A multicenter phase 2 clinical trial of low-dose subcutaneous decitabine in myelofibrosis

Myelofibrosis (MF), including primary myelofibrosis (PMF) and MF after essential thrombocythemia or polycythemia vera, is a clonal hematopoietic stem cell neoplasm that commonly manifests with constitutional symptoms, bone marrow fibrosis, and extramedullary hematopoiesis. MF is characterized by canonical driver mutations in JAK2, CALR, or MPL that drive abnormal activation of the Janus kinase/signal transducer and activator of transcription signaling pathway and subsequent cytokine-driven remodeling and fibrosis of the bone marrow.^1,2 Several JAK inhibitors, including ruxolitinib, fedratinib, pacritinib, and, most recently, momelotinib, have been investigated and have received US Food and Drug Administration approval for the treatment of intermediate and high-risk MF based on the core benefits of ameliorating splenomegaly and constitutional symptoms.^3,4 The impact of JAK inhibitors on the underlying clone is modest, and there is little improvement in cytopenias or reduction in the risk of leukemic transformation.⁵ Given these limitations, there remains an urgent need for clinical and translational investigation of novel approaches that can modify the bone marrow failure state associated with MF.

In addition to aberrant Janus kinase/signal transducer and activator of transcription activation, it is likely that epigenetic modifications of genes involved in cell proliferation, differentiation, and trafficking may also contribute to the malignant phenotype of MF.⁶ For example, hypermethylation of the promoter region of the chemokine receptor type 4 gene (CXCR4) likely plays a role in the constitutive mobilization of CD34⁺ cells, which are characteristically increased in the peripheral blood in MF.^7-9 Higher levels of circulating CD34⁺ cell counts have been associated with more pronounced extramedullary hematopoiesis, higher-risk disease, and worse overall survival.⁸ Decitabine (5-aza-2'deoxycitidine) is a potent DNA methyltransferase (DNMT) inhibitor (DNMTi) that has clinical activity in myelodysplastic syndromes and acute myeloid leukemia (AML). In nonobese diabetic/severe combined immunodeficiency mice, sequential treatment of PMF CD34⁺ cells with decitabine and the histone deacetylase inhibitor, trichostatin A, was shown to upregulate CXCR4 expression and redirect CD34⁺ cell homing toward the bone marrow rather than the spleen.^10,11 This phenomenon was independent of JAK2 V617F mutation status and was not demonstrated when PMF CD34⁺ cells were treated with small-molecule inhibitors of JAK2.¹¹ 

Herein, we report the results of a phase 2 multicenter trial of subcutaneous decitabine in MF using a 10-day schedule based on the hypothesis that the use of subcutaneous low-dose decitabine in MF would lead to an improvement in peripheral blood counts and amelioration of the bone marrow failure state. This dosing schedule of subcutaneous low-dose decitabine was previously found to be safe in sickle cell hemoglobinopathies, and to be associated with increasing fetal hemoglobin (HbF) levels and reduction of vaso-occlusive episodes.^12,13 Alternative dosing schedules of decitabine with a focus on prolonged DNMT depletion and reduced cytotoxicity have also been reported to be safe and effective in the context of other myeloid malignancies.^14,15 

This was an open-label, multicenter, National Cancer Institute–sponsored phase 2 clinical trial conducted through the University of Chicago phase 2 consortium. Patients were eligible for this study if they were adults aged ≥18 years who met diagnostic criteria for MF as defined by the Italian Consensus Criteria and had palpable splenomegaly and/or Hb of <11 g/dL.¹⁶ No cytotoxic therapy (aside from hydroxyurea) or radiation therapy was permitted within 4 weeks of study entry, and no growth factor therapy was permitted within 2 weeks. Important exclusion criteria include prior treatment with decitabine, known central nervous system disease, and inadequate organ function. A full description of study eligibility criteria is in supplemental Appendix 1A. The primary objective of the trial was to assess the efficacy and safety profile of low-dose decitabine in MF. Primary end points include improvements in cytopenias, transfusion dependence, and splenomegaly. Correlative objectives include assessments of the pharmacodynamic effects of the drug on various markers of interest, including circulating CD34⁺ cells, CXCR4 transcript levels, and HbF levels. The protocol was reviewed and approved by each institution’s institutional review board, and all enrolled patients gave written informed consent.

