Considerations for Drug Development in Myelodysplastic Syndromes

Despite increased understanding of the pathobiology and molecular underpinnings of myelodysplastic syndromes (MDS), therapeutic options remain limited. To date, only five drugs have been approved by the FDA for patients with MDS or subsets of MDS: azacitidine, lenalidomide, decitabine, luspatercept, and decitabine-cedazuridine tablets (1–5). Three of these are hypomethylating agents (HMA), which remain the standard of care for higher-risk patients with MDS (HR-MDS) and lower-risk patients with multiple cytopenias requiring therapy. Unfortunately, less than half of patients respond to HMAs, and even the best responses are transient and noncurative without hematopoietic cell transplantation (HCT; ref. 6). This standard of care has not changed for nearly two decades (Fig. 1), underscoring a need for new therapies.

Recognizing this need, in April 2022, the FDA brought together a panel of regulators and academic experts in MDS to discuss approaches to improve MDS drug development. At this meeting, the panel reviewed challenges in MDS clinical trial design and endpoints and outlined considerations for future trial design in MDS to facilitate drug development to meaningfully meet patient needs.

MDS has historically been a challenging disease for drug development, for a variety of reasons. MDS is characterized by significant heterogeneity in pathobiology due to its genetic landscape (with a spectrum of over 50 myeloid mutations identified in patients with MDS; ref. 7). Consequently, outcomes of affected patients, who tend to be older, have comorbidities, and face logistical challenges in trial participation, are also heterogeneous (8, 9).

Unlike many other cancers, presentations can vary from incidentally found low, but relatively stable, blood counts, to a sudden diagnosis requiring urgent intervention, similar to acute myeloid leukemia (AML). Reflecting this, survival ranges from a few months to a decade or more (10). In addition, preclinical testing of drugs and new agents in MDS has been challenging due to the historic lack of animal models that resemble human MDS, or that can faithfully recapitulate the same complex biology to a degree high enough to reliably predict success or failure of an agent in humans (11).

Even when translated to the clinical setting, trial designs may not be amenable to the patients needing the drugs. For example, the standard management of HR-MDS includes repeated monthly cycles of 5 to 7 consecutive days of injectable HMAs, which have been shown prospectively to improve outcomes and are recommended to be continued ad infinitum, as long as the patient is responding and not experiencing prohibitive toxicities. Therefore, most clinical trials of new drugs for patients with HR-MDS typically combine another agent to an HMA backbone. However, as the median age at diagnosis of MDS is typically the eighth decade of life, and patients often have poor functional status, multiple comorbidities, and/or limited social support, trial participation at tertiary centers requiring multiple visits monthly becomes prohibitive (12). From a practical standpoint, nonspecialty settings may also be challenged to manage the adverse events and supportive care needs of vulnerable patients treated aggressively.

Most clinical trials for MDS are conducted in large centers, while the majority of patients with MDS in the United States are treated in community settings. As MDS is rare (with an estimated yearly incidence rate of 4/100,000 people), clinical trials often underaccrue and take a long time to complete, discouraging industry sponsors from conducting trials in patients with MDS until agents have been explored in other cancers traditionally managed at larger centers, such as AML. Add to this complexity the post-HMA failure setting, where patients are typically too sick either to enroll in clinical trials or remain on treatment long enough to realize a potential benefit. Drug development in this setting is further challenged by a lack of understanding of resistance mechanisms that lead to HMA failure, precluding rational drug design targeting specific mechanisms.

The landscape of therapeutics in MDS remains unchanged, as underscored by the failure of multiple large, randomized trials conducted in HR-MDS because the FDA approval of azacitidine due to an inability to demonstrate an improvement in outcomes over azacitidine, or after HMA failure (13–16). Compare this therapeutic lag with other blood cancers: in AML, for example, nine drugs have been approved since 2017. Clearly, we need to rethink clinical trial design specifically for MDS.

Clinical trials for patients with MDS should be designed to reflect the biology of disease evolution, including molecular mutations and overlap with other disorders; the advanced age of patients with MDS and how patients are treated in real-world settings; and to maximize the likelihood of identifying truly active drugs (Table 1).

