Pharmacoeconomics of ruxolitinib therapy in patients with myelofibrosis

Abstract

Background:

Overall survival (OS) and other important clinical trial end-points seem increasingly more elusive in supporting rapid and efficient incorporation of innovative cancer drugs in clinical practice. This study proposes a clinical trial based pharmacoeconomic framework to assess the therapeutic and economic value of ruxolitinib in patients with intermediate-2 or high-risk myelofibrosis.

Methods:

Individual patient level 144 week follow-up data from the COMFORT-II trial was used to account for the crossover effect on overall survival. Lifetime treatment benefits and costs were estimated considering detailed patterns of both ruxolitinib dose adjustments and blood transfusion needs.

Results:

The authors estimate a 3.3 years increment in life expectancy (HR = 0.30; 95% CI = 0.17–0.55; p-value <0.001) and an incremental cost-effectiveness ratio of €40,000 per life year gained with the use of ruxolitinib.

Conclusion:

This study also demonstrates how valuable information from clinical trials can be used to support informed decisions about the early incorporation of innovative drugs.

Keywords::

• Ruxolitinib
• Myelofibrosis
• Cost-effectiveness
• Individual level data

Introduction

The identification of mutations in the Janus kinase (JAK) gene introduced a new era of promising molecular targeted therapy for myeloproliferative neoplasms. Ruxolitinib was the first JAK1/JAK2 inhibitor approved for the treatment of myelofibrosis (MF), with evidence of rapid reductions in splenomegaly and noticeable improvements in disease-related symptoms and overall survival (OS) in controlled clinical trials, contrasting to the limited impact of conventional treatments so far1. Long-term OS assessment continued to favor ruxolitinib both in COMFORT-I and COMFORT-II studies, despite the majority of comparator arm patients crossing over to ruxolitinib1^,2. Also very encouraging are the retention rates of ruxolitinib in the long run, with a significant proportion of patients benefiting from treatment, sometimes even after randomized controlled trial (RCT) pre-defined progression end-points3. These findings have important implications both in terms of benefits to patients and economic impact to health systems.

Patients with myelofibrosis experience substantial impairment in daily activities and quality-of-life due to the burden of accompanying symptoms: bone and muscular pain, fever and night sweats, early satiety and abdominal discomfort, itching, and weight loss4. MF-associated complications are the leading source of morbidity and mortality5. Next to advanced age, leukocytosis, the presence of circulating blasts, anemia, and red blood cell (RBC) transfusion dependency, among others, are established risk factors for survival in MF6^,7. Evolution to acute myeloid leukemia (AML), which occurs in ∼10–20% of patients9, is a major cause of death in MF, together with other co-morbid conditions such as cardiovascular events and consequences of cytopenias such as infection or bleeding9–11.

The economic burden of MF is also staggering. Recent research showed that annual healthcare costs for patients with MF may be 5-times that of matched non-cancer controls12^,13. Outpatient visits represent the largest share (53%, mean cost = $18,395) of annual total costs ($34,690) in a large US commercial claims database analysis13, with inpatient stays (23%, $8106), drugs (22%, $7803) and emergency visits (1%, $386) accounting for the remaining costs. In addition, non-medical and indirect costs (€15,142/year) constitute an important economic burden to patients, their families, and society as a whole14.

Considering the evidence in favor of the long-term use of ruxolitinib in patients with MF, a cost-effectiveness analysis was performed to assess long-term survival benefits of ruxolitinib treatment and estimate the long-term costs associated with the management of the disease.

Materials and methods

Overview

A discrete state cohort model was developed to estimate the costs and effectiveness related to the treatment of MF with either ruxolitinib or best available therapy (BAT). Given the chronic and progressive nature of the disease, the model considers a lifetime time horizon. Cost-effectiveness of ruxolitinib is assessed in terms of the incremental cost per life year (LY) gained, from a Portuguese National Health Service (NHS) perspective. Following the guidelines of the Portuguese Ministry of Health15, an annual 5% discount rate was applied to both costs and effectiveness. A summarized description of the key model components is presented in Table 1. The statistical models that are at the base of most of these components are described below and were based on the 144 week follow-up patient-level data from the COMFORT-II study1. The COMFORT-II study is an open-label phase 3 study that includes patients with a diagnosis of primary MF, post-polycythemia vera MF (PP-MF), or post-essential thrombocythemia MF (PET-MF), classified as intermediate-2 or high risk using the International Prognostic Scoring System (IPSS). Patients (n = 219) were randomized 2:1 to receive ruxolitinib (15 or 20 mg twice daily, based on baseline platelet count; 100–200 or >200 × 10⁹/L, respectively) or BAT1. Probabilistic sensitivity analysis was used to assess parameter uncertainty, when possible.