Decitabine was administered subcutaneously at a dose of 0.3 mg/kg per day on days 1 through 5 and days 8 through 12, previously investigated in sickle cell disease, with cycles repeated every 42 days.¹² The use of filgrastim or granulocyte-macrophage colony-stimulating factor was only permitted in patients with febrile neutropenia. Treatment could continue for a maximum of 9 cycles in the absence of disease progression or unacceptable adverse events. Patients with a complete response (CR) had their therapy discontinued, provided at least 4 cycles of therapy had been administered. Patients who tolerated the therapy but had no response after 2 cycles were permitted to escalate the decitabine dose to 0.4 mg/kg per day. Dose reductions from toxicity were allowed at dose level −1 with 0.2 mg/kg per day and dose level −2 with 0.1 mg/kg per day.

The design and initiation of the study predated the development of standardized response criteria for MF. Response was therefore evaluated primarily as an improvement in blood counts and/or splenomegaly. A CR was defined as the normalization of blood counts (Hb >12 g/dL; platelets >100 x 10⁹/L; white blood cell count >4 x 10⁹/L; and absolute neutrophil count >1.5 x 10⁹/L), transfusion independence, resolution of splenomegaly, and bone marrow blasts of <5%. A partial response (PR) was defined as a Hb increase of at least 2 g/dL in a patient with baseline anemia (Hb <11 g/dL) or a >50% reduction in palpable splenomegaly in patients with baseline splenomegaly as determined by physical examination. A hematologic improvement with platelet response (HI-P) was defined as an increase in platelet count of at least 30x 10⁹/L in patients with baseline thrombocytopenia (platelets of <150 x 10⁹/L) or the achievement of transfusion independence in patients previously dependent on platelet transfusions.¹⁷ A hematologic improvement with erythroid response (HI-E) was defined as achievement of transfusion independence in patients who were previously dependent on red blood cell transfusions. The objective response rate was defined as the rate of CR, PR, HI-E, or HI-P. Finally, because of the update of the International Working Group (IWG) response criteria for MF after study enrollment, responses were also retrospectively recategorized by IWG guidelines.¹⁸ Toxicity was graded using National Cancer Institute Common Terminology Criteria for Adverse Events version 3.0.¹⁹ 

Peripheral blood mononuclear cells (PBMCs) were collected before decitabine administration on day 1 and on days 5 and 12 after therapy in the first 2 treatment cycles. Total RNA was isolated from PBMCs, and CXCR4 expression analysis was performed by quantitative real-time polymerase chain reaction (PCR). Circulating CD34⁺ progenitor cells in PBMC samples were quantified using flow immunostaining (CD34⁺ CD45⁺) and light scatter. The percentage of HbF as a proportion of total Hb was measured by high-performance liquid chromatography on peripheral blood samples. JAK2 V617F mutational status was ascertained using 2 separate methods: the amplification refractory mutational system technique with quantitative PCR, as well as real-time PCR.²⁰ Additional details on correlative analysis protocols are included in supplemental Appendix 1B.

A total of 20 patients were to be enrolled using a single-stage design to test the null hypothesis that the objective response rate is 10% against the alternative hypothesis that it is at least 32%. If ≥5 responders were observed among the 20 patients, the null hypothesis would be rejected. This design would have ∼80% power using a 1-sided test at the 0.05 significance level. The Clopper-Pearson exact method was used to calculate confidence intervals (CI) for the overall response rate. Univariate logistical regression was used to examine factors related to overall response, whereas Student t test was used to assess impact of drug exposure to response. The Kaplan-Meier method was used to estimate overall survival probability. A mixed effects linear model was used to examine whether there was a significant change in circulating CD34⁺ cells and CXCR4 expression over the time points analyzed (days 1, 5, and 12) within the first 2 cycles of therapy and whether there was a significant change in the kinetics of gene expression between cycles 1 and 2 of therapy (cycle 1 day 1 vs cycle 2 day 1). The logarithms of gene expression values were analyzed because the distribution was skewed. Similarly, change in HbF was analyzed using a mixed-effect model, after imputing undetectable values (<1.0%) to 0.5. Statistical analyses were conducted using Stat (StataCorp LLC, College Station, TX), R 4.1.2 (R Core Team, Vienna, Austria), and GraphPad Prism (GraphPad Software, San Diego, CA).

