Cost-effectiveness of Treating Chronic Anemia with Epoetin Alfa among Hemodialysis Patients in the United States Peter Quon, MPH a , Matthew Gitlin, PharmD b , John J. Isitt, MS b , Sumit Mohan, MD c , William M. McClellan, MD, MPH d , Jill Javier, BS a , Gregory de Lissovoy, PhD, MPH e , Christopher S. Hollenbeak, PhD f a UNITED BIOSOURCE CORPORATION,CHEVY CHASE, MD; b GLOBAL HEALTH ECONOMICS,AMGEN INC., THOUSAND OAKS,CALIF; c DEPARTMENT OF MEDICINE,DIVISION OF NEPHROLOGY,COLUMBIA UNIVERSITY COLLEGE OF PHYSICIANS &SURGEONS,NEW YORK, NY; d DEPARTMENTS OF MEDICINE AND EPIDEMIOLOGY,EMORY UNIVERSITY, ATLANTA, GA; e DEPARTMENT OF HEALTH POLICY AND MANAGEMENT,JOHNS HOPKINS BLOOMBERG SCHOOL OF PUBLIC HEALTH,BALTIMORE, MD; AND f DEPARTMENTS OF SURGERY AND PUBLIC HEALTH SCIENCES,COLLEGE OF MEDICINE,PENN STATE UNIVERSITY,HERSHEY, PA - ABSTRACT OBJECTIVES: The objectives of this analysis were to assess health and economic consequences of targeting hemoglobin (Hb) levels around 10-11 g/dL relative to 9-10 g/dL using an economic model and to explore the impact of different assumptions on cost-effectiveness. STUDY DESIGN: Clinical and economic impact of treating anemia in the US hemodialysis population to target Hb levels of 10-11 g/dL and 9-10 g/dL was assessed using a Markov model. A sensitivity analysis assessed the effects of varying assumptions on the model. RESULTS: Our cost-effectiveness analysis suggests that maintaining Hb 10-11 g/dL would result in average reductions of 0.51 hospitalizations and increases of 0.09 quality-adjusted life years per patient, with hospitalization cost offsets of $15,340 over 5 years when compared with Hb of 9-10 g/dL. Over the lifetime of the patient, cost-effectiveness improved with hospitalization cost offsets of $21,450 and increases of 0.12 quality-adjusted life years. Sensitivity analysis of individual parameters showed that mortality, hospitalization, health preference, and time horizon of the model had the most influence on cost-effectiveness. CONCLUSIONS: Our analysis suggests that epoetin alfa use targeting Hb levels of 10-11 g/dL relative to 9-10 g/dL may result in better patient outcomes and lower costs. The sensitivity analysis high- lighted how assumptions affected cost-effectiveness conclusions; the appropriateness of these assumptions will remain uncertain until new research in today’s dialysis population examining the effects of targeting to lower Hb levels is conducted. See funding, conflict of interest, and authorship disclosures at the end of this article. Ó 2012 Elsevier Inc. All rights reserved Health Outcomes Research in Medicine (2012) 3, e79-e89 COMPARATIVE EFFECTIVENESS AND HEALTH POLICY
Transcript
COMPARATIVE
EFFECTIVENESS
AND HEALTH
POLICY
Cost-effectiveness of Treating Chronic
Anemia with Epoetin Alfa among
Hemodialysis Patients in the United States
Peter Quon, MPHa, Matthew Gitlin, PharmDb, John J. Isitt, MSb, Sumit Mohan, MDc,William M. McClellan, MD, MPHd, Jill Javier, BSa, Gregory de Lissovoy, PhD, MPHe,Christopher S. Hollenbeak, PhDf
OAKS, CALIF; cDEPARTMENT OF MEDICINE, DIVISION OF NEPHROLOGY, COLUMBIA UNIVERSITY COLLEGE OF
PHYSICIANS & SURGEONS, NEW YORK, NY; dDEPARTMENTS OF MEDICINE AND EPIDEMIOLOGY, EMORY UNIVERSITY,ATLANTA, GA; eDEPARTMENT OF HEALTH POLICY AND MANAGEMENT, JOHNS HOPKINS BLOOMBERG SCHOOL OF
PUBLIC HEALTH, BALTIMORE, MD; ANDfDEPARTMENTS OF SURGERY AND PUBLIC HEALTH SCIENCES, COLLEGE OF
MEDICINE, PENN STATE UNIVERSITY, HERSHEY, PA
- A B S T R A C T
O B J E C T I V E S : The objectives of this analysis were to assess health and economic consequencesof targeting hemoglobin (Hb) levels around 10-11 g/dL relative to 9-10 g/dL using an economic model and toexplore the impact of different assumptions on cost-effectiveness.