Increasingly, molecular characterization of MDS and incorporation of molecular mutations into clinicopathologic scoring systems has become standard, with increased accuracy of molecular systems in predicting leukemia transformation and overall survival (OS) and almost 50% of patients being restratified into risk categories that differ from those suggested by the International Prognostic Scoring System (IPSS) or Revised IPSS (IPSS-R) (17, 18). Thus, future MDS clinical trials ideally would include next-generation sequencing (NGS)results, both to better risk stratify patients and to help identify genetic or biologic subsets who could benefit from a therapeutic intervention—or subsets who would not benefit. It is also reasonable for trials to be designed independent of MDS risk stratification (with inclusion of patients with secondary MDS), provided disease severity can be assessed in the context of treatment efficacy and toxicity.

In designing trials for treatment-naïve MDS, one real-world consideration is that patients with MDS initially may be evaluated and started on treatment at local clinics, and then referred to specialty centers for a trial. For such trials, patients could still be considered to have “treatment-naïve” MDS if they have received, for example, a single cycle of HMA or erythropoiesis stimulating agent therapy previously. Once a patient is enrolled in a clinical trial, administration of standard-of-care therapies (including blood product transfusions or drugs such as HMAs considered standard) could be allowed at local treatment locations, as frequent study visits to a specialized center may be a barrier to trial participation. Further engagement of community sites should also be pursued to decentralize clinical trials and develop more practicable therapies for a broad population of patients with MDS across the country.

In addition, clinical trials for MDS should be designed to maximize the chance of identifying truly active drugs. The diagnosis of MDS can be challenging, as is reflected in large databases or in the rate of discrepancy between an MDS diagnosis and alternative diagnoses [such as AML, MDS/myeloproliferative neoplasms (MPN), or nonmalignant conditions] >25% of the time (19, 20). Thus, central review of the MDS diagnosis is needed to reduce these discrepancies on trials and should include risk stratification.

It has long been recognized that older adults who develop MDS may have comorbidities associated with anemia, which can complicate efficacy assessments of MDS drugs intended to correct anemia (21). Consequently, randomization in advanced phase trials is needed to adjust for this effect.

Multiple trials in higher-risk patients (including histone deacetylase inhibitors or immunomodulatory drugs combined with HMAs) have demonstrated that dose reductions, delays or premature cessation of HMA-based therapy may have compromised efficacy (13, 14, 22, 23). On the other hand, in the recent randomized trial of the NEDD8 activating enzyme pevonedistat combined with azacitidine versus azacitidine monotherapy, strict protocol language limiting dose adjustments led to a 32% complete response (CR) rate for the control arm (azacitidine monotherapy; ref. 13). Thus, protocols should include instructions to maintain dose, avoid cycle delays, and continue HMA-based therapy for a minimum of six treatment cycles (in the absence of clear progression or unacceptable toxicity) prior to determining failure of response.

For most patients with HR-MDS, the standard backbone of therapy includes an HMA, which as monotherapy results in a typical median response rate (complete + partial) of approximately 20%, and response duration and survival of approximately 18 months (24, 25). The paradigm of recent clinical therapeutic trial interventions has been to introduce combination therapies upfront, one being an HMA, to improve on these mediocre outcomes (i.e., add-on trials). However, a head-to-head comparison versus HMA alone could be considered for agents with compelling monotherapy data in the relapsed or refractory setting. As patients with HR-MDS have a median survival often measured in months, trials should ideally be powered to show differences in treatment arms based on an appropriate response rate, duration, quality of life, and survival in the comparator arm based upon data from control arms of recent clinical trials (12, 26).

MDS is typically divided into lower- and higher-risk (LR- and HR-MDS) categories commonly defined by the IPSS-R, which serves as a default staging system, but only for newly diagnosed patients (10). Consequently, trials are often constructed for patients with either lower- or higher-risk disease, with eligibility for a given trial being mutually exclusive for one or the other.

However, disease severity can evolve as patients acquire additional molecular mutations or progress through multiple lines of therapy (7, 27)—aspects of advanced MDS that may not be captured through existing prognostic scoring systems, and that make HCT or intensive therapies viable options (28, 29). Thus, clinical trials should avoid eligibility language referencing a patient's potential for HCT, as this is variably interpreted.

“Relapsed or refractory” MDS should be defined distinctly from treatment-naïve MDS, and consistently across trials (e.g., for patients with lower-risk disease, previous exposure for a defined time period to erythropoiesis stimulating agents, and/or to prior immunomodulatory or HMAs; for higher-risk disease, previous exposure to hypomethylating agents for a defined time period and/or prior HCT).

