Economic Analysis of Decitabine Versus Best SupportiveCare in the Treatment of Intermediate- and High-RiskMyelodysplastic Syndromes From a US Payer Perspective
Feng Pan, PhD, MA1; Siyang Peng, MS1; Rachael Fleurence, PhD, MBA1;ohn E. Linnehan, MPH, MBA1; Kevin Knopf, MD, MPH2; and Edward Kim, MD, MBA3
1United BioSource Corporation, Bethesda, Maryland; 2California Pacific Medical Center, San Francisco,California; and 3Eisai, Inc., Woodcliff Lake, New Jersey
are blood and bone marrow disorders that occur pri-marily in the elderly population, with 30% of all casesprogressing to acute myeloid leukemia (AML). Redblood cell transfusions—a conventional treatment ofMDS—have been associated with high costs and de-creased quality of life compared with transfusion inde-pendence. Phase III clinical trial data suggest that de-citabine may offer an improved AML-free survivalversus best supportive care (BSC), which consists of redblood cell transfusions, deferoxamine, erythropoiesis-stimulating agents, platelet transfusions, and colony-stimulating factors. The US Food and Drug Adminis-tration has approved a 5-day outpatient decitabinedosing regimen, which might reduce administrationcosts compared with the standard 3-day inpatientregimen.
Objective: The aim of this study was to assess thecost-effectiveness of 5-day dosing of decitabine versusBSC in US patients with intermediate- and high-riskMDS from a US payer perspective.
Methods: A Markov model with 3 health states(MDS, AML, and death) was constructed to simulatenatural disease progression. The model followed pa-tients in 4-week cycles for �5 years. Clinical inputs andpatient characteristics were based on decitabine PhaseIII clinical trial data. Costs of supportive care and ad-verse events were based on trial resource utilizationdata. Drug and AML costs were obtained from pub-lished sources. Deterministic and probabilistic sensitiv-ity analyses were performed to determine the impact ofmodel parameters on results.
Results: In the base-case model, decitabine yielded0.276 additional year of AML-free survival and 0.052more quality-adjusted life-year (QALY) compared
with BSC. Total decitabine and administration costs
2444
over the 5-year time horizon were $28,933. Total di-rect medical costs were $122,940 in the decitabine armand $122,666 in the BSC arm. The incremental cost-effectiveness ratio for decitabine versus BSC was$5277 per QALY gained. Sensitivity analyses indicatedthat decitabine had a higher probability than BSC ofbeing cost-effective despite the uncertainty aroundsome model parameters, including survival.
Conclusion: In this study, decitabine administeredon a 5-day dosing schedule was likely a cost-effectivetreatment option in patients with intermediate- andhigh-risk MDS from a US payer perspective. (Clin
INTRODUCTIONThe term myelodysplastic syndromes (MDS) refers to aheterogeneous group of disorders, all with characteris-tic abnormalities of peripheral blood and bone marrowmorphology and impaired bone marrow function.1 Pa-tients with MDS typically present with symptoms ofanemia, bacterial infections, bleeding or bruising, andfatigue.2–4 MDS are primarily diseases of the elderly;in the United States, �72% of cases occur in patientsaged �70 years, with an incidence of newly diagnosedcases of �20 per 100,000 people.5 Although the clini-al courses of MDS are variable, median survival is 20
The data in this paper were previously presented at the American Society ofClinical Oncology Annual Meeting, June 4–8, 2010, Chicago, Illinois.
Accepted for publication November 4, 2010.
Express Track online publication December 6, 2010.doi:10.1016/j.clinthera.2010.12.0030149-2918/$ - see front matter
months, with about two thirds of deaths due to bonemarrow failure and the remainder due to unrelatedcauses.2 MDS represent a significant economic burdeno health care systems. One recent analysis of datarom a claims database representing 10 million geo-raphically diverse individuals found that the cost ofDS in the United States was $51,066 per patient per
ear.6
Although MDS have historically not been defined asmalignancies, they result from the clonal expansion ofa hematopoietic progenitor and progress to acute my-eloid leukemia (AML) in �30% of patients.5,7 AML is
particularly severe disease in elderly patients. Che-otherapy, the standard treatment of AML, achieves
ower response rates in patients �70 years of age, whoypically have more comorbidities to complicate treat-ent and are less able to tolerate adverse events
AEs).8,9 Patients with AML that develops from MDShave a poorer prognosis than do those with de novoAML. A retrospective analysis of data from 3439 el-derly patients with AML in the Surveillance Epidemi-ology and End Results (SEER)-Medicare database in-dicated a median survival across all patients of 2.4months, with �7% of patients with 2-year survival.8
Treatment of MDS typically includes best support-ive care (BSC), which consists of red blood celltransfusions, deferoxamine, erythropoiesis-stimulat-ing agents, platelet transfusions, and colony-stimulat-ing factors.6 Most patients with MDS are transfusionependent, a characteristic that has been associatedith significantly shorter overall survival than transfu-
ion independence.10 In addition, transfusion depen-ence has been associated with a reduced quality of
ife3 and incremental costs of �$31,000 per patient perear over transfusion independence.6
