Long-term treatment with ruxolitinib for patients with myelofibrosis: 5-year update from the randomized, double-blind, placebo-controlled, phase 3 COMFORT-I trial

Background The randomized, double-blind, placebo-controlled, phase 3 COMFORT-I trial evaluated the JAK1/JAK2 inhibitor ruxolitinib in patients with intermediate-2/high-risk myelofibrosis. The primary and planned 3-year analyses of COMFORT-I data demonstrated that ruxolitinib—the first myelofibrosis-approved therapy—reduced splenomegaly and prolonged overall survival versus placebo. Here, we present the final 5-year results. Methods Patients managed in Australia, Canada, and the USA were randomized centrally (interactive voice response system) 1:1 to oral ruxolitinib twice daily (15 or 20 mg per baseline platelet counts) or placebo. Investigators and patients were blinded to treatment. The secondary endpoints evaluated in this analysis were durability of a ≥35% reduction from baseline in spleen volume (spleen response) and overall survival, evaluated in the intent-to-treat population. Safety was evaluated in patients who received study treatment. Results Patients were randomized (September 2009–April 2010) to ruxolitinib (n = 155) or placebo (n = 154). At termination, 27.7% of ruxolitinib-randomized patients and 25.2% (28/111) who crossed over from placebo were on treatment; no patients remained on placebo. Patients randomized to ruxolitinib had a median spleen response duration of 168.3 weeks and prolonged median overall survival versus placebo (ruxolitinib group, not reached; placebo group, 200 weeks; HR, 0.69; 95% CI, 0.50–0.96; P = 0.025) despite the crossover to ruxolitinib. The ruxolitinib safety profile remained consistent with previous analyses. The most common new-onset all-grade nonhematologic adverse events starting <12 versus ≥48 months after ruxolitinib initiation were fatigue (29.0 vs 33.3%) and diarrhea (27.8 vs 14.6%). New-onset grade 3 or 4 anemia and thrombocytopenia both primarily occurred within the first 6 months, with no cases after 42 months. The most common treatment-emergent adverse event-related deaths in the ruxolitinib-randomized group were sepsis (2.6%), disease progression (1.9%), and pneumonia (1.9%). Conclusion The final COMFORT-I results continue to support ruxolitinib as an effective treatment for patients with intermediate-2/high-risk MF. Trial registration ClinicalTrials.gov, NCT00952289

