Advancing drug development in myelodysplastic syndromes

Myelodysplastic syndromes/neoplasms (MDSs) are a group of heterogeneous stem cell malignancies characterized by ineffective hematopoiesis, progressive peripheral blood cytopenia, bone marrow dysplasia, germ line and/or somatic pathogenic mutations, and an increased risk of transformation to acute myeloid leukemia (AML). Currently, there is no curative therapy for MDS apart from allogeneic hematopoietic stem cell transplantation. The prognosis for MDSs remains poor, with few advances in drug development over the last 2 decades. Notable exceptions include luspatercept for selected patients with lower-risk MDSs,¹ ivosidenib for relapsed or refractory MDS with IDH1 (isocitrate dehydrogenase-1) mutation,² decitabine/cedazuridine, an oral formulation of decitabine³ and imetelstat for low- and intermediate-1 risk MDSs with transfusion-dependent anemia.⁴ There is a pressing need for developing novel and more effective therapeutics for the treatment of MDS. Among major challenges to expedite clinical translation of the current knowledge on MDS biology is improving the design of clinical trials.⁵ Key issues in clinical trial design for MDS include criteria for risk stratification and eligibility, response definitions, time-to-event end points, transfusion end points, functional assessments, and biomarker development. These topics have recently become a major concern and focus of academic experts and stakeholders including government, patient advocates, and the pharmaceutical industry.⁶ Here, we present specific actions to address these challenges and optimize future clinical trial design for patients with MDS (Table 1).

Diagnostic classification systems for MDS have recently been updated from the fourth edition of the World Health Organization (WHO) classification⁷ to the 5th edition (WHO5),⁸ and a new International Consensus Classification (ICC) has been proposed.⁹ These classifications incorporate molecular and genetic data into disease subgroups with diagnostic, prognostic, and treatment-related relevance. Major differences between ICC and WHO5 include the recognition of additional subgroups, including MDS/AML overlap syndrome with 10% to 19% blasts (ICC), variant allele frequency cutoffs for TP53 and SF3B1 pathogenic mutations (none in WHO5 vs 10% in ICC), and blast cutoffs for the diagnosis of AML with specific genetic abnormalities (none in WHO5 vs 10% in ICC).^8,9 In case of discrepancy between the ICC and WHO5 classifications, the Society of Hematopathology recommends including diagnoses according to both systems in the pathology report. This may allow patients with MDS/AML overlap to be eligible for either MDS or AML trials. However, when designing clinical trials, sponsors should choose either the WHO5 or ICC classification to clearly define the population being studied, after carefully considering the specifics of their studied drug, the available data in the patient population, and the clinical context. Regardless, sponsors should gather sufficient information for retrospective classification using either system.

Prognostic classifications, such as the revised International Prognostic Scoring System (IPSS; IPSS-R), demonstrated survival differences based on the presence of excess blasts and specific cytogenetic abnormalities.¹⁰ However, given that ∼94% of patients with MDS have at least 1 genetic abnormality, the molecular IPSS (IPSS-M) added molecular considerations into the IPSS-R, thereby enhancing risk assessment.¹¹ IPSS-M refines patient stratification, resulting in the restratification of 46% of patients.¹¹ Over half of the patients classified as intermediate risk by the IPSS-R were reclassified, with approximately one-fifth being upstaged to the IPSS-M very high-risk category.¹¹ IPSS-M high-risk categories are enriched in secondary or therapy-related MDS. Thus, therapy-relatedness should be deprioritized for trial eligibility, because molecular features may more accurately capture the elevated risk.

IPSS-M may offer superior prognostication compared with IPSS-R.^12,13 Although several retrospective validation studies have been conducted,^12,14 prospective validation, especially in pretreated and transplant populations, is warranted. Standardizing definitions for higher-risk MDS (HR-MDS) for clinical trial enrollment, using either IPSS-R >3.5 or IPSS-M moderate high, high, and very high, to clearly define the population being studied could enhance consistency and reliability across trials. In addition, collecting sufficient information to allow for retrospective prognostication using either IPSS-R or IPSS-M would be beneficial.

