Multicenter Phase I Trial of Ivosidenib as Maintenance Treatment Following Allogeneic Hematopoietic Cell Transplantation for IDH1-Mutated Acute Myeloid Leukemia

This article is featured in Highlights of This Issue, p. 2013

Allogeneic hematopoietic cell transplantation (HCT) is the curative consolidation approach recommended for the majority of clinically robust patients with acute myeloid leukemia (AML). Nevertheless, leukemic relapses remain a frequent challenge (1, 2). The recent discovery of targetable molecular alterations has allowed the development of small molecule inhibitors as novel therapies (3–5), including those of FMS like tyrosine kinase 3 (FLT3) and isocitrate dehydrogenase (IDH1/2). After demonstration of impressive activity in clinical studies, several agents have been approved for use, including the FLT3 inhibitor gilteritinib and the IDH inhibitors ivosidenib and enasidenib (6–8).

Maintenance therapy, which aims to reduce relapse incidence and buttress the potential of cure following HCT (9), has shown marked promise among patients with FLT3-mutated AML. Early-phase clinical studies (10, 11) have been followed by randomized clinical trials of post-HCT FLT3 inhibitors, revealing improved outcomes and survival (12–14). In addition, a recent dose-escalation phase I of study of enasidenib as post-HCT maintenance among IDH2-mutated patients demonstrated promising activity and tolerability (15).

However, to date, there have been no published data on the IDH1 inhibitor ivosidenib as post-HCT maintenance therapy. IDH1 mutations are found in approximately 5% to 10% of patients with AML (16–18). The resulting altered IDH1 protein, through neo-enzymatic activity, catalyzes the formation of the oncometabolite 2-hydroxyglutarate (2-HG) from α-ketoglutarate in the malignant cell (5). 2-HG in turn suppresses enzymes involved in epigenetic regulation, leading to DNA and histone hypermethylation, cessation of normal differentiation, and ultimately, the development of myeloid malignancy (19).

Ivosidenib is a specific and potent inhibitor of the altered IDH1 enzyme. In clinical trials, it led to impressive rates of marrow response and composite remission among relapsed/refractory (R/R; ref. 6) as well as older, newly diagnosed (20) patients with AML, for whom the drug is now approved as monotherapy. A recent placebo-controlled phase III clinical trial, the AGILE Study, demonstrated markedly superior survival outcomes with the combination of ivosidenib and azacitidine among newly diagnosed patients unsuitable for intensive therapies (21), leading to the additional approval of this combination. We sought, with this phase I, multicenter, dose-finding study, to establish the MTD or recommended phase 2 dose (RP2D) of ivosidenib as maintenance therapy following HCT among patients with IDH1-mutated AML and to gather a preliminary assessment of its therapeutic activity in this setting.

This multicenter study was conducted and enrolled patients at Massachusetts General Hospital, Dana-Farber Cancer Institute, and the Johns Hopkins Hospital Cancer Center, with Institutional Review Boards at these respective institutions reviewing and approving the study. Informed written consent was obtained from each study subject. The trial was conducted according to the Declaration of Helsinki and registered at ClinicalTrials.gov (NCT03564821).

The study eligibility allowed those patients who were 18 years and over, with either AML in remission, or with MDS or chronic myelomonocytic leukemia (CMML), and whose disease possessed an IDH1 p.R132 mutation by local testing, established by PCR or next-generation sequencing (NGS) technique. Transplant donors could be: ≥5/6 (HLA-A, B, DR) matched related donors, ≥7/8 (HLA-A, B, DR, C) matched unrelated donors, haploidentical donor of ≥3/6 (HLA-A, B, DR), or umbilical cord blood (UCB) ≥4/6 (HLA-A, B, DR). Conditioning approaches included myeloablative (MAC) or reduced intensity (RIC) regimens. Graft versus host disease (GVHD) prevention was at the discretion of the treating physicians. Patients were screened and enrolled prior to transplant, at which point they were required to demonstrate normal organ and marrow function [neutrophil count (ANC) >1,000/μL and platelet count > 50,000/μL for patients with AML). Prior HCT, relapsed or refractory disease, uncontrolled infections, class 3 or higher New York Heart Association heart failure, and a QTc ≥450 milliseconds were exclusionary. Use of ivosidenib prior to HCT was allowed.

