cost effectiveness of darunavir/ritonavir in highly treatment-experienced, hiv-1-infected adults in...

Cost Effectiveness of Darunavir/Ritonavirin Highly Treatment-Experienced,HIV-1-Infected Adults in the USAJosephine Mauskopf,1 Anita Brogan,1 Silas Martin2 and Erik Smets3

1 RTI Health Solutions, Research Triangle Park, North Carolina, USA

2 Johnson & Johnson Pharmaceutical Services, New Jersey, USA

3 Johnson & Johnson Pharmaceutical Services LLC, Beerse, Belgium

Abstract Introduction: Darunavir is a new protease inhibitor (PI) that is co-adminis-

tered with low-dose ritonavir and has demonstrated substantial efficacy

in clinical trials of highly treatment-experienced patients when combined with

an optimized background regimen (with or without enfuvirtide). This study

estimates the cost effectiveness of darunavir with ritonavir (DRV/r) in this

population over 5-year and lifetime time horizons in the USA.

Methods:AMarkovmodel was used to follow a treatment-experiencedHIV-1

cohort through six health states, based on CD4 cell count: greater than 500,

351–500, 201–350, 101–200, 51–100 and 0–50 cells/mm3, and death. The mag-

nitude of the CD4 cell count increase and duration of increasing and stable

periodswere derived fromweek 48DRV/r clinical trial results (POWER1 and 2).

The treatment pathway assumed one regimen switch following treatment failure

on the initial regimen. The use of antiretroviral drugs was based on usage in

DRV/r clinical trials. US daily wholesale acquisition costs were calculated using

the recommended daily doses. For each CD4 cell count range, utility values,

HIV-1-related mortality rates and costs for medical resources (other than antire-

troviral drug costs) were obtained from published literature. Non-HIV-1-related

mortality rates were calculated by applying a relative risk value to theUS general

population age and gender-specific mortality rates. Costs and outcomes were

discounted at 3% per year. One-way and probabilistic sensitivity analyses and

variability analysis were performed.

Results: In a 5-year analysis, patients receivingDRV/r experienced 3.80 quality-adjusted life-years (QALYs) and incurred total medical costs of US$217 288,

while those receiving control PIs experienced 3.60 QALYs and incurred costs of

US$218962. DRV/r was both more effective and less costly than control PIs.

For the lifetime analysis, the QALYs and lifetime medical costs with DRV/rwere 10.03 and US$565358, compared with 8.76 and US$527 287 with control

PIs. The incremental cost-effectiveness ratio for DRV/r compared with control

ORIGINAL RESEARCH ARTICLEPharmacoeconomics 2010; 28 Suppl. 1: 83-105

1170-7690/10/0001-0083/$49.95/0

ª 2010 Adis Data Information BV. All rights reserved.

PIs was US$30046. One-way sensitivity analyses for both time horizons

indicated that the results were most sensitive to changes in the rate of CD4 cell

count change during stable and declining periods (lifetime only), duration of

stable period (5-year only) andHIV-1-related mortality rates. The results of the

variability analysis weremost sensitive to themodel time horizon. Nevertheless,

for all ranges and scenarios tested in these analyses, the incremental cost perQALY

gained remained below US$50000. The probabilistic sensitivity analysis showed

that there was a 0.921 and 0.950 probability of a cost-effectiveness ratio below

US$50000 per QALY for the 5-year and lifetime time horizon, respectively.

Conclusions:DRV/r is predicted to be cost effective compared with control PI

in highly treatment-experienced patients and is predicted to yield an average

of 0.20 additional QALYs per treatment-experienced patient over 5 years and

1.27 additional QALYs over a lifetime in this population.