S T U D Y D E S I G N : Clinical and economic impact of treating anemia in the US hemodialysispopulation to target Hb levels of 10-11 g/dL and 9-10 g/dL was assessed using a Markov model. A sensitivityanalysis assessed the effects of varying assumptions on the model.
R E S U L T S : Our cost-effectiveness analysis suggests that maintaining Hb 10-11 g/dL would result inaverage reductions of 0.51 hospitalizations and increases of 0.09 quality-adjusted life years per patient, withhospitalization cost offsets of $15,340 over 5 years when compared with Hb of 9-10 g/dL. Over the lifetimeof the patient, cost-effectiveness improved with hospitalization cost offsets of $21,450 and increases of 0.12quality-adjusted life years. Sensitivity analysis of individual parameters showed that mortality, hospitalization,health preference, and time horizon of the model had the most influence on cost-effectiveness.
C O N C L U S I O N S : Our analysis suggests that epoetin alfa use targeting Hb levels of 10-11 g/dLrelative to 9-10 g/dL may result in better patient outcomes and lower costs. The sensitivity analysis high-lighted how assumptions affected cost-effectiveness conclusions; the appropriateness of these assumptionswill remain uncertain until new research in today’s dialysis population examining the effects of targeting tolower Hb levels is conducted.
See funding, conflict of interest, and authorship disclosures at the end of this article.� 2012 Elsevier Inc. All rights reserved � Health Outcomes Research in Medicine (2012) 3, e79-e89
K E Y W O R D S : Anemia; Erythropoiesis-stimulating agent; Hemodialysis; Hemoglobin; Iron
or patients with chronic kidney disease (CKD) receiving hemodialysis, evidence suggests that use oferythropoiesis-stimulating agents (ESAs) to maintain Hb $10 g/dL improves patient-reported
outcomes and leads to marked reductions in risks of transfusion and hospitalization.1 Before the availabilityof ESAs, anemia was a substantial burden in dialysis patients,2-5 with mean hemoglobin (Hb) valuesaround 7 g/dL.2 The majority of patients required transfusions to manage chronic anemia,2 which also placedpatients at risk to allosensitization and elevations in panel reactive antibody levels that can reduce renal graftsurvival rates after transplantation,6,7 as well as the inherent risks associated with transfusion.8,9 After thedevelopment of recombinant human erythropoietin, randomized clinical trial results demonstrated thatpatients previously dependent on transfusions for the treatment of chronic anemia became transfusionindependent.3 Hemodialysis patients now receiving ESAs for treatment of chronic anemia have increasedtheir mean Hb levels to $10 g/dL,10 and there has been a significant reduction in allosensitization.11,12
Until recently, an Hb target of 10-12 g/dL was recommended by the US Food and Drug Administra-tion (FDA) for patients with CKD.13 In June 2011, the FDA replaced this target range with a recommendationto consider initiating ESA treatment when Hb levels are<10 g/dL and to reduce or interrupt ESAs when Hbapproaches or exceeds 11 g/dL.14 The revised label cited risks seen in the Normal Hematocrit Study(NHCT),15 the Correction of Anemia with Epoetin Alfa in Chronic Kidney Disease (CHOIR) study,16
and the Trial to Reduce Cardiovascular Events with Aranesp Therapy (TREAT) study,17,18 whichdemonstrated increased cardiovascular events and mortality when higher Hb levels were targeted($13 g/dL) and achieved ($12.5 g/dL).
Therefore, this study was designed to understand the potential clinical and economic consequencesof targeting to lower Hb levels around 10-11 g/dL. There are no data on the absolute risks of these lower Hbtargets, as they have been previously compared only with higher Hb targets of ($13 g/dL); these studiesdemonstrated no benefit and increased risk for acute myocardial infarction and stroke at the higher targets.A further goal of this study was to identify the gaps in evidence for cost-effectiveness of the recent Hbmanage-ment recommendations using ESAs. There are limited studies examining the cost-effectiveness of ESAs inhemodialysis patients19,20 and they used different assumptions about mortality, hospitalization rates, andquality of life, resulting in varying conclusions.21,22 This constitutes a substantial limitation of healtheconomic models assessing anemia treatment because these parameters have a profound influence onresults.22 The current analysis was designed to further explore the contributions of these model parametersand assumptions and their impact on health economic analyses. The primary objective of the analysis wasto compare the cost-effectiveness of maintaining Hb at 10-11 g/dL, versus 9-10 g/dL in hemodialysis patientsin the US.
