Disease Modification in Myelofibrosis: An Elusive Goal?

Introduction

Janus kinase (JAK) inhibitors represent the cornerstone of myelofibrosis (MF) management, consistent with the ubiquitous activation of JAK-signal transducer and activator of transcription signaling in the Philadelphia chromosome–negative myeloproliferative neoplasms.^1,2 Their broad immunosuppressive and anti-inflammatory actions underlie their efficacy in graft-versus-host disease,^3,4 and rheumatoid^5,6 and psoriatic arthritis.^7,8 However, notwithstanding their undeniable benefits in ameliorating splenomegaly and MF-related symptoms,^9,10 they are viewed as being predominantly anti-inflammatory, with limited disease-modifying effects. The latter remain poorly understood, however. Reductions in bone marrow fibrosis (BMF) and driver mutation (JAK2, MPL, and CALR) expressed allele burden, only modestly achieved by ruxolitinib,^11,12 have generally been considered to represent disease modification. However, BMF reduction should be accompanied by improvement in cytopenias, progression-free survival (PFS), or overall survival (OS) to be clinically meaningful, and MF is genomically a complex disease not addicted to a single molecular driver. Modulation of cytokines has been advanced as an example of disease modification, but clinical correlation is mostly limited to symptom improvement.¹³ No therapy has been demonstrated to thwart leukemic transformation. With a plethora of novel agents¹⁴ in development (Table 1), several new primary end points (beyond the usual ≥ 35% spleen volume reduction [SVR35] and ≥ 50% reduction in total symptom score [TSS50] at 24 weeks), including red blood cell transfusion independence (RBC-TI) and OS, are being adopted in pivotal trials. There is also much discussion of the disease-modifying potential of the newer agents, sometimes in combination with JAK inhibitors, on the basis of preclinical synergism.^15,16

TABLE 1.

Candidate Surrogate Markers for Disease Modification by JAK Inhibitors and Other Novel Therapies in Development for Myelofibrosis

Survival Outcomes in JAK Inhibitor–Naive and JAK Inhibitor–Exposed Patients

An exploratory analysis of pooled 5-year data from the randomized COMFORT trials (N = 528) demonstrated a survival benefit for ruxolitinib, despite crossover.¹⁰ An OS advantage for ruxolitinib was also documented in comparisons of trial patients with matched historical controls^17,18 and in real-world studies.^19,20 Considering the modest impacts of ruxolitinib on BMF and JAK2 V617F expressed allele burden,^11,12 survival prolongation by ruxolitinib may be indirect, ie, through improvements in metabolism, organ function, and overall well-being.⁹ Although spleen size may not be prognostic in MF, multiple studies have demonstrated that the achievement, durability, and degree of spleen response to ruxolitinib correlate with OS.^17,21,22 Spleen response, thus, likely serves as a biomarker of JAK inhibition in MF. However, it is not clear that early intervention with ruxolitinib in patients without symptomatic splenomegaly or significant MF-related symptoms improves survival; therefore, consensus guidelines recommend using ruxolitinib on the basis of the presence of symptoms and/or splenomegaly, rather than purely for extending survival.^23-25

Long-term follow-up data are lacking for the randomized JAKARTA trial of fedratinib²⁶ because of the temporary full clinical hold (all patients had to stop therapy) that was mandated by the US Food and Drug Administration due to concerns over Wernicke's encephalopathy. Nevertheless, despite short follow-up and censoring, median PFS was significantly better for fedratinib 400 mg/d versus placebo.²⁷ Median OS had not been reached in either group. Importantly, PFS is a relatively novel end point in MF trials, and data to establish the surrogacy of PFS for OS are currently lacking.

The randomized SIMPLIFY-1 and -2 trials evaluated the JAK1/2 and activin receptor type 1 (ACVR1) inhibitor momelotinib in the JAK inhibitor–naive and –pretreated settings, respectively.^28,29 In both trials, significantly greater proportions of patients in the momelotinib arms achieved RBC-TI at 24 weeks; interestingly, in SIMPLIFY-1, the achievement of 24-week RBC-TI on momelotinib, but not achievement of SVR35 or TSS50 at week 24, correlated significantly with OS.³⁰ This relationship trended toward statistical significance in SIMPLIFY-2. Requirement for RBC transfusions is a well-established adverse prognostic factor in MF; thus, the correlation of RBC-TI with OS is not unexpected and may not be unique to momelotinib, although the anemia benefit of this drug does appear to be unique among current JAK inhibitors.³¹

