Hypomethylating agents and FLT3 inhibitors as maintenance treatment for acute myeloid leukemia and myelodysplastic syndrome following allogeneic hematopoietic stem cell transplant – a systematic review and meta-analysis

Abstract

Background:

Disease relapse remains the major cause of death among patients with acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS) who receive an allogeneic hematopoietic cell transplant (allo-HCT). Maintenance treatment with FLT3 inhibitors and hypomethylating agents (HMA) has been studied in various clinical trials with mixed results.

Objective:

To synthesize the current evidence on the efficacy and safety of FLT3 inhibitors and HMA for maintenance therapy following allo-HCT in AML and MDS.

Methods:

For this systematic review and meta-analysis Cochrane Library, Google Scholar, Ovid Medline, Ovid Embase, PubMed, Scopus, and Web of Science Core Collection were searched from inception to March 2021 for studies on maintenance therapies following allo-HCT in AML and MDS. Studies were excluded if they were reviews, commentaries, case series with <5 patients, or basic research articles, not published in English, not on post-allo-HCT maintenance with FLT3 inhibitors or HMA in AML or MDS, or if they were clinical trials without published results or duplicate publications from the same patient cohort. Studies with insufficient reporting of the primary endpoint (2-year overall survival [OS]) and studies using FLT3 inhibitors or HMA for pre-emptive treatment of imminent relapse based on positive measurable residual disease testing were excluded.

Random-effects models were used to pool response rates for the primary outcome of 2-year OS. Hazard ratios (HR) for death and relapse were calculated for studies that included a control group. Rates of relapse-free survival (RFS), non-relapse mortality and acute and chronic graft-versus-host-disease (GVHD) were studied as secondary endpoints. Downs and Black checklist and risk of bias assessments were used to gauge the quality of individual studies. The study protocol has been registered on PROSPERO (CRD42020187298).

Results:

Our search strategy identified 5559 studies. Twenty-one studies with a total of 809 patients were included in the meta-analysis. 2-year OS rates were 81.7% (95% confidence interval [CI]: 73.8–87.7%) and 65.7% (95% CI: 55.1–74.9%) among patients treated with FLT3 inhibitors and HMA, respectively. In sensitivity analyses restricted to studies that included a control group, maintenance therapy with FLT3 inhibitors (HR for death: 0.41; 95% CI: 0.26–0.62) or HMA (HR: 0.45; 95% CI: 0.31–0.66) appeared superior to no maintenance therapy. 2-year RFS rates were 79.8% (95% CI: 75.0–83.9%) and 62.4% (95% CI: 50.6–72.9%) among patients treated with FLT3 inhibitors and HMA, respectively. Rates of any grade acute and chronic GVHD were 33.1% (95% CI: 25.4–41.8%; grade 3/4: 16.5%) and 42.5% (95% CI: 26.3–60.4%) among FLT3 inhibitor and 42.7% (95% CI: 33.5–52.4%; grade 3/4: 8.1%) and 41.5% (95% CI: 32.0–51.6%) among HMA-treated patients, respectively.

Conclusion:

Maintenance therapy with either FLT3 inhibitors or HMA following allo-HCT, can lead to prolonged and improved OS and RFS with a favorable safety profile. Additional studies are needed to define the optimal duration of treatment, the role of measurable residual disease status, and transplant characteristics in patient selection.

Introduction:

Allogeneic hematopoietic cell transplant (allo-HCT) has been shown to provide a significant overall survival (OS) benefit in patients with acute myeloid leukemia (AML) and intermediate or poor risk cytogenetics compared to patients that did not undergo allo-HCT in multiple studies.^1–4 Similar OS benefits with allo-HCT have been documented in patients with higher-risk myelodysplastic syndromes (MDS).^{5, 6} The advent of reduced intensity conditioning (RIC) allo-HCT has expanded transplant eligibility for older patients and those with significant comorbidities with favorable outcomes compared to non-transplant strategies.^7–10

Even in AML patients proceeding to allo-HCT the long-term OS remains only around 40–55% in large cooperative group clinical trials.^{1, 2} The prognosis of patients with MDS or AML who relapse after allo-HCT is dismal with 2-year OS rates of 14–26%.^{11, 12} This highlights the need for better treatment options aimed at reducing the rate of or at least delaying disease relapse following allo-HCT.

As disease relapse is an important driver of post-allo-HCT mortality, maintenance therapy following allo-HCT has received increasing attention. Injectable hypomethylating agents (HMA) including azacitidine and decitabine, as well as inhibitors of mutant fms like tyrosine kinase 3 (FLT3) such as sorafenib and midostaurin have been studied in various clinical trials.^13–16 However, it is important to note that the patient populations included in these studies were heterogenous and that the natural history of AML with and without FLT3 mutations is different with the former having a higher incidence of relapse following allo-HCT and an adverse prognosis.^{17, 18} While FLT3 mutations act as driver mutations in AML and targeted inhibitors are available, HMA have been used independent of baseline molecular and cytogenetic alterations.¹⁹ Following intensive induction, consolidation chemotherapy, and allo-HCT itself, the burden on patients and caregivers associated with frequent clinic visits and adverse events is an important consideration that must be balanced against the potential of a prolonged relapse-free survival (RFS) and OS. We conducted a systematic review and meta-analysis to objectively assess the safety and efficacy of post-allo-HCT maintenance therapy with HMA and FLT3 inhibitors in AML and MDS patients. We hypothesized that maintenance therapy following allo-HCT with FLT3 inhibitors or HMA would improve 2-year OS among AML and MDS patients.