Introduction

In the USA, the number of people living withHIV/AIDS has been increasing steadily despite adecrease in the incidence of new cases of HIVinfection, which peaked at 150 000 per year in the1980s and decreased to approximately 40 000 peryear in the 2000s.[1] The increase seen in the numberof people living with HIV/AIDS is attributable toadvances in HIV/AIDS management over the past20 years that have reduced morbidity, decreasedmortality and extended the life expectancy of thosediagnosed with HIV infection.[2-5]

Highly active antiretroviral therapy (HAART)regimens, combinations of at least three drugsrepresenting two or more drug classes, are veryeffective at reducing the plasmaHIV-1-RNA levelto below detectable limits and at increasing pa-tients’ immunity (CD4 cell count levels) to levelsat which the risk of the development of HIV-induced opportunistic infections is rare, therebysignificantly improving quality of life and life ex-pectancy. Nevertheless, patients eventually fail and/or discontinue their initial and subsequentHAARTregimens, generally becoming resistant to currentlyavailable drugs, leading to progressive immune de-ficiency, clinical disease progression and, ultimate-ly, premature death.[6-11] There is thus a continuingneed for new antiretroviral drugs, from new or ex-isting drug classes, which are effective against theHIV virus.

In addition to the human toll, HIV infectionexacts an economic burden in direct healthcare coststhat increases as the extent of the HIV-induced im-mune deficiency progresses. Schackman and col-leagues,[4] using a computer simulation model andpublished data, have estimated the lifetime costs ofcare for HIV infection in 2004 from diagnosis todeath to be US$385200, discounted at 3%, andUS$618900, undiscounted, associated with a life ex-pectancy of 24.2 years. Chen and colleagues[12] esti-mated annual costs for 2001 for the treatment ofHIV, depending on the CD4 cell count range, of be-tween US$13885 (CD4 cell count >350 cells/mm3)and US$36532 (CD4 cell count <50 cells/mm3).Based on the population distribution in the study byChen et al., the overall annual cost per person wasUS$18640, with 56% of the costs attributable toantiretroviral drugs. Gebo and colleagues[13] esti-matedmore recent annual costs for 2003 of betweenUS$21869 (for a patient with a CD4 cell count>500 cells/mm3) and US$57565 (CD4 cell count<50 cells/mm3), with an overall mean cost per per-son ofUS$28594, of which 45%was for antiretrovi-ral drugs.

In June 2006, darunavir, a protease inhibitor(PI) with a high genetic barrier to the develop-ment of resistance, received worldwide approvalfor the treatment of HIV-1 infection in treatment-experienced adults, such as those resistant tomore than one PI.[14] Darunavir co-administeredwith low-dose ritonavir (DRV/r) has been shown

84 Mauskopf et al.

ª 2010 Adis Data Information BV. All rights reserved. Pharmacoeconomics 2010; 28 Suppl. 1

in clinical trials [POWER 1 and 2 (TMC114-C213and C202; Performance Of TMC114/r Whenevaluated in treatment-Experienced patients withprotease inhibitor Resistance)] to have bettervirological and immunological efficacy thaninvestigator-selected control PIs in highly treat-ment-experienced patients [all patients receivedan optimized background regimen (OBR), with orwithout enfuvirtide]. In particular, significantlymore DRV/r patients experienced complete viralsuppression at both 24 and 48 weeks than controlPI patients.[15]

When a new antiretroviral drug, such as dar-unavir, is introduced, it is critical to understandthe impact that the new drug will have on boththe lifetime healthcare costs and the health con-sequences for people with HIV-1. These impactsdetermine its value for money and its likely im-pact on healthcare budgets. An understanding ofthe value for money can help healthcare decision-makers (including both private payers and publicpayers such as Medicaid and the AIDS drug as-sistance programmes) to identify the appropriateplace in the treatment pathway for the new drug,given finite healthcare budgets.

In this paper, the cost effectiveness of DRV/rcompared with control PIs in highly treatment-experienced patients was estimated using a Mar-kov model. The cost-effectiveness ratios werecomputed for both 5-year and lifetime modeltime horizons in the USA. In addition, in order tounderstand fully the impact of uncertainty on theresults, an extensive one-way sensitivity analysis,probabilistic sensitivity analysis (PSA), and ana-lyses demonstrating the impact of variability inpractice patterns, population characteristics andmodelling assumptions, were conducted.

