Enasidenib in patients with mutant IDH2 myelodysplastic syndromes: a phase 1 subgroup analysis of the multicentre, AG221-C-001 trial

Summary

Background Mutations in isocitrate dehydrogenase-2 (IDH2) occur in around 5% of patients with myelodysplastic syndromes. Neomorphic activity of mutant IDH2 proteins results in hypermethylation of DNA and histones, leading to blocked haemopoietic differentiation. Enasidenib, an inhibitor of mutated IDH2 proteins, induces responses in patients with IDH2-mutated, relapsed or refractory acute myeloid leukaemia. We aimed to establish the clinical outcomes of enasidenib monotherapy in a subgroup of patients with myelodysplastic syndromes harbouring mutations in IDH2 from the AG221-C-001 trial.

Methods The multicentre, open-label, phase 1–2 AG221-C-001 trial enrolled patients with advanced haematological malignancies (2008 WHO criteria) harbouring an IDH2 mutation. The present study is a subgroup analysis of patients with IDH2-mutated myelodysplastic syndromes in the phase 1 dose-escalation and expansion portions of the trial. Patients with myelodysplastic syndromes were aged 18 years or older with an ECOG performance status score of 2 or lower, and were relapsed or refractory to, or ineligible for, standard treatments. Patients received oral doses of enasidenib at 60–300 mg per day in repeated 28-day treatment cycles. In this subgroup analysis, we focused on the safety and activity of enasidenib as main outcomes. Overall response rate, duration of response, and overall and event-free survival analyses were by intention-to-treat. Safety was assessed in all participants who received at least one dose of study drug in terms of treatment-emergent adverse events. The AG221-C-001 trial is registered on ClinicalTrials.gov, NCT01915498, status ongoing but closed to recruitment.

Findings 17 patients with myelodysplastic syndromes harbouring an IDH2 mutation (median age, 67·0 years [IQR 60·5–73·0]) were enrolled between Feb 18, 2014, and Sept 1, 2015. At data cutoff (Oct 1, 2018), after a median follow- up of 11·0 months (IQR 6·8–23·0), all patients had discontinued enasidenib, with a median of 3 treatment cycles (2–15) for all patients (five [29%] received ≥12 cycles). At entry, three (18%) patients had relapsed after allogeneic stem-cell transplants, 13 (76%) had previously received therapy with hypomethylating agents, and ten (59%) had received at least two previous therapies. No dose-limiting toxicities were reported. The most common treatment-emergent adverse events were diarrhoea and nausea (in nine [53%] patients each). Most common grade 3–4 treatment-emergent adverse events were indirect hyperbilirubinaemia (in six [35%] patients), pneumonia (in five [29%] patients), and thrombocytopaenia (in four [24%] patients). Serious treatment-emergent adverse events in more than one patient were pneumonia (in five [29% patients); tumor lysis syndrome (in three [18%] patients); and sepsis, atrial flutter, indirect hyperbilirubinaemia, cerebral hemorrhage, and mental status change (in two [12%] patients each). No treatment-related deaths occurred. An overall response was achieved in 9 patients (53% [95% CI 28–77]), with a median duration of response of 9·2 months (95% CI 1·0–not reached). Six (46%) of 13 patients previously treated with hypomethylating agents responded. Median overall survival was 16·9 months (95% CI 1·5–32·3), and median event-free survival was 11·0 months (1·5–16·7).

Interpretation Enasidenib is generally well tolerated and can induce responses in patients with mutant IDH2 myelodysplastic syndromes, including in those who have had previous therapy with hypomethylating agents. Testing for IDH2 mutations in myelodysplastic syndromes is essential for identifying patients who might benefit from enasidenib therapy, including those patients in whom conventional treatments have been unsuccessful.

Introduction

Myelodysplastic syndromes are a diverse group of bone marrow malignancies characterised by ineffective haemopoiesis and cytopaenias, leading to infections,
bleeding events, and risk of progression to acute myeloid leukaemia.1 The increasing availability of panels to test for gene mutations has allowed the identification of recurring somatic mutations with diagnostic and prognostic value in myeloid neoplasms, including myelodysplastic syndromes, and has improved understanding of the molecular biology of these neoplasms.2

