A cost-effectiveness analysis of using azacitidine vs. decitabine in treating patients with myelodysplastic syndromes

Abstract

Objective:

Azacitidine and decitabine are used to treat patients with myelodysplastic syndromes (MDS) in the United States (US). This study sought to assess their relative cost-effectiveness.

Design and methods:

The authors developed a cost-effectiveness Markov model (1-month cycles) tracking hypothetical cohorts of MDS patients treated with azacitidine or decitabine over 2 years. The model used a US payer perspective and 2009 costs. Health states modeled included MDS with Transfusion Dependence, MDS with Transfusion Independence, Progression to Acute Myelogenous Leukemia (AML), and Death. Incremental cost-effectiveness outcomes included cost per quality-adjusted life year (QALY), cost per life year (LY), cost per patient-month of transfusion independence, and cost per case of AML progression avoided. One-way sensitivity analyses were performed on key model parameters.

Results:

Compared to decitabine, azacitidine was associated with better survival (1.512 LYs vs 1.292), more QALYs gained (1.041 vs 0.870), more patient-months with transfusion independence (8.328 vs 6.224), and a greater proportion of patients avoiding progression to AML (50.9% vs 28.5%). Total per-patient costs over 2 years for azacitidine were lower than for decitabine ($150,322 vs $166, 212).

Limitations:

To inform and update the model over time, it will be important that randomized or observational clinical studies be conducted to directly compare azacitidine and decitabine, provide new information on how these medicines are used, and on their relative clinical effectiveness.

Conclusion:

Results demonstrate that azacitidine provides greater clinical benefit and costs less than decitabine across all key outcomes. These results accentuate the positive role of azacitidine in providing cost-effective care for MDS.

Keywords::

• Cost-effectiveness
• Myelodysplastic syndromes
• Azacitidine
• Decitabine
• Economic evaluation

Introduction

Myelodysplastic syndromes (MDS) are a group of clonal hemopathies affecting mainly the elderly. MDS is characterized by ineffective hematopoiesis, leading to peripheral blood cytopenias and progressive bone marrow failure1. Patients with MDS are at higher risk for infection, unusual bleeding, and anemia. MDS patients also are at a higher risk for transformation to acute myeloid leukemia (AML)1–4. Patients’ health-related quality-of-life (HRQoL) can be considerably degraded by MDS, due to fatigue, hemorrhage, infections requiring hospitalization and/or treatment with intravenous medication, and the frequent need for blood transfusions5–8. Patients requiring repeat blood transfusions are likely to experience iron overload, further impacting health and health-related quality-of-life due to the subsequent need for iron chelation therapy.

Overall, 40% of MDS patients die of disease complications such as infection and hemorrhage and 20–30% of patients progress to AML, which may be resistant to treatment1. While prognosis for all MDS patients is poor, MDS risk is inversely associated with survival. MDS patients can be categorized as lower-risk (International Prognostic Scoring System (IPSS) low and intermediate-1) patients or higher-risk (IPSS intermediate-2 and high) patients. Intermediate-2 risk patients have a median overall survival of 1.2 years; the median overall survival for high-risk patients is 0.4 years9.

Therapeutic options for MDS are limited. Conventional care regimens include best supportive care (BSC), low-dose chemotherapy, and standard chemotherapy, but have not shown a survival benefit10–12. Allogeneic stem cell transplants are potentially curative in younger patients, but have a significant risk of mortality1. However, treatment with hypomethylating agents has been shown to improve outcomes in MDS patients13.

In the US, two different hypomethylating agents, azacitidine and decitabine, are approved to treat all French-American-British (FAB) sub-types (refractory anemia, refractory anemia with ringed sideroblasts, refractory anemia with excess blasts, refractory anemia with excess blasts in transformation, and chronic myelomonocytic leukemia) and intermediate-1, intermediate-2, and high-risk IPSS groups of MDS. National Comprehensive Cancer Network (NCCN) guidelines recommend azacitidine or decitabine to treat MDS patients, with azacitidine being the first preferred treatment for intermediate-2 or high-risk patients2.