A total of 21 patients were enrolled between January 2005 and May 2008; 1 additional patient was allowed in the trial beyond the initial recruitment goal. Pretreatment patient characteristics are summarized in Table 1. Of 21 patients, 16 (76%) had PMF, 4 (19%) had MF after polycythemia vera, and 1 (5%) had MF after essential thrombocythemia. Seventeen patients (81%) had intermediate or high-risk disease according to the Lille scoring system, and 16 (76%) had either intermediate-2 or high-risk disease by the International Prognostic Scoring System model.^21,22 Two patients had accelerated phase (AP) MF, whereas an additional patient had MF in the blast phase (BP).²³ Transfusion-dependent anemia was present in 12 patients (57%) at the start of treatment. Eighteen patients (85%) had baseline splenomegaly on examination. A total of 14 patients had received some form of treatment for MF before starting decitabine, including hydroxyurea (n = 9), thalidomide/lenalidomide (n = 7), and imatinib (n = 2).

Overall, 7 patients had a response according to the protocol’s predefined response criteria (Figure 1; supplemental Table 2), resulting in an objective response rate of 33% (95% CI, 14.6-57.0). The median time to response was 2 months (range, 1-21), and the median duration of response was 7 months (range, 3-44). All 7 responders demonstrated an improvement in their preexisting cytopenias after treatment with decitabine. The majority of these responses were multilineage improvements in blood counts and included 1 patient who achieved a CR (supplemental Figure 3A), 2 patients who achieved a PR as defined by an increase in Hb of >2.0 g/dL (supplemental Figure 3B), and 4 patients who achieved either HI-E and/or HI-P. Responses were also retrospectively recategorized within the 2013 revised IWG response criteria, which had not yet been published at the onset of this trial. By the IWG criteria, of the 7 noted responders, 1 would have fulfilled IWG criteria for a PR, whereas 4 would have had an anemia response (supplemental Table 2). Of 4 patients with an anemia response, 2 would also have concurrently fulfilled the criteria for a spleen response. Using a univariate logistic regression model, patients who had a high-risk prognosis indicated by a Lille score of ≥2 were more likely to respond to decitabine (odds ratio, 9.17; 95% CI, 1.30-93.74; P = .038). Two patients tolerated the decitabine but failed to respond after 2 cycles, and thus underwent dose escalation per protocol but did not ultimately have a response. Furthermore, there was no association between drug exposure and response to therapy (mean 12-week drug exposure of 5.00 mg/kg in responders vs 5.01 mg/kg in nonresponders; P = .97).

Of the 2 patients who had AP-MF at the time of enrollment, both experienced significant reductions in peripheral blast counts. The first patient had a baseline absolute blast count of 2.7 x 10⁹/L, which decreased to 0.08 x 10⁹/L during cycle 1, maintaining a peripheral blast count of < 0.3 x 10⁹/L for the remainder of his 4 cycles of treatment (supplemental Figure 4). This patient also achieved a HI-P response by having a sustained recovery of platelet counts. The second patient had an initial improvement in peripheral blasts from 3.24 x 10⁹/L at baseline to complete clearance of peripheral blasts during cycle 1. The circulating blast count returned to 13.46 x 10⁹/L during the second cycle, 10 weeks after initiating therapy.