Table 1. Summary of key model components.

Model component	Description/assumption
Model structure	Discrete state cohort model
Cycle length	4 weeks
Time horizon	Lifetime
Annual discount rate costs and effectiveness	5%
Perspective	National health service
Overall survival and extrapolation	Weibull parametric survival model
Overall survival cross-over correction (BAT)	Rank preserving structural failure time model
Proportion of deaths due to AML	15%
Ruxolitinib treatment persistence and extrapolation	Weibull parametric survival model
Ruxolitinib average daily dose (mg)	Piecewise mixed regression model for continuous longitudinal data
Number of RBC transfusions per cycle	Mixed regression model for ordered multinomial longitudinal data
Drug costs per cycle (€)
Ruxolitinib - 1st year average	€2 408.15
Ruxolitinib - 2nd year average	€2 108.89
Ruxolitinib - subsequent years	€2 007.96
BAT	€252.28
Cost per RBC transfusion (€)	€392.90
Non-hematologic adverse event costs per cycle (€)
Ruxolitinib	€10.93
BAT	€17.72
Other medical costs per event (€)
Disease monitoring	€451.10
Outpatient visit	€114.04
Inpatient visit	€3 913.18
Splenic irradiation	€3 140.84
Splenectomy	€2 384.06
AML	€21,215.06

BAT, Best available therapy; AML, Acute myeloid leukemia.

Mortality and AML transformation

In the COMFORT-II study, after demonstration of the superiority of ruxolitinib treatment at the time of the primary analysis (January 2011), all patients randomized to the BAT arm were allowed to crossover to ruxolitinib treatment, irrespective of their progression status. After 144 weeks of follow-up, 45 (61.6%) of the 73 patients originally randomized to receive BAT had crossed over to ruxolitinib1. For these patients, and for the purpose of cost-effectiveness modeling, OS times were corrected using Rank Preserving Structural Failure Time (RPSFT) models16^,17, resulting in latent survival times that could have been observed if the patients had not crossed over to ruxolitinib treatment. After this adjustment, joint parametric survival models were used to estimate OS curves for both treatment groups, allowing extrapolation of OS for time periods beyond the 144-week follow-up observed in the COMFORT-II study. The Weibull survival model was selected from a pre-defined set of models based on the Akaike Information Criterion.

For ∼10–20% of patients, transformation to AML is part of the natural evolution of MF9. Due to the lack of robust data on how MF treatment, specifically ruxolitinib, affects the incidence of AML, detailed modeling of AML transformation as a function of treatment was not possible. As such, it was assumed that 15% of all deaths were due to AML transformation.

Ruxolitinib persistence and dosing

Up until 48 weeks, most patients randomized to ruxolitinib (97%) in the COMFORT-II study had evidenced a measurable reduction in spleen volume. For patients receiving BAT, this was the case for only 56%. The remaining patients demonstrated an increase in spleen volume, especially within the BAT group, with a median (range) best percentage change from baseline of 6.6% (0.1–57.0%). Furthermore, 28% of the patients randomized to ruxolitinib treatment met the criterion for the primary end-point (≥35% reduction in spleen volume at 48 weeks), as compared to 0% in the BAT group3. As a result, many patients who suffered protocol-defined disease progression (splenectomy or increase in spleen volume of >25% from the nadir during the study period) in the ruxolitinib group could still be deriving a clear clinical benefit, as opposed to patients progressing in the BAT group. For these patients, randomized treatment could be maintained as long as a clinical benefit of ruxolitinib was deemed present. For modeling purposes, in order to capture the full clinical benefit of ruxolitinib, treatment persistence (time to treatment discontinuation or death), rather than protocol-defined progression, was used as a proxy for clinical progression. As for OS, a Weibull parametric survival model was used for extrapolation. Upon ruxolitinib discontinuation, the model assumes that patients initiate BAT until death.