Ivosidenib maintenance was started between days 30 and 90 following HCT. At that time point, the following eligibility was required to be met: no disease relapse, ANC ≥1,000/μL and platelets ≥50,000/μL, direct bilirubin ≤2.0 × upper limit of normal (ULN), chimerism ≥70% of donor origin among blood or marrow cells, no acute GVHD requiring equivalent dose of ≥0.5 mg/kg/day of prednisone, and no active infection requiring intravenous antibiotics. Of note, patients could receive prophylaxis with standard medications for GVHD prophylaxis, such as tacrolimus. However, trough levels were monitored closely, given that some of these agents are CYP3A substrates, with possible interactions with ivosidenib. Of note, the protocol was silent about donor lymphocyte infusions (DLI), but no patients received DLI treatment during the conduct of the study. Standard transplant platforms at the participating institutions did not include planned DLI given the low incidence of mixed chimerism. Also, no patients in this study had any ex vivo graft manipulation such as T-cell depletion. All pre-maintenance assessments, including a bone marrow biopsy, were performed within 14 days of maintenance initiation.

Ivosidenib was to be administered orally and daily in 28-day cycles. The initial dose level was 500 mg daily. A second de-escalated dose level cohort of 250 mg daily would enroll patients if 500 mg daily was deemed intolerable or dose-limiting. The period for dose-limiting toxicity (DLT) evaluation was the first cycle. Dose de-escalation was guided by DLT incidence according to a standard 3+3 design. If either 250 or 500 mg daily were found tolerable, this would be deemed to be the MTD or RP2D, respectively, after which 10 additional patients would be subsequently enrolled in an expansion cohort. Bone marrow biopsies were performed before initiation of treatment, and after cycle 6 and 12, to assess for ongoing response and remission. Patients were monitored for tolerability and relapse throughout. Treatment was to continue until disease progression, intolerable toxicity, or upon receipt of 12 cycles.

Per protocol, if at any time following the DLT period, a participant experienced a ≥grade 3 nonhematologic toxicity or ≥grade 4 neutropenia or thrombocytopenia, at least possibly related to ivosidenib, drug was held until the toxicity resolved to a grade 1 or lower. If the participant was at the 500 mg daily dosing level when the toxicity occurred, and the toxicity resolution to ≤grade 1 extended beyond 14 days, then dosing would be reduced to 250 mg daily at time of resumption. Otherwise, the resumption would be at 500 mg daily if the resolution was sooner. If the participant was at the 250 mg daily dosing level when the toxicity occurred, and the toxicity resolution to ≤grade 1 extended beyond 14 days, then ivosidenib would be discontinued.

As an assessment for MRD, IDH1 mutation clonal burden in the marrow was assessed for a subset of patients prior to HCT, and also following HCT but prior to the start of ivosidenib. During the duration of the accrual period, IDH1 mutational burden as measured by variant allelic fraction (VAF) was measured by three sequential assays, with varying sensitivities. The genomic DNA isolated from the bone marrow samples met the minimum input material threshold (per assay specification) to obtain valid quantification of the fraction of mutant alleles to wild-type alleles in each sample. The two NGS assays were performed at Brigham and Women's Hospital: versions 2 and 3 of the Rapid Heme Panel (RHP v2 and v3). For any given locus, the analytical sensitivity is determined by the ability to call true signal over background noise. For RHP v2, the published analytical sensitivity of the assay was 5% VAF, although for identification of a previously reported variant, the practice was to call variants more than 10-fold above the background noise of that locus (minimum of 10–20 reads if minimal noise at the locus; ref. 22). The mean coverage of the IDH1 p.R132 locus for the samples in this study was 1,740× read depth (range: 1,476–2,471×) read depth, allowing as low as approximately 0.6% to 1.2% analytical sensitivity. RHP v3 uses unique molecular identifiers that removes much of the noise, allowing variants to be called at potentially lower VAFs. For the samples in this study, the mean coverage at the IDH1 hotspot was 1,063× consensus read depth (range: 616–2,186×) with variants reported in as few as five unique molecules (0.5%).