M E T H O D S
AMarkov model assessed the clinical and economic impact of correcting anemia from the Medicareperspective over a 5-year time period (referred to as “time horizon”) in the US hemodialysis population. Thebase case analysis assessed patients targeting Hb 10-11 g/dL versus Hb 9-10 g/dL, levels that pertain to main-tenance with epoetin alfa. These patients were either with or without diabetes and were subject to annualprobabilities of hospitalization, death, and transplantation. Patients receiving transplantation were assumedto cease dialysis. Nevertheless, these patients may experience graft failure and return to dialysis.
The Markov model (depicted in Supplemental Figure S1) used expected estimates of clinical andeconomic parameters (ie, averages) to estimate expected costs and outcomes. A probabilistic sensitivity anal-ysis (PSA) was conducted to verify the robustness of these results, taking into account the uncertainty aroundparameters, and a sensitivity analysis was conducted to explore the impact of different model assumptions.
Health Outcomes Research in Medicine - Vol. 3 / No. 2 / May 2012 e81
Literature Review
A literature review was conducted to identify randomized clinical trials, observational studies,and published models to inform clinical efficacy parameters, and was restricted to studies published nolater than 2000. We included studies of patients with CKD either on or off hemodialysis. We prioritizedfindings from NHCT, CHOIR, and TREAT, key clinical trials that have recently influenced US regula-tory and policy decisions surrounding anemia correction using ESAs in hemodialysis. Data and conclu-sions from these studies were factored into our analysis as either point estimates for parameters ora basis for assumptions.
Cl in ica l Assumptions
Baseline hospitalization rates were determined from the US Renal Data System (USRDS)5 anda longitudinal study of hemodialysis patients in the US.23 We calculated the average rate of hospitalizationand assumed that rate to be associated with the registry average of Hb 12 g/dL.5 Subsequently, relative risksfrom the longitudinal study were applied to estimate the rate of hospitalization for each of the targeted Hblevels (Table 1).23 Refer to the Supplemental Materials for the calculation of hospitalization rates by Hblevel.
We assumed that all patients were administered epoetin alfa at a dose dependent on targeted Hb,which was based on an Hb-ESA dose curve first published by Tonelli et al.21 We updated their dose-response curve by using data from additional randomized trials and plotting the midpoint of each target rangeby average epoetin alfa dose.15,21,24-29 Similar to Tonelli et al,21 we assumed there was an exponential increasein dose required to target higher Hb levels. Further details about the calculation of the dose curve are providedin the Supplemental Materials.
Observational analyses of mortality among dialysis patients at different achieved Hb ranges havesuggested increased mortality at lower Hb.23,30,31 However, these observations are recognized as subject tosubstantial confounding. Thus, data from clinical trials evaluating Hb target ranges remains the onlysource of data on this question. We note that there are no data to distinguish the relative mortality,cardiovascular (acute myocardial infarction), and thromboembolic (stroke) risks among dialysis patientsat the 2 Hb ranges we model in this analysis because 2 Hb ranges are entirely contained within theHb target of the lower arm of the NHCT study (ie, 10� 1 g/dL). However, in the TREAT study,with over 4000 patients randomized to achieve an Hb of 13 g/dL versus rescue therapy in placebopatients at <9 g/dL, no significant differences were found in attributing mortality to Hb level. Basedon these results and in the absence of data from a clinical trial in hemodialysis patients comparingthe 2 Hb ranges we model in this analysis, we assume for the base case no difference in mortalityacross Hb comparators. However, testing the potential for cardiovascular or thromboembolic risksthrough increased mortality was assessed in a sensitivity analysis using data from a past economicmodel by Clement et al.22
Health State Preference Weights
Health state preference weights (health utility) were assigned to each of the 3 clinical states(hemodialysis, transplant, and acute graft failure) and were used to calculate quality-adjusted life years(QALYs) (Table 1). A utility weight is a quantitative expression of an individual’s preference for or desir-ability of a particular state of health under conditions of uncertainty. These preference weight estimates(health utilities) were improved at higher Hb levels. Utility scores were extrapolated from a model
- T A B L E 1 : Model Inputs and Assumptions
Input/Adjustment Source
Base case analysis parametersBaseline population distribution 41% of pts have diabetes USRDS 20105
Mean (SE) annual mortality rate ondialysis per 1000 pt-yrs
Pts without diabetes: 224.0 (44.8)Pts with diabetes: 257.0 (51.4)
USRDS 200810*
Mean (SE) annual mortality rate ontransplant per 1000 pt-yrs
Inclusion of allosensitization Rate of transplantation is 30 per 1000pt-yrs for Hb
9-10 g/dL
Assumptionk
Equal mortality and hospitalizationbetween patients with and withoutdiabetes
Mortality is 257 per 1000 pt-yrs for allpatients
Hospitalization rate per 1000 pt-yrsregardless of diabetes:Target Hb 10-11 g/dL: 2374Target Hb 9-10 g/dL: 2624