Methods: Patients managed in Australia, Canada, and the USA were randomized centrally (interactive voice response system) 1:1 to oral ruxolitinib twice daily (15 or 20 mg per baseline platelet counts) or placebo. Investigators and patients were blinded to treatment. The secondary endpoints evaluated in this analysis were durability of a ≥35% reduction from baseline in spleen volume (spleen response) and overall survival, evaluated in the intent-to-treat population. Safety was evaluated in patients who received study treatment. Results: Patients were randomized (September 2009-April 2010) to ruxolitinib (n = 155) or placebo (n = 154). At termination, 27.7% of ruxolitinib-randomized patients and 25.2% (28/111) who crossed over from placebo were on treatment; no patients remained on placebo. Patients randomized to ruxolitinib had a median spleen response duration of 168.3 weeks and prolonged median overall survival versus placebo (ruxolitinib group, not reached; placebo group, 200 weeks; HR, 0.69; 95% CI, 0.50-0.96; P = 0.025) despite the crossover to ruxolitinib. The ruxolitinib safety profile remained consistent with previous analyses. The most common new-onset all-grade nonhematologic adverse events starting <12 versus ≥48 months after ruxolitinib initiation were fatigue (29.0 vs 33.3%) and diarrhea (27.8 vs 14.6%). New-onset grade 3 or 4 anemia and thrombocytopenia both primarily occurred within the first 6 months, with no cases after 42 months. The most common treatment-emergent adverse event-related deaths in the ruxolitinib-randomized group were sepsis (2.6%), disease progression (1.9%), and pneumonia (1.9%).
(Continued on next page) Background Myelofibrosis (MF) is a Philadelphia chromosomenegative myeloproliferative neoplasm [1] that is often associated with splenomegaly, anemia, and burdensome symptoms that negatively affect quality of life [2,3]. In addition, patients with MF have shortened survival compared with age-and sex-matched members of the general population [4]. Allogeneic stem cell transplantation is the only potentially curative treatment option [5]. However, transplant-related morbidity and mortality are considerable, and many patients with MF are ineligible because of their age or comorbidities.
Ruxolitinib is a JAK1/JAK2 inhibitor approved by the US Food and Drug Administration for patients with intermediate-or high-risk MF, including primary MF (PMF), post-polycythemia vera MF (PPV-MF), and postessential thrombocythemia MF (PET-MF), as well as patients with PV who have had an inadequate response to or are intolerant of hydroxyurea [10]. Ruxolitinib is also approved by the European Medicines Agency for the treatment of disease-related splenomegaly or symptoms in adult patients with PMF, PPV-MF, or PET-MF and for the treatment of adult patients with PV who are resistant to or intolerant of hydroxyurea [11]. Approval for MF was based on two randomized phase 3 clinical trials in patients with intermediate-2 or high-risk PMF, PPV-MF, or PET-MF [12,13]. Controlled Myelofibrosis Study with Oral JAK Inhibitor Treatment (COMFORT)-I was a double-blind, placebo-controlled trial, and COMFORT-II was an open-label trial comparing ruxolitinib with the best available therapy. In both trials, ruxolitinib was superior to control interventions, reducing spleen size and improving MF-related symptoms and quality-of-life (QoL) measures. Spleen volume reductions and improvements in measures of QoL at week 24 in COMFORT-I were observed regardless of MF subtype, age, International Prognostic Scoring System (IPSS) risk score, Eastern Cooperative Oncology Group (ECOG) performance status, and baseline hemoglobin level, platelet count, spleen size, and JAK2V617F mutation status [14].
Long-term follow-up analyses of the COMFORT studies have indicated durable clinical benefit and are suggestive of a survival advantage with ruxolitinib treatment [15][16][17]. Most nonhematologic adverse events in COMFORT-I and COMFORT-II were grade 1 or 2, with the rate generally decreasing with long-term ruxolitinib treatment [15,16]. Dose-dependent cytopenias were the most common hematologic adverse events. These occurred primarily within the first 12 weeks of ruxolitinib treatment and stabilized thereafter in patients continuing therapy, with hemoglobin levels returning to nearbaseline levels [15,16].
This analysis reports the final long-term efficacy and safety results of COMFORT-I after 5 years of ruxolitinib treatment.

Study design and patients
The detailed study design and protocol of the randomized, double-blind, placebo-controlled, phase 3 COMFORT-I trial have been reported previously [12]. The study was conducted in 89 sites across Australia, Canada, and the USA. Briefly, patients with intermediate-2 or high-risk MF and splenomegaly of >5 cm below the left costal margin by palpation were eligible.
The protocol was reviewed and approved by each participating site's institutional review board. All patients provided written informed consent.

Randomization and masking
Patients were randomized 1:1 to ruxolitinib or matching placebo tablets by a centralized interactive voice response system (IVRS). Study investigators and patients were blinded to the treatment. Study treatments were provided in encoded bottles, and patient study drug assignments were provided to site staff by the IVRS.