A third assay for a subset of samples assessed IDH1 R132 VAFs through BEAMing digital PCR technology (OncoBEAM; Sysmex Inostics). This technique has a lower limit of detection of 0.02% VAF for mutant IDH1 alleles c394t, c394g, g395t, and c394a (corresponding to R132C/G/L/S, respectively) and a lower limit of detection of 0.04% VAF for mutant IDH1 allele g395a, corresponding to R132H.

A standard 3+3 study design was used for dose de-escalation, with an expansion cohort. The primary endpoint of this study was to establish the MTD or RP2D of ivosidenib in the post-HCT setting. Nonhematologic DLTs included drug-related, NCI Common Terminology Criteria for Adverse Events (CTCAE v4.0) ≥grade 3 adverse events (AE) not related to underlying disease, with exceptions of grade 3 nausea, vomiting, or diarrhea that lasted ≤72 hours and responded to medical intervention. Hematologic DLTs were defined as any drug-related grade 4 neutropenia (ANC <500/μL) or thrombocytopenia (platelets <25,000/μL) that did not resolve to ≤grade 1 within 14 days or resolved within 14 days but recurred after ivosidenib resumption. Any ≥grade 2 nonhematologic toxicity deemed intolerable or which caused an inability to take ≥75% of doses, any febrile neutropenia not due to underlying disease or other causes, and study drug-related deaths would additionally be considered DLTs. Patients required receipt of at least 50% of dosing during the first cycle to be DLT-evaluable. If a DLT occurred among those receiving <50% of study drug, however, it would remain evaluable.

Among the secondary and exploratory endpoints were progression-free survival (PFS), overall survival (OS), cumulative rate of relapse, cumulative incidence rates of acute GVHD (aGVHD) and chronic GVHD (cGVHD), GVHD-free and relapse-free survival (GRFS), and the proportion of patients who successfully screened prior to HCT but did not reach the maintenance phase. aGVHD grading was conducted as per consensus criteria, and cGVHD grading was conducted as per NIH consensus criteria (23, 24). GRFS was defined as period from stem cell infusion to institution of systemic immune suppression, relapse or death, whichever occurs first.

Descriptive analysis was primarily performed. OS, PFS, and GRFS were estimated using the Kaplan–Meier method, whereas cumulative incidence of (acute/chronic) GVHD, relapse, and non-relapse mortality (NRM) were estimated in the competing risks framework (25) treating relapse or death without developing GVHD as a competing event for GVHD, NRM for relapse, and relapse for NRM as a competing event. Univariable Cox regression analysis was performed to identify potential risk factors for OS and PFS. Although ivosidenib was administered after stem cell infusion, because the trial was not intended to compare with nontreated patients, the time zero was set to the date of stem cell infusion for time-to-event analysis for the purpose of future reference. All testing was two-sided at the significance level of 0.05. All analyses were performed using SAS 9.3 (SAS Institute Inc.), and R v3.4 (the CRAN project).

Data generated in this study are available within the article and its supplementary data files. Any additional data from this study are available upon request from the corresponding author.

A total of 18 patients, all with acute myeloid leukemia, were registered prior to HCT. Two patients did not initiate ivosidenib maintenance following transplant, both due to interval development of acute GVHD requiring concurrent treatment. The remaining 16 patients were initiated on treatment with ivosidenib, with a median time of starting maintenance at day 66 (range 37–109) after HCT. One patient was allowed to initiate maintenance beyond the protocol window, at day 109. This patient had been on a steroid taper for a recent diagnosis of aGVHD, and a request was made to extend the start date so that the patient could conclude the taper and then initiate maintenance, because they had otherwise been eligible.

