Azacitidine maintenance after allogeneic hematopoietic stem cell transplantation in high-risk AML: a single-center retrospective study

In patients with acute myeloid leukemia (AML) undergoing allogeneic hematopoietic stem cell transplantation (allo-HSCT), disease relapse remains the leading cause of transplant failure [ 1 ]. Growing evidence indicates that immune escape represents a major mechanism driving post-transplant relapse [ 2 ]. Hypomethylating agents have gained interest due to their dual properties, antileukemic and immunomodulatory [ 3 ]. Specifically, azacitidine has been evaluated as post-transplant maintenance in a prospective randomized trial in AML and MDS, showing no OS and RFS benefit [ 4 ]. However, a subgroup analysis showed that azacitidine was associated with a reduced risk of disease progression and improvement in PFS among high-risk AML and MDS [ 5 ]. Therefore, with the aim of evaluating the impact of post-transplant azacitidine in high-risk AML, we retrospectively analyzed 29 adult patients (AZA cohort) undergoing allo-HSCT from 2020 to 2025 and receiving post-transplant azacitidine maintenance. Next, we compared the AZA cohort with a control cohort (NO AZA cohort, 2018-2024) of 21 adult patients, retrospectively selected by applying the exact same eligibility criteria used for the AZA cohort and not receiving any post-transplant maintenance due to personal preference or physician’s discretion, largely due to concerns regarding treatment compliance, as well as labeling restrictions. Eligibility criteria included: adverse risk according to ELN2017 and/or ELN2022 criteria [ 6 , 7 ]; primary refractory disease [ 8 ]; ≥5% blasts or ≥1% measurable residual disease by multiparameter flow cytometry (MFC-MRD) and/or droplet digital PCR (PCR-MRD) at time of conditioning start; second allo-HSCT or beyond; ≥0.1% MFC-MRD and/or PCR-MRD detected within 120 days after allo-HSCT. Patients were eligible after engraftment with an absolute neutrophil count ≥1.0 × 10⁹/L, platelet count ≥50 × 10⁹/L and complete remission (<5% marrow blasts, no evidence of extramedullary leukemia and absence of peripheral blasts) documented within 120 days after allo-HSCT and at the time of azacitidine initiation. ECOG score 0–2, adequate organ function, no uncontrolled infection, and no active grade 2–4 acute Graft-versus-Host Disease (GvHD) were also required. Patients with FLT3-ITD mutations and those who received post-transplant maintenance other than azacitidine were excluded. All patients in the AZA cohort received azacitidine 37.5–75 mg/m²/day (dosing details in Supplemental Methods and Supplemental Table 1 ) given subcutaneously on days 1–5 every 28 days, starting between 35 and 120 days after allo-HSCT, administered for at least 12 cycles and until disease relapse and/or unacceptable toxicity or according to patient and/or physician’s choice. All patients provided written informed consent. The AZA cohort included 29 patients (median age 61 years, range 28–74). At the start of conditioning, 15 (52%) were in CR1, while 14 (48%) had advanced disease (AD; 1 patient in CR2 and 13 with active disease). Pre-transplant MRD was evaluable in 22 patients (76%): ≥0.1%, and ≥1% MFC-MRD were detected in 62% and 38%, respectively. Post-transplant MRD was evaluable in 24 patients (83%): 7 (24%) displayed MFC-MRD at the time of azacitidine start. Notably, the AZA cohort and the NO AZA cohort were largely comparable except for pre-transplant ≥1% MFC-MRD, which resulted in more frequent occurrence in the AZA cohort. Detailed patient characteristics are reported in Supplemental Table 1 . With a median follow-up of 18 months, 2-year OS and RFS in the AZA cohort were 69% (95% CI, 48–98%) and 60% (95% CI, 40–88%), respectively (Fig. 1a, b ). Two-year CIR was 40% (95% CI, 16–65%), and NRM was 0%. Nine (31%) patients relapsed, and the median time to relapse was 419 days (95% CI, 126–930). Patients received a median of 12 cycles (range, 2–25) of azacitidine, with discontinuation in 13 patients (45%) due to toxicity ( n = 2: grade 3 pneumonia; hematological toxicity), maintenance completion ( n = 3), and disease relapse ( n = 8). Relapse occurred after a median number of 9 cycles and mostly (67%) in patients with baseline TP53 alterations (TP53 alt) and/or complex or monosomic karyotype (CMK). The first bone marrow assessment, performed within day +60, documented MFC-MRD clearance (<0.1%) in 12/18 patients (67%): 4/12 then experienced