A randomized trial13 comparing decitabine plus supportive care to supportive care alone (with supportive care defined as the use of red blood cell (RBC) and platelet transfusions for patients where clinically indicated) found decitabine-treated patients had a significantly higher overall response rate (p < 0.001). The overall response rate consisted of a complete response (defined as normalization of peripheral counts and bone marrow for at least 8 weeks with serial bone marrow blasts <5% without dysplastic changes, a neutrophil count of ≥1.5 × 10⁹/L, a platelet count of ≥100 × 10⁹/L, and hemoglobin level >11 g/dL), and a partial response (defined similarly to the complete response except a reduction of ≥50% of blasts that remained above 5%, or a reduction in MDS severity as per the FAB categories)13. Compared with supportive care, decitabine did not significantly delay median time to AML or death (p = 0.16)13.

Azacitidine, meanwhile, has been shown to be effective in delaying progression to AML or death. A trial comparing patients randomized to azacitidine or conventional care (the conventional care regimen was preselected by investigator and could be BSC, low-dose cytarabine, or intensive chemotherapy) found azacitidine-treated patients had significantly longer median survival (24.5 months vs 15.0 months for the conventional care group, p < 0.0001) and significantly longer median time to progression to AML (17.8 months vs 11.5 months for the conventional care group, p < 0.0001)14. Azacitidine is the first drug to demonstrate a significant overall survival benefit in higher-risk MDS patients14.

While the clinical benefits of azacitidine have been demonstrated, no study to date has evaluated its cost-effectiveness relative to decitabine. The objective of this study was to assess the cost-effectiveness of azacitidine vs decitabine using a decision-analytic model involving hypothetical cohorts of MDS patients from the perspective of a US third-party payer.

Design and methods

Overview

Cost-effectiveness models are used to assess the relative value of two or more treatment strategies. A Markov model is a specific type of cost-effectiveness model which simulates changes in health states (for example, disease progression) over time. In a Markov model, a patient or a cohort of patients exists in one of a set of mutually exclusive health states, including death. Over time, patients either remain in a health state or move among health states at regular intervals (e.g., monthly cycles) based on transition probabilities (e.g., monthly risk of disease progression or death). Transition probabilities can be static (Markov chain) or change over time (Markov process). Patients in each state are assigned a cost value and a health-effect value to reflect the economic and health experience of spending time in that condition. The overall expected value of each treatment strategy is calculated by multiplying the number of patients in each state by the value of being in that state, and summing these across states and then across all cycles. Incremental differences in economic and health outcomes between treatment strategies are determined and incremental cost-effectiveness ratios (ICERs) are calculated, using change in cost as the numerator and change in health-effect as the denominator.

Model structure

The authors developed a Markov-process model with 1-month cycles using TreeAge Pro Suite 2009 (TreeAge Software, Inc., Williamstown, MA) decision-analysis software. The model structure was identical across the two treatment arms (azacitidine or decitabine). During each model cycle, patients could either remain in or transition among four health states: State 1: MDS and Transfusion Independence; State 2: MDS and Transfusion Dependence; State 3: Progression to AML; State 4: Death. Patients who advanced to AML either could remain in that state or progress to death (Figures 1 and 2).

Figure 1.  Markov health states in the model. This figure depicts the health states among which patients transition in the model. Patients can begin the model in a state of MDS with transfusion independence or MDS with transfusion dependence. They can either stay in that state (indicated by the recursive arrow) or move on to another health state. A double-sided arrow indicates that patients can move back and forth between health states. For example, a patient can start the model as MDS with transfusion independence, become sicker and transition to MDS with transfusion dependence and, at a later point, become healthier and transition back to MDS with transfusion independence. A one-sided arrow indicates that patients can only move from one health state to the next. In this case, patients can only move to AML; they cannot transition from AML to MDS. Once patients have AML, they can continue to exist in that health state, or become sicker and die. Naturally, patients cannot move out of death; it is an absorbing state.