Key demographic and clinical data for treated patients are provided in Table 1. The median age was 62 years (range 28–76). Nine (56%) participants were male and 7 were female (44%). Fourteen subjects (87.5%) were Caucasian, and 2 were Black (12.5%). Among the 16 treated patients, 13 (81%) had de novo AML and 3 (19%) had AML with MDS-related changes (AML-MRC). Twelve (75%) and 4 (25%) patients received one and two lines of therapy prior to HCT, respectively. The second line of treatment was for relapsed or refractory disease. Before HCT, 15 of 16 patients (94%) received at least one line of intensive induction chemotherapy, although 4 of these patients received subsequent nonintensive treatment (with hypomethylating agent and/or ivosidenib-based therapies) for relapsed or refractory disease. One patient received only nonintensive treatments with hypomethylating agent-based therapy prior to HCT. Among all patients receiving maintenance therapy, 5 of 16 (31%) had received ivosidenib prior to HCT. Ten patients (63%) received reduced-intensity, and 6 patients (38%) received myeloablative conditioning prior to transplantation. Ten patients (63%) underwent an 8/8 matched unrelated, 4 (25%) haploidentical, and 2 (13%) matched related transplant. Of note, the representativeness of study participants is detailed in Supplementary Table S1.

Table 1 also demonstrates the molecular data for patients. Data on the type of IDH1 mutation were available for 15 of 16 treated patients. Seven participants (47%) displayed a R132C, 4 (25%) had R132H, 3 (20%) had R132G, and 1 had an R132L (6.7%) alteration. Twelve patients (75%) had intermediate risk cytogenetics, with 10 (63%) having a normal karyotype, and 4 patients (25%) had adverse-risk cytogenetics. Of these 4, 1 had inversion of chromosome 3, 1 had a 6;9 translocation, and 2 patients displayed a complex karyotype. Figure 1A displays the concurrent mutations seen among treated patients at time of initial diagnosis, with the most common mutations being DNMT3A (69%), NPM1 (56%), SRSF2 (24%), PTPN11 (19%), BCOR (19%), and FLT3-TKD (19%). Risk categorization by cytogenetics and ELN, along with concurrent mutations, according to patients who subsequently relapsed versus those remaining in remission, is demonstrated in Fig. 1B.

Six total patients were enrolled at the 500 mg daily (the first dose level) of maintenance ivosidenib. Only one DLT was detected, a case of grade 3 prolonged QTc interval. Dose de-escalation was therefore not deemed necessary. An additional 10 patients were enrolled in an expansion cohort at the same dose of 500 mg daily, the RP2D. Adverse events, at least possibly attributable to study treatment, are shown in Table 2. Grade 3 or higher attributable adverse events were uncommon, and included prolonged QTc interval (in 2 patients, both grade 3) and paresthesia (in 1 patient, grade 3). For the two cases of grade 3 QTc interval prolongation, both had their drug held, and upon resolution of the grade 3 event, were resumed ultimately on a lower dose, 250 mg daily, without recurrent grade 3 prolongation. During the course of maintenance treatment, 3 patients (19%) experienced transient dose interruption on 6 separate occasions. These pauses lasted a median duration of 20 days (5–36 days) and were triggered by QTc prolongation (2 instances), skin GVHD (1 instance), pruritis (1 instance), neuropathy (1 instance), and azotemia (1 instance). Three patients (19%) patients underwent a dose reduction to 250 mg, on which they continued treatment. Eight participants (50%) discontinued study treatment prior to reaching 12 cycles; 3 due to disease relapse, 3 to pursue therapies for chronic GVHD (cGVHD), 1 due to adverse events (pruritis), and 1 due to patient preference. The patients who discontinued treatment for cGVHD either required other experimental therapies or higher doses of steroids than allowed per protocol for their GVHD. The median number of completed cycles was 11 (3–12 cycles). Eight patients received all 12 planned cycles of maintenance. One patient each received 10, 8, 7, 6, and 4 cycles, respectively, and 3 patients received three cycles of ivosidenib maintenance.