Epigenetic mechanisms related to DNA methylation and histone acetylation can alter gene transcription.3 Altered DNA methylation, including abnormal methy­ lation of transcription promoter sites, is prevalent in patients with myelodysplastic syndromes, and the number of involved loci is increased in high­risk disease and during disease progression.1 Mutations in genes involved in DNA methylation, including DNA­ methyltransferase 3 alpha (DNMT3A), tet methylcytosine dioxegenase 2 (TET2), and isocitrate dehydrogenase (IDH) 1 and 2, appear to occur early in the genesis of myeloid malignancies (sometimes years before diagnosis4). Epigenetic changes resulting from these mutations might contribute to the development of myelodysplastic syndromes.1
Mutations in the mitochondrial IDH2 gene occur in around 5% of patients with myelodysplastic syndromes.5 The IDH1 and IDH2 proteins (isocitrate dehydrogenases [NADP], mitochondrial) catalyse the oxidative decar­ boxylation of isocitrate to α­ketoglutarate. Mutant IDH2 proteins have neomorphic enzymatic activity, catalysing the reduction of α­ketoglutarate to the oncometabolite, R­2­hydroxyglutarate.6,7 Pathogenic mechanisms of R­2­hydroxyglutarate include the induction of a hyper­ methylation phenotype, in part via inhibition of TET family enzymes involved in DNA demethylation and inhibition of Jumonji­C domain histone demethylases.6 The resulting hypermethylation of DNA and histones leads to altered gene expression and blocked differen­ tiation of hematopoeitic progenitor cells.6,7

Enasidenib is an oral, small­molecule, allosteric inhibitor of mutated IDH2 proteins that is approved in the USA for the treatment of adult patients (≥18 years) with mutant­IDH2 relapsed or refractory acute myeloid leukaemia. Although myelodysplastic syndromes and acute myeloid leukaemia might lie on a continuum of haematological malignancies, some data suggest they differ in important respects; for example, chromosomal deletions and amplifications are more common than chromosomal translocations in myelo­ dysplastic syndromes, whereas translocations are the more common abnormality in acute myloid leukeamia.8 Altered DNA methylation patterns also differ between acute myeloid leukaemia and myelo­ dysplastic syndromes.9

In this Article, we report a final phase 1 analysis of the safety and clinical and pharmacodynamic activity of enasidenib monotherapy in patients with advanced myelodysplastic syndromes with mutations in IDH2. Patients were from the AG221­C­001 trial, a large phase 1–2 study of orally administered enasidenib in patients with advanced haematological malignancies with an IDH2 mutation.

Methods

Study design and participants

The AG221­C­001 trial was a phase 1–2, multicentre, open­label study. The study design and procedures have been reported in detail previously.10,11 Participants were from 24 academic medical centres and cancer centres in France (five sites) and the USA (19 sites). The present subgroup analysis was of the phase 1 dose­escalation and expansion portions of AG221­C­001. The AG221­C­001 study protocol and amendments were approved by institutional review boards or ethics committees at all participating sites. All patients provided written informed consent before study participation.

The phase 1 portions of the AG221­C­001 trial enrolled patients aged at least 18 years presenting with an IDH2 mutation and presenting with advanced haematological malignancies according to 2008 WHO criteria (newly diagnosed or relapsed or refractory acute myeloid leuk­ aemia; or relapsed or refractory [or treatment­intolerant] myelodysplastic syndrome), or other advanced haema­ tological malignancies that met the eligibility criteria on a case­by­case basis. Eligible patients with myelodysplastic syndromes were those presenting with refractory anaemia with excess blasts (subtype RAEB­1 or RAEB­2), or those classified as having a high­risk prognosis by the revised International Prognostic Scoring System (IPSS­R),12 with an Eastern Cooperative Oncology Group (ECOG) performance status score of 2 or lower. The presence of an IDH2 mutation was evaluated by the local laboratory at each site and confirmed by the treating physician. Patients with myelodysplastic syndrome were in relapse or refractory to previous treatment, or were considered by the treating physician to be intolerant to regimens known to provide clinical benefit, including hypomethylating agents. Patients were excluded if they had: received haemopoietic stem­cell transplant within 60 days, were on immunosuppressive therapy post­transplant at the time of screening, or had clinically significant graft­versus­host disease; or received systemic anticancer therapy, radiation, or small­molecule investigational drugs within 14 days before their first dose of enasidenib. Hydroxyurea was allowed before enrollment and after the start of enasidenib for the control of peripheral leukaemic blasts in patients with leucocytosis. We also excluded patients with uncontrolled hypertension (systolic blood pressure
>180 mm Hg or diastolic blood pressure >100 mm Hg), unstable angina, active CNS leukaemia, life­threatening complications of leukaemia (eg, uncontrolled bleeding, pneumonia with hypoxia or shock, or disseminated intravascular coagulation), or any condition that would limit the ingestion or gastrointestinal absorption of drugs administered orally (eg, dysphagia, short­gut syndrome, or gastroparesis). The complete eligibility criteria are provided in the appendix (pp 1–4).