Figure 2.  Decision-tree diagram of model structure. This figure depicts the pathways that patients may take throughout the model. The square box indicates a decision node, meaning the model will compare both azacitidine and decitabine-treated patients. The circle with an M indicates a Markov node and signifies that patients can transition among the four health states of interest; each of which has its own branch in the decision tree. The health states are mutually exclusive and collectively exhaustive; in each model cycle, the probability of being in any one of the four states must add up to 100%. While the pathways are identical for patients treated with azacitidine and decitabine, the probabilities of being in the four health states will differ across the two treatments. An empty circle denotes a chance note, meaning that patients can transition from the health state to the left of the circle to any one of the health states to the right of the circle. The text to the right of the branches indicates the health state in which the patient begins the next model cycle.

Health outputs generated by the model include: (1) quality-adjusted life years (QALYs); (2) life years (LYs); (3) number of patient-months with transfusion independence; and (4) number (and percentage) of patients progressing to AML. The primary economic output was treatment cost. The ICERs for azacitidine vs decitabine were calculated as: (1) cost per QALY gained; (2) cost per LY saved; (3) cost per additional patient-month of transfusion independence; and (4) cost per additional case of AML progression avoided.

The model used a US third-party payer perspective, which only considers direct medical care costs. All estimated costs were adjusted to 2009 US dollars using the Medical Care Services component of the Consumer Price Index15. The model was run for a time horizon of 2 years, with LYs, QALYs, and costs discounted at 3% annually16.

Patient population

Patients entering the model replicate the demographics of patients enrolled in Phase III clinical trials for azacitidine and decitabine. The hypothetical cohort of patients in the azacitidine arm of the model had a median age of 69 years, was predominantly male (74%), and almost exclusively had higher-risk MDS (89%)14. While the hypothetical cohort of patients in the decitabine arm had a similar median age of 70 years, and was also predominantly male (66%), it included both lower- (31%) and higher-risk (69%) MDS patients13.

Costs

The model considered the direct medical costs of treatment, including drug acquisition, administration and monitoring, RBC transfusions, iron chelation treatment, and treatment of patients who progressed to AML (Table 1). Costs of azacitidine and decitabine were based on 2009 wholesale acquisition costs (WAC). Dosing schedules were obtained from product labels at the time of analysis (January 2010)17,18. Drug administration costs were based on method of drug delivery, with azacitidine administered through subcutaneous injection and decitabine administered through intravenous infusion, to reflect the mode of drug administration used in clinical trials13,14. Per the product label, patients treated with azacitidine should have liver function and serum creatinine monitored prior to each treatment cycle17; costs of these tests were included in the azacitidine treatment cost estimates. Costs of drug administration and tests reflect 2009 Medicare reimbursement rates19,20.

Table 1.  Model inputs and parameters.