Procedures
Study treatments, administered orally twice daily, were ruxolitinib (Incyte Corporation, Wilmington, DE; dosing based on baseline platelet counts: 100-200 × 10 9 /L, 15 mg; >200 × 10 9 /L, 20 mg) or placebo. Dose modification was allowed for efficacy and safety. Crossover from placebo to ruxolitinib was permitted before week 24 for protocol-defined worsening splenomegaly. After week 24, patients with protocol-defined worsening symptomatic spleen growth either received unblinded ruxolitinib or discontinued the study; patients with protocoldefined asymptomatic spleen growth were given the option to unblind, after which they were required to receive ruxolitinib or discontinue the study.
This final analysis occurred when all the patients reached the 5-year visit or discontinued participation. Changes from baseline or crossover baseline in spleen volume were assessed by magnetic resonance imaging or computed tomography every 12 weeks from weeks 12 to 72 and every 24 weeks thereafter. Patients who had a spleen volume measurement at baseline and each time point of interest were evaluable to determine if a ≥35% reduction from baseline in spleen volume was achieved; all patients who withdrew before the time point were considered nonresponders.

Outcomes
The primary endpoint was the proportion of patients who achieved spleen response (≥35% reduction from baseline in spleen volume) at week 24. Secondary endpoints reported in this analysis included duration of spleen response and overall survival (OS).
Hematologic adverse events were based on laboratory abnormalities. Because the majority of the anemia and thrombocytopenia events occurred early in the study, the incidence of new-onset or worsening grade 3 or 4 anemia or thrombocytopenia was assessed at 6-month intervals in patients originally randomized to ruxolitinib. The placebo group was included in only the first 6month interval because all the patients receiving placebo discontinued or crossed over to ruxolitinib within 3 months of the primary analysis. Nonhematologic adverse events were assessed per National Cancer Institute Common Terminology Criteria for Adverse Events [18]. The incidence of nonhematologic adverse events was assessed in yearly intervals for patients originally randomized to ruxolitinib.

Statistical analysis
Changes from baseline or crossover baseline in spleen volume were summarized with descriptive statistics. Durability of spleen response and OS were calculated with the Kaplan-Meier method in the intent-to-treat population. OS was calculated based on randomized treatment. Spleen response was considered lost at the first measurement that was no longer a ≥35% reduction from baseline and was also a >25% increase from the nadir. Hazard ratio with 95% confidence interval and P values were calculated with the Cox proportional hazards model and the log-rank test. A subgroup analysis of OS was conducted in patients with intermediate-2 or high-risk MF per IPSS criteria [19].
Safety analyses were conducted in all patients who received ≥1 dose of study treatment. The incidence of new-onset or worsening grade ≥3 anemia and thrombocytopenia (based on laboratory data) and of new-onset or worsening all-grade and grade ≥3 nonhematologic adverse events was calculated using the lifetable method. The time to the first event censored at the date of the last laboratory evaluation was used for anemia and thrombocytopenia; the earlier discontinuation or date of data cutoff was used for nonhematologic adverse events. Per the life-table method, the incidence of each adverse event was based on the effective sample size of the time interval, which was the number of patients at risk at the beginning of the interval minus half of the censored patients during the time interval.
Statistical analyses were conducted using SAS version 9.2 (SAS Institute, Cary, NC).
The trial was overseen by a data monitoring committee and is registered at ClinicalTrials.gov (NCT00952289).

Role of the funding source
Conduct of this study and editorial assistance were funded by Incyte Corporation. Incyte Corporation employees worked with external investigators in designing the study, analyzing data, and confirming accuracy of this report. The authors had full access to all the data in the study and had final responsibility for the decision to submit.

Patient disposition
Patients were recruited between September 2009 and April 2010 and randomized to ruxolitinib (n = 155) or placebo (n = 154; Fig. 1). All patients were included in the intent-to-treat population; three patients in the placebo group were not evaluable for safety. By the time of the 3-year analysis, all evaluable patients in the placebo group had discontinued (40/151 [26.5%]) or crossed over to ruxolitinib (111/151 [73.5%]) [15]. The median (range) time to crossover was 39.9 (5.0-65.3) weeks. At study termination (i.e., the 5-year data cutoff ), 27.7% (43/155) of patients originally randomized to ruxolitinib and 25.2% (28/111) of those who crossed over to ruxolitinib were receiving treatment in the study. An additional four patients in the ruxolitinib-randomized group who discontinued the study transitioned to commercial ruxolitinib.