During dose escalation in AG221­C­001, eligible patients were enrolled into sequential cohorts of increasing doses of enasidenib. Each dose cohort enrolled a minimum of
three patients, up to a maximum of five to account for patients in screening at the time the third patient initiated treatment. The expansion portion enrolled patients into four non­randomised arms of a minumum of 25 patients per arm, as detailed previously.10,11 Three groups comprised patients with acute myeloid leukaemia; patients with myelodysplastic syndrome were assigned to the remaining arm for all other IDH2­mutated advanced haematological malignancies. Neither participants nor clinicians were masked to dose allocation throughout the study.

Procedures

Patients with myelodysplastic syndrome enrolled in the dose­escalation phase received oral doses of enasidenib at 60–300 mg per day, via a standard 3+3 design. Patients receiving a dose of enasidenib that was found to be safe (dose­limiting toxicities [DLTs] in <2 of 6 [or 0 of 3] patients) could be escalated to a higher safe dose pending review with a designated site medical monitor, with no limit to the number of dose increases.
Patients with myelodysplastic syndrome in the phase 1 expansion portion of the study received oral enasidenib at 100 mg once a day. During dose escalation and expansion, enasidenib was administered in repeated 28­day treatment cycles. A clinical study team (composed of a sponsor designee, study medical monitor, and treating investi­ gators) monitored and defined all adverse events according to the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE]), version 4.03, including DLTs and serious adverse events. DLTs were evaluated during the first cycle of treatment at each dose and defined as all clinically significant toxicities of CTCAE grade 3 or higher, with dosing discontinued permanently for any DLT unless the patient showed improvement with study drug as judged by local investigators and approved by the medical monitor. Adverse events were monitored throughout all treatment cycles. Dose reductions or interruption of dosing in the event of any­grade adverse event possibly related to study drug were approved by the medical monitor. If the time to recover from toxicity (ie, a return to at least baseline state) was longer than one treatment cycle, the risks and benefits of a patient’s continuation was assessed by the medical monitor. The enasidenib dose could be reduced in multiples of 50 mg; any patient unable to tolerate 50 mg was discontinued from study treatment.

Patients were followed up until withdrawal of consent; unacceptable toxicity; the appearance of any medical condition that put patients at risk and led to discontinuation of enasidenib at the decision of the investigators; disease progression (unless benefits were also evident and continuation was approved by the medical monitor); confirmed pregnancy; development of an intercurrent medical condition that precluded participation; removal from the trial at the investigator’s discretion for any reason in the best interests of the patient; use of a prohibited concomitant therapy; protocol violation; loss to follow­up; or death. The end of the study was defined as the time at which all patients had discontinued treatment and had been followed up for survival for at least 12 months, or had died, been lost to follow­up, or withdrawn consent before 12 months.

During dose­escalation and expansion, serial blood and urine sampling was done locally on treatment days 15, 29, and 57, and every 56 days thereafter to measure concentration­time profiles and pharmacokinetic para­ meters of enasidenib and its metabolite AGI16903. At these intervals we also sampled blood, bone marrow, and urine to measure R­2­hydroxyglutarate concentration and characterise the pharmacodynamic effects of enasidenib. Plasma R­2­hydroxyglutarate concentrations at study entry and during treatment were evaluated for patients with baseline and post­baseline assessments by liquid chromatography–mass spectrometry, as previously described.13

Peripheral blood and bone marrow aspirates were collected at screening for local IDH2 mutation analysis by PCR and next­generation sequencing according to institutional protocols, and for central next­generation sequencing with the FoundationOne Heme panel (Foundation Medicine, Cambridge, MA, USA) at Celgene (Summit, NJ, USA) to identify co­mutations in purified mononuclear cells. The final prospective study evaluation occurred on Oct 1, 2018.

Haematological response to study drug was assessed from peripheral blood and bone marrow aspirate or biopsy samples and objectively reported by site investigators. Response parameters were defined according to modified International Working Group 2006 criteria for myelo­ dysplastic syndromes.14 Overall response included complete remission, partial remission, marrow complete remission, and haematological improvement of the erythroid, platelet, and neutrophil lineages. Patients needed 5% blasts or more in the bone marrow at baseline to be eligible to attain complete remission, partial remission, or marrow complete remission. Haematological improvement was assessed by retrospective review of laboratory parameters. Overall survival was defined as the time from first enasidenib dose to death by any cause. Event­free survival was defined as the time from the first enasidenib dose to the date of relapse, progression, or death, whichever occured first.