Methods

POWER 1 and 2 Design Details

In line with the latest US and international treat-ment guidelines and current standard of care,[9-11,16]

investigators in the POWER 1 and 2 trials wereasked to select, before randomization, what theyconsidered to be the best available PI-based regi-men and the best available OBR [nucleoside reverse

transcriptase inhibitors (NRTIs) with or withoutenfuvirtide] based on the previous antiretroviralhistory, genotypic resistance testing (determinedby VirtualPhenotype) and patient preference. Fortrial participants randomly assigned to the DRV/rtreatment group, the investigator-selected PI wasreplaced by DRV/r. PIs that could be selectedincluded lopinavir/ritonavir (LPV/r), amprenavir,atazanavir, saquinavir, indinavir, nelfinavir andboosting dosages of ritonavir. The recently ap-proved ritonavir-boosted PI tipranavir (TPV/r) wasnot included as a treatment option in the POWER1 and 2 clinical trials because it was not commer-cially available at the time the trials were initiated.

Model Overview

A disease progression model with a Markovstructure and a 3-month cycle time was chosen toassess the cost effectiveness of DRV/r comparedwith currently available, investigator-selected con-trol PIs used in the POWER studies in a patientcohort that had failed to respond to previous anti-retroviral therapy (ART), including NRTIs, non-nucleoside reverse transcriptase inhibitors (NNRTIs)and PIs.

Themodel simulates disease progression in thisHIV-1-infected population, by simulating theirevolution over different health states defined byCD4 cell count ranges (>500, 351–500, 201–350,101–200, 51–100 and 0–50 cells/mm3) and death(figure 1). These health states/CD4 cell countranges were chosen based on their clinical re-levance in terms of impact on the risk of oppor-tunistic infections, other symptoms and death andclinical decision-making in the therapeutic man-agement of HIV-1-infected individuals.[11,16-19]

The modelled highly treatment-experiencedpatients entered the model in a mix of CD4cell count ranges that were reflective of the base-line CD4 cell count distribution of the pooledPOWER1and 2 study populations. Every 3months,participants either remained in their current state ormoved to any other state of the model, includingdeath. The arrows in the figure display the possiblemovements of participants with CD4 cell counts be-tween 201 and 350cells/mm3. Similar movementswere possible for participants in the other states of

Cost Effectiveness of DRV/r in the USA 85


the model, except death, which was an absorbingstate. Depending on the model input parameters, theprobability of some transitions may be zero.

AUS societal perspective was used in the mod-el with direct medical costs in the numerator andquality-adjusted life-years (QALYs) in the de-nominator. The indirect costs are assumed to becaptured in the QALY estimates as recommen-ded for the reference case byGold et al.[20] In eachCD4 cell count range, participants incurred directmedical costs based on estimates of healthcareservices used during that health state, taken frompublished and unpublished US costing studiesusing US databases. Each CD4 cell count rangein the model also had an associated utility valuethat quantified mean individual wellbeing for thatCD4 cell count range on a scale ranging from zero(worst possible health) to one (perfect health) ta-ken from published US data sources. These utilityvalues were used to convert the time spent in eachhealth state over an individual’s remaining life-time into estimates of QALYs.

There were no significant differences in ad-verse events (AEs) between the DRV/r and con-trol PI groups in the POWER 1 and 2 clinicaltrials.[15] Therefore, the model did not make spe-cific adjustments to costs and quality of life toaccount for AEs beyond those that were alreadyaccounted for by the cost and utility estimates forthe CD4 cell count ranges.

The model allowed for death from any of theCD4 cell count ranges. Mortality rates were in-cluded in the model for both HIV-1-related andnon-HIV-1-related deaths. TheHIV-1-related deathswere assumed to vary by CD4 cell count rangeand were taken from published database studies.[21]

Non-HIV-1-related deaths were assumed to vary byage andwere taken fromUS life tableswith a relativerisk value applied to account for higher generalmortality rates among those withHIV infection thanamong those without such infection.[22]

As tipranavir was not included as a treatmentoption in the POWER trials and in recognition ofthe fact that TPV/r is currently being used in theUSA in the population also eligible for DRV/r, ascenario analysis was performed in which TPV/rwas included in the comparator/control group toassess the impact of this alternative treatmentstrategy on the cost-effectiveness results.