Assumption
Model time horizon is lifetime Time horizon set to lifetime AssumptionAddition of transfusion costs Target Hb 10-11 g/dL: $28/yr
Target Hb 9-10 g/dL: $54/yrIbrahim 200837;
assumption{
Hb¼ hemoglobin; Pt¼ patient; SE¼ standard error; pt-yrs¼patient-years; USRDS¼United States Renal Data System.*Derived from total patient deaths and annual mortality rates for hemodialysis patients never on transplant wait list (Tables H.13,H.14) and hemodialysis patients on transplant list (Tables H.15, H.16). We assumed equivalent mortality across Hb targets and thatSE was 20% of mean. Mortality for hemodialysis patients without diabetes was not available from USRDS; therefore, this rate wascalculated from mortality of all hemodialysis patients and hemodialysis patients with diabetes. For face validity we calculatedaweightedmortality rate based on proportion of diabetics of 239 per 1000 pt-yrs, which is reasonably close to the 5th-year moralityon hemodialysis (245 per 1000 pt-yrs) in 2003. In Table D.11, point prevalent end-stage renal disease patients I 2008 by treatmentmodality, diabetics comprised 44% of “center hemo” patients. In our approach, we determined that diabetics comprised 45% of allhemodialysis patients.†SE was not available from source; therefore, a SE of 20% of input value was assumed.‡Codes are for 2011 and were accessed in April 2011.§Extrapolated from Clement et al22 assuming linear trend. Actual mortality rates were 200 per 1000 pt-yrs for Hb 9-10.9; 212 for Hb11-12 g/dL; and 219 for Hb>12 g/dL. Rates for diabetic patients were calculated by applying a hazard ratio calculated from the differ-ence between patients with and without diabetes from base case.kAssumption based on parallel approach in Canadian Agency for Drugs and Technologies in Health report on erythropoiesis-stimulating agents,30 which assumed a potential 25% reduction for lower Hb target ranges.{Assumed 1 unit of blood per transfusion.
Health Outcomes Research in Medicine - Vol. 3 / No. 2 / May 2012 e83
published in clinical guidance on anemia management in CKD for the National Health Service in theUnited Kingdom.32 Using the midpoints of each target range and assuming a linear relationship, utilitieswere estimated for our model’s Hb targets.
Costs were estimated based on Medicare payments to providers, adjusted to 2011 US dollars and dis-counted 3.5% annually. Patients on hemodialysis were assigned an annual cost encompassing hemodialysis,ambulance, and outpatient care. Costs for epoetin alfa and iron were calculated separately. Hospitalizationcosts were calculated from USRDS estimates for length of stay and average daily cost.5 Patients with a func-tioning kidney transplant were assigned an annual cost that covered outpatient care and immunosuppressivetreatment. Further details on the calculation of the annual cost of patients on hemodialysis, length of stay, andaverage daily cost are provided in the Supplemental Materials. Costs for patients with a functioning kidneytransplant were taken directly from USRDS 2010.
Sensi t iv i ty Analys is
We conducted a sensitivity analysis to explore the effects of different assumptions on model results.This was performed by adjusting model parameters according to the assumption being tested and measuringthe degree of change in model results from the base case. We tested different assumptions on allosensitization,transplantation, time horizon, and transfusions. The rationale and source for different assumptions areprovided in Table 1. Prior health economic models have made varying assumptions about the effects of cor-recting anemia on survival, hospitalization, and quality of life, subsequently arriving at different conclusions.These assumptions were tested in the sensitivity analysis as a test of validity of our model design and to eluci-date how these assumptions are the primary causes of difference in model conclusions.
We included assumptions to test from 3 prior health economic models that assessed the cost-effectiveness of targeting higher Hb.21,22,32 The model from Clement et al22 was based on findings fromameta-analysis and associated poorer survival with Hb 11-12 g/dL versus Hb 9-10.9 g/dL. Such an assumptionmay be driven by the potential for increases in cardiovascular or thromboembolic risks as seen in clinical trialsassessing higher Hb levels. However, this meta-analysis did not incorporate recent observational studies suchas Regidor et al31 and Messana et al,33 which found lower survival with lower Hb concentrations. An increasein risk of cardiovascular events at the Hb targets in our analysis has not been reported, and the results of ouranalysis would not change even if a corresponding increase in events were assumed. Clement et al22 assumedthat hospitalization and utility would be the same across Hb ranges, and also explored an assumption of nodifference in survival between Hb groups in their sensitivity analysis. In the original model from this groupof investigators, reported by Tonelli et al,21 there was no difference assumed in survival and hospitalizationacross Hb groups, and utility improved at higher Hb. Another model was presented in a clinical guidanceon anemia management in CKD for National Health Services in the UK.32 In this model, survival was basedon an observational study and improved at higher Hb. Hospitalization was assumed to be equivalent across Hbgroups and the utility improved at higher Hb.