Three patients (19%) experienced aGVHD following HCT; 2 grade I and 1 grade II. Of these, two cases occurred before initiation of maintenance ivosidenib, including a subject who developed grade I aGVHD before maintenance therapy and then progressed to grade II aGVHD after therapy. The overall cumulative incidence of grade II to IV aGVHD at 6 months post-HCT was thus 6.3% (95% CI, 0.4–25), but no cases of grade III or IV aGVHD were detected. Eleven instances of cGVHD were reported; 1 mild, 6 moderate, and 4 severe, and 10 patients with cGVHD required systemic steroids. The 2-year cumulative incidence of all grades of cGVHD was 63% (95% CI, 32%–82%), and the 2-year cumulative incidence of moderate/severe cGVHD was 56% (95% CI, 28%–77%; Table 3; Fig. 2). Chimerism data prior to start and after conclusion of maintenance were available for 14 patients. No substantial change in chimerism between these timepoints, as measured by T cell and bone marrow / peripheral blood analysis, was noted (median 100% vs. 100%).

Median follow-up following HCT among surviving patients was 29 months (range 17–44 months). Three patients (19%) have relapsed during follow-up, at 115, 168, and 175 days following HCT. Among the 16 treated patients, 14 (88%) remain alive. To date, 2 patients (13%) have died, both due to disease relapse, at 253 and 265 days following transplantation. The 2-year cumulative incidence of disease relapse was 19% (95% CI, 4.0%–41%) and the 2-year cumulative incidence of nonrelapse mortality was 0%. At 2 years, OS was 88% (95% CI, 59%–97%), PFS was 81% (95% CI, 52%–94%), and GRFS was 25% (95% CI, 8%–47%) for treated patients (Table 3; Fig. 3).

Baseline demographic, disease, and transplant factors were evaluated for association with PFS and OS. All 3 patients who relapsed, which included the 2 who subsequently died in follow-up, were among the 5 patients who had received transplantation following second-line therapy for their AML (OS, P = 0.027, Fig. 4; Supplementary Fig. S1). None of the participants undergoing consolidative HCT in first response have to date relapsed or died. An additional analysis specifically assessed NPM1-mutated patients, and no significant difference in either OS or PFS between NPM1-mutated and NPM1-wild-type patients was noted (Supplementary Fig. S2). Otherwise, statistically significant associations with outcome were not detected. This is likely largely due to the small number of events and participants.

As a marker of measurable residual disease, bone marrow IDH1 mutational burden prior to transplant was available in a limited number of patients, 5. Among these, 3 had detectable IDH1 mutations, of whom 2 (67%) subsequently relapsed and have both died. The 2 patients who did not have detectable IDH1 mutations prior to HCT remain alive and in remission. Bone marrow IDH1 mutational burden prior to start of ivosidenib maintenance was available for a larger group, 11 patients. Among these, IDH1 mutations were detectable in 2 patients, of whom 1 (50%) relapsed and subsequently died. Among the 9 remaining participants who did not have detectable IDH1 mutations prior to ivosidenib maintenance, 2 patients (22%) have relapsed. These data along with other baseline information is highlighted in the Swimmer plot in Fig. 4, along with IDH1 variant allelic fractions provided in Supplementary Table S2.

Of the 13 patients who have not relapsed on study, none had detectable IDH1 mutations at 12 months. Of note, mutational data at time of relapse was available for the 3 relapsing patients. For two of the patients, no new alterations were noted at relapse, and the same IDH1 mutation was again detected. However, for the remaining patient, a newly emergent IDH2 R140Q mutation was noted on relapse, and likely contributed mechanistically to relapse and therapeutic resistance.

The last decade has seen a transformation of the therapeutic landscape for acute myeloid leukemia. A variety of novel and targeted therapies have demonstrated tolerability and efficacy in clinical trials, and several have subsequently been approved for use. Among these are IDH inhibitors, which in single-arm phase I and II studies of predominantly R/R patients were associated with a very promising rate of marrow response and marked tolerability (6, 8, 26). IDH inhibitors have also demonstrated substantial clinical activity in combination with hypomethylating therapies as upfront treatment of older patients with AML (21, 27). Ivosidenib, a potent and selective IDH1 inhibitor, in particular, when combined with HMA, demonstrated a marked survival advantage over HMA monotherapy in a phase III, placebo-controlled clinical trial (21).