The COMFORT-II study protocol included a dosing regimen which required dose adjustments for safety reasons (development of neutropenia or thrombocytopenia) and dose escalation to increase efficacy. At 144 weeks of follow-up, adverse events requiring ruxolitinib dose reductions or interruptions had occurred in 72% of patients (47% due to thrombocytopenia). As a result, average daily dose intensity decreased over time, from an initial 36.0 mg/day over 29.1 mg/day at 24 weeks, to a stable 27.0 mg/day after 2 years (patient-level data from the COMFORT-II study1). To adequately capture the evolution of the average daily dose and its related uncertainty, a piecewise mixed regression model for continuous longitudinal data was applied. Resulting average ruxolitinib drug cost for the first, second and subsequent years of treatment are presented in Table 1.

Red blood cell transfusions

In the COMFORT-II study, thrombocytopenia and anemia occurred more frequently in patients receiving ruxolitinib than in patients receiving BAT. Irrespective of the treatment group, these adverse events were generally managed with dose reductions (see above) and/or packed RBC transfusions3. Although more ruxolitinib randomized patients required at least one RBC transfusion than BAT randomized patients (51% vs 38% up until 48 weeks), the average number of transfusions per month per patient was slightly lower in the ruxolitinib group (0.86 vs 0.91)3. The evolution of the average number of RBC transfusions per patient over time was studied using a mixed regression model for ordered multinomial longitudinal data, considering the percentage of patients with zero, one and two or more transfusions per cycle, for both treatment arms of the COMFORT-II study (BAT patients were modeled until cross-over). The average number of RBC transfusions for patients with two or more transfusions per cycle was estimated to be 2.4 and 2.5, for ruxolitinib and BAT patients, respectively. The average number of transfusion post-ruxolitinib was assumed to be fixed and equal to 0.52 per patient and cycle (patient-level data from the COMFORT-II study1).

Resource use and costs

For BAT treatment, an average annual number of 4.38 inpatient and 12.75 outpatient visits were considered18. Outpatient visits are assumed to include disease monitoring. Given the lack of information on resource use for patients treated with ruxolitinib in clinical practice, and the fact that ruxolitinib allows a better disease control (spleen volume and occurrence of constitutional symptoms), a reduction to 1/3 of the BAT in- and outpatient rates was assumed. This is in line with the NICE technology appraisal guidance on ruxolitinib in MF, in which a reduction to 30% of resource use costs was considered for the responder state as compared to the non-responder state19.

At the 48 week analysis of the COMFORT-II study, no cases of splenic irradiation or splenectomy were reported for ruxolitinib randomized patients3. Assuming a continued clinical benefit in terms of spleen volume until treatment discontinuation, splenic irradiation or splenectomies were not considered while on ruxolitinib. Estimates for the frequency of these events under BAT were determined from an unpublished US based chart review of 180 patients, of whom 102 had splenomegaly20: 1.85% per patient-year for splenic irradiation; 3.23% per patient-year for splenectomy. In terms of non-hematologic adverse events, only those clinically and economically relevant were considered (e.g., dyspnea, pneumonia). Unit costs are detailed in Table 1 and are based on the official price list from the Portuguese Ministry of Health21.

Results

Overall survival

After adjusting for cross-over effects on OS, using RPSFT models, we estimated a reduction of 70% in the risk of death for ruxolitinib patients compared to BAT patients (HR = 0.30; 95% CI = 0.16–0.55; stratified log-rank p-value <0.001) (Figure 1). This estimate compares favorably to the survival effect of ruxolitinib at 144 weeks of follow-up before taking into account the treatment switching in the BAT arm of the COMFORT-II study (HR = 0.48; 95% CI = 0.28–0.85; stratified log-rank p-value = 0.009)1. Median OS was reached in neither of the original treatment arms. After cross-over correction, median survival for BAT was reached at 2.7 years (Figure 1). To obtain survival estimates beyond the 144 week follow-up of the COMFORT-II study, a Weibull parametric survival model was applied, resulting in a predicted median survival for ruxolitinib treated patients of 5.7 years, as compared to a predicted median survival of 2.8 years for BAT patients.

Figure 1. Kaplan–Meier estimates of OS for ruxolitinib and BAT with (RPSFT) and without (original) cross-over correction.

Patterns of RBC transfusions

The temporal pattern of RBC transfusions observed (points) and estimated (shadowed area under the curve) from individual patient level data is presented in Figure 2 for the proportion of patients with zero, one or two or more transfusions per 4-week cycle. In the long run, patients treated with ruxolitinib are expected to be in a lesser need of transfusions compared to patients treated with BAT. Up to week 96, the estimated proportion of patients with zero transfusion per cycle is 82% with ruxolitinib as opposed to 77% with BAT. For the same period, only 5% of ruxolitinib patients are expected to need two or more RBC transfusions per cycle as compared to 13% with BAT treatment. These differences are expected to increase further over time (Figure 2).