Methods:

This systematic review and meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA)²⁰ and Meta-Analysis of Observational Studies in Epidemiology (MOOSE) guidelines (Supplementary Table 1).²¹ The study protocol was registered on PROSPERO (CRD42020187298) prior to the start of data extraction.

Search strategy:

A systematic search of the literature was conducted by a medical librarian in the Cochrane Library, Google Scholar, Ovid Embase, Ovid Medline, Pubmed, Scopus, and Web of Science Core Collection databases to find relevant articles published from inception of each database to March 2021. The final searches were performed in all the databases on March 28, 2021. The search was peer-reviewed by a second medical librarian using PRESS (Peer Review of Electronic Search Strategies).²² Databases were searched using a combination of controlled vocabulary and free text terms for AML/MDS, stem cell therapy, and maintenance therapy. Details of the full search strategies are listed in Supplementary Materials. Reference lists of included studies were reviewed for additional relevant literature.

Study selection:

Titles and abstracts were reviewed for inclusion in the meta-analysis by two reviewers independently (JPB and CA or ASM) and conflicts were resolved by consensus discussion. Studies were excluded if they were review or basic research articles, not on post-allo-HCT maintenance in AML or MDS, or if they were clinical trials without published results. Full texts were reviewed and excluded for any of the following reasons: (I) no English full text was available, (II) duplicate publications from the same patient cohort, (III) insufficient reporting of endpoints, (IV) studies not on FLT3 inhibitors or HMA, (V) case series with <5 patients, or (VI) commentaries without original independent data. Studies using FLT3 inhibitors or HMA for pre-emptive treatment of imminent relapse based on positive measurable residual disease (MRD) testing were excluded.

Data Extraction and Quality assessment:

Data extraction and quality assessment was performed using a standardized data-extraction sheet in Microsoft Excel by two investigators independently (JPB and CA) with cross-check for accuracy performed by a third reviewer (MS) in case of discrepant results. The Downs and Black checklist was used by one author to assess study quality of included studies as published previously with cross-check of a random subset of studies by a second author.^23–25 The risk of bias of individual studies for the primary outcome was assessed using tools developed by the Cochrane collaboration.²⁶

Definition of endpoints:

The primary endpoint was the rate of 2-year OS with the 2-year RFS rate and pooled hazard ratios for death and relapse as secondary outcomes. Incidence rates of acute and chronic graft-versus-host disease (GVHD) as well as ≥grade 3 acute GVHD and extensive stage chronic GVHD and non-relapse mortality were defined as safety outcomes. Analyses were performed separately for the type of post-allo-HCT therapy used (HMA and FLT3 inhibitor). Sensitivity analyses were conducted among studies including an internal control group to compare the efficacy of maintenance therapy relative to observation. In order to avoid selection bias, only studies that reported both the primary endpoint of 2-year OS and the 2-year hazard ratio for death and/or relapse were included in sensitivity analyses.

Statistical analysis:

Random-effects models were used to pool the 2-year OS, RFS, GVHD rates and hazard ratios (HR) for death and relapse. Due to the differences in baseline patient populations and disease characteristics we did not compare maintenance therapy with HMA to FLT3 inhibitors. Cochran Q and I² indices were used to assess heterogeneity across included studies. Study heterogeneity was graded based on I² indices as low (I² <30%), moderate (I² 30–60%), or high (I² >60%). The same analytic approach was applied for sensitivity and univariate meta-regression analyses. Comprehensive Meta-Analysis (CMA version 2.2, Biostat) was used for all analyses.

Results:

Results of the literature search:

Our search strategy identified 5559 studies after duplicate removal. As illustrated in the PRISMA diagram (Figure 1), 142 studies were assessed as full texts for eligibility using the prespecified inclusion and exclusion criteria outlined in the methods section yielding a final set of 21 studies that were included in the meta-analysis. Studies that were reviewed as full texts but not included in the meta-analysis with specific reason for exclusion are provided in Supplementary Table 2.

Figure 1: Study selection flow chart.