With traditional supportive treatments for MDScarrying a significant burden in terms of shorter overallsurvival, high costs, and decreased quality of remain-ing life, novel treatments with the capacity to reducetransfusion dependence and prolong time to AMLhave the potential to reduce costs and improve health-related quality of life. One such treatment is decitabine,one of a few US Food and Drug Administration�ap-proved agents for the treatment of MDS. Decitabine isa cytosine nucleoside analogue that, through hypo-methylation of DNA promoter regions, can activatetumor-suppressor genes, thus slowing the progressionof the disease. Decitabine has demonstrated activity in
a broad array of hematologic disorders, including
December 2010
MDS and AML. In a Phase III randomized clinical trialof decitabine versus BSC (N � 170), decitabine wasassociated with a significantly higher overall responserate (30% vs 7%, respectively) and had a trend towardimproved time to AML progression or death comparedwith BSC (12.1 vs 7.8 months; P � NS by the log-ranktest).11 Decitabine was initially approved with a
-day intravenous dosing schedule,12 which necessi-ated hospitalization. However, a revised 5-day out-atient dosing schedule was approved in 201013 thatay reduce the high administration costs associatedith hospitalization.Given the possible benefits of an MDS treatment
ith the potential to increase AML-free survival andecrease transfusion dependence, decitabine with a-day dosing schedule has the potential to be cost-ef-ective. The present study reports findings from a
arkov model that assessed the cost-effectiveness of-day dosing of decitabine versus BSC in patients withntermediate- and high-risk MDS from a US payer per-pective. BSC was the comparator in this analysis, aspposed to other novel MDS treatments, because het-rogeneity in patient populations and clinical trial de-igns for these treatments prevents accurate compari-ons of treatment efficacy. From a clinical perspective,SC continues to represent the most commonly usedreatment for MDS.14
PATIENTS AND METHODSPatient Population
The patient population in the model was matched tothe population characteristics observed in the Phase IIIdecitabine randomized controlled clinical trial.11 Pa-tients in the trial were adults aged �18 years with aconfirmed diagnosis of MDS who fit any of the recog-nized French-American-British classifications1 and hadan International Prognostic Scoring System (IPSS)score �0.5 (ie, Intermediate Risk 1, Intermediate Risk2, and High Risk). Diagnosis of MDS could be eitherde novo or secondary (arising as a toxicity of chemo-therapy or radiation treatment of another disease). Pa-tients with a diagnosis of AML or other progressivemalignant disease were excluded from the trial andhence were not considered in the model.
Model DescriptionThe model used in the analysis was a Markov model,
programmed in Excel (Microsoft Corporation, Red-
mond, Washington), that simulated patients’ natural pro-
2445
uiPamhcttos
tempoch
smaptcc
avsepeaf
Clinical Therapeutics
gression from MDS to AML to death (Figure 1). As such,these 3 stages represented the 3 distinct model healthstates. Patients’ transitions between these health stateswere determined using data on survival without AMLand overall survival from the Phase III decitabine trial.In the model, patients with MDS remained in the MDSstate until they progressed to AML or died. Patientswho progressed to AML may have died in any cycle.Patients who died remained in the death health statefor the duration of the model.
With abnormal peripheral blood cell morphologyand bone marrow function, many patients with MDShave hematopoietic disorders, such as anemia, throm-bocytopenia, and neutropenia, thus requiring bloodproduct transfusions.6 Because transfusion depen-dence has been reported to be a significant cost driverand to negatively impact patients’ quality of life, themodel differentiates transfusion-independent andtransfusion-dependent patients within the MDS healthstate. Similar to the 3 primary health states, patientscould switch between these transfusion status sub-states based on their response to treatment. All patientswho showed a complete treatment response, partialtreatment response, or hematologic improvement wereconsidered responders and transfusion independent.Therefore, patients who were transfusion dependent at
Patients with MDS
AML Death
MDSTransfusion
free/dependent
Figure 1. Markov model health states and tran-sitions in a cost-effectiveness analysisof decitabine versus best supportivecare in the treatment of myelodysplas-tic syndromes (MDS). Patients in theMDS state are classified according tored blood cell transfusion dependency.AML � acute myeloid leukemia.
baseline and responded to treatment became transfu- w
2446
sion independent in the model. Patients who weretransfusion dependent at baseline and did not respondto treatment remained transfusion dependent untilprogression. The model assumed that all patients whowere transfusion independent at baseline remained assuch until progression to AML.
The model compared decitabine to BSC in a 5-dayoutpatient dosing regimen. Patients in both the decit-abine and BSC treatment arms received standard med-ical care for MDS. In addition, patients in the decit-abine arm received decitabine intravenous infusion at adose of 20 mg/m2 over 1 hour, every day for 5 consec-tive days, in 4-week cycles. Compared with the 3-daynpatient decitabine dosing regimen assessed in thehase III trial, 5-day outpatient decitabine dosing wasssociated with favorable overall improvement in aulticenter single-arm study.15 Because the 2 regimensave not been compared head-to-head in a randomizedlinical trial, the model assumed that the 5-day outpa-ient regimen had the same efficacy as the 3-day inpa-ient dosing regimen presented in the trial. The impactf using efficacy results from the 5-day outpatienttudy was investigated in the sensitivity analysis.