Probabi l i s t ic Sens i t iv i ty Analys is (PSA)
We conducted a PSA using a Monte Carlo simulation to test the robustness of our conclusions touncertainty of input parameter values. The simulation varied all inputs simultaneously and calculated andplotted expected costs and outcomes of each iteration. One thousand draws were taken from the parameterdistribution to reach convergence where the probability of being cost-effective did not change further.
R E S U L T S
In the base case cost-effectiveness analysis, targeting an Hb level of 10-11 g/dL compared with9-10 g/dL resulted in lower costs and improved outcomes (QALYs) over 5 years. There was also an
Health Outcomes Research in Medicine - Vol. 3 / No. 2 / May 2012 e85
average reduction of 0.51 hospitalizations per patient, resulting in cost offsets of $15,340 (Table 2).Figure 1A shows the results of the PSA and each iteration’s incremental costs and outcomes. ThePSA showed consistent results to the base case deterministic model with incremental costs of
- T A B L E 2 : Expected Costs and Effectiveness of Hb10-11 g/dL and 9-10 g/dL per Patient Over 5 Years*
Hb 10-11 g/dL Hb 9-10 g/dL
Life years 2.94 2.94QALYs 1.71 1.63CostsOutpatient care for patients
without transplant52,512 52,512
Outpatient care for patientswith functioning graft
16,201 16,201
Outpatient care for patientswith failed transplant
Hb 10-11 g/dL vs 9-10 g/dLIncremental costs Cost savings
�$13,105 and incremental QALYs of0.10 years, while the deterministicmodel showed costs of �$13,149and QALYs of 0.09 years. Fifty-eightpercent of the iterations showed tar-geting Hb 10-11 g/dL resulted in incre-mental cost-effectiveness ratios (ratioof the difference in cost between Hb10-11 g/dL and Hb 9-10 g/dL to thedifference in QALYs gained) thatwere below the willingness-to-paythreshold of $50,000 per QALY gained(the generally accepted thresholdrange of cost-effectiveness in theUS), and 60% below $100,000 perQALY gained. Thus, Hb 10-11 g/dL islikely to be cost-effective relative to9-10 g/dL, if our assumptions aboutsurvival, hospitalization, and healthutility are accurate. Figure 1B showsa cost-effectiveness acceptability curve,which plots the probability that Hb 10-11 g/dL is cost-effective at differentwillingness-to-pay thresholds.
Hb¼ hemoglobin; QALYs¼quality-adjusted life years.*All costs and effects were discounted at 3.5% per year. Incremental cost-effectiveness ratio was not calculated because Hb 10-11 g/dL cohort wascost saving compared with Hb 9-10 g/dL.
Sensi t iv i ty Analys is
Figure 2 presents the results ofthe sensitivity analysis and shows thedifference in expected costs andoutcomes between Hb 10-11 g/dL
versus 9-10 g/dL. Assumptions involving parameters for mortality, hospitalization, health utility, and timehorizon had the most influence on cost-effectiveness of Hb 10-11 g/dL (40%-166% change in incrementalcosts, 33%-122% change in incremental QALYs).
Major changes in model results and conclusions were observed when testing assumptions used inprior models. Assuming higher mortality with higher Hb levels with stable hospitalization rates and utilityacross Hb levels, as suggested by Clement et al,22 resulted in Hb 10-11 g/dL being more costly and less effectivethan 9-10 g/dL. Assuming equal hospitalization rates in the 2 groups, as suggested by Tonelli et al,21 resulted inHb 10-11 g/dL being more costly and more effective than 9-10 g/dL. However, the incremental cost-effectiveness ratio was $24,667 per QALY gained, indicating it was cost-effective relative to widely citedUS standards.
Assuming mortality and utility improved at higher Hb levels, while hospitalization remained thesame across Hb levels, as done by the National Health Services,32 resulted in Hb 10-11 g/dL being morecostly and more effective than 9-10 g/dL, with an incremental cost-effectiveness ratio that remained belowthe upper threshold for cost-effectiveness in the US ($62,171 per QALY gained). Results of the tested
F I G U R E 1 : Cost-effectiveness analyses. (A) Cost-effectiveness plane showing Monte Carlo probabilisticsensitivity analysis simulations when comparing Hb of10-11 g/dLwith 9-10 g/dL. The quadrants are interpretedas 1) upper right represents higher cost and higherQALY; 2) lower right represents lower cost and higherQALY; 3) upper left represents higher cost and lowerQALY; and 4) lower left represents lower cost and lowerQALY. The number of iterations reached convergence at1000 iterations with each data point on the figurerepresenting a single iteration of the 1000 iterationssimulated. (B) Cost-effectiveness acceptability curveshowing Monte Carlo probabilistic sensitivity analysisresults relative to different willingness to pay thresholdswhen comparing Hb of 10-11 g/dL with 9-10 g/dL.
e86 Epoetin Alfa for Treating Chronic Anemia
assumptions were all consistent withthe conclusions of prior models whencomparing higher Hb versus lowerHb levels.