morphologic relapse, and 8/12 had durable CR without MFC-MRD. At the time of maintenance initiation, 7 patients displayed ≥0.1% MFC-MRD: 5 obtained durable MFC-MRD clearance, while 2 ultimately experienced morphologic relapses. During maintenance, 4 patients (13%) received additional therapy due to either MFC-MRD relapse or persistence (Fig. 1c and Supplemental Table 1 ). Univariate analyses identified HCT-CI, TP53 alt, CMK, disease status at transplant, conditioning regimen, and donor type as significantly associated with OS (Supplemental Fig. 1A–F ) and RFS (Supplemental Fig. 2A–F ). Furthermore, the azacitidine dose had no impact on OS and RFS (Supplemental Fig. 3A–D ) or toxicity (Supplemental Table 2 ) in the AZA cohort as well as in the TP53 alt subgroup. The landmark analysis showed no significant differences in OS ( p = 0.74) nor RFS ( p = 0.98) between AZA and NO AZA cohorts (Fig. 1d, e ); similarly, comparable two-year CIR ( p = 0.50) and NRM ( p = 0.16) were found. In the overall population ( n = 50), multivariable Cox analysis confirmed no survival benefit from azacitidine; moreover, the final model showed independent association of AD and TP53 alt with OS (AD: HR = 2.91, p = 0.06; TP53 alt: HR = 6.68, p = 0.004) and RFS (AD: HR = 3.41, p = 0.02; TP53 alt: HR = 4.61, p = 0.01). Interaction analyses did not reveal any significant modification of the effect of azacitidine maintenance on OS or RFS according to TP53 alt, CMK, disease status at transplant, or conditioning intensity. To further explore the impact of TP53 alt on azacitidine efficacy, we additionally performed a Restricted Mean Survival Time (RMST) analysis, suggesting a modest survival benefit from azacitidine in TP53 wild-type patients only (Supplemental Results; Supplemental Table 3). Azacitidine was overall well-tolerated. The most relevant hematologic adverse events (AEs) [ 9 ] were grade 3-4 neutropenia (41%) and thrombocytopenia (14%). Non-hematologic AEs were mainly grade 1–2 gastrointestinal toxicities: constipation (62%), nausea (55%), and diarrhea (10%). Infections occurred in 14 patients (48%) (Supplemental Table 2 ). No significant differences were observed between AZA and NO AZA cohorts in acute and chronic GvHD (Supplemental Results). The 2-year CIR of 40% and RFS of 60% observed in the AZA cohort can be grossly explained by the high risk of this population. Approximately half of the patients underwent allo-HSCT with AD, nearly one-third harbored TP53 alt and/or CMK, and more than 60% had detectable pre-transplant MFC-MRD. Accordingly, despite frequent early post-HSCT MRD clearance, almost one-third of patients ultimately relapsed. The comparison between the AZA cohort and the NO AZA control cohort did not show significant differences in terms of 2-year OS nor RFS. When considering both cohorts together, TP53 alterations and advanced disease status at transplant were confirmed as major adverse factors, while conditioning intensity did not. Among patients with persistent MFC-MRD at azacitidine initiation, only one-third achieved MFC-MRD negativity, while the remaining ones required additional therapies with conflicting results, confirming the limited role of MFC-MRD clearance and lack of effective MRD-eradicating strategy in high-risk AML [ 10 , 11 , 12 ]. Overall, these data confirm post-transplant azacitidine maintenance to be feasible and safe, although with no improvement in OS or RFS in high-risk AML. The continuous evolution of transplant practices may represent a confounding factor of retrospective comparisons, although both cohorts were treated within the same transplant center and during two overlapping time frames. Despite these limitations, small sample size and variability in dosing and escalation approaches, our results highlight that post-HSCT maintenance in high-risk AML remains an unmet need. Prospective randomized studies on tailored maintenance strategies, including novel agents and immunotherapeutic combinations, are warranted, with the aim of reducing post-transplant relapse without compromising safety and tolerability. Fig. 1: The impact of azacitidine maintenance after allo-HSCT. Full size image Kaplan–Meier survival curves showing a overall survival (OS) and b relapse-free survival (RFS) of AZA cohort. c Swimmer plot of the clinical course over time after allo-HSCT in each patient of the AZA cohort. Kaplan–Meier survival curves (log-rank test) showing d OS and e RFS comparison between AZA and NO AZA cohorts.