Improving clinical trial accrual requires adopting clinically relevant eligibility criteria¹⁵ and eliminating overly strict criteria related to organ function and performance scores.¹⁶ Patients with clonal cytopenia of undetermined significance and specific molecular abnormalities exhibit outcomes similar to patients with MDS without excess blasts.^17,18 Therefore, evaluating risk assessment tools, such as the Clonal Hematopoiesis Risk Score is advisable.¹⁹ Including patients with clonal cytopenia of undetermined significance in exploratory cohorts within lower-risk MDS (LR-MDS) trials may be warranted, especially for those who are symptomatic, transfusion dependent, previously treated with cytotoxic therapies, or have higher clonal hematopoiesis risk scores.^19,20 

1. Sufficient information should be collected to allow for retrospective classification using either ICC or WHO5, but 1 classification should be selected to clearly define the study population.

2. Sufficient information should be collected to allow for retrospective prognostication using either IPSS-R or IPSS-M, but 1 of them should be used to risk stratify the target population.

3. Minimizing exclusions and barriers to patient accrual in clinical trials is critical.

Response criteria have also been updated in recent years, raising questions about their implementation in clinical trials. The 2018 and 2023 International Working Group (IWG) response criteria represent significant improvements over the 2006 IWG response criteria for MDS.^21-23 The 2023 criteria for HR-MDS addressed shortcomings of the 2006 response definitions by introducing provisional “less-than–complete remission (CR)” responses and reducing the emphasis on marrow CR in the absence of hematologic improvement (HI).²³ For CR and Partial Remission (PR), the 2023 criteria lowered the required hemoglobin level from 11 to 10 g/dL.²³ Provisional “less-than-CR” responses include CR with partial hematologic recovery (CR_h), akin to the criteria used for the approval of nonmyelosuppressive targeted therapies in AML.²⁴ In 1 retrospective analysis, CR_h was associated with improved survival in patients with MDS compared with lesser responses.²⁵ Recognizing the impact of HI on clinical outcomes as an adjunct to marrow CR,²⁶ the 2023 criteria introduced CR with limited count recovery (which includes unilineage and bilineage) as a replacement for marrow CR.²³ CR_h improves upon the previously accepted “marrow CR without HI,” which has been a less reliable indicator of benefit.²⁶ Incorporating the 2023 response criteria as end points in clinical trials for HR-MDS may better capture clinically meaningful responses than the 2006 criteria and facilitate prospective validation, particularly for CR_h and CR with limited count recovery. Collection of end points per both response criteria would help with prospective validation.

In HR-MDS, overall survival (OS) remains the gold standard, and clinical responses predictive of improved survival should be considered meaningful.^27,28 Durable CR + PR has been viewed as a direct measure of clinical benefit, reflecting the restoration of trilineage hematopoiesis, and supported the regular approval of decitabine for MDS. In the phase 3 trial of decitabine vs supportive care, although durable CR + PR was achieved, the median time to AML progression or death was not significantly delayed.²⁹ In contrast, the AZA-001 trial comparing azacitidine with conventional regimens demonstrated benefits in both CR + PR and OS.³⁰ CR has been suggested to correlate with survival in responder analyses, as seen in the S1117 trial of azacitidine with or without lenalidomide or vorinostat, and in an MDS Clinical Research Consortium analysis of Hypomethylating agents (HMA)-treated patients.^27,31 However, promising CR rates observed in early-phase trials do not necessarily translate into CR or OS benefits in phase 3 trials, as evidenced by the results of the pevonedistat,³² eprenetapopt,³³ and magrolimab³⁴ trials. Earlier randomization in phase 2 trials, particularly for combination therapies, stand to provide more interpretable preliminary data on CR rates and OS.