Figure 2. Observed (points) and estimated (shadowed area under the curve) proportion of patients with zero, one or two or more transfusions per cycle of 4 weeks for MF patients treated with ruxolitinib or BAT.

Figure 3. Cost-effectiveness acceptability curve showing the probability of ruxolitinib being cost-effective over a range of values for the maximum willingness-to-pay per life year gained.

Cost-effectiveness

The main results obtained from the cost-effectiveness analysis are presented in Table 2. These results are discounted at a 5% annual rate according to the guidelines from the Portuguese Ministry of Health15. Undiscounted life expectancy after ruxolitinib initiation was estimated to be 6.5 years relative to 3.2 years for BAT, with an estimated benefit of 3.3 years in favor of ruxolitinib. Discounted incremental lifetime healthcare cost amounted to €97,052 per patient starting treatment with ruxolitinib. Of those, roughly 90% is for drug therapy costs (€87,267/patient), whereas the remaining €9785/patient is attributable to patients’ longer survival, needing additional RBC transfusions (€3257€/patient), and consuming more heathcare resources (€6738/patient). This increment was only partially offset by higher AML transformation and splenic irradiation/splenectomy costs in BAT (−€484/patient), resulting in an incremental cost-effectiveness ratio of €40,000 per life year gained with ruxolitinib. A probabilistic cost-effectiveness analysis assessing the uncertainty in results revealed that a willingness-to-pay up to €50,000 per additional life-year was associated with a greater than 95% probability of ruxolitinib being cost-effective against BAT (Figure 3).

Table 2. Cost-effectiveness results of ruxolitinib in comparison to best available therapy in patients with MF.

	Ruxolitinib	BAT	Incremental
Total Costs	€188,967	€91,915	€97,052
Total Life Years	5.39 LY	2.96 LY	2.43 LY
ICER			€40,000
Total Cost Breakdown
Treatment Costs	€96,980	€9 713	€87,267
Adverse Event Costs	€10,466	€6 935	€3 531
PRBC Transfusions	€9 509	€6 252	€3 257
Other Adverse Events	€957	€682	€274
AML Transformation	€2 419	€2 795	−€376
Splenic Irradiation/ Splenectomy	€292	€400	−€108
Resource Utilization	€78,809	€72,072	€6 738

BAT, Best available therapy; ICER, Incremental cost-effectiveness ratio; PRBC, Packed red blood cells; AML, Acute myeloid leukemia.

Discussion

In this study, the cost-effectiveness analysis aimed at assessing the long-term therapeutic benefits and costs of ruxolitinib for the treatment of patients with intermediate-2 or high-risk myelofibrosis. Our analysis was focused on adequately modeling OS, considering that the median survival time has not yet been reached in either arm of the COMFORT-II study (follow-up of 144 weeks) and taking into account the potential bias in survival times of BAT due to substantial cross-over to ruxolitinib1. Two additional issues required special statistical modeling due to their long-term consequences. The patterns of RBC transfusions were modeled given that repeated measurements were available for the same individual over time. Average ruxolitinib daily dose was modeled assuming not only dose adjustments for efficacy and safety reasons, but also to account for the fact that patients may derive some benefit from ruxolitinib continuation after protocol-defined progression3.

After adjusting for cross-over using RPSFT models we estimated a reduction of 70% in the risk of death for ruxolitinib patients as compared to BAT patients (HR = 0.30; 95% CI = 0.17–0.55). This estimate is in line with the results from a recent pooled analysis of OS in the COMFORT-I and COMFORT-II studies, reporting a cross-over corrected hazard ratio of 0.29 (95% CI = 0.13–0.63, between patients who received ruxolitinib and patients who received either placebo or BAT22.

As a result of this 70% reduction in mortality risk, the use of ruxolitinib for the treatment of patients with MF and splenomegaly is estimated to increase life expectancy by 3.3 years in comparison to managing patients with best available therapy. Corresponding discounted (5%/year) incremental lifetime healthcare costs (€97,052) and life years (2.43 LY) result in an incremental cost of €40,000 per life year gained with ruxolitinib treatment over BAT.

There is only one study in the literature reporting cost-effectiveness estimates of ruxolitinib for the treatment of MF23. This paper was based on NICE single technology appraisal for the UK. The results from the NICE assessment and ours cannot be directly compared because of the differences in the effectiveness measures (life years vs quality-adjusted life years), costs measurement, and modeling assumptions. For example, in the NICE assessment, progression to AML was not explicitly modeled, nor were related costs. Also discount rates recommended in the UK and Portugal differ between 3.5% and 5%, respectively.