Each model cycle was 4 weeks, to reflect the decitabinereatment cycle and accurately account for risks of keyvents. The model followed patients for up to 5 years, asost patients with MDS die within this period. A USayer/managed care perspective was taken; therefore,nly direct medical costs (ie, physician costs, hospitalosts, and medication costs) were considered. Cost andealth effects were discounted at 3% per annum.
Each model health state was associated with repre-entative costs, as well as utility values needed to esti-ate quality-adjusted life-years (QALYs). These costs
nd QALYs were summed over the model time horizoner comparator arm, allowing the determination ofotal costs and total QALYs as well as incrementalost-effectiveness ratios (ICERs), so that comparisonsould be made between treatments.
Before final results were determined, the structurend programming of the completed Excel model werealidated by a modeling expert not involved in thetudy, and a variety of stress tests were performed tonsure that the model results were reflective of the in-uts entered. For example, both extreme values andqual values across treatment arms were input, andctual results were compared against results expectedrom a properly functioning model. In situations in
hich actual results diverged from expected results,
Volume 32 Number 14
fichw(s
eAtTwqAst
F. Pan et al.
debugging was performed to investigate and remedythe discrepancy.
Clinical InputsPatients with MDS may die as a direct result of
MDS/bone marrow failure or progress to AML, whichis almost always fatal. Hence, survival without AMLand overall survival were the key clinical inputs. Themodel used survival data directly from the intent-to-treat population from the Phase III decitabine trial toinform these inputs. Weibull survival functions were
tted to the AML-free survival and overall survivalurves of the BSC arm of the trial. Cox proportionalazard ratios (HRs) were estimated for treatmentith decitabine versus BSC for AML-free survival
ure 2 presents the projected overall survival andAML-free survival with decitabine and BSC.
Baseline transfusion dependency status for both re-sponders and nonresponders was obtained from thePhase III decitabine trial.
Utility InputsHealth state utilities are presented in Table I. Utility
values are represented on a scale of 0 to 1, with a utility
0
Prop
ortio
n Su
rviv
ing
6724482240
D
1.000
0.900
0.800
0.700
0.600
0.500
0.400
0.300
0.200
0.100
784560336112
Figure 2. Projected overall and acute myeloid leukemiabest supportive care [BSC]) in the treatment ofitted to the AML-free survival and overall survproportional hazard ratios were estimated fo
December 2010
value of 0 representing death and 1 representing fullhealth. Utilities for both transfusion-independent andtransfusion-dependent MDS were determined in thestudy by Szende et al3 in patients with MDS in theUnited States and Europe, using the time-tradeoffmethod.
While some studies have reported utility values ofpatients with AML directly,16,17 these patients are gen-rally healthier than those with MDS progressing toML and consequently have utility values higher than
hose of patients with transfusion-dependent MDS.o assign a more realistic utility value for patientshose AML has progressed from MDS, publisheduality-of-life measures among older adults withML using the European Organization for the Re-
earch and Treatment of Cancer core 30-item ques-ionnaire18 were converted to utility values via map-
ping to the preference-based EQ-5D utility scale19
using a published algorithm.20
Resource Utilization and Cost InputsAll model cost inputs are presented in year-2009
US $ (Table I). Model costs were divided into 3 com-ponents: drug and administration costs, BSC costs, andAML costs.
1792156813441120 1680145612321008
Survival without AML, decitabineSurvival without AML, BSCOverall survival, decitabineOverall survival, BSC
1904
L)-free survival by model treatment arm (decitabine vslodysplastic syndromes. Weibull survival functions wererves of the BSC arm of a Phase III decitabine trial.11 Cox
tment with decitabine versus BSC.
896
ays
(AMf myeival cur trea
2447
d
Ps
Clinical Therapeutics
Drug and Administration CostsDecitabine costs were obtained at wholesale acqui-
sition costs from the October 2009 edition of the RedBook,21 and were based on the defined 5-day dosingregimen and an assumed normal adult body surfacearea of 1.73 m2. Administration costs associated with
Table I. Base-case model inputs for the analysis of deof myelodysplastic syndrome (MDS).
Parameter
Survival, decitabine vs BSC, hazard ratio (95% CI)AML-freeOverall
sion-independent patients separately. For the period in
Volume 32 Number 14
t
iptaAfmatR
sop00trdApyHvAcd
F. Pan et al.
which patients were AML-free, publicly available unitcost data were applied to resource utilization collectedalong the Phase III decitabine trial to inform the costsof BSC and AE-related cost inputs. Specifically, hospi-tal per diem costs were assigned based on reason forhospitalization using cost data provided by the Health-care Cost and Utilization Project database.23 Medica-ion costs were obtained from the Red Book,21 with
administration frequencies obtained from drug indica-tions for the most general adult-specific use. Unit costsof physician visits, laboratory tests, and bone marrowaspirations occurring during the trial were obtainedusing Current Procedural Terminology codes and theMedicare physician fee schedule.22
AML CostsOn progression to AML, AML-related costs were
assigned in each model cycle. As the Phase III decit-abine clinical trial did not collect detailed resource uti-lization data after progression to AML, these costs(Table I) were derived from the SEER-Medicare data-base analysis by Lang et al.8 That analysis also foundthat although 34% of AML patients received chemo-therapy, the probability of receiving chemotherapywas significantly reduced in the group with a history ofMDS. Therefore, the model assumed that 25% of pa-tients who progressed to AML received chemotherapyin the base case. This assumption was supported by theclinical experience of one of the authors (K.K.), a prac-ticing oncologist/hematologist who treats MDS andAML in a community setting, as well as 3 other clini-cians (his partners). The true impact of this parameteron model results was ascertained using sensitivityanalysis.