Achievement of HI and complete cytogenetic remissions may also be beneficial in HR-MDS; however, prospective validation in randomized clinical trials is necessary.^35,36 The IWG 2023 criteria recommend the collection of data regarding measurable residual disease (MRD). However, due to variability in individual mutation clearance, context (eg, transplantation or continued therapy), and lack of testing standardization (see “MRD and biomarker development” below), MRD-negative response was introduced only as a provisional response category, pending prospective validation.²³ 

In LR-MDS, the 2018 criteria revised the hematologic erythroid response criteria to better capture clinically meaningful improvements in anemia. These revisions included lowering the hemoglobin threshold for treatment initiation from <11 to <10 g/dL, extending the screening period for transfusion burden and baseline hemoglobin evaluation from 8 to 16 weeks, extending the response evaluation period from 8 to 16 to 24 weeks, and incorporating subgroups based on transfusion burden.²² Implementing the updated 2018 HI-E criteria in clinical trials will allow for prospective validation.

1. Collection of responses per IWG 2023 response criteria, in addition to earlier response criteria, will facilitate prospective validation.

2. In HR-MDS, CR + PR with durability remains a clinically meaningful end point that should be supported by improvements in OS.

3. In LR-MDS, data on HI-E per IWG 2018 criteria will allow for prospective validation.

OS is considered the gold standard time-to-event measure of clinical benefit in randomized clinical trials for patients with MDS.^21,37 However, to expedite drug development, it is advantageous to identify alternative, clinically appropriate surrogate end points to evaluate drug efficacy sooner than OS. Event-free survival (EFS), leukemia-free survival, relapse-free survival, and progression-free survival (PFS) have all been used as end points in MDS clinical trials. However, their definitions lack standardization and may not always correlate with clinically meaningful outcomes. It is critical that the specific definitions of events remain consistent between trials, although the application of these end points may vary depending on specific MDS risk populations.

For HR-MDS, standardized definitions for EFS and PFS including specific event criteria were proposed in the IWG 2023 criteria.²³ However, these definitions still require prospective validation. Early results from a trial-level meta-analysis demonstrated a moderate association between EFS and OS on 9 randomized, controlled trials submitted to the US Food and Drug Administration (FDA) between 2000 and 2024.³⁸ EFS was defined as the time from randomization to the earlier occurrence of transformation to AML (given the availability of this information across all trials) or death from any cause. Confidence interval for R² was wide, and the analysis was limited by the small number of trials; thus, EFS as defined requires further study.

For LR-MDS, transfusion-free survival is a reasonable time-to-event end point to evaluate.^39,40 Red blood cell (RBC) transfusion independence (RBC-TI) has been considered a direct measure of clinical benefit and has supported drug approvals.^21,37 However, defining a durable RBC-TI period that translates to OS and quality of life (QoL) benefits is essential. Although a fixed RBC-TI period (eg, 8 or 16 weeks) may be reasonable for assessing the early activity of novel agents, more extended responses evaluated with a time-to-event measure should be investigated in advanced studies, along with their correlation with OS.

For select patients with LR-MDS with higher-risk disease features, including severe cytopenia(s) or a higher risk of progression to HR-MDS or AML, EFS may be a reasonable time-to-event end point. EFS could include death, progression to HR-MDS/AML, or the need for next-line treatment based on well-defined criteria, but this requires prospective validation.⁴¹ Notably, OS is a critical efficacy and safety end point to include in all randomized studies of MDS therapies to ensure that new treatments do not result in harm to patients.

1. In HR-MDS, OS remains the gold standard time-to-event end point. EFS and PFS may be considered but require prospective validation.

2. In LR-MDS, transfusion-free survival is a potential end point that requires a standard definition and prospective validation.

3. OS is an important efficacy and safety end point to include in all MDS trials.

Most patients with LR-MDS are anemic.⁴² As MDS progresses, anemia can worsen leading to RBC transfusion dependency. A low transfusion burden is defined as requiring 3 to 7 units of RBCs within 16 weeks (with a maximum of 3 in 8 weeks), whereas a high transfusion burden entails ≥8 units in 16 weeks (or ≥4 in 8 weeks).²² Anemia in MDS and RBC transfusion dependency are associated with worse QoL, increased morbidity, and decreased life expectancy as the RBC transfusion burden increases. Accordingly, transfusion burden reflects disease severity.^43,44 However, achievement of RBC-TI has not been predictive of improved OS in clinical trials, as evidenced by data from the MEDALIST and IMerge studies.^45,46 Regardless, minimizing transfusions mitigates risks such as infection, alloimmunization, and iron overload. Drugs that eliminate the need for RBC transfusions have the potential to improve QoL and reduce morbidity, making RBC-TI a meaningful early clinical end point for trials.^5,47,48 