A systematic search of the literature was conducted in the Cochrane Library, Google Scholar, Ovid Embase, Ovid Medline, Pubmed, Scopus, and Web of Science Core Collection databases to find relevant articles published from inception of each database to March 28, 2021. Included studies reference lists were reviewed for additional relevant literature. Titles and abstracts were reviewed for inclusion in the meta-analysis by two reviewers independently and conflicts were resolved by consensus discussion. Studies were excluded at the abstract and title screening stage if they were identified as review or basic research articles, not on post-allo-HCT maintenance in AML or MDS, or if they were clinical trials without published results. Full texts were reviewed and excluded for any of the following reasons: (I) no English full text was available (n=3), (II) duplicate publications from the same patient cohort (n=23), (III) insufficient reporting of endpoint (n=55), (IV) studies not on FLT3 inhibitors or HMA (n=9), (V) case series with <5 patients (n=5), or (VI) commentaries without original independent data (n=15). Studies using FLT3 inhibitors or HMA for pre-emptive treatment of imminent relapse based on positive minimal residual disease (MRD) testing were excluded (n=12). The citations of those 20 studies were reviewed and one additional study was found and included in the meta-analysis yielding a final sample of 21 studies.

Description of included studies:

We included 21 studies with a total of 809 patients who received either FLT3 inhibitors or HMA as maintenance therapy following allo-HCT in the meta-analysis. Eleven studies with 366 patients reported outcomes of patients treated with FLT3 inhibitors, among which sorafenib was used in nine studies ^{16, 27–34} and quizartinib ³⁵ and midostaurin ³⁶ in one study each. HMA were used in ten studies with a total of 443 patients with seven studies on azacitidine,^{14, 37–42} two on decitabine,^{43, 44} and one study on both azacitidine and decitabine.⁴⁵

Among the included studies, there were eight retrospective studies,^{27, 30, 31, 34, 38, 40, 41} four prospective cohort studies,^{29, 32, 37, 45} two phase I clinical trials,^{35, 43} six phase II clinical trials,^{14, 16, 36, 39, 42, 44} and one phase III clinical trial.³³ Four studies were randomized of which only the study by Burchert et al. used a placebo control.^{16, 33, 36, 44}. Among the non-randomized studies, three studies included a matched ^{27, 28, 30} and four studies an unmatched control group.^{31, 34, 38, 41}

Studies on FLT3 inhibitors exclusively enrolled AML patients, while 349 and 94 patients (21.2%) receiving post-allo-HCT maintenance therapy with HMA had AML and MDS, respectively. The median patient age ranged from 24 to 62 years. The proportion of patients with R/R disease was reported in 14 studies and ranged from 7.0% to 36.0%.^{16, 29, 32–35, 39, 41–45} Patients with R/R disease were excluded in the studies by Brunner et al. and Maziarz et al.^{27, 36} Table 1 summarizes key patient and treatment characteristics of the included studies. All patients treated with FLT3 inhibitors had a FLT3 mutation; studies using HMA maintenance therapy included a genetically heterogenous patient population with baseline genetic characteristics being reported by seven studies (Supplementary Table 3).^{14, 39, 41–45}

Table 1:

Patient and treatment characteristics of included studies

Author	Treatment and treatment schedule	Study population	Transplant characteristics	Outcomes
Ahmad et al.28	Sorafenib (dosing not reported); started within 101 days post allo-HCT	13 FLT3-ITD mut AML and 26 matched controls; 31 % transplanted with active disease	Conditioning: 100% MAC Donor source: not reported GVHD prophylaxis: not reported	Efficacy: 2-year OS 100% vs 60% control; 2-year PFS: 82% vs 45% Safety: not reported
Ali et al.41	Azacitidine 16–50 mg/m2 5-days per 28-day cycle; started median 62 days post allo-HSCT	45 AML, 14 MDS patients; 90 historic controls	Conditioning: 30% MAC, 70% RIC Donor source: 19% MRD, 54% MUD, 20% haploidentical, 7% umbilical cord GVHD prophylaxis: 56% CNI + MTX +/− MMF; 34% CNI + MMF +/− PTCY; 10% other	Efficacy: 2-year OS 67% vs 42% control (HR: 0.47); 2-year EFS: 71% vs 55% (HR: 0.52) Safety: 64% aGVDH (3% grade III/IV); 44% cGVHD (27% moderate/extensive); 27.1% AE-related treatment discontinuation, 33.9% grade ≥3 AEs
Americo et al.38	Azacitidine maintenance (dosing not reported)	17 AML patients; 51 historic controls	Conditioning: not reported Donor source: not reported GVHD prophylaxis: not reported	Efficacy: 2-year OS 74% vs 57% control (HR: 0.41); 2-year EFS: 57% vs 39% Safety: not reported
Battipaglia et al.29	Sorafenib 200mg daily – 400mg b.i.d. based on tolerability; started median 70 days post allo-HCT (range 29 –337 days)	27 FLT3- mut AML; no control group	Conditioning: 74% MAC, 19% RIC Donor source: not reported GVHD prophylaxis: not reported	Efficacy: 2-year OS 80%; 2-year PFS: 73% Safety: GVHD not reported; 18.5% AE-related treatment discontinuation, grade ≥3 AEs not reported
Bazarbachi et al.30	Sorafenib 200–800mg daily based on tolerability; started median 55 days post allo-HCT (range 1 – 173 days)	28 FLT3- mut AML; 26 matched historic controls	Conditioning: 75% MAC Donor source: not reported GVHD prophylaxis: not reported	Efficacy: 2-year OS 83% vs 62% for controls (HR: 0.32); 2-year LFS: 79% vs 54% Safety: not reported
Brunner et al. 27	Sorafenib 200–400 mg b.i.d. for 12–24 months; started median 68 days post allo-HCT (range 36 – 193 days)	26 FLT3-ITD-mut AML; 55 controls	Conditioning: 54% MAC; 46% RIC Donor source: 81% matched, 19% mismatched GVHD prophylaxis: 65% CNI + MTX	Efficacy: 2-year OS 81% vs 62% for controls (HR: 0.26); 2-year PFS: 82% vs 53% (HR: 0.25) Safety: 27% aGVHD; 55.5% cGVHD; 30.8% AE-related treatment discontinuation, grade ≥3 AEs not reported
Burchert et al.16	Sorafenib 200–400 mg b.i.d. vs. placebo for 24 months; dose adjustments based on tolerability; started 60–100 days post allo-HCT	43 FLT3-ITD-mut AML in sorafenib arm, 40 in placebo arm	Conditioning: 45% MAC; 55% RIC Donor source: 76% MUD., 24% MRD GVHD prophylaxis: not reported	Efficacy: 2-year OS 91% vs 66% for controls (HR: 0.24); 2-year RFS: 85% vs 53% (HR: 0.26) Safety: 24% aGVHD (≥ grade 2): 62% cGVHD (19% severe); 20.9% AE-related treatment discontinuation, grade ≥3 AEs not reported
Craddock et al.42	AZA 36 mg/m2 5 days per 28-day cycle for up to 12 cycles; started 40–194 days post allo-HCT	37 AML patients	Conditioning: 100% RIC Donor source: 35% MUD, 65% MRD GVHD prophylaxis: 100% cyclosporine	Efficacy: 2-year OS 49%; 2-year RFS: 49% Safety: 46% aGVHD (0% ≥ grade 3); 27% cGVHD (0% extensive); 16.2% AE-related treatment discontinuation
Gao et al.44	G-CSF 100 mg/m2 on days 0–5 + 5 mg/m2 of Dec on days 1–5 vs observation every 6–8 weeks for up to 6 cycles; started 60–100 days post allo-HCT	202 AML (100 G-CSF + decitabine; 102 observation)	Conditioning: 100% MAC Donor source: 5% MUD, 20% MRD, 75% haploidentical GVHD prophylaxis: not reported	Efficacy: 2-year OS 86% vs 70% with observation (HR: 0.45); 2-year LFS: 82% vs 61% (HR: 0.38) Safety: 31% aGVHD, 34% cGVHD (4% severe); 0% AE-related treatment discontinuation, grade ≥3 AEs not reported
Gill et al.37	Azacitidine 100mg 3-days per 28-day cycle for up to 8 cycles	31 AML, 3 MDS patients; 22 patients with first allo-HCT; no control group	Conditioning: not reported Donor source: not reported GVHD prophylaxis: not reported	1st transplant patients only Efficacy: 2-year OS 73%; 2-year PFS: 66% Safety: 32% aGVHD; 77% cGVHD; AE-related treatment discontinuation and grade ≥3 AEs not reported
Guillaume et al.39	Azacitidine 32 mg/m2 5-days per 28-day cycle for up to 12 cycles + donor- lymphocyte infusion for up to 3 cycles	40 AML, 19 MDS patients	Conditioning: 9% MAC, 78% RIC, 14% sequential Donor source: 51% MUD, 34% MRD; 2% 9/10 UD; GVHD prophylaxis: Cyclosporine +/− MMF	Efficacy: 2-year OS 71%; 2-year PFS: 67% Safety: 32% aGVHD (15% grade III/IV); 39% cGVHD; AE-related treatment discontinuation and grade ≥3 AEs not reported
Joris et al.40	Azacitidine 37.5mg/m2/day 5 days every 4 weeks for 12 cycles with 3 donor lymphocyte infusions	48 R/R-AML, 5 MDS; no control group; 60.5% with active disease at time of transplant	Conditioning: 17% RIC, 83% sequential Donor source: 51% MUD, 25% MRD; 8% MMUD; 17% haploidentical GVHD prophylaxis: Cyclosporine + MMF	Efficacy: 2-year OS 52%; 2-year GRFS: 34% Safety: 53% aGVHD (11% grade III/IV); AE-related treatment discontinuation and grade ≥3 AEs not reported
Maziarz et al.36	Midostaurin 50mg b.i.d. vs standard of care for up to 12 4-week cycles	30 FLT3-ITD-mut AML in each arm	Conditioning: 100% MAC Donor source: 67% MUD., 33% MRD GVHD prophylaxis: not reported	Efficacy: 2-year OS 85% vs 76% for controls (HR: 0.58); 2-year RFS: 85% vs 76% (HR: 0.6) Safety: 50% aGVHD (0% grade III/IV); 30% cGVHD (24% moderate/severe cGVHD); 26.7% AE-related treatment discontinuation, 57% grade ≥3 AEs
Morin et al.31	Sorafenib (dosing and schedule not reported)	20 FLT3-ITD-mut AML; 13 controls	Conditioning: not reported Donor source: not reported GVHD prophylaxis: not reported	Efficacy: 2-year OS 88% vs 45% for controls Safety: not reported
Pratz et al.32	Sorafenib 200 every other day – 400mg b.i.d. based on tolerability for up to 24 months; started median 65 days post allo-HCT (range 30 – 119 days)	44 FLT3-ITD-mut AML	Conditioning: 36% MAC, 64% non-myeloablative Donor source: 16% MUD., 34% MRD, 43% haploidentical, 7% UC GVHD prophylaxis: not reported	Efficacy: 2-year OS 76%; 2-year EFS: 74% Safety: 25% grade III/IV aGVHD; 15.9% AE-related treatment discontinuation, grade ≥3 AEs not reported
Pusic et al.43	Decitabine 2.5–15 mg/m2 5-days per 6-week cycle for up to 8 cycles; started median 95 days post allo-HCT	17 AML, 5 MDS patients	Conditioning: 91% MAC, 9% non-myeloablative Donor source: 50% MUD., 27% MRD, 18% MMUD, 5% MMRD GVHD prophylaxis: 100% MTX + tacrolimus	Efficacy: 2-year OS 56%; 2-year DFS: 48% Safety: 41% aGVHD, (9% grade III/IV), 55% cGVHD (55% moderate/severe); 20.8% AE-related treatment discontinuation, 75% grade ≥3 AEs
Sandmaier et al.35	Quizartinib dose escalation study 40–60mg daily for up to 24 cycles; started 30–60 days post allo-HCT	13 FLT3- mut AML	Conditioning: not reported Donor source: 69% unrelated, 31% related donor GVHD prophylaxis: not reported	Efficacy: 2-year OS 31% Safety: 69% combined aGVHD and cGVHD; 30.8% AE-related treatment discontinuation, grade ≥3 AEs not reported
Shi et al.34	Sorafenib 200mg b.i.d. or 400mg daily; start median 83 days (range 37 – 222 days)	24 FLT3- mut AML	Conditioning: not reported Donor source: 8% unrelated, 92% related donor GVHD prophylaxis: not reported	Efficacy: 2-year OS 96%; 2-year LFS: 96% Safety: 38% aGVHD (grade II-IV), 48% cGVHD; 12.5% AE-related treatment discontinuation, grade ≥3 AEs not reported
Sizemore et al.45	Azacitidine 25–75 mg/m2 5-days per 28-day cycle for up to 10 cycles; started median 40 days post allo-HCT (10 patients); 1 patient with decitabine	10 AML, 1 MDS patients	Conditioning: 91% MAC, 9% RIC Donor source: 8% unrelated, 92% related donor GVHD prophylaxis: not reported	Efficacy: 2-year OS 76%; 2-year PFS: 51% Safety: not reported
Vij et al.14	Azacitidine 32mg/m2 5-days per 28-day cycle for up to 6 cycles; started median 61 days post allo-HCT (range 42 – 90 days)	13 AML, 50 MDS patients	Conditioning: 100% RIC Donor source: 64% MUD, 37% MRD GVHD prophylaxis: 100% MTX + tacrolimus	Efficacy: 2-year OS 46%; 2-year PFS: 41% Safety: 37% aGVHD (13% grade III/IV), 30% cGVHD (14% extensive); 15.9% AE-related treatment discontinuation, 100% grade ≥3 AEs
Xuan et al.33	Sorafenib (dose not reported) started 30–60 days post allo-HCT for 180 days	202 FLT3-ITD AML patients (100 sorafenib, 102 to no maintanence)	Conditioning: 100% MAC Donor source: 8% MUD, 44% MRD; 48% haploidentical GVHD prophylaxis: cyclosporin + MTX + MMF +/− ATG	Efficacy: 2-year OS 82% vs 68% placebo (HR: 0.48); 2-year LFS: 79% vs 57% (HR: 0.37) Safety: 31% aGVHD (23% grade III/IV), 23% cGVHD (18% extensive); 5% AE-related treatment discontinuation, 50% grade ≥3 AEs