To date, despite its impressive tolerability and safety profile, ivosidenib has not been studied as a maintenance strategy following HCT. In recent years, FLT3 inhibiting therapies have been studied in this setting, and have demonstrated substantial therapeutic promise in FLT3-mutated patients. Sorafenib demonstrated improved survival and significantly lower rates of relapse in multiple recent randomized studies (12, 13). The recently reported Quantum-First study evaluating quizartinib in the frontline setting for AML, also incorporated this potent FLT3 inhibitor as maintenance therapy following transplant, again revealing therapeutic promise (14). Another phase III study focusing on maintenance with FLT3 inhibitor, gilteritinib, recently finished accrual. This progress has reinvigorated interest in other post-HCT maintenance therapies, with several ongoing studies of various novel therapies. We recently published data on the use of the IDH2 inhibitor enasidenib as post-HCT maintenance therapy (15), and this agent was well-tolerated, with promising activity.

With this study, we sought to gain insights on the tolerability and an early estimation of efficacy for ivosidenib as a maintenance strategy following HCT in IDH1-mutated patients. We found that the agent was well-tolerated at a dose of 500 mg daily and deemed this the RP2D. 500 mg daily is also the approved dose of ivosidenib in the frontline and R/R settings. Grade 3 or higher attributable adverse events were uncommon, with the most common being a prolonged QTc, occurring in two patients, and in neither case leading to episodes of tachyarrhythmia. Although 8 patients did not complete all protocol-assigned cycles, only 1 patient stopped treatment due to adverse events deemed at least possibly related to treatment. Transient dose interruptions did occur in a minority (19.1%) of patients, in whom a reduction to 250 mg daily allowed them to resume and tolerate treatment.

Cases of aGVHD were minimal on study, and all three noted cases were grade I or II. Only one case of aGVHD developed following start of ivosidenib maintenance. The cumulative rate of aGVHD was lower than, and that of cGVHD approximated, that which has historically been seen in patients undergoing allogeneic HCT. The cumulative incidence of NRM was 0, which is very reassuring and much better than historical data, although our dataset is obviously small (28, 29). Four patients did experience severe cGVHD, and 10 required initiation of steroids for treatment of cGVHD. This too is aligned with what is typically expected in patients with AML undergoing allogeneic transplant in this setting.

Prolongation of the QTc interval, attributable to ivosidenib, occurred in 5 patients on study, with two cases being grade 3. For those patients with grade 3 QTc prolongation, dosing pause followed by resumption at a lower dose allowed continuation of therapy. QTc interval prolongation is a known adverse event associated with ivosidenib and has been seen in clinical trials of the agent (6, 21). The rate of occurrence was similar in our study, and none of the events led to dangerous arrhythmias. Differentiation syndrome (DS) due to IDH inhibitors has also been identified in prior clinical trials of these agents (30–32). DS is caused by differentiation of myeloid blasts through targeted IDH inhibition, and its manifestation is clinically heterogeneous and potentially lethal if not addressed promptly (30). We mandated close monitoring and management guidance for DS, but no cases were reported, likely due to the fact that ivosidenib-treated patients were in presumed remission, with suppression of leukemia, at time of study enrollment.