Sensitivity AnalysesBoth deterministic and probabilistic sensitivity anal-
yses were conducted using the model base-case param-eter values (Table I). One-way sensitivity analyses wereconducted on parameter values to determine their im-pact on the ICER. This study included a sensitivityanalysis that considered, as an alternative to trial-based costing, a costing approach using data from apublished claims database analysis that summed thetotal annual health care costs in a subgroup of patientswith MDS and a mean age of 70 years. The decitabinePhase III clinical trial data indicated a slightly higherrisk for death in the decitabine treatment group on
progression to AML.6 The sensitivity analysis exam-
December 2010
ned the effects of assuming an equal risk for death onrogression. The one-way sensitivity analysis assessedhe effects of discount rate, time horizon, utility valuessociated with AML, percentage of patients withML who received chemotherapy, and clinical inputs
rom the decitabine 5-day outpatient dosing study onodel results. The cost-effectiveness of decitabine
mong trial subgroups�patients with de novo MDS,reatment-naive patients, and patients with IPSS Highisk or Intermediate Risk 2�was analyzed.The Phase III decitabine trial data did not indicate a
tatistically significant difference in time to AML orverall survival between decitabine and BSC (Cox pro-ortional HRs: time to AML, HR � 0.736; 95% CI,.477–1.136; overall survival, HR � 1.054; 95% CI,.725–1.532). To investigate the uncertainty aroundhese inputs and their impact on the cost-effectivenessesults, probabilistic sensitivity analyses were con-ucted by varying most model parameters (eg, time toML, survival, response, costs) randomly across theirotential distributions over 1000 iterations. The anal-sis here assumed log-normal distribution for survivalRs; � distributions for treatment response and utility
alues; and � distribution for costs of MDS, AML, andEs. For HRs, treatment response, and costs, thehoice of the distribution used was based on how theata were presented or distributed. For utility values, �
distributions are common and were therefore used inthis model.
RESULTSBase-Case Results
The base-case analysis assessed the cost-effective-ness of a 5-day outpatient dosing regimen of decitabineversus BSC for the treatment of MDS. The base-caseanalysis was deterministic, in which only point esti-mates for each parameter were considered. The deter-ministic model results are shown in Table II. At base-line, treatment with decitabine was estimated to yield0.276 additional year of AML-free survival, 0.332fewer year with AML, and 0.260 additional transfu-sion-independent life-year compared with BSC. Whilethe point estimate for overall survival was higher withBSC, decitabine was associated with 0.052 moreQALY, or 19 additional quality-adjusted days, thanBSC because transfusion dependence and AML are as-sociated with progressively lower quality of life com-pared with transfusion independence. Total decitabine
costs over the 5-year model time horizon were
2449
Clinical Therapeutics
$22,596, with an additional $6337 for administration.Total direct medical costs were $122,940 in the decit-abine trial arm and $122,666 in the BSC arm. There-fore, the ICER for decitabine versus BSC was $5277per QALY gained.
Sensitivity Analysis ResultsOne-way sensitivity analysis results are presented in
Table III. The results of the model were particularlysensitive to the discount rate and time horizon settings.At a discount rate of 0%, the decitabine treatment wasdominant over BSC, whereas at a discount rate of 5%,the ICER was increased to $24,356. The ICER alsoprogressively increased as the model time horizon wasshortened. At 1 year, the ICER was $804,319, while at3 years it decreased to $108,181. One significant costdriver in the model was the percentage of patients whoreceived chemotherapy on progression to AML. Whenthis percentage was 15%, the ICER was $29,465, butwhen the percentage increased to 35%, decitabine wasdominant over BSC. When the risk for death on pro-gression to AML was set to be equal between the 2treatment arms, the ICER was $183,695. When the
Table II. Base-case results in an analysis of the cost-efin the treatment of myelodysplastic syndrom
Health Outcome De
Survival, yAML-freeWith AMLQALYs accrued
Costs, year-2009 US $Decitabine drug costs 2Decitabine administration costsBSC costs (transfusion-independent) 1BSC costs (transfusion-dependent) 5AML costs 2Total direct medical costs 12
Incremental cost-effectivenessQALYsCosts, year-2009 US $ICER, year-2009 US $/QALY
AML � acute myeloid leukemia; QALY � quality-adjusted li
source of model cost data was based on the published
2450
claims analysis rather than trial-based resource utiliza-tion, the ICER was dominant. The ICER was also dom-inant when data from the decitabine 5-day outpatientdosing study were used as clinical inputs in place of thePhase III clinical trial data. In this sensitivity test, de-citabine was estimated to yield 0.276 additional year ofAML-free survival, 0.332 fewer year with AML, and0.540 additional transfusion-independent life-yearcompared with BSC. Although the point estimate foroverall survival was again higher in the BSC group,decitabine was associated with 0.119 more QALY, or44 quality-adjusted days, compared with BSC. Totaldecitabine costs over the 5-year model time horizonwere $29,166, with an additional $8179 for adminis-tration. Total direct medical costs were $125,039 inthe decitabine arm and $129,732 in the BSC arm. Inthe sensitivity test in which the utility associated withAML was increased to equal that of transfusion-depen-dent MDS (ie, 0.60), the ICER was increased by�$4000. The ICER also varied by subgroup. TheICERs were $56,476, $90,177, and dominant, respec-tively, in the MDS de novo, treatment-naive, and IPSS
eness of decitabine versus best supportive care (BSC)
Results of the probabilistic sensitivity analysis areshown in Figures 3 and 4. Because the HRs of decit-abine versus BSC for time to AML and survival werenot statistically significant, incremental QALYs and in-cremental costs varied widely (from �0.3 to 0.5 andfrom �$60,000 to $83,000, respectively) among the000 model iterations performed (Figure 3). However,ased on the cost-effectiveness acceptability curve (Fig-re 4), decitabine had a higher probability than BSC ofeing cost-effective at all payer willingness-to-payhresholds, despite a lack of statistical significance ofRs between the 2 treatments.