AE – adverse events; Allo-HCT – allogeneic hematopoietic cell transplant; AML – acute myeloid leukemia, ATG – anti-thymocyte globulin; AZA – azacitidine; b.i.d. – twice daily; GRFS – graft-versus-host disease and relapse-free survival; GVHD – graft-versus-host disease; HR – hazard ratio; LFS – leukemia-free survival; MAC – myeloablative conditioning; MDS – myelodysplastic syndrome; MMF – mycophenolate; MMUD – mismatched unrelated donor; MRD – matched related donor; MTX – methotrexate; MUD – matched unrelated donor; OS – overall survival; PFS – progression-free survival; RFS – relapse-free survival; RIC – reduced intensity conditioning; UC – umbilical cord blood

Assessment of study quality:

Based on the heterogeneity of study design, the quality of the included studies varied with scores between 15 and 28 points on the Downs and Black checklist with the study by Burchert et al. scoring highest.¹⁶ A detailed quality assessment of each individual study based on the Downs and Black checklist is provided in Supplementary Table 4. There was some concern for risk of bias in each of the four included randomized, controlled trials.^{16, 33, 36, 44} Among the single-arm clinical trials and cohort studies, four studies were found to be of low risk of bias,^{28, 30, 41, 46} one study with unknown risk,³⁴ and 12 studies with high risk;^{14, 29, 31, 32, 35, 37–40, 42, 43, 45} primarily due to the absence of a control group. Supplementary Table 5 and 6 show risk of bias assessments for individual studies.