Of 18 patients with palpable splenomegaly at baseline, 2 patients had complete resolution of splenomegaly on physical examination from baseline spleen sizes palpated to 8 cm and 10 cm below the left costal margin; the duration of spleen response was 24 and 44 months, respectively. Ultrasonographic data of the largest spleen dimension were available for 16 patients, including for the 7 responders. A mixed-effects model showed a statistically significant reduction in spleen length from baseline (mean ± standard deviation, 21.2 ± 4.3 cm) to cycle 2 (20.1 ± 4.8) and cycle 4 (17.3 ± 4.3) (P = .029). Seven patients, including 4 responders, had sufficient ultrasonographic data to assess for spleen volume changes. These patients had an average reduction in spleen volume of 13% from baseline after cycle 2, and 28% after cycle 4 (P = .014).

The protocol was subsequently amended to allow for indefinite continuation of therapy in responding patients because of the excellent response experienced by a subset of patients. One patient who was transfusion dependent at baseline remained transfusion independent while on therapy for 23 cycles. Another patient who achieved their best Hb response (increase of 2 g/dL from baseline) after the sixth treatment cycle, aside from transient anemia due to the drug, maintained that response for 41 months; this patient also experienced resolution of palpable splenomegaly for 44 months. Of 20 patients who did not have baseline BP disease, 3 eventually transformed to AML while on study at 5, 12, and 32 months, respectively. Furthermore, 3 patients underwent allogeneic stem cell transplantation. The 2-year overall survival rate for the entire cohort was 38% (95% CI, 15-61).

A total of 101 cycles were administered to 21 patients. The median number of cycles administered was 4 (range, 1-23). Of 21 patients, 18 completed at least 2 cycles of therapy. Of the 3 patients who completed only 1 cycle of treatment, the reasons for early discontinuation included progression and neutropenic infection, whereas 1 patient did not have a cited reason for discontinuation. Drug-induced myelosuppression occurred in most patients: 71% of patients (n = 15) experienced grade 3/4 neutropenia, 76% (n = 16) experienced grade 3/4 anemia, and 62% (n = 13) experienced grade 3/4 thrombocytopenia (Figure 2A). Febrile neutropenia was noted in 48% of patients (n = 10). Prolonged marrow suppression necessitating a dose reduction occurred in 5 patients (24%), 3 for whom the dose was reduced to 0.2 mg/kg per day (dose level −1) and 2 for whom it was reduced to 0.1 mg/kg per day (dose level −2). Of these 5 patients, 3 ultimately had a protocol-defined response to therapy. The majority of myelosuppression events were otherwise transient and recovered by 6 weeks after the start of therapy. Apart from drug-induced myelosuppression, nonhematologic adverse events were relatively infrequent. Overall, the most common nonhematologic adverse events were fatigue (n = 13 [62%]), hypocalcemia (n = 8 [38%]), anorexia (n = 8 [38%]), elevated alkaline phosphatase (n = 7 [33%]), hyperglycemia (n = 7 [33%]), hypoalbuminemia (n = 7 [33%]), and infection (n = 7 [33%]) (Figure 2B). Grade 3 and 4 nonhematologic toxicities were uncommon and included infection (n = 4) and bleeding (n = 3). There was 1 death due to sepsis reported in a patient with baseline MF-BP.

Molecular testing of driver mutations such as JAK2, CALR, and MPL was not initially performed because the trial design preceded the first reports of these mutations in PMF. JAK2 mutational status was retrospectively ascertained in 20 of 21 patients. Of these, 11 were found to carry the JAK2 V617F mutation. Twelve patients had enough remaining sample to test for additional pathogenic mutations. Additional canonical driver mutations in those 12 patients included MPL (n = 2), and CALR (n = 3). Two patients had confirmed triple-negative MF. Pathogenic mutations in epigenetic modifiers, including DNMT3A, TET2, ASXL1, and EZH2 were detected in 8 of 12 patients (67%).

Nine patients (43%) had known cytogenetic abnormalities. Seven patients (33%) had known unfavorable karyotype and 11 patients (52%) had favorable karyotypes, including normal karyotypes in 9. Cytogenetic analysis was not performed in 3 patients. Of the 7 patients who responded, 4 had a favorable-risk karyotype at baseline, 2 had unfavorable risk, and 1 had an insufficient sample for cytogenetic analysis.