Parameters	Azacitidine	Decitabine	Sources
Costs
Drug acquisition (monthly costs based on treatment cycle of 75 mg/m^2/day for 7 days, every 4 weeks for azacitidine and 15 mg/m^2/day every 8 hour for 3 days, every 6 weeks for decitabine)	$6695	$9243	WAC provided by manufacturer. Azacitidine WAC: $478.23/100 mg Decitabine WAC: $1422.00/50 mg Assumes patient body surface area of 1.7 m^2 and no vial splitting
Cost of administration (monthly costs based on treatment cycle of 1 hour for 7 days, every 4 weeks for azacitidine and 3 hour every 8 hour for 3 days, every 6 weeks for decitabine)	$472	$1395	2009 Physician Fee Schedule, CMS, HCPC codes 96401, 96413, and 96415
Liver function tests and serum creatinine tests (frequency: once per treatment cycle)	$26	—	2009 Clinical Diagnostic Lab Fee Schedule, CMS
Transfusion of blood or blood components	$252	$252	Goss et al.21
Transfusion related lab tests (compatibility tests, irradiation of blood)	$94	$94	Goss et al.21
Pooling of platelets or other blood	$23	$23	Goss et al.21
RBC (2 units)	$236	$236	Goss et al.21
Physician office visit	$97	$97	Goss et al.21
Complete blood count	$13	$13	Goss et al.21
Comprehensive metabolic panel	$18	$18	Goss et al.21
Iron chelation	$5168	$5168	Dosing regimen of deferasirox: Deferasirox Product Label. Dosing Regimen of deferoxamine: National Comprehensive Cancer Network, 2010. Market share of patients on deferasirox and deferoxamine: Greenberg et al.24. Drug Cost: RED BOOK23. Cost of subcutaneous administration: 2009 Physician Fee Schedule, CMS.
Treatment of AML	$5633	$5633	Menzin et al.25
Utilities
Transfusion Independent	0.91	0.91	Goss et al.21
Transfusion Dependent	0.5	0.5	Goss et al.21
AML	0.524	0.524	Dalziel et al.26
Probabilities
Transfusion Independence (initial probability)	0.634	0.74	Azacitidine: Fenaux et al.14 Decitabine: Kantarjian et al.13
Transfusion Dependence (initial probability)	0.366	0.26	Azacitidine: Fenaux et al.14 Decitabine: Kantarjian et al.13
Transfusion Independent to Transfusion Independent	0.992	0.992	Azacitidine: Fenaux et al.14 Decitabine: Due to lack of data, assumed to be the same as azacitidine
Transfusion Independent to Transfusion Dependent	0.008	0.008	Azacitidine: Fenaux et al.14 Decitabine: Due to lack of data, assumed to be the same as azacitidine
Transfusion Dependent to Transfusion Dependent	0.9706	0.9514	Azacitidine: Fenaux et al.14 Decitabine: Kantarjian et al.13
Transfusion Dependent to Transfusion Independent	0.0294	0.0486	Azacitidine: Fenaux et al.14 Decitabine: Kantarjian et al.13
Iron chelation	1.000 (after 12 months and every month thereafter)	1.000 (after 12 months and every month thereafter)	National Comprehensive Cancer Network, 2010
Progression to AML	0.007	0.0173	Azacitidine: Fenaux et al.14; Celgene Corporation, 2009 Decitabine: Kantarjian et al.13
Death*	Survival curve	Survival curve	Azacitidine: Fenaux et al.14 Decitabine: Kantarjian et al.13

WAC, wholesale acquisition cost; CMS, Centers for Medicare & Medicaid Services; HCPC, healthcare common procedure code; RBC, red blood cell; MDS, myelodysplastic syndromes; AML, acute myeloid leukemia. *Survival curve data, probabilities change per model cycle.

In accordance with NCCN guidelines, all patients who receive 24 units of blood transfusions require iron chelation therapy2. Therefore, treatment costs also included the costs of blood transfusions and iron chelation therapy, if needed. Components of blood transfusion costs included cost of RBC units, cost of transfusion of blood or blood components, transfusion-related laboratory tests including compatibility tests, and irradiation of blood, pooling of platelets or other blood, a physician office visit, a complete blood count, and a comprehensive metabolic panel21. Iron chelation costs were calculated as a weighted average, assuming 90% of patients were treated with deferasirox, an oral iron chelator, and 10% of patients used deferoxamine, an infused iron chelator2,22–24. If patients progressed to AML, their treatment costs included the cost of standard chemotherapy, as well as cost of inpatient stays, outpatient visits, use of skilled nursing facility services, home healthcare, and hospice use25. Patients who progressed to AML did not incur cost of treatment with azacitidine or decitabine.

Utilities

A utility is a measurement that incorporates HRQoL associated with a certain health state with the preference for that health state. In essence, such measurement provides a common scale to compare health-related impact of different diseases. By convention, a utility value of 1.0 connotes perfect health, while a utility state of 0.0 indicates death. Using utility values and information about years of life lived, one can calculate a QALY.