Overall survival
At the time of the final 5-year analysis, median followup time for the OS analysis was 268.4 weeks for the ruxolitinib group and 269.0 weeks for the placebo group. Patients randomized to ruxolitinib experienced prolonged OS compared with those in the placebo group. Fig. 1 Patient disposition. *Three patients in the placebo group were not evaluable for safety (n = 151); these patients were excluded from the calculation of the percentage of patients who discontinued. (dagger) Limited to patients whose study discontinuation dates matched their dates of death. (double dagger) Including but not limited to the following: received a different therapy, transitioned to commercial ruxolitinib, and loss of response Median OS was not reached in the ruxolitinibrandomized group. Among patients randomized to placebo, median OS was 108 weeks for patients censored at crossover and 200 weeks for all patients (HR, 0.69; 95% CI, 0.50-0.96; P = 0.025; Fig. 4). There were a total of 69 deaths (regardless of cause) in the ruxolitinibrandomized group and 82 deaths among those randomized to placebo. In a subgroup analysis by IPSS risk status, there was a nonsignificant trend toward longer OS among patients in the ruxolitinib group compared with the placebo group for both intermediate-2 and high-risk patients (Fig. 5).
Anemia and thrombocytopenia (per abnormal hematologic laboratory values) occurred in most patients in the ruxolitinib-randomized group (98.7 and 83.9%, respectively). The rates of new or worsening grade 3 or 4 anemia, thrombocytopenia, and leukopenia were the highest within the first 6 months of treatment, decreasing thereafter (Fig. 6). No patients in the ruxolitinibrandomized group had new or worsening grade 3 or 4 anemia, thrombocytopenia, or leukopenia after month 42. In agreement with the hematologic laboratory abnormalities over time in the ruxolitinib-randomized group, mean hemoglobin levels decreased during the first 12 weeks of treatment with randomized or crossover ruxolitinib but increased toward baseline levels and stabilized thereafter (Fig. 7). Similarly, mean platelet and white blood cell counts also decreased during the first 12 weeks of treatment with ruxolitinib, after which they remained stable (Fig. 7). In agreement with these blood count dynamics, the mean (SD) number of transfusions per month in the ruxolitinib group peaked between weeks 4 and 8 (1. Serious adverse events occurred at any time during treatment with ruxolitinib in 103/155 (66.5%) patients in the ruxolitinib-randomized group and 74/111 (66.7%) patients in the ruxolitinib crossover group. The most frequent serious adverse events, occurring in ≥4% of patients in the ruxolitinib-randomized or crossover groups, were pneumonia (randomized, 15.5%; crossover, 10.8%), anemia (11.0%; 11.7%), sepsis (4.5%; 3.6%), and congestive cardiac failure (3.2%; 4.5%).
Throughout the course of the study, adverse events resulted in a ruxolitinib dose decrease in 88/155 (56.8%) patients in the ruxolitinib-randomized group and 45/111 (40.5%) patients in the ruxolitinib crossover group. Thrombocytopenia was the most frequent cause of dose decreases, occurring in 75/155 (48.4%) and 36/111 (32.4%) patients in the ruxolitinib-randomized and crossover groups, respectively.
Herpes zoster infections occurred at higher rates in patients treated with ruxolitinib compared with placebo ( Table 3). All herpes zoster events in the ruxolitinib-  *Occurring in ≥2 patients in the ruxolitinib group in any yearly interval randomized group were grade 1 or 2, occurring in three or four patients each year. Among all patients treated with ruxolitinib, the majority of cases were single episodes that were grade 2 or lower and resolved without long-term sequelae. The only serious event of herpes zoster infection occurred in a patient randomized to placebo after crossing over to ruxolitinib. Sepsis occurred at similar rates between patients treated with ruxolitinib and those receiving placebo ( Table 3). All sepsis events were grade 3 or 4, with the exception of one grade 2 event in the placebo group. Serious events of sepsis and septic shock occurred at rates of 1.5 and 0.4 per 100 patient-years of exposure in the ruxolitinib-randomized group and 1.5 each in the ruxolitinib crossover group.
Nonmelanoma skin cancers, including basal cell carcinoma and squamous cell carcinoma of the skin, occurred at similar rates between patients treated with ruxolitinib and those receiving placebo (Table 3). Basal cell carcinoma occurred at a rate of 2.7 per 100 patientyears of exposure in the ruxolitinib-randomized group, 4.0 in the ruxolitinib crossover group, and 3.9 among patients during treatment with placebo (Table 3). There were two cases of basal cell carcinoma in the ruxolitinib crossover group; in both cases, patients had a history of skin cancer.  Disease transformation to AML occurred in five patients each in the ruxolitinib-randomized and ruxolitinib crossover groups; no patients developed AML during treatment with placebo (Table 3). Overall, AML occurred in five male and five female patients. The median (range) time from the first ruxolitinib dose to AML diagnosis was 838 (157-1150) days in the ruxolitinib-randomized group and 376  days in the ruxolitinib crossover group; median (range) time from MF diagnosis to AML diagnosis was 1190 (699-1708) days and 1015 (372-11,971) days, respectively. Prior medications for the treatment of MF in patients who developed AML were hydroxyurea (ruxolitinib-randomized group, n = 2; ruxolitinib crossover group, n = 2) and lenalidomide, and corticosteroids (all in one patient in the ruxolitinib crossover group); three patients in the ruxolitinibrandomized group and two in the ruxolitinib crossover group had no prior treatments for MF.
Overall, 28/155 (18.1%) patients in the ruxolitinibrandomized group and 39/151 (25.8%) in the placebo randomized group experienced a treatment-emergent adverse event that resulted in death while on study or within 28 days of the last dose of study drug. Among the patients randomized to placebo, a treatment-emergent adverse event led to death in 11/151 (7.3%) patients   (Table 4).