To ensure consistency across MDS trials, the group recommended establishing a standard hemoglobin transfusion threshold of 7.5 to 8 g/dL.⁴⁹ Notably, although this threshold is generally considered “safe,” physicians should not dogmatically adhere to it. Some patients may prefer transfusions at higher thresholds to optimize symptoms and functionality.⁵⁰ It is important to note that lower transfusion thresholds commonly used for hospitalized patients (eg, <7 g/dL) are considered too austere for older, active outpatients with MDS. Transfusion density should be assessed over a 16- to 24-week period before study enrollment, when practical, categorizing patients as having either a low or high transfusion burden. The duration of RBC-TI should be assessed for periods ≥16 weeks.²² 

Of note, attribution of transfusions to intercurrent illnesses (eg, gastrointestinal bleeding) or drug-induced thrombocytopenia may prove problematic and lead to biases in assessments of RBC-TI. Therefore, the primary analysis of efficacy should be based on raw transfusion data in the intention-to-treat population, followed by sensitivity analyses considering the causality of intermittent transfusion needs. A “time without transfusion reliance” analysis⁵¹ may also capture the cumulative benefit of an intervention. Of note, hemoglobin improvement alone does not suffice as an end point in MDS trials. End points that directly measure clinical benefits in patients with MDS and disease impact on overall health should be used.

1. Standard hemoglobin transfusion threshold of 7.5 to 8 g/dL for most patients should be considered.

2. Transfusion density assessed over 16 weeks before enrollment defines low or high transfusion burden.

3. TI of ≥16 weeks should be considered clinically meaningful.

4. Primary analysis should be based on all transfusions in intention-to-treat population, followed by sensitivity analyses considering causality of transfusions.

5. A "time without transfusion reliance" type of analysis should be considered.

MDS is associated with impaired functioning and debilitating symptoms in patients compared with age-matched controls.⁵² Of particular relevance, many patients with cancer prioritize improvement in disease symptoms and functioning over merely extending their life span.⁵³ Unfortunately, health-related QoL (HRQoL) is typically neglected in MDS clinical trials, often relegated to a secondary or exploratory outcome. Prospective, systematic, and comprehensive measurement of patient-reported outcomes (PROs) in MDS clinical trials is recommended, including prespecified PROs. International guidelines and FDA guidance provide direction on the analysis of PRO data in randomized controlled trials.^54,55 Recommendations include establishing a clear PRO research hypothesis, using appropriate statistical methods for data analysis, implementing methodologies for handling missing data, and using standardized terminology. Sample sizes should be statistically modeled with HRQoL end points in mind, even as secondary end points.⁵⁶ Analyses should include assessments of minimally clinically important differences, and in cases in which these are not known, efforts should be made to investigate and define them. An international multidisciplinary working group has provided guidelines for using item libraries in PRO measurements in oncology, enabling standardized assessments of key HRQoL concepts relevant to the population.⁵⁷ The initial HRQoL assessment should occur before treatment, and the subsequent timing of assessments should consider specific treatments and confounding variables such as blood transfusion and treatment schedules.