MDS-related utilities were obtained from a study involving face-to-face interviews of MDS patients in the US21. Responses were elicited using a time trade-off method. Respondents were queried on dimensions of time spent in blood transfusions, impaired normal functioning due to fatigue or disease, worry about future due to health condition, discomfort due to health condition and its treatment, reliance on others, feelings of being a burden, and feeling sad or helpless due to the health condition. AML-related utility was obtained from a study of chronic myeloid leukemia (CML) patients completing the EQ-5D26; the AML utility used in the model (0.524) was the mean value for patients in the blast phase of CML, which is considered to be similar to AML (Table 1)27.

Health state transition: probability of progression to AML

The likelihood of progressing to AML was assessed using data from Phase III clinical trials of azacitidine and decitabine14,28. The rate of azacitidine-treated patients progressing to AML was 12.8% over 20.05 months14,29. This rate was transformed into a monthly probability of 0.00730 (Table 1). Data from Phase III trials for decitabine indicates that the rate of progression to AML is 27% over 18 months27; therefore the monthly probability of progressing to AML is 0.0173 for the decitabine arm of the model.

Probability of death

The probability of death is also drug-specific and was derived from Kaplan-Meier survival curves reported in Phase III clinical trial publications13,14. In instances where the simulation time horizon of the model exceeded the time frame of published survival curves, the last reported value of death was applied. The probability of death was assumed to be the same for patients dying from MDS or AML, in accordance with how the underlying clinical trial data were reported.

Probability of transfusion dependence and transfusion independence

The probability of beginning the model simulation in a state of transfusion dependence was obtained from published literature13,14; the probability of initially being transfusion independent was calculated as its complement (Table 1). As the model progresses, patients can become healthier and no longer require blood transfusions; the monthly probability of this transition was calculated based on transfusion dependence to independence rates reported in the individual clinical trials13,14. The complements of these probabilities were the probabilities of remaining transfusion dependent. Also, in accordance with azacitidine clinical trial data, patients could become sicker and transition from transfusion independence to transfusion dependence14,17. Due to lack of data for decitabine, the monthly probability of transitioning from transfusion independence to transfusion dependence for decitabine was assumed to be the same as that for azacitidine. The complements of these probabilities were the probabilities of staying transfusion independent. The probabilities were readjusted for each model cycle according to the increasing probability of death, so that, in each model cycle, the probability of being in any of the health states equals 1.0.

Model assumptions

Patients requiring blood transfusions were assumed to receive two units of RBCs per transfusion; platelet transfusions were not included in the model. All patients receiving more than 12 transfusions (greater than 24 units of RBCs) were assumed to require iron chelation therapy2. Drug cost inputs were calculated based on an average patient body surface area of 1.7 m², derived from the mean patient body surface area reported in the azacitidine clinical trial. Consistent with the on-label use of azacitidine and decitabine, vial splitting was prohibited17,18. Adverse event rates were assumed to be the same across the two treatment arms.

Sensitivity analyses

Sensitivity analyses were performed on all key model parameters, including the time horizon, costs, and transition probabilities. Acquisition and administration costs of decitabine were varied to reflect an alternative approved dosing regimen of 20 mg/m² of drug administered intravenously for 1 hour daily for 5 days every 4 weeks31. Sensitivity analyses also were run to examine results when the assumption of medication wastage (i.e., “no vial splitting”) was violated.

Results

Clinical outcomes

Results demonstrate that azacitidine conferred greater clinical benefit than decitabine in four key outcomes (Table 2). Azacitidine-treated patients experienced a survival benefit of 0.22 additional LYs compared with decitabine-treated patients (1.51 vs 1.29). The total number of QALYs over 2 years for azacitidine-treated patients exceeded those for decitabine-treated patients (1.04 vs 0.87). The total number of patient-months with transfusion independence was also higher for azacitidine vs decitabine-treated patients (8.33 vs 6.22). Decitabine-treated patients were more likely to progress to AML at the end of the 2-year model simulation (71.5% vs 49.1%).