Survival outcomes with maintenance therapy following allo-HCT:

Overall survival:

2-year OS rates were 81.7% (95% confidence interval [CI]: 73.8–87.7%) and 65.7% (95% CI: 55.1–74.9%) among patients treated with FLT3 inhibitors and HMA, respectively (Figure 2A). Study heterogeneity was significant for both types of maintenance therapy and was rated as moderate for FLT3 inhibitors (Cochran’s Q=22.63; p=0.012; I²=55.81%) and as substantial for HMA (Cochran’s Q=36.14; p<0.001; I²=75.09%). Four studies using FLT3 inhibitors (n=199 patients)^{16, 27, 33, 36} and three studies of HMA (n=176 patients) included a control group.^{38, 41, 44} In sensitivity analyses restricted to the aforementioned studies that included a control group, maintenance therapy with either FLT3 inhibitors (HR for death: 0.41; 95% CI: 0.26–0.62) or HMA (HR: 0.45; 95% CI: 0.31–0.66) appeared superior to no maintenance therapy (Figure 3A).

Figure 2: Pooled outcomes for post allo-SCT maintenance therapy with either a FLT3 inhibitor or an HMA.

A: Pooled 2-year overall (OS) rate with maintenance therapy

B: Pooled 2-year relapse free survival (RFS) rate with maintenance therapy

Figure 3: Comparative outcomes for post allo-SCT maintenance therapy with either a FLT3 inhibitor or an HMA vs. no post allo-SCT maintenance therapy (for studies directly comparing maintenance vs. no maintenance).

A: Hazard ratio (HR) for death with maintenance therapy vs. no maintenance therapy

B: Hazard ratio (HR) for relapse of disease with maintenance therapy vs. no maintenance therapy

Relapse-free survival:

RFS was reported by nine studies on FLT3 inhibitors and eight studies on HMA. 2-year RFS rates were 79.8% (95% CI: 75.0–83.9%) and 60.9% (95% CI: 49.8–71.0%) among patients treated with FLT3 inhibitors and HMA, respectively (Figure 2C). Study heterogeneity was not significant for studies on FLT3 inhibitors and was rated as minimal (Cochran’s Q=6.03; p=0.644; I²=0%), while it was significant and substantial for HMA (Cochran’s Q=30.1; p<0.001; I²=73.41%). Four studies using FLT3 inhibitors (n=199 patients)^{16, 27, 33, 36} and two studies of HMA (n=159 patients) included a control group.^{41, 44} In sensitivity analyses restricted to studies that included a control group, maintenance therapy with either FLT3 inhibitors (HR for relapse: 0.35; 95% CI: 0.23–0.52) or HMA (HR: 0.45; 95% CI: 0.30–0.65) appeared superior to no maintenance therapy (Figure 3B).

Safety:

Incidence rate of graft-versus-host disease:

Regimens used for GVHD prophylaxis were variable across included studies and were not reported by 13 studies (Table 1). Incidence rates of acute and chronic GVHD were reported by five and seven studies on FLT3 inhibitors^{16, 27, 33, 34, 36} and HMA, respectively.^{14, 37, 39, 41–44}. Rates of any grade acute GVHD were 33.1% (95% CI: 25.4–41.8%) and 42.7% (95% CI: 33.5–52.4%) among FLT3 inhibitor and HMA treated patients, respectively (Figure 4A). Study heterogeneity was not significant for studies on FLT3 inhibitors and was rated as moderate (Cochran’s Q=6.14; p=0.189; I²=34.85%), while it was significant and substantial for studies using HMA (Cochran’s Q=23.67; p=0.001; I²=70.43%). Rates of grade 3/4 acute GVHD were 16.5% (95% CI: 7.6–31.9%) and 8.1% (95% CI: 4.3–14.7%) among FLT3 and HMA treated patients, respectively (Figure 4C). Study heterogeneity was significant for studies on FLT3 inhibitors and was rated as substantial (Cochran’s Q=7.86; p=0.049; I²=61.82%), while it was non-significant and moderate for studies using HMA (Cochran’s Q=12.03; p=0.061; I²=50.14%).

Figure 4: Acute and chronic graft versus host disease (GVHD).