The very definition of MDS is evolving, as data suggest that patients with clonal cytopenias of undetermined significance (CCUS) with discrete molecular abnormalities have similar outcomes to patients with LR-MDS, and that patients with HR-MDS with excess blasts have similar outcomes to patients with oligoblastic AML (30, 31). This is now reflected in recent revisions in classification of myeloid neoplasms from the World Health Organization and the International Consensus Classification, which further blur the line between MDS and AML, and identify discrete categories of MDS based on molecular profile (31, 32). In contradistinction, patients with certain MDS/MPN, such as chronic myelomonocytic leukemia, may not have similar outcomes or drug responsiveness as pure MDS (13, 14). Thus, trials for patients with LR-MDS may consider allowing enrollment of patients with CCUS, and trials for patients with HR-MDS may consider allowing enrollment of patients with oligoblastic leukemia (particularly if targeting a common mutation; ref. 31), provided patients in either scenario are otherwise eligible, can be stratified, and treatment and endpoints can be harmonized.

In addition, patients in HR-MDS trials should reflect the real-world MDS population—the majority of whom are over age 60 years and with managed medical comorbidities. Patients should not be excluded from trial eligibility based on prior therapy for lower-risk disease that is distinct from the therapy being evaluated for HR-MDS. Patients potentially eligible for HCT should also be included as noted above.

Patients with LR-MDS remain a heterogenous group with a variable degree of presenting cytopenias and clinical outcomes. Morbidity and mortality in this group are mainly related to cytopenia complications or interplay with other comorbidities. In addition, approximately one-third of lower-risk patients progress to have HR-MDS or AML (33, 34). Thus, clinically meaningful endpoints would capture important treatment effects such as alleviating cytopenias associated with symptoms and morbidity, preventing disease progression, or improving OS. Moreover, such endpoints should be practical, with applicability both to pivotal clinical trials and to clinical practice to assess patient benefit objectively. As such, we propose that the response criteria for patients with LR-MDS should be based on the predominant cytopenia or clinical feature—similar to a “target lesion” in solid tumor trials.

In LR-MDS, the clinical manifestations of anemia dominate the initial presentation in 85% of patients and drive the need for therapeutic intervention (35). Regular red blood cell (RBC) transfusion requirements are associated with a worse prognosis, and hemoglobin (Hgb) levels below 9 g/dL in men and 8 g/dL in women are associated with reduced OS and a higher risk of nonleukemic and cardiac death (36). However, existing criteria defining achievement of transfusion independence or lack of transfusion dependence are problematic.

For example, despite the arbitrary threshold for transfusion of RBCs for a Hgb ≤ 9 g/dL specified in International Working Group (IWG) 2006, most clinicians do not give a transfusion until patients are symptomatic or for a Hgb < 7–8 g/dL, likely adopting thresholds from trials in patients with septic shock, surgical considerations, or inpatient therapy, and from concerns for iron overload, transfusion burden, and product availability (37–40). Fairness and use of limited resources are issues that have been highlighted during the COVID-19 pandemic where global blood shortages affected when transfusions could occur.

A proposal enumerating stricter criteria for defining hematologic improvement (HI) has provided initial guidance for improving definitions, and for making them more clinically meaningful (41). The IWG 2006 criteria require receipt of at least 4 units of packed red blood cells for a Hgb <9 g/dL within an 8-week period to define transfusion dependence. The 2018 modified criteria called for this period to be extended to 16 weeks and to stratify patients as not transfusion dependent, low transfusion dependent and high transfusion dependent—to better reflect the heterogeneity of anemia burden in patients with MDS, and to set the stage for defining achievement of transfusion independence that must last 16 weeks or longer.

For patients who are not transfusion dependent (NTD; defined as no RBC transfusion in the 16 weeks prior to starting treatment), initiation of therapy can be considered if patients are symptomatic and have a Hgb ≤ 9 g/dL. Following the IWG 2018 response criteria (41), an objective Hgb increase of at least 1.5 g/dL for 16 weeks or more may be considered clinically meaningful only if associated with objective improvement in quality of life using patient reported outcomes (PRO) tools such as the MDS-QUALMS (42) or significant prolongation of time-to-transfusion dependency/transfusion-free survival.