Following the initial screening for IDH2 mutations, the frequencies of variant mutant alleles, detected with the FoundationOne Heme panel, were recorded. An exploratory endpoint of sequencing was the presence of co­mutations in other genes at screening in patients with an IDH2 mutation, in patients included in the co­ mutation analyses with available samples, to assess associations between baseline co­mutations and response.
Safety was assessed by the reporting of treatment­ emergent adverse events, classified according to the Medical Dictionary for Regulatory Activities and graded according to the CTCAE v4·03. Treatment­emergent adverse events were defined by site investigators as any adverse events that began or worsened at or after the start of enasidenib treatment, until 28 days (inclusive) after the last dose. Treatment­emergent events considered by the investigators as at least possibly drug­related were reported as such. Serious adverse events were recorded as those that resulted in death; patient hospitalisation or prolonged hospitalisation; persistent or substantial incapacity to perform life functions; congenital anomaly or birth defects; or were life­threatening.

Outcomes

The primary endpoints from the phase 1 portion of the AG221­C­001 trial were the safety and maximum tolerated dose of enasidenib, reported previously for the overall study population.10,11 Secondary endpoints included DLTs, pharmacokinetics, the pharmacokinetic­ pharmacodynamic relationship between enasidenib and R­2­hydroxyglutarate, and clinical activity. Safety and pharmacokinetic data have been reported for the overall study population of the AG221­C­001 trial,10,11 and overall response, duration of response, overall survival, and event­free survival have been reported for patients with relapsed or refractory acute myeloid leukaemia11 and patients with newly diagnosed acute myeloid leukaemia.15 The present report focuses on safety and activity outcomes in participants with myelodysplastic syndromes. We present data on safety (treatment­ emergent adverse events), overall response, duration of response, and overall and event­free survival in these patients.

Additionally, we assessed transfusion independence post­hoc among patients who were transfusion­ dependent at baseline, defined as having received at least one platelet or red blood cell (RBC) transfusion within the 28 days before and 28 days after the first dose of enasidenib. Transfusion independence was defined as having no RBC or platelet transfusions for at least 56 consecutive days while on the study drug.

Statistical analysis

Sample size and power calculations were reported previously for the AG221­C­001 trial.10,11 Demographic and disease characteristics at entry and response outcomes are reported descriptively. The primary, secondary, and exploratory analyses were done in the intention­to­treat population, except for the co­mutation analyses, done in patients with available data at baseline. We estimated overall survival and event­free survival with Kaplan­Meier methods, with patients without relapse, progression, or death censored at the last response assessment date. Analysis of event­free survival included patients who discontinued enasidenib treatment to receive bone­ marrow transplantation. p values were generated with Student’s t­tests or Fisher’s exact tests, as appropriate. All participants who received at least one dose of study drug were assessed for safety. Clinical endpoints were assessed with SAS (version 9.2 or higher), and translational analyses (ie, R­2­hydroxyglutarate concentration, co­ mutations, and mutant IDH2 variant allele frequency) were done with GraphPad Prism (version 7·0). All participants who received at least one dose of study drug were included in the safety analysis of treatment­emergent adverse events. An independent data monitoring committee oversaw the study. The AG221­C­001 trial is registered with ClinicalTrials.gov, NCT01915498.

Role of the funding source

The funders were involved in study design, data collection, and data analysis. The funders had no role in data interpretation or writing of the report. The lead authors (EMS, ATF, and CDD) prepared the initial draft with assistance from a medical communications agency, funded by Celgene. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Results

17 patients with myelodysplastic syndromes involving a mutation in IDH2 (in Arg140 or Arg 172 at local screening) were enrolled in phase 1 of the AG221­C­001 trial between Feb 18, 2014, and Sept 1, 2015, from seven sites in the USA and one site in France (appendix p 13). Ten patients were enrolled in the dose­ escalation portion and seven in the expansion portion (figure 1). Among patients in the dose­escalation group, three received enasidenib at 100 mg once a day (other daily doses were 60 mg [one patient], 75 mg [one patient], 150 mg [one patient], 200 mg [three patients], and 300 mg [one patient]). No DLTs were reported in any of the 17 patients with myelodysplastic syndromes.

Patients received a median of 3 treatment cycles (IQR 2–15), with five patients receiving 12 cycles or more. At the final prospective study evaluation, all patients had discontinued enasidenib treatment (two patients were still being followed up for survival). The most common reason for treatment discontinuation was disease progression, reported in five [29%] patients (figure 2). After discontinuing enasidenib, five patients received one or more subsequent drug therapies for myelo­ dysplastic syndrome; treatments received were cytarabine (three patients), the investigational drug AG­881 (two patients), decitabine (one patient), azacitidine (one patient), and quizartinib (one patient). Information regarding responses to subsequent therapies was not available.