Discussion
This final analysis of the COMFORT-I trial demonstrated that treatment with ruxolitinib was associated with rapid and durable reductions in splenomegaly and significantly longer OS compared with patients originally randomized to placebo. Patient risk of death was approximately 30% lower in the ruxolitinib group compared with placebo, despite the crossover from placebo to ruxolitinib. Given that COMFORT-I was restricted to patients with intermediate-2 or high-risk MF with splenomegaly, OS data suggest that delaying treatment with ruxolitinib may worsen outcomes and that studies evaluating ruxolitinib in patients with earlier MF disease states may be warranted. The exact mechanism by which ruxolitinib prolongs survival and ameliorates splenomegaly remains unclear, but it is rational to hypothesize that the downstream effects of ruxolitinib confer changes in cytokines, metabolic properties, and JAK2V617F allele burden that may play a role. Ruxolitinib has been associated with reductions in inflammatory cytokines and markers of inflammation [12], improvements in measures of metabolic and nutrition status [20], reduced fibrosis in some patients [17], and reductions in JAK2V617F allele burden [21]. In COMFORT-I patients receiving long-term treatment with ruxolitinib, relationships have been identified between reductions in spleen volume and (1) increases in body weight and normalization of serum albumin and total cholesterol levels [20] and (2) reductions in JAK2V617F allele burden in some patients [21]. In addition, ruxolitinib has been associated with improvements in spleen volume and OS in a wide variety of patient subgroups stratified by MF subtype, age, IPSS risk score, ECOG performance status, and baseline hemoglobin level, platelet count, spleen size, and JAK2V617F mutation status [14]. Future work will be required to elucidate the mechanism by which ruxolitinib is efficacious and if there are any related disease markers or patient characteristics that could be helpful in identifying the types of patients who may benefit the most from ruxolitinib treatment.
Overall, the safety profile was supportive of long-term treatment with ruxolitinib, with no unexpected safety signals. The nonhematologic adverse event rates generally remained stable or decreased with prolonged ruxolitinib treatment duration and were consistent with those reported in previous analyses of the COMFORT-I study [12,15]. As expected, based on the mechanism of action of ruxolitinib as a JAK1/JAK2 inhibitor [22,23], thrombocytopenia and anemia occurred in most patients treated with ruxolitinib. Anemia and thrombocytopenia can be managed with dose adjustments and, for some patients with anemia, red blood cell transfusions [24]. Indeed, although thrombocytopenia was the most common cause for ruxolitinib dose reduction in COMFORT-I, thrombocytopenia and anemia resulted in relatively few discontinuations (each ≤3.6% in the ruxolitinib- Fatal treatment-emergent adverse events that occurred following the first dose of ruxolitinib (ie, after crossover from placebo) were included in the ruxolitinib crossover group randomized and crossover groups). Mean hemoglobin, platelet, and white blood cell levels stabilized after 12 weeks of treatment, with hemoglobin levels returning to near-baseline levels thereafter; however, this finding must be interpreted taking into account the positive selection