DISCUSSIONThe model presented here assessed the cost-effective-ness of decitabine versus BSC for the treatment of in-termediate-to high-risk patients with MDS. Before theresults were interpreted, model survival outputs for the
Table III. Decitabine 1-way sensitivity and subgroupsupportive care in the treatment of myelod
BSC arm were externally validated against publicly t
December 2010
available data sources to verify that the results werereflective of actual survival of patients with MDS andAML. Overall survival (survival without AML � sur-vival with AML) comparison data were obtained froma study in 816 patients with MDS stratified by IPSSsubgroup.24 AML survival data were obtained fromEER cancer statistics available through the Nationalancer Institute.25 Both overall survival and survivalith AML were comparable to, but slightly lower
han, the comparison data. The reason for this discrep-ncy may be the differences in patient populations be-ween the Phase III decitabine clinical trial and otherublished data. Patients in the trial had a particularlyigh-risk profile that likely was not matched exactly inther studies.
In the deterministic base-case analysis, decitabineepresented a cost-effective alternative to BSC at5277 per QALY gained. Although there is no defini-
ses of the cost-effectiveness of decitabine versus beststic syndromes (MDS).
alysis Sensitivity AnalysisICER, Year-2009
US $/QALY
– $5277
0% Dominant5% $24,356
1 year $804,3193 years $108,181
d Claims database analysis Dominant
0.6 $9554
15% $29,46535% Dominant
l data Equal $183,695
MDS de novo $56,476Treatment naive $90,177IPSS High Risk andIntermediate Risk 2
Dominant
ntl trial
5-Day outpatient dosingsingle-arm study
Dominant
justed life-year; AML � acute myeloid leukemia; IPSS �
analyyspla
se An
–
%
years
l-base
.53
5%
n tria
patielinica
lity-ad
ive cost-effectiveness threshold among US payers, an
2451
cwWvwds
Clinical Therapeutics
ICER of this magnitude typically represents cost-effec-tiveness. Establishing the significance of this ICERamong the spectrum of treatments for MDS is difficultbecause although many studies have assessed the costsassociated with MDS,6,26–30 few available pharmaco-economic analyses have combined the costs of MDStreatment with data on effectiveness and quality of lifeto determine cost-effectiveness.31,32 Based on the find-ings from a literature search, this is the first such anal-ysis of decitabine for the treatment of MDS. The searchfor studies of MDS treatments similar to decitabineidentified an abstract by Miller et al32 of a Markovost-effectiveness model that compared decitabineith azacitidine, a similar hypomethylating agent.hile the abstract alone does not allow a detailed re-
iew of the methods, a number of confounding factorsere not explicitly addressed in this model. The 2rugs were not directly compared. Rather, results fromeparate decitabine and azacitidine33 Phase III trials
were used to inform clinical inputs. The patient popu-lation in the azacitidine trial was at a higher risk thanthat in the decitabine trial and included only patients
-
-2020,00000
0
00
Incr
emen
tal C
ost
($)
Increm
100,000
80,000
60,000
40,000
20,000
0
–20,000
–40,000
–60,000
–80,000
–0.2–0.4
Figure 3. Incremental quality-adjusted life-years (QAmyelodysplastic syndromes with decitabineruns of the model, with selected parameterbutions for each replication.
with de novo diagnoses; thus, these patients may have
2452
been more responsive to treatment. In addition, theprotocol design for the azacitidine trial specified thatpatients were to be treated until disease progression,while in the decitabine trial patients were discontinuedafter 2 cycles after response was achieved. The mediantreatment durations were 9 cycles in the azacitidinetrial and 6 cycles in the decitabine trial. The modelconstructed by Miller et al had a shorter time horizonand did not include costs of AEs, complicating anyuseful comparison between models. As such, a robustand valid inference of comparative cost-effectivenessamong hypomethylating agents requires either directcomparative clinical data or more comparable indirectcomparisons.