Notably, patient-related factors (PRFs) such as age, comorbidities, disability, and frailty⁵⁶ significantly affect the QoL of patients with MDS. Comorbidities are prevalent in patients with MDS^58-60 and can affect PRO assessments,⁵² their ability to tolerate MDS therapies, and OS.⁶¹ Disability refers to impairments in activities of daily living (ADL) and instrumental ADL. In patients with HR-MDS treated with azacitidine, those experiencing any instrumental ADL impairment often exhibit shorter survival, receive fewer treatment cycles, and are more prone to complications that may lead to premature treatment discontinuation.⁶² Frailty, a multidimensional syndrome characterized by diminished physiological reserves, predisposes patients to disability, hospitalization, and death. It can be assessed using screening tools such as the Rockwood frailty scale and the MDS-specific frailty index, which have been shown to enhance prognostic accuracy by ∼35% in survival models.^63-66 Thus, information on PRFs should be collected in MDS clinical trials. PRFs can serve as stratification factors to ensure balanced randomization or be used to identify patients suitable for investigational treatments. Categorically defined PRFs could guide clinical trial design, potentially allowing for lower dose approaches with dose escalations as tolerated in specific patient groups. This strategy aims to maximize the benefits of novel therapies while minimizing potential risks, based on varying degrees of frailty and comorbidity.

Additionally, patient-reported symptoms (eg, dyspnea and fatigue) and functional impairments (eg, social, physical, and role functioning) can capture the impact of disease and therapy-specific effects. These dimensions are captured by domains included in instruments such as the Quality of life questionnaire core 30 items (QLQ-C30),⁶⁷ Quality of life in Myelodysplasia Scale (QUALMS),⁶⁸ Functional assessment of cancer therapy: Fatigue (FACT-F),⁶⁹ and 36-item Short Form Survey (SF36).⁷⁰ Symptom scales such as MD Anderson Symptom Inventory (MDASI)-AML/MDS⁷¹ or Patient-Reported Outcome (PRO)-Common Terminalogy Criteria for Adverse Events (CTCAE)⁷² may not always correlate with HRQoL scales during therapeutic interventions.

There are many tools available for incorporating into clinical trials to assess QoL,^67-69,73 comorbidity,^58,74 vulnerability/frailty,^75-77 and disability.^78,79 For example, tools to assess fatigue could be the Functional assessment of chronic illness therapy (FACIT)-fatigue subscale of Functional assessment of cancer therapy-anemia (FACT-AN) or fatigue measured by European Organisation for Research and Treatment of Cancer (EORTC)-QLQ-C30. Tools to assess physical functioning include the Quality of life in Myelodysplasia Scale-Physical functioning subscale (QUALMS P) subscale or EORTC-QLQ-C30 function score. Global Quality of life (QoL) can be measured by the EuroQoL (which provides health utilities for different states) and QLQ-C30. Comorbidity can be measured using either the Charlson comorbidity index or the MDS comorbidity index. Disability, vulnerability, and frailty tools include the Lawton instrumental activities of daily living scale (Lawton-Brody SIADL), 13-item Vulnerable elders survey (VES-13) and Rockwood Frailty Scale, or MDS-15 respectively. These measures are examples of commonly used PRO measures and not specific endorsements. The choice of tools may vary depending on the drug and patient population, but integrating PROs and PRFs in MDS clinical trials is essential for tailoring treatment approaches to individual needs.

1. A prespecified, comprehensive assessment of PRO end points in MDS clinical trials should be included.

2. Data on comorbidities, symptoms, disability, and physical functioning should be collected using well-defined and reliable tools.

3. Using PRFs in clinical trial design should be considered.

There is growing focus on developing assays to assess MRD in MDS; however, their use remains controversial due to uncertainties with available assays. Flow cytometry is a quick and accessible method but requires a detectable aberrant immunophenotype on malignant cells to assess MRD and can be user dependent. Next-generation sequencing (NGS) can monitor specific genetic mutations over time and is more objective but comes with technical challenges and higher costs. The lack of effectiveness of NGS in measuring residual MDS could be attributed to the limited sensitivity of available assays and the grouping of different disease subtypes.⁸⁰ Droplet digital polymerase chain reaction (ddPCR) is a highly sensitive and affordable method for MRD assessment that has shown promise in prognostication for posttransplant patients with AML and MDS.⁸¹ A recent effort combined NGS to identify patient-specific mutations with ddPCR to trace these mutations after allogeneic hematopoietic stem cell transplantation. MRD assessment using the combination of NGS and ddPCR proved feasible and useful in predicting early relapse and OS in 1 study.⁸² Ultimately, further exploration of both NGS and flow cytometry assays in MDS clinical trials is encouraged, and DNA samples should be stored for future analyses.

