relapse are shown. There was significant difference between the groups (P-values of 0.002, <0.0001 and <0.0001, respectively, by log-rank test).

Full size image

MRD assays by MFC were performed at enrollment and at approximately 3–6-month intervals during the study on bone marrows of 38 patients, including 16 patients on the DAC and 22 on the CC arm. MFC identified MRD in 9 of 20 (45%) patients with subsequent relapse, at levels ranging from 0.04 to 1% of the total number of cells. In 14 patients remaining alive in CR, no MRD was identified. Relapses occurred 1.5–10 months after the first positive MFC.

A multivariate Cox regression model was used to assess simultaneously the effect of two or more variables on OS or EFS (Table 1c). On adjusting for treatment, age and cytogenetics, positive MFC at study entry was found to be associated with short survival (hazard ratio (HR)=5.24; P=0.0004). There was no significant difference in survival by treatment groups (HR=0.63; P=0.32), age (HR=1.013; P=0.51) and cytogenetics (HR=2.32; P=0.104). After adjusting for treatment, age and cytogenetics, positive MFC was also found to be associated with short EFS ((HR)=14.24; P<0.0001). There was no significant difference in EFS by treatment groups (HR=0.7; P=0.4), age (HR=1.003; P=0.85) and cytogenetics (HR=1.75; P=0.23).

Table 1c Multivariate Cox regression model for OS and EFS with FC as a time-change covariate adjusting for treatment, age and cytogenetics
Full size table

Eight patients had positive MFC (five in the CC arm and three in the DAC arm) at baseline. One patient on the DAC arm changed the MFC status from negative to positive at 3 months after treatment; two patients on the CC arm changed the MFC status from negative to positive at 4 and 18 months, respectively. Therefore, we included MFC as a time-dependent variable in the Cox regression model. Median survival was 46 months for patients with negative MFC and 10 months for patients with positive MFC at the time of enrollment (Figure 1d). The OS at 3, 6, 9 and 12 months was 88%, 75%, 50% and 38%, respectively, in patients with positive MFC, in comparison with 100%, 93%, 83% and 80%, respectively, in patients with negative MFC at enrollment (P=0.002, log-rank test). Median EFS was 31 months for patients with negative MFC and 3 months for patients with positive MFC at enrollment (Figure 1e). The EFS at 3, 6, 9 and 12 months was 50%, 13%, 13% and 0%, respectively, in patients with positive MFC at enrollment, in comparison with 93%, 87%, 77% and 73%, respectively, in patients with negative MFC (P<0.0001, log-rank test). Median RFS was >50 months for patients with negative MFC and 3 months for patients with positive MFC at enrollment (Figure 1f). The RFS at 3, 6, 9 and 12 months was 50%, 19%, 19% and 0%, respectively, in patients with positive MFC, in comparison with 93%, 90%, 83% and 79%, respectively, in patients with negative MFC at enrollment (P<0.0001, log-rank test). No significant differences were found in OS, EFS and RFS for patients who were MFC positive or negative at the time of study entry between the two treatment arms.

DNA methylation was studied in patients receiving DAC, for LINE-1 global methylation marker, two genes (CDH13 and PDLIM4) and one microRNA (Mir124a) (previously reported to be hypermethylated in AML with prognostic impact).5, 7, 8, 9 There was a statistically significant higher LINE-1 methylation at baseline between the relapsed patients and those remaining in CR (78% versus 73%, P=0.02), whereas no differences were observed for methylation of CDH13 (10% versus 12%, P=0.93) and Mir124a-1 (9% versus 9%, P=0.98). No statistically significant differences were observed for LINE-1, CDH13, Mir124a-1 and PDLIM4 methylation levels measured at cycles 1–5 for available samples.

In summary, this randomized trial demonstrated the safety and feasibility of maintenance treatment with DAC for patients with AML in first or subsequent CR. Treatment with standard-dose DAC was well tolerated, with the most common adverse events being uncomplicated grade 3 neutropenia and/or thrombocytopenia. After a median follow-up of 44.9 months, fewer patients in the DAC arm relapsed (50% versus 60%) and OS rate was 45% in the DAC versus 36% in the CC group; these differences were not statistically significant. Using multivariate Cox regression model, after adjusting for treatment, age and cytogenetics, positive MFC was associated with shortened EFS and OS survival, as has been reported by others.10, 11, 12

We have previously reported that high levels of LINE correlate with high levels of methylation genome-wide, as reflected by the increased cumulative number of methylated genes and by CpG island methylator phenotype.13 Thus, patients who relapse may have higher global methylation, as reflected by the higher LINE-1 levels. Indeed, we and others have previously reported that higher methylation levels are associated with poor prognosis in MDS and AML.7, 14 Whether LINE-1 DNA methylation levels at baseline truly predict relapse remains to be studied and confirmed in larger trials. The small size of our study prohibits us from making more definitive conclusions. However, these data support the need for larger randomized studies with hypomethylating agents in the maintenance setting in patients with AML.