We also explored the impactof higher allosensitization by loweringthe rate of transplantation for the Hb9-10 g/dL cohort by 25% (changefrom 40 per 1000 patient-years to 30per 1000 patient-years), which hada minor impact on cost-effectiveness(4% change in incremental costs,11% change in incremental QALYs).This is likely due to the small startingrate of transplantation in the Hb9-10 g/dL cohort.
Cost-effectiveness of main-taining Hb 10-11 g/dL was also shownto improve as the time horizon of theanalysis increased. Time horizon wasset to patient lifetime, which resultedin average reductions of 0.78 hospitali-zations and increases of 0.12 QALY perpatient, with hospitalization costoffsets of $21,450 compared with Hbof 9-10 g/dL.
D I S C U S S I O N
Our base case analysis sug-gested that Hb levels of 10-11 g/dLusing epoetin alfa may result in betterpatient outcomes and lower Medicarecosts than lower Hb levels. Under base-line model assumptions, cost savingsfrom Hb levels of 10-11 g/dL wasmostly attributed to improvements inhospitalization and health state (healthutility) preference. However, as shownin our sensitivity analyses, even ifhospitalization were equal across Hblevels, health utility alone would leadto the conclusion that Hb 10-11 g/dLversus 9-10 g/dL would be cost-effective.
The sensitivity analysis high-lighted the uncertainty of assumptionsinvolving survival, hospitalization,
Health Outcomes Research in Medicine - Vol. 3 / No. 2 / May 2012 e87
and quality of life made in prior models and how these relatively influential parameters can affect cost-effectiveness conclusions. Until the effects of lower Hb levels are examined in today’s dialysis population,there will remain uncertainty in cost-effectiveness analyses and, more importantly, in real-world benefits oftargeting specific Hb levels.
F I G U R E 2 : Sensitivity analysis. Effects of differentmodel assumptions on cost-effectiveness whencomparing Hb of 10-11 g/dL with 9-10 g/dL are shown;base case results indicated by intersection of dashedlines. Gray shaded region represents conditions in whichHb 10-11 vs 9-10 g/dL was no longer cost-effective.Hb¼ hemoglobin; QALYs¼ quality-adjusted life years.
The sensitivity analysisshowed that the effect of epoetin alfain decreasing risk of allosensitizationand improving transplantation rateshad a minor influence on cost-effectiveness because of a low propor-tion of hemodialysis population thatundergoes transplantation. In addition,in a higher Hb target era (datacurrently used in model), it is likelythat patients with very low Hb areless likely to be transplanted, althoughage is likely a major predictor.However, if overall Hb targets arelower, low Hb may very well be less ofa predictor of transplantability. Thus,the low transplantation rate assump-tions may not be applicable in today’spopulation.
Our findings should notundervalue the clinical and economicbenefits of ESA use in patients ontransplant waiting lists, because trans-fusions increase the risk of allosensitiza-tion, thus increasing the risk for acuterejection and shortened graftsurvival.34
The primary limitations ofthis analysis stem from a reliance on
observational studies to inform major drivers of the model, however, we have considered randomized clinicaltrial data where possible for the base case analysis to limit this potential bias. The epoetin alfa dose curve wasextrapolated from a mixture of data in which the dose response has not been formally established. Moreover,we recognize that this relationship is based on averages from populations of patients with varying responses toepoetin alfa and thus may not reflect treatment at the patient level.
The results of our analysis have several potential policy implications. Although our analysis showsthe benefits of targeting Hb levels of 10-11 g/dL, there are strong economic incentives to target lower Hblevels. Moreover, several factors, specifically Medicare’s bundling of ESA use into a capitated payment,have raised uncertainty about the economic value of ESA use. Since the initiation of the bundled paymentsystem in the US, there has also been a trend for targeting lower Hb levels, increasing use of iron anddecreasing ESA use.35,36 The long-term clinical and economic consequences of such practice pattern changesrequire evaluation, especially the effects on Hb levels and subsequent patient outcomes.
Clinical trials have shown the danger of targeting Hb level$13 g/dL, but very little is known aboutthe clinical consequences of Hb to levels <10 g/dL in today’s population. Thus, possible consequences ofthese policy changes are unclear at best.