Beyond the choice of assay, key questions remain regarding the optimal MRD thresholds and time points for assessment. It is also important to specify whether blood or bone marrow is preferable for MRD analysis, considering accessibility and informativeness. Furthermore, for MRD-negative response to be interpretable, it should be tied to disease response criteria (ie, CR per IWG 2023) and should be defined by clear thresholds (ie, <1 × 10^–x).⁸³ Standardizing assay development and defining the optimal sensitivity and specificity of analyte measurements are necessary before MRD can be effectively used in MDS management or as a primary end point on clinical trials. Thus, MRD should be added as an exploratory end point in MDS clinical trials for now.

Further, despite our increasing knowledge of disease pathobiology, there is a paucity of validated biomarkers for MDS.⁸⁴ An important step forward was the recent approval of ivosidenib for patients with relapsed or refractory MDS with an IDH1 mutation, as detected by an FDA-approved test.² Other biomarkers could also play a crucial role in disease diagnosis, prognostication, and predicting or monitoring therapeutic response. Biomarkers can guide treatment decisions, assess drug efficacy, and elucidate the mechanisms of action of therapies.⁸⁵ Documentation of quantitative MRD results over the course of treatment with a targeted therapy could help characterize treatment response over time.

Because MDS is driven in part by inflammation, there is potential to exploit this biological aberrancy to define inflammatory markers. These may include plasma inflammatory cytokine levels, activation of intracellular signaling pathways such as NF-κB, or the presence of inflammatory gene signatures such as the recently published iScore.⁸⁶ Additionally, the extensive hypermethylation of gene promoters and the presence of epigenetic mutations in MDS suggest the potential utility of methylation arrays to identify patterns predictive of treatment response.⁸⁷ Similarly, mutations in RNA splicing genes, which are highly prevalent in MDS, may confer drug vulnerabilities and could be investigated as treatment indicators. Thus, unbiased drug screens or composite artificial intelligence algorithms identifying convergent pathway susceptibilities could be evaluated. It is advisable that samples (both cells and plasma) be immediately analyzed or banked in all clinical trials to allow for retrospective analysis in an effort to define clinically valuable biomarkers based on the specific therapeutic or clinical context.

1. Both NGS and flow cytometry should be evaluated for MRD detection, with DNA samples stored for future analysis.

2. Clear thresholds for MRD response are needed to interpret MRD data and ensure consistency across trials.

3. Assay development should be standardized, with optimal sensitivity and specificity of analyte measurements defined.

4. Documentation of MRD results quantitatively over the treatment course, particularly in patients with HR-MDS, should be considered.

5. Biomarkers specific to different therapeutic or clinical contexts should be collected.

The need for new therapies in MDS cannot be overstated. In contrast to the major expansion in knowledge of MDS biology, advances in drug development over the past decade have been limited, with little to no impact on patient survival. The key considerations summarized in Table 1 provide a pointed road map for improvements to MDS clinical trials, aiming to optimize clinical trial design and expedite new drug development. Trials should leverage the growing understanding of MDS biology to achieve improvements in patients’ outcomes. These considerations, through collaborative efforts of all stakeholders, should lead to greater consistency in clinical trial design and jumpstart progress in MDS therapy.