The deterministic base-case analysis suggested thatdecitabine delays progression to AML and prolongstransfusion independence compared with BSC. Whileoverall survival based on the parameter point estimateswas numerically greater in the BSC arm, prolongedtransfusion independence in patients who received de-citabine was associated with increases in quality of lifeand therefore more quality-adjusted life-years. In-
QALYs (y)
0.40.2 0.6
plotted against incremental cost for the treatment ofs best supportive care. Each dot represents 1 of 1000d randomly within their respective probability distri-
ental
LYs)versus varie
creased treatment costs associated with decitabine are
Volume 32 Number 14
dtetsttMidwptc
atgt
F. Pan et al.
partially offset by cost savings incurred due to less timespent with AML. However, in the Phase III decitabinetrial, the differences in both time to AML and overallsurvival between decitabine and BSC did not achievestatistical significance. Therefore, to assess the robust-ness of the model results in light of base-case uncer-tainty, the impact of these key model drivers (ie, qualityof life, costs, AML-free survival, and overall survivalamong treatment arms) were analyzed using sensitivityanalyses.
The base-case assumed that utility associated withAML was lower than MDS regardless of transfusionstatus. A case in which utility of AML was equal to thatof transfusion-dependent MDS was tested. The ICERwas increased to $9554 per QALY, a modest increasethat suggests cost-effectiveness.
Model costs are based on resource utilization datacollected together with the Phase III decitabine trial.Although this method had the benefit of application toa small, targeted population of patients with MDS andcollected data on treatment of decitabine-related AEs,these data may not be attributable to all patients withMDS, even in the same risk category. This limitationwas tested by applying alternative cost data based on a
0
Prob
abili
ty o
f Cos
t-Ef
fect
iven
ess
Payer Willingness to P
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
10,000
20,000
40,000
60,000
30,000
50,000
7
0
Figure 4. Probability of cost-effectiveness of decitabmyelodysplastic syndromes at increasing p[QALY]) thresholds, based on 1000 runs oftheir respective probability distributions fo
published claims database analysis of total annual
December 2010
health care costs for a subgroup of patients with MDSand a mean age of 70 years.6 Using these cost data,ecitabine treatment was dominant over BSC, meaninghat it was less expensive in addition to being moreffective. The benefits of this costing approach werehat it used data from clinical practice and had a largeample size that may be more generalizable to all pa-ients with MDS. The limitations of this approach werehat it included all resources utilized by patients with
DS, including non-MDS�related utilization. It alsoncluded less severe MDS than those considered in theecitabine trial and did not include AE costs associatedith decitabine use. Regardless, the use of this ap-roach did not increase the ICER, indicating that therial-based costing approach was the conservativehoice.
Perhaps the most significant cost driver was the costssociated with AML. Because the decitabine clinicalrial did not collect detailed resource utilization on pro-ression to AML, these model costs were derived fromhe analysis by Lang et al8 of SEER-Medicare data in
3439 elderly patients, stratified by chemotherapy use.The investigators found that 34% of patients receivedchemotherapy, but that a history of MDS significantly
st per QALY), Year–2009 US $
DecitabineBSC
80,000
100,000
130,000
90,000
110,000
120,000
140,000
150,000
rsus best supportive care (BSC) in the treatment ofillingness-to-pay (cost per quality-adjusted life-year
odel with selected parameters varied randomly withinreplication.
ay (Co
0,000
ine veayer wthe mr each
decreased the probability of receiving chemotherapy.
2453
Clinical Therapeutics
In the base-case model, it was assumed that 25% ofpatients with AML would receive chemotherapy. Thisvalue was supported by the clinical expert opinion ofone of the authors (K.K.), a practicing oncologist/he-matologist who treats MDS and AML in a communitysetting, and 3 other clinicians (his partners). Becausethe experience of just one hematology practice may notbe indicative of broader clinical practice, this limita-tion was tested in the sensitivity analysis, and it wasfound that if 15% of patients received chemotherapy,the ICER was $29,465 per QALY, and if 35% of pa-tients received chemotherapy, decitabine would domi-nate BSC. Given that Lang et al reported a value of34%, 25% may represent a conservative estimate. Ifthe true value is lower, the resulting ICER, whilehigher, may still be deemed cost-effective by many USpayers.
In the Phase III clinical trial, BSC was associatedwith numerically but not statistically longer mediansurvival compared with decitabine after progression toAML. Longer survival would lead to higher AML-re-lated costs in the BSC group and may have overesti-mated the cost-effectiveness of decitabine. This effectwas tested using sensitivity analysis with the assump-tion that patients in the decitabine and BSC treatmentarms had the same risk for death on progressionto AML. Under this assumption, the ICER was$183,695—significantly higher than that of the basecase, suggesting that the lack of statistical significancein overall survival calculations between trial armsadded a fair amount of uncertainty to the model.