Posttransplant outcomes for IDH-mutated patients with AML have been reported in a few recent publications. These outcomes appear to approximate historical outcomes of the general AML population. One study assessed 23 patients, with a very brief follow-up of 7.8 months, and reported a 1-year OS of 68% and a relapse rate of 29%. These outcomes were not further categorized by IDH1 and IDH2 subtypes (29). A larger multicenter study retrospectively evaluated transplanted IDH-mutated patients, with a longer median follow-up of 25.2 months. IDH1-mutated patients experienced a 2-year PFS and OS of 58% and 74%, respectively. The incidence of relapse and NRM at 2 years was 31% and 11%, respectively (28). In another recently published multicenter study of 3,234 patients, allogeneic HCT improved relapse-free survival in IDH1-mutated AML at 5-years when compared with those undergoing chemotherapy consolidation (51% vs. 30%). OS (47% vs. 40% at 5 years) was not significantly different between the two groups, however. The ultimate question remains as to whether maintenance therapy with ivosidenib will improve outcomes among transplanted patients further (33). In our study, 3 patients have experienced relapse in follow-up, two of whom have subsequently succumbed to their leukemia. Among 16 patients treated with post-HCT maintenance therapy, the estimated 2-year PFS was 81% and 2-year OS was 88%. The 2-year cumulative incidence of relapse was 19% and that of NRM was 0%. Despite the real limitation of low numbers in this phase I study, these results compare very favorably with the reported historical experience. Of note, almost a third of our patients had undergone HCT following therapy beyond the first-line setting. Additional study, in larger trials, is needed to confirm these promising results.

NPM1-mutated patients, considered to have more favorable-risk disease, were relatively overrepresented in our cohort. We performed an additional analysis specifically looking at the NPM1-mutated patients, and albeit impacted by small numbers, there was no significant difference in survival outcomes between NPM1-mutated or wild-type patients, suggesting this was not the driver of the favorable outcomes as a whole. We also reviewed and analyzed other comutations noted among these patients, but no definitive prognostic trends for outcomes can be established, given the small number of data points.

Clinical studies to date have shown that MRD detection in transplanted patients predicts for worse outcomes in AML (34–38). In our study, a detectable marrow IDH1 mutation prior to maintenance appeared to be associated with a higher rate of subsequent relapse and death. Admittedly, however, interpretation of our MRD data, and the impact of ivosidenib maintenance on MRD, are limited by the number of assessed patients. Therefore, larger randomized studies are needed to more fully assess the role of MRD in this setting, particularly its implications for use of maintenance therapy. Intriguingly, one of three relapsing patients had a newly emergent IDH2 mutation at time of relapse, as a possible explanation for therapeutic resistance and disease progression. It is noteworthy to further mention that IDH1 and IDH2 mutations have been uncommonly associated with alterations thought to be premalignant, rather than malignant, as markers of clonal hematopoiesis of indeterminate potential (CHIP), particularly among older patients. It is also known that persistence of mutations prior to transplant as MRD is associated with relapse (39–41). These considerations add an extra layer of complexity among patients diagnosed with AML and found to have IDH mutations, in terms of monitoring the disease with and relying on IDH mutations as an MRD marker. Therefore, for such patients, more broad approaches to MRD analysis, such as with flow cytometry or with NGS may be more suitable.

Study limitations included a small study sample, as expected in a phase I trial. As mentioned, true estimations of efficacy, impact of MRD, and long-term outcomes are best assessed in larger studies, which we hope to pursue in the near future. This was a multicenter study, but given that only three academic sites enrolled patients, the effects of demographics, patient selection, and regional treatment and transplant patterns may not be fully representative of the larger population. Also, our study population included both newly diagnosed and R/R disease, and in that sense was heterogeneous, and therefore appropriate contextualization of the results are appropriate. Larger studies of ivosidenib maintenance therapy are needed to overcome some of these shortcomings. The current HOVON-150 study (ClinicalTrials.gov Identifier: NCT03839771), a large phase III study conducted in Europe and assessing the promise of IDH inhibition with induction chemotherapy, also incorporates maintenance with ivosidenib for IDH1-mutated patients. This may also shed further light on its promise in this setting.

In sum, to our knowledge, this is the first report of a clinical trial to focus on the role of IDH1 inhibition as maintenance following allogeneic transplantation for myeloid malignancies. We found that ivosidenib was well-tolerated, with a RP2D of 500 mg daily. The cumulative incidence of relapse was low, we did not note any NRM, and the estimated 2-year PFS and OS was highly promising. Larger randomized studies are needed to more fully capture the role and promise of IDH inhibition as maintenance therapy following HCT among patients with AML.

