Relapse of myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML) remains the leading cause of treatment failure after allogeneic hematopoietic stem cell transplantation (alloHSCT). Relapses often occur early, typically within the first year post-transplant, although approximately 30% of patients experience later relapses characteristic of less proliferative disease. Historically, options included intensive salvage chemotherapy, low-dose cytarabine, donor lymphocyte infusions (DLI), or second alloHSCT; these approaches were associated with limited efficacy and substantial toxicity.
Over the last two decades, hypomethylating agents (HMAs)—in particular azacytidine—have transitioned from a novel option to the current backbone for relapse therapy after alloHSCT. This review summarizes clinical evidence for HMAs as monotherapy and in combination with immunomodulatory approaches, examines mechanisms of relapse and HMA action, and describes prognostic factors that influence response and long-term outcomes.
Relapse after alloHSCT is primarily driven by clonal selection under therapeutic and immune pressure. Molecular studies document frequent dysregulation of immune-relevant pathways, including reduced MHC-II expression and, in some settings, complete HLA loss after haploidentical transplantation, enabling leukemic escape from donor T-cell recognition. Adaptive immune-escape mechanisms are apparent, and features of T-cell exhaustion—activation of PD-1/PD-L1, TIM-3, or TIGIT pathways, diminished cytokine production, and reduced TCR diversity—are described. NK-cell dysfunction and immunosuppressive bone marrow microenvironment factors (for example, TGF-β1, IDO1, regulatory macrophages) further facilitate persistence of residual leukemic cells.
Persistence often involves clones with mutations in epigenetic regulators or CHIP-associated lesions, while proliferative signaling mutations show more dynamic evolution. Leukemic stem cells and early progenitors are central to relapse biology; they express immune-escape molecules and are protected by bone marrow niche interactions. Residual stem cell populations may be detectable long before morphological relapse and can undergo clonal selection during prior HMA therapy, supporting the concept of quiescent, niche-protected clones as a reservoir for later relapse.
HMAs are azanucleoside prodrugs with multiple dose-, exposure-, and cell cycle–dependent effects. Following phosphorylation and S-phase incorporation, they form covalent adducts with DNA methyltransferases (DNMTs), resulting in DNMT depletion, replication stress, and passive demethylation over time. Low doses favor demethylation and transcriptional modulation; higher doses elicit direct cytotoxicity.
Azacytidine and decitabine share core DNMT-related mechanisms but differ biologically. A substantial fraction of azacytidine incorporates into RNA, affecting rRNA/tRNA processing and translation, whereas decitabine is more DNA-focused. These differences lead to overlapping DNA methylation markers but divergent transcriptional and cell-cycle consequences.
HMAs exert complex immunological effects. They can increase tumor-antigen expression and antigen-presentation programs (for example, re-expression of cancer-testis antigens), and induce a “viral mimicry” state via demethylation of endogenous retroviral elements, activating type I interferon pathways. T-cell effects are context-dependent: azacytidine has been associated with TCR repertoire reconstitution and Treg expansion after alloHSCT alongside tumor-antigen–specific CD8 responses, but prolonged exposure in other settings has been linked to reduced Th1/Th2 cells and increased PD-1 expression. NK-cell and MDSC effects are heterogenous across studies. Overall, immunomodulation by HMAs can be beneficial, neutral, or inhibitory depending on timing, compartment, and therapeutic combinations.
Importantly, experimental and clinical evidence suggests azacytidine can attenuate graft-versus-host disease (GvHD) while preserving graft-versus-leukemia (GvL), an attribute relevant when combining HMAs with immunotherapies such as DLI.
Prospective studies of azacytidine alone after alloHSCT are limited but informative. The RELAZA program is central to this evidence. In RELAZA, MRD was monitored using CD34+ donor chimerism in peripheral blood; a decline below 80% triggered azacytidine (75 mg/m² days 1–7 q28). In the single-center Phase II RELAZA study, 10/20 patients achieved a major response and 6/20 a minor response; hematologic relapse was delayed in many but ultimately occurred in a majority, and cytopenias were common adverse events. RELAZA 2 expanded MRD criteria to include RT-qPCR targets such as NPM1 and applied MRD-guided azacytidine in a multicenter setting for up to 2 years. Long-term results show that sustained MRD negativity predicts excellent prognosis and that achieving a second molecular response improves relapse-free and overall survival. Whether azacytidine alone routinely cures molecular relapse after alloHSCT remains unclear from available data.
Following early case reports, the combination of azacytidine plus DLI has become the most commonly used salvage strategy in clinical practice for relapsed myeloid neoplasms after alloHSCT. Multiple single-arm and retrospective series have examined azacytidine with or without DLI; population-level surveys indicate widespread adoption of this combination. Across studies, better responses and longer survival correlate with lower disease burden at intervention, molecular rather than hematological relapse, and longer interval from transplant to relapse. Rates of de-novo acute GvHD vary across cohorts but are generally manageable, and some studies report preserved GvL with acceptable GvHD rates. Published tables summarize key studies, response rates, and findings but also underscore heterogeneity in patient selection, prior therapies, and use of DLI.
Clinical and biological predictors of response to HMA-based relapse therapy include disease burden (molecular vs hematologic relapse), timing of relapse relative to transplant, and pre-transplant therapies. Minimal residual disease (MRD) monitoring is crucial to enable early, preemptive interventions; MRD-guided approaches can delay hematologic relapse and improve outcomes in responders. Prognostic models such as the APSS-R score integrate relapse timing, disease burden, and pre-transplant therapy to stratify patients and support biologically tailored treatment decisions.
Beyond azacytidine ± DLI, combination regimens pairing HMAs with targeted agents (FLT3 and IDH inhibitors), the BCL2 inhibitor venetoclax, or immunomodulatory drugs such as lenalidomide have shown promising activity in early reports. These combinations aim to improve direct anti-leukemic activity and to modulate immune responses favorably. Future strategies will likely emphasize individualized, biology-guided approaches, selective use of second alloHSCT, and rational integration of targeted and immunotherapeutic options with HMAs.
HMAs—particularly azacytidine—constitute the current therapeutic backbone for relapse management after alloHSCT in MDS and AML. MRD-guided, preemptive azacytidine can delay hematologic relapse and, when combined with immunological strategies such as DLI, produce meaningful remissions in selected patients. HMA immunomodulation is complex and context-dependent; nevertheless, the balance of data supports HMA-based combinations for salvage therapy. Prognostic tools and MRD monitoring are essential to optimize timing and selection of interventions. Open questions remain about the curative potential of azacytidine monotherapy for molecular relapse and how best to combine HMAs with emerging targeted and immunotherapies. Continued clinical research and biologically informed treatment personalization are needed to improve durable outcomes in this high-risk population.