Prognosis after ruxolitinib discontinuation is poor (median OS, 11-14 months).^32-35 Ruxolitinib failure (variously defined, and may include patients who exhibit intolerance, suboptimal, lack, or loss of response) is an area of intense new drug development. In a phase II trial of imetelstat (9.4 mg/kg intravenously once every 3 weeks) in patients with MF who had failed JAK inhibitor therapy, the median OS was 29.9 months after a median follow-up of 27.4 months.³⁶ By contrast, the rate of SVR35 at 24 weeks was a modest 10.2%. The apparent OS advantage for imetelstat in the ruxolitinib failure setting was, however, validated in a comparison with a closely matched cohort of patients seen at a major academic center.³⁷ Imetelstat is now being studied in a registrational, phase III trial (IMpactMF, ClinicalTrials.gov identifier: NCT04576156) in the JAK inhibitor failure setting with OS as the primary end point, unprecedented in MF pivotal trials. Interestingly, however, in an Italian study (N = 218), the median OS of patients who remained in chronic phase at ruxolitinib discontinuation (n = 167) was 27.5 months, whereas that of patients whose disease transformed to blast phase was 3.9 months.³⁴ Furthermore, there were no survival differences between the chronic phase patients who discontinued ruxolitinib because of resistance versus intolerance. These observations underscore the likely heterogeneity of ruxolitinib failure populations in various reports. It is possible, for instance, that ruxolitinib is discontinued much later in the disease course at certain centers than at others, eg, those highly experienced in management of this drug. This could significantly affect survival after ruxolitinib discontinuation.

Candidate Biomarkers of Disease Modification: Problems and Pitfalls

Bone marrow fibrosis.

Long considered reactive, there is evidence that the fibrocytes that drive BMF in MF are clonal (neoplastic).³⁸ BMF improvement has been reported in a number of recent studies evaluating novel therapeutics (Table 1); however, the significance of this per se remains doubtful, unless a clear correlation with improvement in cytopenias (often confounded by myelosuppression from the drugs themselves), or prolongation of PFS or OS, can be shown. BMF improvement, although modest in comparison to spleen and symptom responses, has also been observed with JAK inhibitor therapy, particularly in the setting of prolonged treatment.^39,40 Of note, successful allogeneic hematopoietic cell transplantation, to date the only potential cure for MF, often results in complete regression of BMF, the speed of attainment of which may predict OS independently of prognosis at transplantation.⁴¹ Although current BMF grading systems mainly consider reticulin fibrosis, there is increasing recognition that assessment of collagen and the grade of osteosclerosis provides important additional information.⁴² With the exception of PRM-151,^43,44 most contemporary clinical trials reporting rates of BMF improvement have not separately reported changes in reticulin and collagen. In general, reproducibility of BMF assessment in expert hands is high (approximately 90%),⁴² emphasizing the importance of central expert pathology review of samples from clinical trial participants.

Driver mutation allele burden reduction.

Reduction in the variant allele frequencies (VAFs) of mutated JAK2, MPL, and CALR has intuitive appeal as a measure of disease modification, but MF is characterized by a high incidence of other, non–phenotype-driving mutations.^45,46 In a small study of pegylated interferon alfa-2a in patients with early-stage MF (N = 30), the presence of such mutations predicted poorer responses.⁴⁷ Conversely, JAK2 V617F, in particular, is one of the most common forms of clonal hematopoiesis of undetermined significance (CHIP).^48-50 Thus, the presence or persistence of the mutation at a low level on or after effective therapy may simply reflect background CHIP. Also, the attainment of complete molecular response is itself a moving target. The lower limit of detection of most next-generation sequencing–based mutation panels in use is approximately 5%. More sensitive assays, eg, droplet digital or allele-specific oligonucleotide polymerase chain reaction, may be able to detect much lower VAFs.⁵¹ Long-term follow-up of COMFORT-1 demonstrated greater JAK2 V617F VAF reductions in ruxolitinib-treated MF patients with shorter disease duration, suggesting a benefit of earlier treatment,¹¹ as alluded to above. Conversely, the odds of spleen response to ruxolitinib are lower and time to discontinuation shorter in the presence of additional, nondriver mutations.⁵² However, early intervention trials are challenging to undertake because of the long-term follow-up necessary to demonstrate favorable clinical outcomes. The ReTHINK trial,⁵³ which randomly assigned patients with lower-risk MF without clinically relevant problems but high-risk molecular features (ie, mutations in ASXL1, EZH2, IDH1/2, or SRSF2) to low-dose ruxolitinib (10 mg twice a day) or placebo for the purpose of delaying clinical progression, could not be completed because of poor accrual, likely related in part to the placebo component and also concerns about long-term ruxolitinib adverse events in clinically well patients.