Table 2.  Economic and clinical results.

	Azacitidine	Decitabine	Difference
Total costs	$150,322	$166,212	−$15,890
QALYs	1.041	0.870	0.171
LYs	1.512	1.292	0.220
Months of transfusion independence	8.328	6.224	2.105
Percentage of patients with AML progression	49.1%	71.5%	−22.4%

QALYs, quality-adjusted life years; LYs, life years; AML, acute myeloid leukemia.

Economic outcomes

Azacitidine-treated patients incurred fewer costs than decitabine-treated patients. Over 2 years, total per-patient costs were $150,322 for azacitidine vs $166,212 for decitabine (Table 2).

Cost-effectiveness

Model results indicate that azacitidine was cost-saving compared with decitabine across all key outcomes. Treatment over 2 years with azacitidine cost $15,890 less than decitabine, and conferred a clinical benefit of 0.171 additional QALYs (Table 2). ICERs demonstrate that azacitidine dominated decitabine; meaning azacitidine was less expensive and provided greater clinical benefit.

Sensitivity analyses

Model ICER results were most sensitive to variations in time horizon, costs of decitabine and azacitidine, and cost of AML treatment. However, the conclusion that azacitidine was cost saving compared with decitabine was robust to variations in ICERs. Azacitidine was cost-saving compared with decitabine across variations in all parameters, except when a 5-day dosing schedule was used for decitabine and when the assumption of medication wastage was violated (i.e., vial splitting did occur) (Table 3). When a 5-day dosing schedule was used for decitabine, an ICER of $121,152 per QALY was obtained. Assuming that vial splitting occurred resulted in an ICER of $15,528 per QALY.

Table 3.  Results of sensitivity analyses.

Analysis		ICER	Conclusion
Base-case		−$92,901	Azacitidine dominant*
5-day dosing schedule, decitabine		$121,152	Azacitidine not cost-effective
Initial probabilities: 36.6% Transfusion independent, 63.4% Transfusion dependent		−$102,211	Azacitidine dominant*
Initial probabilities: 26% Transfusion independent, 74% Transfusion dependent		−$101,276	Azacitidine dominant*
Initial probabilities: 50% Transfusion independent, 50% Transfusion dependent		−$103,082	Azacitidine dominant*
Vial splitting		$15,528	Azacitidine cost-effective
	ICER, increase	ICER, decrease
Transfusion dependence to independence, azacitidine, cycle probability ±10%	−$93,178	−$92,538	Azacitidine dominant in both scenarios*
Transfusion dependence to independence, decitabine, cycle probability ±10%	−$94,889	−$91,737	Azacitidine dominant in both scenarios*
Transfusion independence to dependence, azacitidine, cycle probability ±10%	−$92,419	−$93,379	Azacitidine dominant in both scenarios*
Transfusion independence to dependence, decitabine cycle probability ±25%	−$93,818	−$91,989	Azacitidine dominant in both scenarios*
Total drug cost, azacitidine, ±10%	−$22,056	−$163,747	Azacitidine dominant in both scenarios*
Total drug cost, decitabine, ±10%	−$174,506	−$11,297	Azacitidine dominant in both scenarios*
Cost of transfusion ±25%	−$91,111	−$94,692	Azacitidine dominant in both scenarios*
Cost of iron chelation ±25%	−$91,111	−$94,692	Azacitidine dominant in both scenarios*
Cost of AML treatment ±25%	−$101,880	−$83,926	Azacitidine dominant in both scenarios*
Time horizon: ±12 months	−$28,638	−$481,446	Azacitidine dominant in both scenarios*

QALY, quality-adjusted life year; AML, acute myeloid leukemia; *Azacitidine costs less and is more effective relative to decitabine.