A: Pooled acute GVHD rate (all grade)

B: Pooled chronic GVHD rate (all grade)

C: Pooled grade 3/4 acute GVHD rate

D: Pooled moderate/extensive chronic GVHD rate

Any grade chronic GVHD occurred in 42.5% (95% CI: 26.3–60.4%) and 41.5% (95% CI: 32.0–51.6%) among FLT3 and HMA treated patients, respectively (Figure 4B). Study heterogeneity was significant and substantial for studies on both FLT3 inhibitors (Cochran’s Q=23.71; p<0.001; I²=83.13%) and HMA (Cochran’s Q=18.10; p=0.006; I²=66.86%). Rates of extensive stage chronic GVHD were 19.4% (95% CI: 14.2–26.0%) and 15.0% (95% CI: 5.0–37.2%) among FLT3 inhibitor and HMA treated patients, respectively (Figure 4D). Study heterogeneity was non-significant and minimal for studies on FLT3 inhibitors (Cochran’s Q=0.53; p=0.767; I²=0%), while it was significant and substantial for studies using HMA (Cochran’s Q=32.41; p<0.001; I²=87.67%).

Incidence of non-relapse mortality:

The incidence rate of 1-year and 2-year non-relapse mortality (NRM) was reported by three^{39, 40, 42} and four studies, respectively.^{14, 27, 33, 44} Rates of 2-year NRM ranged from 3.4% in the study by Gao et al.⁴⁴ to 33.4% in the study by Vij et al.¹⁴ Due to the small number and heterogeneity of studies we were unable to perform a meta-analysis of NRM rates.

Cumulative incidence rate of relapse:

The 2-year cumulative incidence of relapse (CIR) rate was reported by three studies on FLT3 inhibitors and ranged from 4.2% to 11.9%.^{27, 33, 34} Four studies with HMAs reported 2-year CIR rates, which ranged from 5.9% to 38.6%.^{14, 38, 43, 44}

Selected other adverse events:

Treatment discontinuation rates due to adverse events (AE) were reported by five studies using HMA^{14, 41–44} and eight studies on FLT3 inhibitors, respectively.^{16, 27, 29, 32–36} Median rates of AE-related treatment discontinuation were 16.2% (interquartile range [IQR]: 15.8 – 20.8%) and 19.7% (IQR: 15.1 – 27.7%) for HMA- and FLT3 inhibitor-treated patients, respectively. Dose adjustments related to AE were required at a median rate of 42.0% (IQR: 20.0 – 48.8%) among nine studies of FLT3 inhibitors.^{16, 28, 29, 31–36} No studies reported AE-related dose adjustments for studies on HMA.

Reporting and grading of adverse events across included studies was heterogenous precluding a formal meta-analysis (Table 1). Incidence rates of any ≥3 grade adverse events were reported by four studies on HMA^{14, 41, 43, 44} and two studies on FLT3 inhibitors,^{33, 36} and ranged from 33.9% in the study by Ali et al. to 100% in the study by Vij et al.^{14, 41} Among studies of HMA maintenance therapy, the most common ≥3 grade AEs were thrombocytopenia (reported by n=3 studies),^42–44 neutropenia (n=3),^42–44 and infection (n=3)^41–43 with median incidence rates of 35.1%, 35.1%, and 33.3%, respectively. Most common ≥3 grade AEs among sorafenib-treated patients were infection (n=3),^{16, 33, 35} GI toxicity (n=5),^{16, 32, 33, 35, 36} and thrombocytopenia (n=4)^{16, 32, 33, 35} with median incidence rates of 25.0%, 14.0%, and 14.2%, respectively.

Discussion:

Disease relapse remains the most common cause of death of AML patients undergoing allo-HCT and strategies to reduce relapse risk are needed.⁴⁷ In addition to optimizing conditioning regimens and reducing pre-transplant disease burden, post-transplant maintenance therapies have been studied in various trials providing the background for this systematic review and meta-analysis.^48–50 We included 21 studies with 809 AML and MDS patients receiving maintenance therapy with either FLT3 inhibitors or HMA following allo-HCT, which makes this the largest and most comprehensive meta-analysis on this topic to date. We found 2-year OS rates of 81.7% and 65.7% among patients treated with FLT3 inhibitors and HMA, respectively, with relative benefits compared to observation in the studies that included a control group (HR for death: FLT3 inhibitors: 0.41; HMA: 0.45). Rates of acute and chronic GVHD and 2-year NRM appeared comparable to other studies on allo-HCT in AML and MDS that did not administer maintenance therapies.⁵¹

Our results for the FLT3 inhibitor cohort are in line with both another meta-analysis and a phase II study by Schlenk et al, which was not included in our analysis, as it did not report data on our primary outcome of 2-year OS.⁵² These studies reported HR for death of 0.48 and 0.60 favoring maintenance therapy over observation, respectively.^{52, 53} While sorafenib maintenance therapy following allo-HCT has demonstrated significant OS benefits in randomized trials, this was not the case for midostaurin, and it remains to be seen if survival benefits also apply to other FLT3 inhibitors such as gilteritinib.^{16, 33, 36} The ongoing randomized phase III trial of the FLT3 inhibitor gilteritinib (NCT02997202) will be important to characterize the efficacy and safety profile of other FLT3 inhibitors in this setting further.⁵⁴