At study entry, the median age was 67·0 years (IQR 60·5–73·0; table 1). The median time from diagnosis to trial enrolment was 14·1 months (IQR 5·4–25·3). At study entry, nine patients (53%) had intermediate­2 or high­risk prognoses according to the IPSS.16 Similarly, nine patients (53%) had high or very­high risk prognoses per IPSS­R12 classification.

13 patients (76%) had previously received hypo­ methylating agent therapy for myelodysplastic syn­ dromes, including ten patients who had received two or more prior regimens (table 1). All 13 patients had received at least two cycles of hypomethylating agent therapy, which might have been used in combination with other drugs. 11 (85%) of the 13 patients were refractory to hypomethylating agents, and two (15%) had relapsed (appendix p 5). In all 13 patients, other previous drug treatments for myelodysplastic syndromes were lenalidomide (two patients), sorafenib (two patients), vosaroxin (two patients), pracinostat (two patients), clofarabine and low­dose cytarabine, ruxolitinib, and rigosertib (one patient for each). Three patients (18%) treated with hypomethylating agents had also undergone an allogeneic stem­cell transplant.

On next­generation sequencing, hotspot mutations affecting IDH2-Arg140 or IDH2-Arg172 were detected in ten (77%) of 13 patients with adequate samples available for central review. Two of the three patients without hotspot mutations had IDH2 variants of unknown significance (IDH2 p.Val1471 and IDH2 p.Pro23Arg); the third patient had an IDH2­Arg140Gln mutation, detected by local testing and not the FoundationOne test. The median variant allele frequency for IDH2­Arg140 and IDH2­Arg172 in the ten patients was 39·5% (range 11·0–47·0). The most common co­mutations (in addition to IDH2) were in ASXL transcriptional regulator 1 (ASXL1; nine patients), serine and arginine rich splicing factor 2 (SRSF2; six patients), and stromal antigen 2 (STAG2; five patients; appendix p 7). The median number of co­ mutations was 3·0 (IQR 2·0–4·5) per patient.

An overall response was observed in nine patients (53% [95% CI 28–77]; table 2). Of ten patients who began the study with at least 5% bone marrow blasts, five (50%) had a marrow response, including one patient who attained partial remission in the marrow and four who attained marrow complete remission. No patient attained complete remission. Of the seven patients who entered the study with less than 5% bone marrow blasts and were only eligible to attain haematological improvement as a best response, four (57%) patients achieved haemato­ logical improvement per International Working Group criteria for myelodysplastic syndromes. The median time to onset of a first response occurred within the first treatment cycle (18·0 days; IQR 13·5–30·0; table 2). For patients who entered the study with 5% blasts or more and achieved a marrow response, the median time to a first response was 28·0 days (IQR 18·0–59·5). The estimated median duration of any response was 9·2 months (95% CI 1·0–not reached).
Three patients (18%) proceeded to allogeneic stem­cell transplant, two of whom had attained marrow complete remission during enasidenib treatment. The other patient entered the study with less than 5% bone marrow blasts and received the transplant before assessment for a haematological response (included in response assessments as a non­responder).

Among the 13 patients in whom hypomethylating agents had been unsuccessful, six (46%) attained a response to enasidenib: two of seven patients with at least 5% bone marrow blasts at study entry attained marrow complete remission, and four patients with less than 5% blasts at study entry achieved a haematological response (three patients attained a trilineage haematological response and one patient attained responses in the erythroid and platelet lineages but was not eligible to show a neutrophil response according to the required baseline count (<1·0 × 10⁹/L;14 appendix p 5).

Mean haemoglobin and platelet concentrations increased over the course of treatment (appendix p 9). Five (42%) of 12 patients who were RBC transfusion­ dependent at baseline became RBC transfusion­ independent; all five patients were also refractory or had relapsed to previous treatment with hypomethylating agents. Of seven patients who were platelet transfusion­ dependent at baseline, one (14%) achieved transfusion independence; this patient had also received prior therapy with hypomethylating agents. Mean white blood cell count increased during initial treatment, then decreased to lower than the baseline level with continued treatment (appendix p 9).

No difference was found in the median concentration of R­2­hydroxyglutarate at baseline between responding and non­responding patients (appendix p 10). On­target activity was evidenced by reductions in R­2­hydroxy­ glutarate during treatment in all evaluable patients (appendix p 11). Additionally, no difference was evident in the magnitude of R­2­hydroxyglutarate reduction between responding and non­responding patients (appendix p 10). Baseline variant allele frequencies for mutated IDH2 was also similar between responders (median 21% [IQR unavailable]) and non­responders (39% [unavailable]; p=0·41).