The authors thank their colleagues who contributed to thoughtful discussions and provided the key recommendations provided here. These include the following: for risk stratification and eligibility, Peter Greenberg (Stanford University), Robert Hasserjian (Massachusetts General Hospital, Harvard Medical School), Aziz Nazha (Incyte), Elli Papaemmanuil (Memorial Sloan Kettering Cancer Center), and Anupam Verma (Inova Schar Cancer Institute); for response criteria, Uma Borate (The Ohio State University), Noa Holtzman (National Cancer Institute [NCI]), Richard Little (NCI), Bart Scott (Fred Hutchinson Cancer Center, University of Washington), and David Steensma (formerly Novartis); for time-to-event end points, Maria Diez-Campelo (Universidad de Salamanca), Liz Garrett-Mayer (American Society of Clinical Oncology), Steven Gore (NCI), Emma Groarke (National Heart, Lung, and Blood Institute), Valeria Santini (University of Florence), Jonathon Vallejo (US Food and Drug Administration [FDA]), and Erica Warlick (Syros Pharmaceuticals); for transfusion end points, Yasmin Abaza (Northwestern University), Rafael Bejar (Aptose Biosciences), Uwe Platzbecker (University Hospital in Leipzig, Germany), and David Swoboda (Cancer Center of South Florida, Tampa General Hospital); for functional assessments and clinical trial integration, Andrew Artz (City of Hope Cancer Center), Victoria Barghout (Viver Health), Vishal Bhatnagar (FDA), Tracey Iraca (MDS Foundation), Tito Mendoza (NCI), Shannon McCurdy (University of Pennsylvania), Lisa Pleyer (Paracelsus Medical University, Austria), and Fabio Efficace (Italian Group for Adult Hematologic Diseases [GIMEMA]); for biomarker and measurable residual disease assays and development, Omar Abdel-Wahab (Memorial Sloan Kettering Cancer Center), Eric J. Duncavage (Washington University), Pamela Ebrahimi (FDA), Maria “Ken” Figueroa (Sylvester Cancer Center, University of Miami), Sylvie D. Freeman (University of Birmingham), Jarek Maciejewski (Cleveland Clinic), Thomas Prebet (Bristol Myers Squibb), Eduard Schulz (NCI), Dan Starczynowski (Cincinnati Children's Hospital); and for expert administrative support, Laura Wisch (FDA) and Amy Swanson (NCI). The visual abstract was created with BioRender.com

This work was supported, in part, by the Immune Deficiency Cellular Therapy Program and the Myeloid Malignancies Program within the Intramural Research Program of National Institutes of Health.

This publication reflects the views of the authors and should not be construed to represent FDA’s views or policies.

Contribution: All authors were active contributors and were involved in both writing and editing and provided the final approval of this manuscript.

Conflict-of-interest disclosure: A.M.Z. participated in advisory boards of, consulted for, and/or received honoraria from AbbVie, Agios, Pfizer, Astellas, Astex, ALX Oncology, Amgen, Akeso Pharma, BeiGene, Bristol Myers Squibb (BMS)/Celgene, Boehringer-Ingelheim, BioCryst, Chiesi, Daiichi Sankyo, Epizyme, Faron, Geron, Gilead, Genentech, Glycomimetics, Hikma, Ionis, Janssen, Kura, Keros, Karyopharm, Kyowa Kirin, Lava Therapeutics, Mendus, Notable, Novartis, Otsuka, Orum, Taiho, Takeda, Treadwell, Syndax, Sumitomo, STCube, Schrodinger, Servier, Syros, Vincerx, Regeneron, Rigel, and Zentalis. J.S.G. participated in advisory boards of, consulted for, and/or received honoraria from AbbVie, BMS, Genentech, Sanofi, and Servier; and has received institutional (trial) funds from AbbVie, AstraZeneca, Genentech, New Wave, and Pfizer. M.A.S. participated in advisory boards of BMS, Kurome, and Schröedinger. P.D.A. receives royalties from the National Institutes of Health Office of Technology Transfer for the invention of NUP98-HOXD13 mice with MDS. R.B. receives funding for her myelodysplastic syndrome/neoplasm registry and honoraria from BMS. A.V. has received research funding from Prelude, BMS, GlaxoSmithKline, Incyte, MedPacto, Curis, and Eli Lilly; is a scientific adviser for Stelexis, Novartis, Aurigene, Acceleron, and Celgene; receives honoraria from Stelexis and Janssen; and holds equity in Stelexis and Throws Exception. B.S. is on the advisory boards of BMS, Incyte, and Alexion; and receives research funding from BMS and Novartis. The remaining authors declare no competing financial interests.

Correspondence: Alain Mina, National Institutes of Health, 10 Center Dr, Building 10 Room 6N119A, Bethesda, MD 20892; email: alain.mina@nih.gov.