For patients with low transfusion burden (LTB; defined as 3–7 RBCs transfusions in 16 weeks with at least two separate transfusion episodes and a maximum of three in 8 weeks), a clinically meaningful response is defined as complete elimination of RBC transfusion dependence for 16 weeks or more. Assessment of PRO benefit may be more challenging in this group as RBC transfusions in a comparison group not receiving active therapy may alleviate many anemia-related symptoms, thus potentially leveling PRO assessments between study arms.

For patients with high transfusion burden (defined as ≥8 RBCs in 16 weeks, ≥4 in 8 weeks), complete elimination of RBC transfusions for 16 weeks or more is a meaningful endpoint. A ≥50% reduction in transfusion burden sustained for 16 weeks or more may be clinically meaningful if objectively associated with improvement in quality of life as measured by validated PRO tools.

In patients with isolated or predominating thrombocytopenia, platelet improvement should be associated with no bleeding events and/or platelet transfusions. The definition of a complete platelet response includes a platelet count of at least 100 × 10⁹/L without bleeding events.

For patients with isolated neutropenia, treatment should not be initiated merely based on neutrophil count. Clinically meaningful response could include a reduction in recurrent infections, an endpoint that would need to be validated in a randomized trial.

In LR-MDS trials, PROs also have the potential to identify therapies with biological activity that results in clinical benefit. How a patient feels (symptom relief) and functions (increased daily activity) are critical indicators of clinical benefit that are not required by established response criteria alongside achievement of HI. Yet, capturing symptom improvement from a patient's perspective was a key secondary endpoint that contributed to the approval of ruxolitinib in the COMFORT I–II trials in myelofibrosis, and this endpoint was also included in the prescribing information (43, 44). Employing a tool that accurately captures symptom improvement tracking with disease improvement, and that can distinguish improvement between treatment and control trial arms, would be useful.

Future clinical trials should take advantage of advances in technology for assessing PRO improvement using tools such as wearables and make efforts to assess objective data on the impact of disease through health care resource use endpoints. Furthermore, clinical trials using targeted therapies need to assess cytogenetic and molecular responses by measuring minimal residual disease (MRD) to develop this endpoint as a potential surrogate marker of clinical benefit.

Of note, HI and PRO endpoints have not been used to support efficacy for new MDS therapies to date. Thus, the proposed endpoints would need to be validated in the setting of prospective, randomized controlled trials along with appropriate stratification by baseline cytopenias and transfusion status, as applicable. Clinical trial designs would need to carefully define the intended patient population as well as stipulate an analysis plan that rigorously tests the selected efficacy endpoints. For HI endpoints targeting a single cell line, care would need to be taken to “do no harm,” ensuring that other cell lines are not negatively impacted by the therapy, particularly for patients who are NTD or LTB at baseline where the margin of benefit is smaller.

In addition, among patients risk stratified as LR-MDS, the CR definition applied in HR-MDS can be used similarly as a reasonable marker of benefit. This use of CR with full count recovery may be used as an efficacy endpoint to capture responses while preserving trilineage hematopoiesis.

The ultimate clinically meaningful endpoint—improvement in OS, is more challenging to achieve in LR-MDS, where no approved therapies have prospectively achieved this benefit. As progression to AML is less frequent and survival is longer compared with those with HR-MDS, OS may be a more relevant efficacy endpoint for those whose “low-risk” disease features an adverse risk mutation (biologically contradicting the low-risk designation), such as those in EZH2, RUNX1, TP53, and ASXL1 (45). Furthermore, OS may serve as a critical safety endpoint for randomized LR-MDS studies to ensure there is no increased mortality due to toxicities from the experimental agent.

A close look at the initial 2000 and subsequent 2006 IWG-modified response criteria for MDS highlights some challenges in drug development (46, 47). One overarching problem related to these response criteria lies in how cumbersome they are to apply or interpret compared with those for AML (48–50).

Meaningful clinical endpoints specific to HR-MDS should be identified starting with earlier-phase studies (such as randomized phase II trials). Primary endpoints should reflect true modification of the natural history of the disease, such as OS. One challenge with event-free survival (EFS), defined by the IWG 2006 as “failure or death from any cause,” is that it is seldom used due to its ambiguity. An “event” in HR-MDS could be defined to signify a demonstrable change in disease biology that is associated with worse outcomes. Future revisions to the IWG response criteria ideally would reflect meaningful clinical benefit to a patient with enhanced practicability, reduced ambiguity, and enriched interobserver consistency. While there is experience (13) using a more specific and alternative EFS definition of “death or progression to AML with >50% increase in blasts to >20%,” which conveys a metric of the lack or loss of benefit from an investigational intervention, use of EFS by any definition requires further evaluation and validation in prospective randomized, controlled trials.