A.T. Fathi reports grants and other support from Agios and Servier and personal fees from Agios and Servier during the conduct of the study; personal fees from BMS, Celgene, Abbvie, Forma, Genentech, Takeda, Astellas, Kura Oncology, Novartis, Pfizer, Rigel, Daiichi Sankyo, Ipsen, and Immunogen outside the submitted work. R. Soiffer reports grants from Agios (now Servier) during the conduct of the study. R. Soiffer also reports personal fees from Cugene, Takeda, Jasper, Jazz Pharmaceuticals, Precision Biosciences, Alexion, Kiadis, Juno Therapeutics/BMS/Celgene, Rheos Therapeutics, Gilead, Vor Biopharma, Smart Immune, Daiichi Sankyo, Neovii, CSL Behring, Bluesphere Bio, and Astella and nonfinancial support from NMPD - Be The Match outside the submitted work. M.J. Levis reports personal fees from Abbvie, Amgen, Bristol-Myers Squibb, Daiichi-Sankyo, Jazz, Novartis, and Takeda; grants and personal fees from Astellas; grants from FujiFilm outside the submitted work. S. Li reports other support from Takeda outside the submitted work. A.S. Kim reports grants from Multiple Myeloma Research Foundation and personal fees from LabCorp, Inc. outside the submitted work. Z. DeFilipp reports personal fees from Syndax Pharmaceuticals, Kadmon, Omeros, Incyte, and MorphoSys outside the submitted work. A.M. Brunner reports other support from Agios during the conduct of the study; personal fees from Agios, Novartis, Takeda, BMS/Celgene, Taiho, Keros, Acceleron, and Gilead outside the submitted work. P.C. Amrein reports grants from Takeda, Amgen, and AstraZeneca outside the submitted work. A.S. Mims reports personal fees from Servier Pharmaceuticals during the conduct of the study. A.S. Mims also reports other support from Leukemia and Lymphoma Society and personal fees from AbbVie Pharmaceuticals, Syndax Pharmaceuticals, Astellas Pharmaceuticals, Jazz Pharmaceuticals, Daiichi Sanyko, BMS, Ryvu Therapeutics, Genetench, and Foghorn Therapeutics outside the submitted work. D.M. Marchione reports employment with Servier Pharmaceuticals LLC. Y. Chen reports personal fees from Incyte, Abbvie, Daiichi, Equilium, Actinium, Celularity, and Magenta outside the submitted work. No disclosures were reported by the other authors.

A.T. Fathi: Conceptualization, data curation, formal analysis, investigation, methodology, writing–original draft. H.T. Kim: Data curation, formal analysis, writing–review and editing. R.J. Soiffer: Data curation, writing–review and editing. M.J. Levis: Data curation, investigation, writing–review and editing. S. Li: Data curation, formal analysis. A.S. Kim: Data curation, investigation, writing–review and editing. Z. DeFilipp: Data curation, investigation, writing–review and editing. A. El-Jawahri: Data curation, investigation, writing–review and editing. S.L. McAfee: Investigation, writing–review and editing. A.M. Brunner: Investigation, writing–review and editing. P.C. Amrein: Investigation, writing–review and editing. A.S. Mims: Writing–original draft. L.W. Knight: Data curation, formal analysis, writing–review and editing. D. Kelley: Data curation, writing–review and editing. A.S. Bottoms: Data curation, writing–review and editing. L.H. Perry: Data curation, writing–review and editing. J.L. Wahl: Data curation, writing–review and editing. J. Brock: Data curation, writing–review and editing. E. Breton: Data curation, writing–review and editing. D.M. Marchione: Data curation, investigation, writing–review and editing. V.T. Ho: Investigation, writing–review and editing. Y. Chen: Data curation, investigation, writing–original draft, writing–review and editing.

Agios Pharmaceuticals (now Servier) kindly provided funding and supply of ivosidenib for the conduct of this study, as well as kindly providing funding for correlative IDH1 mutational analysis.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.