Anemia improvement and transfusion independence.

Anemia is common in MF, and both anemia and RBC transfusion requirement have long been recognized as risk factors for leukemic transformation and death,^54-56 featured prominently in prognostic models^57,58 and important for transplant decision making.⁵⁹ The correlation between 24-week RBC-TI and OS in SIMPLIFY-1³⁰ lends additional credibility to this end point, linking anemia improvement to survival benefit. However, transfusion decisions in clinical practice are inherently subjective. Ruxolitinib-induced anemia does not carry the adverse prognosis of disease-induced anemia⁶⁰ and is likely biologically distinct, perhaps explaining the higher RBC-TI rates in the add-on cohorts of the phase II trials of luspatercept⁶¹ and the bromodomain and extraterminal protein inhibitor, pelabresib (Table 1).⁶² Besides RBC-TI, amelioration of anemia in non–RBC-transfusion-dependent individuals may possibly represent biologic evidence of disease modification, as exemplified by pelabresib in the MANIFEST trial,⁶² where clinical improvement in anemia appeared to be associated with increased bone marrow erythropoiesis. In the context of JAK inhibitor therapy, anemia is predominantly viewed as an adverse effect of on-target JAK2 inhibition and resultant suppression of erythropoiesis.⁶³ However, even with JAK inhibitors other than momelotinib, which improves anemia via ACVR1 inhibition and suppression of hepatic hepcidin production,⁶⁴ some patients do achieve RBC-TI⁶⁵; the mechanism behind this (eg, improved BM function, correction of hypersplenism, and others) remains unclear. Nevertheless, this is a small subset of patients, and combinations of traditional, more myelosuppressive JAK inhibitors with agents that improve anemia, eg, pelabresib or luspatercept, will likely continue to be attractive.

Cytokines and symptoms/SVR.

A wide array of proinflammatory cytokines are elevated in patients with MF, and some, like interleukins 8, 12, and 15 and the circulating interleukin-2 receptor, retain independent prognostic significance.⁶⁶ In the phase I and II study of ruxolitinib, the drug broadly suppressed cytokines, which correlated with symptom benefit.¹³ Achievement of SVR35 at 24 weeks on both ruxolitinib⁶⁷ and fedratinib⁶⁸ has also been correlated with reductions in tumor necrosis factor alpha, and increases in leptin and adiponectin. Several novel agents, both alone^69,70 and in combination with ruxolitinib,^71-73 have demonstrated cytokine modulation. However, the lack of strong correlation coefficients and validation across multiple large studies precludes the use of one or more cytokines as end points. Specific cytokine levels and/or changes that may correlate with SVR, improvement in TSS and/or individual symptoms, OS, and other disease parameters, eg, cytopenias, need further study. TSS50 at 24 weeks has been a coprimary or key secondary end point of nearly all phase III trials of JAK inhibitors in MF, and is the sole primary end point of the pivotal MOMENTUM trial of momelotinib.⁷⁴ Although of obvious importance to patients, there is likely a subjective component to this questionnaire-based end point, unlike SVR35. With a number of novel agents, for example, there is a noticeable discrepancy between rates of spleen and symptom response, eg, itacitinib, imetelstat, navtemadlin, parsaclisib, bomedemstat, and PRM-151 (Table 1).^36,44,75-78 Given the limitations of cytokines as disease biomarkers, it is likely that patient-reported outcomes are here to stay for the foreseeable future as important clinical trial end points.

In conclusion, cure and/or improvement in survival remains the holy grail of MF treatment.⁷⁹ The profusion of novel agents under investigation and the acceptance of new end points in trials, including OS, are welcome developments. However, the candidacy of biomarker-driven end points as surrogates for OS has yet to be convincingly demonstrated. By contrast, clinical end points, such as SVR35 and RBC-TI, have been shown to correlate with OS. Furthermore, end points tailored to JAK inhibitors may not lend themselves well to other novel agents—this is now being recognized in trial design. It is critical to harmonize inclusion criteria and response assessment across trials, particularly those being conducted in the ruxolitinib failure and suboptimal response settings, to allow for a fair appraisal of investigational agents. For now, it is important not to lose sight of hard clinical outcomes in our quest for disease modification.

AUTHOR CONTRIBUTIONS

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Disease Modification in Myelofibrosis: An Elusive Goal?

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jco/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).