1. Amgen ReInc; 2011
2. Churchillmorbidity
3. Eschbachropoietinmulticent
4. Crowleylong-term
5. US RenaKidney DMD: NatDigestive
6. Aalten J,transplancome: a2559-256
7. CamposPestana Jbodies as2006;6(10
8. Custer Bmedicine18(2):127
9. Despotisrelated pr21(1):147
10. US RenaKidney DMD: NatDigestive
11. Vella JP,human leetin. Nep
e88 Epoetin Alfa for Treating Chronic Anemia
Our analysis suggests that a target Hb 10-11 g/dL may result in better patient outcomes withlower overall costs for Medicare than Hb levels below 10 g/dL. Using different assumptions about theeffects of various Hb ranges on patient outcomes, hospitalization rates, and health preference fromearlier models, different conclusions were found regarding patient outcomes and cost-effectiveness.Uncertainty about the cost-effectiveness of Hb targets will remain until there is new research in today’sdialysis population that assesses the effects of targeting to lower Hb levels and validates our modelinputs and assumptions.
SUPPLEMENTARY DATASupplementary data associated with this article can be found, in the online version, at doi:10.1016/
j.ehrm.2012.03.004.
Corresponding Author: Christopher S. Hollenbeak, PhD, Penn State College of Medicine, 600 Centerview Drive,A210, Hershey, PA 17033.E-mail address: [email protected] study was funded by Amgen Inc.P.Q., W.M.M., S.M., J.J., G.D.L., and C.S.H. are consultants for Amgen. M.G. and J.I. are employees of Amgenand may own stock or stock options.
R E F E R E N C E S
12. Amgen Mamgen.coAccessed
13. US Food[Aranesp(11/2006Health &DrugSafeucm1264
14. US FoodRecommenof HealtNewsEvencessed Jul
15. Besarab Acomparedwho are r584-590.
16. Singh AKanemia w355(20):2
17. Pfeffer Mof darbepMed. 200
18. SolomonEvents wresponseMed. 201
19. Menzin J,tiveness oPharmaco
20. Lee CP, Cupdating12(1):80-
sponse to 2011 CMS QIP Proposal. Thousand Oaks, CA: Amgen.DN, Taylor DW, Cook RJ, et al. Canadian hemodialysisstudy. Am J Kidney Dis. 1992;19(3):214-234.JW,AbdulhadiMH, Browne JK, et al. Recombinant human eryth-inanemicpatientswithend-stage renal disease.Results of a phase IIIer clinical trial. Ann Intern Med. 1989;111(12):992-1000.JP, Nealey TA, Metzger J, Pono L, Chazan JA. Transfusion andhemodialysis. Arch Intern Med. 1987;147(11):1925-1928.
l Data System.USRDS 2010 Annual Data Report: Atlas of Chronicisease and End-Stage Renal Disease in the United States. Bethesda,ional Institutes of Health, National Institute of Diabetes andand Kidney Diseases; 2010.Bemelman FJ, van den Berg-Loonen EM, et al. Pre-kidney-
t blood transfusions do not improve transplantation out-Dutch national study. Nephrol Dial Transplant. 2009;24(8):6.EF, Tedesco-Silva H, Machado PG, Franco M, Medina-O, Gerbase-DeLima M. Post-transplant anti-HLA class II anti-risk factor for late kidney allograft failure. Am J Transplant.):2316-2320.. Economic analyses of blood safety and transfusioninterventions: a systematic review. Transfus Med Rev. 2004;-143.GJ, Zhang L, Lublin DM. Transfusion risks and transfusion-o-inflammatory responses. Hematol Oncol Clin North Am. 2007;-161.l Data System.USRDS 2008 Annual Data Report: Atlas of Chronicisease and End-Stage Renal Disease in the United States. Bethesda,ional Institutes of Health, National Institute of Diabetes andand Kidney Diseases; 2008.O’Neill D, Atkins N, Donohoe JF, Walshe JJ. Sensitization toukocyte antigen before and after the introduction of erythropoi-hrol Dial Transplant. 1998;13(8):2027-2032.