Among the 13 patients with samples screened by next­generation sequencing, eight (62%) achieved a haematological response. Patients who responded showed a numerically, but not statistically significant, lower number of co­mutations at baseline compared with patients who did not respond (appendix p 12).

Three patients entered the study after receiving an allogeneic stem­cell transplant in addition to prior hypomethylating agents. One patient who began the study with less than 5% bone marrow blasts attained haematological responses in the erythroid and platelet lineages, and became transfusion­independent before study discontinuation. Another patient showed increased bone marrow blasts, to 49% by day one of cycle three, and the patient discontinued enasidenib due to disease progression. The other patient attained haematological improvement in all three myeloid lineages and became transfusion­independent before discontinuing the study due to disease progression after 18 treatment cycles.

Seven patients (41%) progressed to acute myeloid leukaemia, four while receiving enasidenib (at 1 month, 2 months, 1·8 years, and 2·3 years after the first dose), and the other three during post­treatment survival follow­ up (at 1·5 years and 2·5 years after the first dose of enasidenib for two patients who discontinued enasidenib due to disease progression; and at 2·8 years after the first dose for the third patient who discontinued enasidenib to proceed to a haematopoietic stem­cell transplant for myelodysplastic syndrome, and later developed acute myeloid leukaemia). Compared with patients who did not progress to acute myeloid leukaemia, the patients who developed acute myeloid leukaemia had lower absolute neutrophil count (median 0·6 × 10⁹/L vs 1·3 × 10⁹/L) and platelet count (median 42 × 10⁹/L vs 76 × 10⁹/L ) at baseline, and were more likely to have received prior treatment with hypomethylating agents (all seven patients [100%] vs six [60%] of ten patients).

With a median follow­up of 11·0 months (IQR 6·8–23·0), we measured a median overall survival for all patients of 16·9 months (95% CI 1·5–32·3; figure 3). For the
13 patients who had received prior treatment with hypomethylating agents, median overall survival was also 16·9 months (1·6–32·1). Median event­free survival among all patients was 11·0 months (1·5–16·7; data not shown).

The most common treatment­emergent adverse events (any grade or cause) were diarrhoea and nausea (in nine [53%] patients each; appendix p 6). The most common treatment­related adverse events of any grade to emerge during treatment were diarrhoea, nausea, and increased blood bilirubin (in six [35%] patients each; table 3). Treatment­related adverse events led to dose interruptions in five (29%) patients. No treatment­related adverse events led to discontinuation of enasidenib.

The most frequent treatment­emergent grade 3–4 adverse events were increased blood bilirubin (in six [35%] patients), pneumonia (in five [29%] patients), and thrombocytopaenia (in four [24%] patients; appendix p 6). While receiving enasidenib, seven (41%) patients had treatment­related adverse events of grade 3–4 severity (table 3). Increased blood bilirubin (in four [24%] patients) and tumour lysis syndrome (in two [12%] patients) were the only treatment­related events of grade 3–4 severity to be reported in more than one patient, and no treatment­ related cytopaenias or infections of grade 3–4 were reported. Serious treatment­related events were reported in five (29%) patients. These were tumour lysis syndrome (in two [12%] patients), and asthenia, increased blood bilirubin, and increased aminotransferase (in one [6%] patient each). 48 serious treatment­emergent adverse events were reported in 13 patients. Serious treatment­ emergent events in more than one patient were pneumonia (in five [29%] patients); tumour lysis syndrome (in three [18%] patients), and sepsis, atrial flutter, increased blood bilirubin, cerebral haemorrhage, and mental status change (in two [12%] patients each).

11 patients died during the study; reasons for death were disease progression (in four [24%] patients), and sudden death, death (not otherwise specified), sepsis, failure to thrive, cerebral haemorrhage, intracranial haemorrhage, and respiratory failure (in one [6%] patient each]. Five (24%) patients died within 60 days of the first dose of enasidenib. Three of these patients had discontinued enasidenib and had received no subsequent therapies. In the two patients still receiving enasidenib, causes of death were sudden death and cerebral haemorrhage. None of the deaths were considered to be related to enasidenib treatment.