However, a closer look at the original evidence review report24 reveals that our discounted life year estimated for ruxolitinib (5.39 LY) is within the same range of NICE estimate (5.03 LY). The major difference is related to our LY estimate for BAT (2.96 LY) and the one performed in the UK (3.99 LY). This is the main reason why we estimate a 2.43 LY benefit in favor of ruxolitinib as compared to a more modest benefit of 1.04 LY reported by ERG in the UK24. We argue that not adjusting BAT survival outcomes for cross-over to ruxolitinib can potentially dilute the differences between treatments. In fact our OS estimates are in line with the OS estimates from the study of Passamonti et al.25 assessing the impact of ruxolitinib on the natural history of MF by comparing a sub-set of COMFORT-II patients to a matched sample of patients from the DIPSS multi-center database not receiving any experimental drug at data cut-off and censored at the time of hematopoietic stem cell transplantation.

Like others, our study is subject to limitations. Long-term outcomes are being derived through prediction models and, hence, subject to uncertainty. Although modeling in cost-effectiveness analysis is a recommended standard practice26^,27, in particular when survival data from RCTs is not yet mature, extrapolation beyond the RCT time horizon is an exercise based on assumptions about the underlying distribution of the end-points studied that may not hold in the future28.

We used life years to measure MF treatment effectiveness instead of quality-adjusted life years. Quality-of-life in patients with MF is severely compromised by several constitutional symptoms29. Ruxolitinib has been shown to greatly improve the presence of symptoms and the quality-of-life in patients with MF30. These are good motives to support the use of QALYs in this context. The main reason we have not considered the calculation of QALYs is because no consistent utility values related to MF were available in the literature and, hence, they would have to be mapped from the EORTC QLQ-C30 scores of the COMFORT-II ruxolitinib study to one preference-based questionnaire like the EQ-5D, SF-6D or HUI31.

The presence of MF associated symptoms has been shown not only to affect quality-of-life, but also to severely impact productivity and activities of daily living32. In reducing the presence of constitutional symptoms, ruxolitinib also has the potential of reducing indirect costs to society due to presenteeism and absenteeism. However, since indirect costs are difficult to quantify and no evidence is available yet on exactly how ruxolitinib impacts patients’ work ability—especially from a Portuguese perspective—they have not been included in the current analysis. As more information becomes available on the impact of ruxolitinib on work productivity, the inclusion of indirect costs will certainly improve the current analysis.

Due to a lack of robust data on how MF treatment, specifically ruxolitinib, affects the incidence of AML, detailed modeling of AML transformation as a function of treatment was not possible. Since in ∼10–20% of patients, transformation to AML is part of the natural evolution of MF8, it was assumed that 15% of all deaths were due to AML transformation. Despite the fact that AML transformations have been considered the most expensive medical cost in the model (€21,215.06), they only represent €376 in favor of ruxolitinib, of the total incremental costs of €97,052. Following the NICE single technology appraisal for the UK23, excluding AML transformation and its cost from the model, results in an ICER of €40,155 per life year gained.

After the appearance of JAK inhibitor therapy in myelofibrosis, controversies about the use of stem-cell transplantation (SCT) arose. Early data on using ruxolitinib before transplant to decrease symptomatic splenomegaly and improve the general status of patients revealed that it may improve the negative prognostic impact of high-risk MF on outcomes after SCT33. On the other hand, it is still unclear what overall long-term benefits will result from JAK inhibitor use prior to transplantation34, delaying or even preventing the only potentially curative approach of SCT for MF, despite its risks and costs35. We have not taken into account these effects in our analysis, as data is still too scarce. As more relevant data is released, future research should assess both the costs and benefits of using ruxolitinib in different timings of the disease.

Conclusions

Overall, our study demonstrates that ruxolitinib is expected to deliver substantial long-term OS benefits in patients with intermediate-2 or high-risk MF while holding good value for money in comparison to other innovative treatments for hematologic malignancies used worldwide36.

Transparency

Declaration of funding

This research project was funded by Novartis.

Declaration of financial/other relationships

Novartis contracted with Exigo for the development of the model. BV, VA, JA, and JF are employed by Exigo Consultores. Exigo Consultores provided support in the form of salaries for the authors, but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.