EDCAC briefing. Amgen Inc.; 2011. Available at: http://www.m/pdfs/misc/AMGEN_MEDCAC_Document_02_22_2010.pdf.August 23, 2011.& Drug Administration. Erythropoiesis Stimulating Agents (ESA)(darbepoetin), Epogen (epoetin alfa), and Procrit (epoetin alfa)] e) - Healthcare Professional Sheet text version. US Department of
Human Services. Available at: http://www.fda.gov/Drugs/ty/PostmarketDrugSafetyInformationforPatientsandProviders/88.htm; 2006. Accessed July 26, 2011.& Drug Administration. FDANews Release: FDAModifies Dosingdations for Erythropoiesis-stimulating Agents. US Departmenth & Human Services. Available from: http://www.fda.gov/ts/Newsroom/PressAnnouncements/ucm260670.htm; 2011. Ac-y 26, 2011., Bolton WK, Browne JK, et al. The effects of normal aswith low hematocrit values in patients with cardiac disease
eceiving hemodialysis and epoetin. N Engl J Med. 1998;339(9):
, Szczech L, Tang KL, et al.; CHOIR Investigators. Correction ofith epoetin alfa in chronic kidney disease. N Engl J Med. 2006;085-2098.A, Burdmann EA, Chen CY, et al.; TREAT Investigators. A trialoetin alfa in type 2 diabetes and chronic kidney disease. N Engl J9;361(21):2019-2032.SD, Uno H, Lewis EF, et al.; Trial to Reduce Cardiovascularith Aranesp Therapy (TREAT) Investigators. Erythropoieticand outcomes in kidney disease and type 2 diabetes. N Engl J0;363(12):1146-1155.Lines LM,Weiner DE, et al. A review of the costs and cost effec-f interventions in chronic kidney disease: implications for policy.economics. 2011;29(10):839-861.hertow GM, Zenios SA. An empiric estimate of the value of life:the renal dialysis cost-effectiveness standard. Value Health. 2009;87.
Health Outcomes Research in Medicine - Vol. 3 / No. 2 / May 2012 e89
R E F E R E N C E S Continued
, Winkelmayer WC, Jindal KK, OwenWF, Manns BJ. The cost-ess of maintaining higher hemoglobin targets with erythropoietinialysis patients. Kidney Int. 2003;64(1):295-304.FM, Klarenbach S, Tonelli M, Wiebe N, Hemmelgarn B,J. An economic evaluation of erythropoiesis-stimulating agentsAm J Kidney Dis. 2010;56(6):1050-1061.N, Labrecque J, Lacson E, Keen M, Lazarus JM. The effects ofmoglobin levels on mortality and hospitalization in hemodialysisKidney Int. 2003;63(5):1908-1914.D, Keown P, Laupacis A, et al. Association between re-t human erythropoietin and quality of life and exerciseof patients receiving haemodialysis. BMJ. 1990;300(6724):
, Parfrey PS, Morgan J, et al. Effect of hemoglobin levels in hemo-atients with asymptomatic cardiomyopathy. Kidney Int. 2000;5-1335.A, Keown P, Pus N, et al. A study of the quality of life and cost-renal transplantation. Kidney Int. 1996;50(1):235-242.t. Healthcare Cost and Utilization Project. Agency for Health-arch andQuality, Rockville,MD. 2008National Statistics. Statis-ecific procedures by CCS. Available at: http://hcupnet.ahrq.gov/.March 14, 2011.S, Foley RN, Wittreich BH, Sullivan DJ, Zagari MJ, Frei D.lind comparison of full and partial anemia correction in incidentysis patients without symptomatic heart disease. J Am SocNephrol.):2180-2189.M, Lobo PI, Taylor RP, et al. Prospective study of the immunenormalizing the hemoglobin concentration in hemodialysis
ho receive recombinant human erythropoietin. J Am Soc Neph-15(5):1339-1346.J. Influence of target hemoglobin in dialysis patients on morbidityality. Kidney Int Suppl. 2002;80:44-48.L, Kopple JD, Kovesdy CP, et al. Associations between changeslobin and administered erythropoiesis-stimulating agent andin hemodialysis patients. J Am Soc Nephrol. 2006;17(4):1.Collaborating Centre for Chronic Conditions. Anaemia
ent in Chronic Kidney Disease: National Clinical Guideline forent in Adults and Children. London, UK: National ClinicalCentre, National Institute for Health and Clinical Excel-
06.JM, Chuang CC, Turenne M, et al. Association of quarterlychieved hematocrit with mortality in dialysis patients: a time-t comorbidity-adjusted model. Am J Kidney Dis. 2009;53(3):
Improved kidney graft survival in nontransfused recipients.Trans-. 1987;19(1 Pt 1):149-152.ayer WC. Potential effects of the new Medicare ProspectiveSystem on drug prescription in end-stage renal disease care. Blood1;31(1-3):66-69.ractice Monitor. Dialysis Outcomes and Practice Patterns Study. Arbor Research Collaborative for Health; 2011. Available at:w.dopps.org/DPM/. Accessed August 16, 2011.N, Ishani A, Foley RN, Guo H, Liu J, Collins AJ. Temporalred blood transfusion among US dialysis patients, 1992-2005.ey Dis. 2008;52(6):1115-1121.