During this first in­human study, signs and symptoms of an isocitrate dehydrogenase (IDH) differentiation syndrome emerged. Throughout the trial period, a differentiation syndrome review committee comprising four treating investigators (EMS, CDDiN, ATF, and SdB) retrospectively reviewed and adjudicated events potentially associated with IDH differentiation syndrome in the study. The committee judged that three (18%) patients with myelodysplastic syndromes showed treatment­ emergent signs and symptoms (dyspnea, unexplained fever, pulmonary infiltrates, and hypoxia) consistent with IDH differentiation syndrome. Signs and symptoms of IDH differentiation syndrome began on treatment day 7 for one patient and on treatment day 23 in the other two patients. Durations of the syndrome were 3 and 6 days (later showing patients) and 41 days (early showing patient). Corticosteroids were used to manage the differentiation syndrome events. The two later showing patients recovered from the events without sequelae; the third event was ongoing when the patient discontinued from the study on day 41 of treatment. The early showing patient had leucocytosis at the same time as IDH differentiation syndrome, and received treatment with hydroxyurea in addition to corticosteroids.

Discussion

To our knowledge, this study is the first to report outcomes of enasidenib monotherapy in patients with myelodysplastic syndromes involving a mutation in IDH2. As a group, our patients had poor prognostic features at baseline: half (53%) had high­risk or very high­ risk disease according to IPSS­R, and three­quarters (76%) were relapsed or refractory to hypomethylating agent therapy and other treatments for myelodysplastic syndromes. Nevertheless, daily oral enasidenib mono­ therapy was well tolerated and induced haematological responses and transfusion independence in approximately half of treated patients.

A sizeable proportion of patients with myelodysplastic syndromes respond to treatment with hypomethylating agents, but response durations are brief (~6–8 months).17,18 Of the 17 patients in this analysis, at study entry, 11 were refractory and two had relapsed after treatment with hypomethylating agents. Patients with myelodysplastic syndromes for whom hypomethylating agents have been unsuccessful have a particularly poor prognosis, and no standard of care treatment is available after treatment failure.19 In a study including 435 patients with higher risk myelodysplastic syndromes (IPSS16 intermediate­2 and high­risk classifications; median age 69 years), median overall survival was 5·6 months from the time of treatment failure with azacytidine.20 Patients who received non­intensive salvage treatment after azacitidine (n=32) had a response rate of 0% (0 of 18 evaluable patients) and a median overall survival of 7·3 months, and patients who received intensive salvage treatment had a response rate of 14% (3 of 22 total patients) and a median overall survival of 8·9 months.20 In a study of 67 patients with higher risk (IPSS) myelodysplastic syndromes for whom treatment with decitabine had failed, median overall survival after decitabine was only 4·3 months.21

Outcomes are also poor in patients with lower risk myelodysplastic syndromes (IPSS low and intermediate­1 classifications) after treatment failure with hypo­ methylating agents; in one study including 27 patients with lower risk myelodysplastic syndromes who had primary or secondary resistance to azacitidine and who subsequently received active salvage treatment, median overall survival was 12·8 months.22 In our study, nearly half (6 of 13 patients [46%]) of patients with mutant IDH2 myelodysplastic syndromes who had previously received hypomethylating agents achieved a morpho­ logical response or showed haematological improvement with enasidenib monotherapy, including two patients who had undergone allogeneic stem­cell transplant. Median overall survival in patients who had received prior therapy with hypomethylating agents was 16·9 months.

The onset of a marrow morphological response in patients with at least 5% blasts in the bone marrow at baseline was faster (median 18·0 days) in our population than that reported in patients with IDH2­mutated, relapsed or refractory acute myeloid leukaemia (of around 2 months)10, possibly because the patients with myelodysplastic syndromes had fewer bone marrow blasts to differentiate to achieve a morphological response. Responses were also durable, with a median duration of any response of 9·2 months (95% CI 1·0–not reached).

Similar to findings reported in patients with relapsed or refractory acute myeloid leukaemia who received enasidenib,11,13 reductions in R­2­hydroxyglutarate during treatment were observed in both responding and non­ responding patients with myelodysplastic syndromes, and responding patients had nominally fewer co­ mutations than non­responding patients. Although enasidenib suppresses R­2­hydroxyglutarate production associated with IDH2 mutations, in heterogenous diseases such as myelodysplastic syndromes and acute myeloid leukaemia, additional cellular contexts (eg, genetic, epigenetic, or cytogenetic abnormalities) might cooperate to promote disease development and prevent haematological responses even with suppression of R­2­hydroxyglutarate.