Discussion

The results of the model and analyses shown here present evidence that azacitidine’s clinical effectiveness translates into favorable cost-effectiveness compared to decitabine. Model results indicate that, in addition to being associated with longer survival, greater number of QALYs saved, more months of transfusion independence, and fewer patients progressing to AML, azacitidine costs less than decitabine. Therefore, when evaluating standard treatment regimens across a time horizon of 2 years, azacitidine is cost-saving compared with decitabine in the treatment of MDS. Overall, these results show that azacitidine confers greater clinical benefit at a lower cost than decitabine, indicating that azacitidine dominates decitabine.

Any cost-effectiveness model may be evaluated on the strength of its inputs and whether modifying them varies the results that it produces. In this case, the sensitivity analyses that were conducted focused on the various issues that could possibly affect results, especially those related to the cost difference between the two treatment arms. Given that the same results were reproduced with two exceptions, this analysis suggests that this conclusion is generally robust: azacitidine dominated conventional decitabine regimens in terms of clinical and economic value.

As there are no head-to-head trials comparing azacitidine with decitabine, data from Phase III clinical trials comparing each drug with other care regimens were used in this analysis. Inputs for the azacitidine treatment arm were taken primarily from a Phase III clinical trial14 comparing azacitidine with conventional care (BSC, low-dose cytarabine, or intensive chemotherapy), while inputs for the decitabine arm were taken primarily from a Phase III clinical trial13 comparing decitabine plus BSC to BSC alone. The lack of head-to-head comparisons of azacitidine to decitabine necessitated this approach. The authors feel this is an example of the need for clinical trials to compare one active treatment to another active treatment, rather than compare an active treatment to placebo or conventional care. As decisions made in clinical practice generally require choosing between one of multiple active treatments, clinical trials would do well to include clinically relevant comparators. The same key outcomes were collected in both trials, which facilitated the development of the model. However, the trials had somewhat different patient populations and different follow-up periods (except for assessment of mortality). Approximately 30% of patients in the decitabine trial were lower-risk, compared to roughly 10% of the azacitidine-trial patients; more patients in the azacitidine trial were classified as having higher-risk MDS. It was not possible to disaggregate outcomes in each of the trials by sub-cohorts defined by risk. It is hypothesized that if more lower-risk subjects were included in the azacitidine trial, thereby matching the cohort in the decitabine trial, results may have been even more favorable for azacitidine. However, as noted above, the true comparative efficacy of the two treatments can best be determined through head-to-head comparative studies.

The issue of different units of time for the rates of reported clinical events, other than death, relates to the transition probabilities. The probability of death was not affected by the trial time horizons as these data were reported on survival curves. However, because the decitabine trial had a shorter follow-up period, and data on other event rates were transformed into monthly probabilities for use in the model, the monthly transition probabilities tended to be higher for the decitabine-treated arm than the azacitidine-treated arm. Thus, decitabine-treated patients had a higher likelihood of progressing to AML, and a higher likelihood of transitioning from transfusion dependence to transfusion independence. That said, these probabilities were adjusted in the sensitivity analysis without changing the overall results. In a similar manner, on the issue of transfusion dependence, it is important to note that the higher decitabine transition probability coincided with the decitabine trial patients requiring more transfusions (which in turn helped contribute to the greater treatment costs among those patients). In this case, initial transfusion status was varied in sensitivity analyses such that either the initial transfusion status of azacitidine-treated patients duplicated that of the decitabine clinical trial data, or that 50% of azacitidine-treated patients and 50% of decitabine-treated patients began the model in a state of transfusion dependence. Again, in these scenarios, the azacitidine regimen was still cost-saving compared with decitabine.