While the one-way sensitivity analysis results indi-cated that the model results were—relative to the pref-erences of particular payers—generally resistant to theuncertainty surrounding individual parameters, theymight still have underestimated decision uncertainty.Given that the differences in AML-free and overall sur-vival between the decitabine and BSC arms were notstatistically significant in the Phase III clinical trial,probabilistic sensitivity analysis, a more robust andcommonly used approach to assessing the effects ofparameter uncertainties,34 was performed in thismodel. When the uncertainty around most model pa-rameters was considered simultaneously and at ran-dom according to the probability distribution of eachparameter, the cost-effectiveness acceptability curveindicated that decitabine was more likely to be cost-effective than BSC at all payer willingness-to-pay
thresholds. This finding offers support to the claim of
2454
cost—effectiveness of decitabine despite inherent un-certainty stemming from sources or lack of statisticalsignificance of point estimates of model parameters.One possible explanation for this finding is that 90%of simulated HRs of decitabine versus BSC in time toAML are �1 given that the 95% CI spans from 0.477to 1.136.
The design of the Phase III decitabine clinical trialhad limitations that may have affected the results ofthis analysis. The trial protocol mandated that patientsbe discontinued from decitabine treatment 2 cycles af-ter showing treatment response. This design may haveserved to underestimate the treatment effect attainablewith continuous decitabine treatment. Evidence from aretrospective, observational study in 137 patients witha diagnosis of MDS and treated with a hypomethylat-ing agent (decitabine or azacitidine) indicated that agreater number of decitabine treatment cycles were as-sociated with a greater probability of clinical re-sponse.35 Therefore, the Phase III trial design may haveprovided a conservative estimate of the clinical benefitof decitabine. From a modeling perspective, while alonger duration of decitabine treatment is likely to in-crease treatment costs, the cost increase might be offsetby improved quality of life related to prolonged trans-fusion independence and time to AML progression.
CONCLUSIONSThis article presented a Markov model that assessedthe cost-effectiveness of 5-day dosing of decitabineversus BSC in US patients with intermediate- andhigh-risk MDS from a US payer perspective. In themodel deterministic analysis, decitabine was cost-effective compared with BSC, with an ICER of$5277 per QALY gained. Probabilistic sensitivityanalysis indicated that a 5-day dosing schedule ofdecitabine was more likely to be cost-effective thanBSC as a treatment option for US patients with in-termediate- and high-risk MDS.
ACKNOWLEDGMENTSThis research was funded by Eisai, Inc. (Eisai). Eisaideveloped decitabine for the treatment of MDS. Dr.Pan, Ms. Peng, Mr. Linnehan, and Dr. Fleurence arepaid employees of United BioSource Corporation,which received the funding for the research and man-uscript preparation. Dr. Knopf is a paid consultant forUnited BioSource Corporation. Dr. Kim is a former
employee of Eisai.
Volume 32 Number 14
F. Pan et al.
The authors have indicated that they have noother conflicts of interest regarding the content ofthis article.
REFERENCES1. Bennett JM, Catovsky D, Daniel MT, et al. Proposals for the
classification of the myelodysplastic syndromes. Br J Haema-tol. 1982;51:189–199.
2. Oscier DG. ABC of clinical haematology. The myelodys-plastic syndromes. BMJ. 1997;314:883–886.
3. Szende A, Schaefer C, Goss TF, et al.Valuation of transfu-sion-free living in MDS: Results of health utility interviewswith patients. Health Qual Life Outcomes. 2009;7:81.
4. Vardiman JW, Harris NL, Brunning RD. The World HealthOrganization (WHO) classification of the myeloid neo-plasms. Blood. 2002;100:2292–2302.
5. Rollison DE, Howlader N, Smith MT, et al. Epidemiology ofmyelodysplastic syndromes and chronic myeloproliferativedisorders in the United States, 2001–2004, using data fromthe NAACCR and SEER programs. Blood. 2008;112:45–52.
6. Frytak JR, Henk HJ, De Castro CM, et al. Estimation ofeconomic costs associated with transfusion dependence inadults with MDS. Curr Med Res Opin. 2009;25:1941–1951.
7. Disperati P, Ichim CV, Tkachuk D, et al.Progression ofmyelodysplasia to acute lymphoblastic leukaemia: Implica-tions for disease biology. Leuk Res. 2006;30:233–239.
8. Lang K, Earle CC, Foster T, et al. Trends in the treatment ofacute myeloid leukaemia in the elderly. Drugs Aging. 2005;22:943–955.
9. Oberg G, Killander A, Björeman M, et al. Long-termfollow-up of patients �or�60 yr old with acute myeloidleukaemia treated with intensive chemotherapy. Eur JHaematol. 2002;68:376–381.
10. Malcovati L. Impact of transfusion dependency and second-ary iron overload on the survival of patients with myelodys-plastic syndromes. Leuk Res. 2007;31(Suppl 3):S2–S6.
11. Kantarjian H, Issa JP, Rosenfeld CS, et al. Decitabine improvespatient outcomes in myelodysplastic syndromes: Results of aphaseIII randomizedstudy.Cancer.2006;106:1794–1803.
12. Dacogen (decitabine) for injection [product information].http://www.accessdata.fda.gov/drugsatfda_docs/label/2006/021790lbl.pdf.AccessedNovember17,2010.