Overall, there was a statistically significant reduction in circulating CD34⁺ cell levels over time (P = .0001; Figure 3A). Over the first cycle, CD34⁺ cell levels decreased by 71% and 65% in responders and nonresponders, respectively. Furthermore, in responders, this reduction in circulating CD34⁺ cells was largely sustained from cycle 1 to cycle 2 (P < .001, comparing day 1 values between cycles), whereas in nonresponders, the change was not sustained into the second cycle (P = .45, comparing day 1 values between 2 cycles; Figure 3B-C).

Because the decline of CD34⁺ cells after exposure to chromatin-remodeling agents in vitro has been linked to an upregulation of CXCR4 levels, we measured CXCR4 transcript levels in vivo via real-time quantitative PCR. Overall, there was no significant change in CXCR4 levels over time (P = .29; Figure 3D). There was also no significant interaction between cycle and response (P = .17). There was no correlation between change in CD34⁺ cells and change in CXCR4 (P = .68).

There was no significant change in HbF levels over the first 2 cycles in either responders or nonresponders. However, responders had higher levels of HbF both at baseline and throughout the first 2 cycles compared with nonresponders (P < .01).

The arsenal of currently available treatments for MF is limited and primarily supportive. Our results add to the prior limited experience with decitabine in MF in the chronic phase and provide, to our knowledge, the first prospective evidence supporting the feasibility and efficacy of a low-dose, extended schedule of subcutaneous decitabine in this population.^24,25 This study demonstrated the clinical activity of decitabine with responses noted in 33% of patients, especially those with high-risk prognostic scores at baseline. Prior studies have suggested an association between spleen response and survival outcomes.^26,27 In our study, the effect on splenomegaly was modest, with 2 of 18 patients with baseline splenomegaly achieving a sustained ≥50% reduction in spleen size, thus highlighting a limitation of this strategy. However, the observed benefits of decitabine in improving anemia and transfusion dependence may nonetheless address a significant need in the field, one that is now exploring alternative surrogate end points beyond spleen response in clinical studies.^28,29 Furthermore, combinatorial approaches using a hypomethylating agent and novel JAK inhibitors are being evaluated (ClinicalTrials.gov identifier: NCT04282187). Novel JAK inhibitors including momelotinib and pacritinib have both demonstrated improvements in cytopenias, which may allow for synergistic activity with decitabine.^30,31 

DNMTi's such as decitabine and azacitidine have demonstrated some activity in MF in recent studies, particularly in AP and BP MF.^25,32-34 The combination of ruxolitinib and decitabine has been prospectively investigated in AP or BP myeloproliferative neoplasms (MPNs) with an overall response rate of 44%.³⁵ A retrospective analysis of DNMTi-venetoclax combinations in AP or BP MPNs has also been reported, with a response rate of 44%.³⁶ In contrast to the retrospective experience of decitabine monotherapy in AP/BP MF by Badar et al, our prospective trial accrued a larger proportion of patients with chronic phase MF, and we observed a sustained improvement in anemia in the chronic phase, which is a challenging aspect of this disease to manage.²⁵ 

Of note, we used a 10-day schedule of decitabine, which is longer than the 5-day schedule implemented in the recent MF studies.^25,32 Given the short half-life of decitabine, it is conceivable that an extended schedule of treatment may allow for greater time for the incorporation of the drug into DNA and modulation of gene expression. Investigations using a 10-day decitabine schedule in AML have led to conflicting results, and similar studies in MF are lacking.^37-39 Although this decitabine regimen was associated with manageable nonhematologic adverse events, the rates of drug-related myelosuppression and febrile neutropenia were high and remain a concern in this population.⁴⁰ An alternative approach may be to use metronomic rather than sequential schedules for decitabine, such as reduced but biologically active doses of 0.2 mg/kg administered weekly to promote DNMT depletion while avoiding the cytotoxicity that leads to myelosuppression.^15,41 