Lack of data was a challenge in other areas. The probability of transitioning from transfusion independence to transfusion dependence was not available for decitabine, so data from the azacitidine clinical trial was applied to both azacitidine and decitabine model cohorts. Additionally, because attributable-mortality was not reported in the clinical trials, probability of death from AML was equivalent to the probability of death from MDS. While these challenges were mitigated with sensitivity analyses and application of best modeling practices, the authors recognize them and recommend that future clinical studies be conducted to overcome them and supply the relevant data. Additionally, more information on use of platelet transfusions and their impact would be valuable. As more studies are conducted of azacitidine and decitabine, mixed treatment comparison analyses of the relative clinical effect of each drug should also be performed.

A cost-effectiveness model should present the best available data given current treatment patterns at the time. This acknowledges, however, that treatment patterns are rarely static; they tend to change over time given advances in physician knowledge and available technologies and processes of care. This analysis was conducted of the current standard regimens employing azacitidine and decitabine in the US. Accordingly, costs of administration were substantially higher for intravenous injection of decitabine compared to subcutaneous injection of azacitidine. The FDA-approved dosing schedule for decitabine was 15 mg/m² via a continuous intravenous infusion for a period of 3 hour every 8 hour for 3 days every 6 weeks18. Recently, decitabine was approved for a second dosing schedule of 20 mg/m² intravenously over 1 h once daily for 5 days every 4 weeks30. It will be important to determine whether this regimen becomes more popular over time than the original dosing regimen. Results of the sensitivity analysis using the recently approved dosing schedule indicate that azacitidine may not be cost-effective compared to decitabine at this dosing schedule (ICER = $121,152).

Another issue with respect to practice patterns is the possibility of vial splitting. Both azacitidine and decitabine are available as single-use vials; on-label use indicates that any opened and unused medication must be discarded. Therefore, the base-case analysis was conducted assuming no vial splitting. Although the authors have no quantitative data to report on this issue, they have heard qualitatively that vial splitting may occur. Therefore another sensitivity analysis was conducted in which vial splitting occurs. The results of that analysis suggested that azacitidine no longer dominates, but is cost-effective compared to decitabine. Additional studies of practice patterns should yield data that will inform future versions of this model.

Other studies are also recommended to inform this comparison of azacitidine and decitabine. For example, it would be valuable to take the societal perspective rather than the perspective of the payer focused only on direct medical costs. Expanding the model and analysis to the societal perspective would include, for example, issues such as unpaid caregiver time and time lost from work (for patients or caregivers). One could hypothesize that an analysis of indirect costs of the treatment regimens evaluated here would further show the advantages of azacitidine, due to differences in treatment administration.

Additionally, the authors recommend the collection of more robust data regarding patient utility for MDS and AML. Neither clinical trials examined patient-reported outcomes; therefore, utility values were obtained from the only available published study. Utilities for AML were taken from a study of patients in the blast phase of CML26, while utility data used for the MDS state were taken from a small study of eight subjects21. A more representative and larger sample would improve the utility inputs.

Conclusions

Using the best available published data, these results suggest that azacitidine’s clinical effectiveness translates into favorable cost-effectiveness, and, in particular, cost-savings. Use of azacitidine vs decitabine is associated with longer survival, more QALYs gained, more months of transfusion independence, and fewer patients progressing to AML, as well as less cost.

Transparency

Declaration of funding

Funding for this study was provided by Celgene Corporation, Summit, NJ, which manufactures azacitidine. Risha Gidwani was the principal investigator and takes primary responsibility for the paper. Chris Pashos, Zeba Khan, and Risha Gidwani conceptualized the project and wrote the manuscript. C. L. Beach and Pierre Fenaux provided useful inputs into conceptualization and into drafts of the manuscript.

Declaration of financial/other relationships

At the time of the study, Risha Gidwani was an employee of United BioSource Corporation. Chris Pashos is still an employee of United BioSource Corporation, which received research funding from Celgene Corporation. Zeba M. Khan and C. L. Beach are employees of Celgene Corporation. Pierre Fenaux has received consultancy fees, honoraria, and research funding from Celgene Corporation.

Acknowledgments

The authors extend their appreciation to Janet Dooley for her administrative assistance.