Background: Relapse is the main cause of treatment failure after allogeneic stem cell transplant (SCT) in acute myeloid leukemia (AML) and myelodysplastic syndromes (MDS). Based on encouraging preliminary reports for S.C. 5-Azacitidine (AZA) in the post-transplant setting, we investigated whether AZA given as maintenance might improve relapse free survival (RFS).

Methods: Between 2009 and 2017, > 900 AML/MDS aged 18-75 were screened for the randomized controlled study of post-transplant AZA if they were in complete remission after SCT.

The randomization was planned with the treatment arm to receive AZA given as 32 mg/m2/day S.C. for 5 days in each of 12 cycles. The control arm (Ct.) had no intervention. The first cycle of AZA was started within 42-100 post SCT. Each cycle consisted of 28 days allowing next cycle to be started within 56 days following the previous one in the presence of toxicity.

The primary outcome for treatment evaluation was RFS. The goal was to test the hypothesis that AZA, given for 12 cycles, provided at least a 50% improvement compared with Ct. in median RFS. An accrual rate of 6 pts. per month was assumed. The final analysis was to be done after 213 to 246 pts. were enrolled based on the number of events observed. However, the study was closed early due to slow accrual 8 years.

Results: 187 pts were randomized to either AZA (n=93) Ct. (n=94). Of 93 patients, 87 started treatment with AZA; with a median of 62 days.

There were no significant differences between the arms in age (median, 57y), race, cytogenetics, donor type, conditioning intensity, hematopoietic stem cell source, except disease status in AML and comorbidity (Table 1).

The majority of the pts.in AZA arm (74.6%) did not receive the planned 12 cycles of treatment, with the number of cycles varying with a median of 4 (range, 1-12). The most common causes of discontinuation of AZA were relapse/death on AZA (46.9%), hematologic/non-hematologic toxicity (26.5%), or patient request/logistical reasons (26.5%).

Median follow-up time was similar between arms; 4.6 years in AZA vs. 4.06 years in Ct. Median RFS was comparable, with 2.07 yrs. (AZA) vs. 1.28 yrs. (obs.) (p=0.43). Due to bias induced by the discontinuation of AZA, any conventional comparison analysis, such as a logrank test, would be biased. Therefore, to obtain unbiased comparisons of RFS, separate landmark analysis using logrank tests were conducted, with each including only the AZA pts. who were compliant up to the landmark time, and the comparable subset of Ct. pts. defined as those at risk of failure at the landmark time (Table 2). All landmark analysis showed no significant AZA effect on RFS. A Cox model regression analysis indicated that there was improvement in RFS with AZA beyond day 180, but this was not statistically significant.

Additionally, three ad hoc comparisons of AZA to Ct. were done by stratifying AZA pts. into those who received: 1-4 cycles (group 1); , 5-8 cycles (group 2) ; ,or 9-12 cycles (group 3). Pts.in.Ct. were included in the comparative analysis for group 2 if they were at risk of failure at day 150, and for group 3 at day 270. Cox regression analyses showed improvement in RFS in AZA group if they received ≥9 cycles of AZA, but the effect was not significant (p=0.18).

Overall, AZA maintenance was well-tolerated with only 1 grade 5 adverse event (AE) reported with AZA vs. 10 grade 5 AEs in the Ct . arm. Twenty grade 4 AEs were reported in AZA arm, 13 were attributed to AZA; all but one was bone marrow (BM) toxicity. All AZA attributable grade IV AEs resolved with a median of 23 days. In obs. arm, 8 grade 4 AEs were reported with 3 BM toxicity.

Conclusion: The SC AZA vs. observation randomized trial conducted over 8 years demonstrated that a prospective trial in the post-transplant setting was feasible and safe but challenging. In this trial, although the aim was compare RFS between arms after 12 cycles of AZA, less than 30% of AZA pts. completed the planned 12 cycles for various reasons, including disease relapse, toxicity, compliance, and logistics for receiving the drug. Because accrual was slow, the study was terminated early. While RFS was comparable between the two arms, stratification by number of cycles of AZA received showed a trend towards improvement in RFS in AZA pts. receiving more cycles of therapy. This finding supports the notion that AZA given for 9-12 cycles as maintenance therapy warrants further study for prevention of relapse after allogeneic HSCT for AML and MDS.


Oran:AROG pharmaceuticals: Research Funding; Celgene: Consultancy, Research Funding; ASTEX: Research Funding. Shpall:Affirmed GmbH: Research Funding. Rezvani:Affirmed GmbH: Research Funding. Champlin:Sanofi: Research Funding; Otsuka: Research Funding.

Author notes


Asterisk with author names denotes non-ASH members.