Predictive biomarkers for DNMTi therapy in MF are lacking. We demonstrate that response to decitabine in MF was associated with a sustained early decline of circulating CD34⁺ cells, validating our hypothesized effect of the agent in MF based on results of preclinical studies.^10,11 CXCR4 transcript levels did not change significantly in this trial, despite the decrease in circulating CD34⁺ cells, in contrast to prior observations from in vitro studies. Because the CXCR4 levels were isolated from PBMCs, this may not be reflective of CXCR4 levels in CD34⁺ cells. Furthermore, alternative mechanisms, such as direct cellular cytotoxicity independent of CXCR4 expression, could be responsible, wholly or in part, for the sustained decrease in CD34⁺ cells. We also demonstrate higher baseline levels of HbF as a potential predictive biomarker for DNMTi response in MF. Similar findings have been observed in myelodysplastic syndromes and AML but will require further validation in MF.^42,43 

Patient-reported outcomes are essential in the evaluation of disease response to investigational therapies, and have been shown to associate with overall survival in a number of cancers.^44-46 However, validated patient-centered instruments that assessed symptom burden were not widely available at the time of study design, and therefore represents a limitation of this study. It is essential for future prospective trials to incorporate validated instruments of patient-reported outcomes such as the MPN–symptom assessment form total symptom score to better assess patient perceptions of drug impact and tolerability, as well as to guide clinician decisions on therapy selection.⁴⁷ Another limitation of this study was its limited sample size and single-arm design. The small sample size and response rate limited the feasibility of multivariable analyses to determine predictors of response. Larger randomized controlled studies are needed in the future to evaluate the benefit of these therapies.

In summary, the results of this prospective study demonstrate the potential of the 10-day schedule of low-dose subcutaneous decitabine therapy to treat cytopenias associated with MF. This therapy is associated with low rates of nonhematologic toxicities, but myelosuppression was a common side effect and suggests the need for future investigation of alternative doses and schedules of administration. Such approaches are already being investigated in myelodysplastic syndromes.¹⁴ The availability of an oral decitabine-cedazuridine preparation may also facilitate future trials in this regard including combinations with other agents that target biologically relevant pathways and have the potential to ameliorate cytopenias in this disease.^48-52 

The authors acknowledge the support of the University of Chicago Phase 2 Consortium and National Cancer Institute grant #NO1-CM-62201. Additionally, the authors express their gratitude to the patients, family, caretakers, and staff who participated in this clinical trial.

Contribution: C.L., A.A.P., and O.O. were responsible for data analysis, interpretation, and manuscript writing; O.O. and W.S. were responsible for study conception and design; D.H. and T.K. performed statistical analysis; M.G., H.W., and D.S. were responsible for data and sample collection; D.S. was responsible for real-time polymerase chain reaction for CXCR4 transcript expression; K.v.B., J.G., J.L.W., R.K., M.R.B., R.A.L., W.S., and O.O. contributed patients to the study; and all authors contributed to editing of the manuscript and approved the final version.

Conflict-of-interest disclosure: A.A.P. reports honoraria from AbbVie and Bristol Myers Squibb, and research funding from Kronos Bio and Pfizer. O.O. is a consultant or advisor to AbbVie, Blueprint Medicines, Bristol Myers Squibb, Celgene, CTI, Incyte, Impact Biomedicines, Kymera, Novartis, Servier, Taiho Pharmaceutical, and Treadwell Therapeutics, and reports research funding, paid to their institution, from AbbVie, Agios, Aprea AB, Astex Pharmaceuticals, AstraZeneca, Bristol Myers Squibb, Celgene, CTI BioPharma Corp, Daiichi Sankyo, Incyte, Janssen Oncology, Kartos Therapeutics, Loxo, Novartis, NS Pharma, and OncoTherapy Science. C.L. is a consultant to Rigel and served on the advisory board of Autolus. The remaining authors declare no competing financial interests.

The current affiliation for C.L. is Division of Hematologic Malignancies and Cellular Therapy, Department of Medicine, Duke University School of Medicine, Durham, NC.

Correspondence: Olatoyosi Odenike, The University of Chicago, 5841 S Maryland Ave, MC 2115, Chicago, IL 60637; email: todenike@medicine.bsd.uchicago.edu.