IDH2 mutations have shown variable prognostic significance in myelodysplastic syndromes.5,23 A study of 1042 patients with myelodysplastic syndromes showed no influence on overall survival in patients harbouring an IDH2 mutation (4·1% of all patients), compared with the overall survival of patients without an IDH2 mutation.5 By contrast, a 2018 multivariate analysis of samples from 426 patients with myelodysplastic syndromes showed IDH2 mutations, as well as mutations in the Cbl proto­ oncogene (CBL), ASXL1, DNMT3A, and tumor protein p53 (TP53) genes, were independently associated with reduced survival and increased risk of leukaemic transformation.23 The authors of the analysis proposed adding these mutations as prognostic factors to the IPSS­R, because within each IPSS­R risk subgroup (ie, very low, low, intermediate, high, and very high), patients who harboured one or more of the mutations had poorer overall survival than patients without the mutations within the same risk group, and similar overall survival to those in the next overall higher risk subgroup.23

The prognostic effect of IDH2 mutations might depend on the co­mutation milieu. As shown in our cohort of patients with myelodysplastic syndromes, and in patients with acute myeloid leukaemia,24 mutant IDH2 is frequently associated with co­mutations in the ASXL1 and SRSF2 genes, which are generally associated with poor prognosis.23,24 A multivariate analysis of outcomes in 193 patients with myelodysplastic syndromes that controlled for karyotype, transfusion dependence, and mutant IDH1 status suggested that the presence of an ASXL1 mutation was independently associated with poorer survival (hazard ratio vs absence of ASXL1 mutation, 1·85 [95% CI 1·03–3·34], p=0·04) and faster time to acute myeloid leukaemia progression (2·39 [1·12–5·09], p=0·02).25 Similarly, in an analysis of 233 patients with myelodysplastic syndromes, mutations in SRSF2 were significantly associated with the presence of IDH2 mutations and reduced overall survival in patients with lower risk (IPSS) myelodysplastic syndromes (median 32·0 months vs 69·3 months in patients without SRSF2 mutation; p<0·01), and appeared to be implicated in disease progression.24 The small number of patients with mutation data in the current analysis prevents inferences regarding relationships between specific co­ mutations and response.

Enasidenib was well tolerated in our patient population. No treatment­related grade 3–4 cytopaenias or infections were reported during therapy. Non­cytotoxic therapies, including enasidenib, might be particularly important in older patients, many of whom might be affected by comorbidities and effects of previous cytotoxic treatments. In our patients, the frequency of IDH differentiation syndrome (18%) was similar to that reported in older patients (≥60 years) with newly diagnosed acute myeloid leukaemia receiving enasidenib,15 and that in patients with relapsed or refractory acute myeloid leukaemia.26 The frequency we observed was also lower than case numbers reported in patients receiving all­trans retinoic acid for acute promyelocytic leukaemia.27 IDH differentiation syndrome is a potentially lethal, but recognisable, event that can occur in patients with mutant IDH myeloid malignancies treated with enasidenib26 or ivosidenib,28 possibly owing to the drug mechanisms of action.7 Recognition of the signs and symptoms of IDH differentiation syndrome via the framework of the Montesinos criteria for differentiation syndrome associated with all­trans retinoic acid or arsenic trioxide for the treatment of acute promyelocytic leukemia29 might lead to early diagnosis and treatment. Onset of IDH differentiation syndrome can occur as early as a week to as late as months after initiation of enasidenib or ivosidenib, and is sometimes accompanied by leucocytosis.26 The signs and symptoms of IDH differentiation syndrome, which include dyspnoea, unexplained fever, pulmonary infiltrates, oedema, weight gain, and hypoxia, can mimic those of leukaemic progression or other acute comorbidities common in patients with marrow malignancies.26 IDH differentiation syndrome should be suspected when other potential causes can be reasonably excluded. IDH differentiation syndrome can be effectively managed with prompt administration of systemic corticosteroids, and cytoreduction per local practice is recommended if IDH differentiation syndrome is accompanied by leucocytosis.26 Limitations of our report include the small number of patients with myelodysplastic syndromes enrolled the AG221­C­001 trial, limiting the generalisability of results. We also did not record changes in the variant allele frequency of mutant IDH2 during treatment. Additionally, follow­up was restricted in patients who proceeded to transplant. A larger phase 2 study in patients with mutant IDH2 myelodysplastic syndromes is currently investi­ gating enasidenib as a monotherapy (in patients who have previously received hypomethylating agents) and in combination with azacitidine (in patients who have not had previous therapy with hypomethylating agents; NCT03383575). This trial is expected to elucidate potential synergistic effects of combining these two regimens in augmenting differentiation and hypomethylating activity. In the present study, enasidenib was well tolerated and induced responses in over half of patients with myelodysplastic syndromes harbouring a mutation in IDH2, of whom three­quarters had previously received treatment with hypomethylating agents. Assessment of IDH2 mutations, along with other gene mutations for which targeted therapies are available (eg, IDH1 and fms related receptor tyrosine kinase 3), could identify patients who might benefit from targeted therapy, AG-221 including those patients for whom conventional treatments have been unsuccessful.