of patients remaining on study. Mean blood transfusion rates were in agreement with these trends. Nevertheless, ruxolitinib may provide a survival benefit even in the presence of anemia. In a pooled analysis of COMFORT-I and COMFORT-II 3year data, treatment with ruxolitinib was associated with a survival advantage regardless of anemia at baseline (3-year OS probability: ruxolitinib, 0.66; control, 0.57) or after initiating study treatment (3-year OS probability: ruxolitinib, 0.87; control, 0.66) [25]. Herpes zoster infections occurred at higher rates among patients treated with ruxolitinib compared with placebo. The incidence of herpes zoster infections increased with longer exposure to ruxolitinib (0-12 months' exposure, 2.1%; ≥48 months' exposure, 10.3%). However, all but one case was grade 1 or 2, and it is unclear if this increase was clinically relevant. Other infections, including pneumonia, sepsis, upper respiratory tract infection, and urinary tract infection, occurred at similar or lower rates with ruxolitinib compared with placebo; however, pneumonia was the most common new-onset grade 3 or 4 adverse event observed after 48 months of treatment with ruxolitinib. Nonmelanoma skin cancers were observed in patients treated with ruxolitinib; however, these occurred at rates that were similar to or lower than those observed during treatment with placebo. Finally, the incidence of AML transformation in the ruxolitinib-randomized and crossover groups was consistent with previous reports in patients with MF [26,27]. Although no patients developed AML during treatment with placebo, the median exposure (37.1 weeks) may not have been long enough to observe AML transformations considering that the median time from the first ruxolitinib dose to AML diagnosis was 119.7 weeks in the ruxolitinib-randomized group.
Overall, 48.9% of the COMFORT-I patient population had died by the time of the final 5-year analysis. Causes of death were generally consistent with expected morbidities resulting from MF progression and/or other underlying disease processes (e.g., infections, transformation to AML), particularly in elderly and chronically ill patients. The most common adverse events leading to death in COMFORT-I were sepsis or septic shock, followed by disease progression, pneumonia, and transformation to AML. Eleven patients treated with ruxolitinib died because of a cardiovascular, thrombotic, or hemorrhagic event. In comparison, an international retrospective analysis of 1131 patients with PMF enrolled between 1980 and 2007 reported that the leading causes of death were transformation to AML, disease progression, thrombosis and cardiovascular complications, and infection [19].

Conclusions
This final analysis of the COMFORT-I study included 5 years of treatment duration and demonstrated that long-term ruxolitinib treatment in patients with intermediate-2 or high-risk MF was associated with durable reductions in spleen size and significantly longer OS compared with placebo. The safety profile continued to remain consistent with previous COMFORT-I and COMFORT-II analysis [12,13,[15][16][17], with no new or unexpected adverse events identified with long-term treatment. Collectively, these data and similar findings in the 5-year analysis of the COMFORT-II study [17] support ruxolitinib as an effective long-term treatment option for patients with intermediate-2 or high-risk MF.