Earlier clinical endpoints could also be considered. CR, as currently defined, lacks distinctness and has not regularly translated to an OS benefit from the phase II to the phase III setting (51). Study drugs added to an azacitidine backbone may be myelosuppressive, making therapy-related cytopenias in the setting of disease resolution difficult to discern from lack of drug benefit. In addition, certain requirements of an MDS CR are stricter than those for an AML CR: a Hgb ≥11 g/dL is required for an MDS CR, whereas no Hgb requirement is specified for AML CR; requiring CR confirmation is also unique to HR-MDS. Finally, the designation of a marrow CR may lead to misleading clinical conclusions. Without concomitant HI, achievement of marrow CR has similar clinical benefit to that of stable disease (52).

We suggest that studies prospectively collect data to evaluate alternative CR definitions using: (i) a decrease in the mandatory Hgb threshold for CR determination to 10 g/dL or consideration of RBC transfusion independence; (ii) CR lasting a clinically meaningful duration of time. For overall response rate (ORR) reporting, this would be limited to true CR patients, along with partial response (PR) and potentially HI (given its inconsistent translation to a meaningful benefit). Marrow CR particularly without HI can be avoided, as the reduction of marrow blasts to less than 5% has no true clinical correlation.

PROs to capture quality of life or symptom improvement are similarly critical for defining clinical benefit in HR-MDS. Real-life data consistently show how clinical outcomes fall short of the those achieved within the context of a clinical trial (25, 53). Patient-specific factors such as frailty and comorbidities have yet to be consistently included in response assessments or adequately captured on clinical trials. Investigational therapies that do not improve disease-related symptom burden may speak to both the drug's ability to truly modify the disease, and to the feasibility of using it in the real world (54).

The MDS field is sensitive to the necessity of interim markers of benefit and secondary endpoints to better reflect the spectrum of improvement. Incorporation of PROs specific to MDS (55) is key. The durability of time without transfusion and improved PRO are also secondary endpoints with clinical relevance, particularly in combination with CR and even PR to demonstrate altered natural history. An endpoint including patients who undergo HCT should also be evaluated. This would be annotated as alive and with disease control that allows the HCT procedure to occur. This endpoint has been evaluated in the AML setting (56) and is increasingly viewed as a measure of modified goals of care, as HCT potentially extends OS. Furthermore, outcomes in HR-MDS trials need not be censored at the time of HCT, as both the bridge to HCT as well as ability to survive past HCT are meaningful.

An area of burgeoning clinical and translational research in HR MDS includes robust specimen banking, requirement to document full molecular panels (NGS), and measurable residual disease assessments. These should be standard in trials to correlate with EFS, OS, and ORR, and to illustrate role the role of certain phenotypes and genotypes more likely to respond to a given treatment.

MDS is a complicated collection of heterogeneous disorders with highly variable presentations and outcomes, and that predominantly affects a vulnerable population of older adults. Trials should be designed from the outset to allow the practicable application of new therapies in this high-needs population, with drugs that can be administered and tolerated in community settings, and with endpoints that meaningfully improve patients’ lives over existing therapies, rather than simply providing supportive care. Trials should also capitalize on our expanding understanding of the biology of these disorders to effect substantive change in patient outcomes.

M.A. Sekeres reports personal fees from BMS, Novartis, and Kurome outside the submitted work. A.E. DeZern reports personal fees from Geron, Novartis, Sobi, and BMS outside the submitted work. A.M. Zeidan reports personal fees from Novartis, AbbVie, Genentech, BMS, Geron, Agios, Syros, Aprea, ASTEX, Kura, Taiho, Chiesi, Takeda, Regeneron, Gilead, Orum, Schrodinger, Syndax, and Ionis outside the submitted work. R.S. Komrokji reports grants and personal fees from BMS and personal fees from Novartis, JAZZ, AbbVie, CTI Biopharma, Geron, PharmaEssentia, Servier, Taiho, Takeda, Gilead, and Rigel outside the submitted work. No disclosures were reported by the other authors.

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