While FLT3 mutations are a well-defined driver in AML that can be directly targeted, preclinical evidence suggests that HMAs might enhance the graft-versus-leukemia effect via the increased expression of leukemia antigens and cytotoxic T-cell responses and potential reversal of loss of human leukocyte antigen (HLA) expression.^55–59 Among the nine studies using HMAs included in our meta-analysis, only the study by Gao et al. had a randomized design.⁴⁴ Recently, Oran et al. reported the results of the randomized, open-label phase III trial of azacitidine maintenance vs observation in AML and MDS patients, which did not show a benefit in RFS and OS with maintenance therapy.⁶⁰ We were unable to include this study in our meta-analysis as it did not report data for our pre-defined primary outcome of 2-year OS and we were unable to obtain the missing information from the corresponding author. However, several limitations of the trial by Oran et al. have been pointed out such as the very heterogenous patient population (e.g. remission status at time of transplant), the high rates of screening failure, the underrepresentation of reduced-intensity conditioning regimens and MRD-positivity (patients who might benefit the most from maintenance therapy), and early discontinuation of the study due to slow accrual.^{60, 61} Therefore, additional results from ongoing studies using oral azacitidine (NCT04173533) or venetoclax + azacitidine (NCT04161885) are necessary to better define the role of HMA-based maintenance therapy following allo-HCT.

Several open questions regarding the optimal use of maintenance therapy following allo-HCT remain such as the timing of maintenance treatment initiation and optimal duration of treatment, baseline disease characteristics in patient selection, the role of MRD status prior to transplant, and the role of transplant characteristics (e.g., conditioning regimen intensity, donor selection, and GVHD prophylaxis).^{62, 63} Randomized clinical trials are needed to address those areas of uncertainty.

The planned duration of maintenance therapy among included studies ranged between 6–24 months for FLT3 inhibitors and 6–12 cycles for HMA, with dose adjustments and treatment discontinuation due to AEs occurring in around 40% and 20%, respectively. While we were unable to perform a formal meta-analysis, neither the duration of therapy nor the need for dose adjustment appeared to impact therapeutic efficacy. However, the optimal duration and dosing needs to be further evaluated in clinical trials. While additional studies are necessary, 40% in the SORMAIN trial relapsed after discontinuation of sorafenib suggesting that a longer duration of therapy might be advantageous.¹⁶

As maintenance therapy can be associated with a substantial burden on patients and caregivers, selecting patients most likely to benefit from maintenance therapy is essential. One potential approach to identifying patients at highest risk for disease relapse and who are thus likely to experience the greatest benefit from maintenance therapy, could be assessment of MRD status either directly prior to or following allo-HCT. MRD status has been shown to be predictive of disease relapse following allo-HCT.^{64, 65} Additionally, patients with MRD-positivity after transplant gained a significant RFS advantage from sorafenib maintenance in the trial by Burchert et al.¹⁶ Similar findings were reported by Xuan et al. with a HR for relapse among MRD-positive, sorafenib-treated patients of 0.25 (95% CI: 0.06–0.94).³³ However, additional studies are needed to validate MRD-based patient selection for maintenance therapy.

Despite its robust methodology and the inclusion of 21 studies with 809 patients, limitations of our analysis exist. First, the patient populations, transplant characteristics, and interventions across the included studies were variable which led to moderate or substantial heterogeneity for most of the analyses presented here. Therefore, we were unable to evaluate the influence of prior treatment (e.g., pre-transplant use of FLT3 inhibitors), donor source, and post-transplant immune suppression on outcomes. Second, genetic data and MRD status were not reported consistently across studies precluding a formal analysis. Third, we were not able to qualitatively assess the safety profile other than GVHD due to the heterogeneity in AE data reporting by the primary studies. While this study used an exhaustive search of the literature, unpublished literature might have not been discovered, which could have caused a publication bias. Additionally, three non-English language articles were excluded from this study and regional and language specific databases were not searched, which may have resulted in potential language and location bias.

Conclusion:

In this systematic review and meta-analysis of 21 studies with 809 AML and MDS patients receiving maintenance therapy with either FLT3 inhibitors or HMA following allo-HCT, we found 2-year OS rates of 81.7% and 65.7% among patients treated with FLT3 inhibitors and HMA, respectively. Subgroup analyses revealed an improved OS and RFS with maintenance therapy compared to observation with a favorable safety profile. Additional studies to define the optimal duration of treatment, the role of MRD status in patient selection, and the role of transplant characteristics (e.g., conditioning regimen intensity, donor selection, and GVHD prophylaxis) are needed.

Highlights.

• Meta-analysis of 21 studies on post-allo-HCT maintenance in AML and MDS

• 2-year OS rates of 81.7% with FLT3 inhibitors and 65.7% with hypomethylating agents

• Rates of GVHD comparable to other studies on allo-HCT in AML and MDS

Data sharing statement:

Original data from the studies included in this meta-analysis have been previously published. Primary data and research methodology can be requested from the corresponding author.