13. Dacogen (decitabine) for injection [product information].http://www.accessdata.fda.gov/drugsatfda_docs/label/2010/021790s006lbl.pdf. Accessed November 17, 2010.
14. Valent P, Horny HP, Bennett JM, et al. Definitions andstandards in the diagnosis and treatment of the myelodys-plastic syndromes: Consensus statements and report froma working conference. Leuk Res. 2007;31:727–736.
15. Steensma DP, Baer MR, Slack JL, et al. Multicenter study ofdecitabine administered daily for 5 days every 4 weeks to
adults with myelodysplastic syndromes: The alternative
December 2010
dosing for outpatient treatment (ADOPT) trial. J Clin Oncol.2009;27:3842–3848.
16. Slovacek L, Slovackova B, Jebavy L, Macingova Z. Psychoso-cial, health and demographic characteristics of quality oflife among patients with acute myeloid leukemia andmalignant lymphoma who underwent autologous hemato-poietic stem cell transplantation. Sao Paulo Med J. 2007;125:359–361.
17. Uyl-de Groot CA, Löwenberg B, Vellenga E, et al. Cost-effectiveness and quality-of-life assessment of GM-CSF asan adjunct to intensive remission induction chemotherapyin elderly patients with acute myeloid leukemia. Br JHaematol. 1998;100:629–636.
18. Alibhai SM, Leach M, Kermalli H, et al. The impact of acutemyeloid leukemia and its treatment on quality of life andfunctional status in older adults. Crit Rev Oncol Hematol.2007;64:19–30.
19. Brooks R. EuroQol: The current state of play. HealthPolicy. 1996;37:53–72.
20. Kontodimopoulos N, Aletras VH, Paliouras D, Niakas D.Mapping the cancer-specific EORTC QLQ-C30 to thepreference-based EQ-5D, SF-6D, and 15D instruments.Value Health. 2009;12:1151–1157.
21. Thomson PDR. Red Book for Windows. Version 61127.Montvale, NJ: Thomson PDR; 2009.
23. Agency for Healthcare Research and Quality. The Health-care Cost and Utilization Project (HCUP). http://www.ahrq.gov/data/hcup/. Accessed November 2009.
24. Greenberg P, Cox C, LeBeau MM, et al. Internationalscoring system for evaluating prognosis in myelodysplasticsyndromes [published correction appears in Blood. 1998;91:1100]. Blood. 1997;89:2079–2088.
25. Surveillance Research Program (SEER). Fast Stats: Aninteractive tool for access to SEER cancer statistics.http://seer.cancer.gov/faststats. Accessed August 6,2010.
26. Casadevall N, Durieux P, Dubois S, et al. Health, economic,and quality-of-life effects of erythropoietin and granulocytecolony-stimulating factor for the treatment of myelodys-plastic syndromes: A randomized, controlled trial. Blood.2004;104:321–327.
27. Gattermann N, Hofmann WK, Meessen A, et al. Myelodys-plastic syndromes: Aspects of current medical care andeconomic considerations in Germany. Onkologie. 2008;31:477–484.
28. Greenberg PL, Cosler LE, Ferro SA, Lyman GH. The costs ofdrugs used to treat myelodysplastic syndromes followingNational Comprehensive Cancer Network Guidelines. J Natl
29. Gupta P, LeRoy SC, Luikart SD, et al.Long-term blood product transfu-sion support for patients with myelo-dysplastic syndromes (MDS): Costanalysis and complications. Leuk Res.1999;23:953–959.
30. O’Brien JA, Caro JJ. Managing myelo-dysplastic syndromes with support-ive therapy in elderly persons. DrugBenefit Trends. 2005;17:341–351.
31. Goss TF, Szende A, Schaefer C, et al.Cost effectiveness of lenalidomide inthe treatment of transfusion-depen-dent myelodysplastic syndromes inthe United States. Cancer Control.2006;(Suppl 13):17–25.
32. Miller JD, Fenaux P, Beach CL, et al.Comparative cost-effectiveness ofazacitidine versus decitabine for thetreatment of myelodysplastic syn-dromes (MDS). Abstract presentedat: 51st ASH Annual Meeting andExposition; December 5, 2009; NewOrleans, La.
33. Fenaux P, Mufti GJ, Hellstrom-Lindberg E, et al, for the Interna-tional Vidaza High-Risk MDS Sur-vival Study Group. Efficacy ofazacitidine compared with that ofconventional care regimens in thetreatment of higher-risk myelodys-plastic syndromes: A randomised,open-label, phase III study. LancetOncol. 2009;10:223–232.
34. Claxton K, Sculpher M, McCabe C,et al. Probabilistic sensitivity analy-sis for NICE technology assessment:Not an optional extra. Health Econ.2005;14:339–347.
35. Bordoni R, Feinberg B, Gilmore J, etal. Hematologic outcomes of MDStreatment with hypomethylatingagents in community practice. Posterpresented at: 2010 American Soci-ety of Clinical Oncology (ASCO)Annual Meeting; June 4–8, 2010;Chicago, Ill.
2456
Address correspondence to: Feng Pan, PhD, MA, 7101 Wisconsin Avenue,