The model compared DRV/r plus an OBRwith control PIs plus OBR as the first treatmentregimen in the base-case analysis (figure 2).

The evolution of the CD4 cell counts (and thusthe transition through the different health statesof the model) of modelled individuals is depen-dent on the extent of the treatment-induced virol-ogical suppression and is programmed separatelyin the model for different categories of viral re-sponse. The model defined three levels of virolo-gical response at 24 weeks from the beginning of anew regimen: complete suppression (<50 copies/mL), partial suppression (between 50 copies/mLand a ‡1 log10 reduction in HIV-1 RNA, referredto hereafter as 50 copies/mL to ‡1 log10 reduc-tion) and no suppression (<1 log10 reduction inHIV-1 RNA). These virological response cate-gories were chosen because they were correlatedwith both the magnitude of increase in the CD4cell count and the likely duration of response totreatment.[9,23,24]

CD4201−350

CD4101−200

CD451−100

CD40−50

CD4 >500

CD4351−500

Death

Fig. 1. Schematic diagram of the cost-effectiveness model: illus-tration of the transitions for individuals in the 201–350 cells/mm3 CD4cell count range.

86 Mauskopf et al.


To estimate the transitions between the CD4cell count ranges for the remaining lifetime for in-dividuals in each virological response category,the model assumed that after starting a new treat-ment each individual would experience an initialperiod of increasing CD4 cell count, followed bya period of either more slowly increasing or stableCD4 cell count, followed by a period of decreas-ing CD4 cell count. This is in line with the short-term and long-term kinetics of the increase of theCD4 cell count that is usually observed after theinitiation of effective HAART and the progres-sive immune deficiency that eventually developswhen these regimens no longermanage tomaintaintheir inhibitory effect on viral replication.[23,25]

For each virological response category, boththe duration of the three time periods of CD4 cellcount change and the rate of CD4 cell count changeduring those time periods were estimated using theobserved CD4 cell count response to the trial regi-mens during the first 48 weeks of their use.

Individuals on the first treatment regimen in themodel were assumed to change treatment to theswitch regimen after a period of decreasing CD4cell count. In the model, the primary indicator forswitching from the initial regimen was assumed tobe a decline in CD4 cell count. This has been thetreatment practice for the treatment-experiencedpopulation in recent years because of the limitednumber of effective alternative regimens.[23,26]With

the increased number of treatment options, how-ever, switching may now occur more rapidly aftervirological failure and before the CD4 cell countbegins to decline. Themodel structure was designedto allow for the possibility of switching before theCD4 cell count begins to decline. Individuals on theswitch regimen were assumed to remain on thatregimen until death, despite declining CD4 cellcounts. In the base-case analysis, all individualswere assumed to switch to a regimen of tipranavir(500mg twice daily; bid) boosted with ritonavir(200mg bid) plus OBR (figure 2).

Using the clinical trial data, the clinical efficacyinputs (viral response andCD4 cell count increases)for each of the treatment options included in themodel have been calculated separately for patientsreceiving enfuvirtide and patients not receiving en-fuvirtide. The model calculates the weighted aver-age viral response rates and CD4 cell transitionprobabilities for each comparator based on the useof enfuvirtide observed in the clinical trials.

Because enfuvirtide is likely to be reimbursedby third-party payers only while continuing to beeffective, it was assumed in the base-case analysisthat the drug would be discontinued in the initialor switch regimen after a 6-month period of de-creasing CD4 cell count. For the initial regimen,in the base-case analysis scenario, this time ofdiscontinuation of enfuvirtide coincides with thetime of therapy switch.

First regimen Switch regimen

Treatmentfailure

andswitch

Darunavir/r

Death

Tipranavir/r

Tipranavir 500 mg bid+

Ritonavir 200 mg bid+

OBR

Darunavir/r 600 mg bid+

Ritonavir 100 mg bid+

OBR

Control

Control PIs based onresistance testing

+Ritonavir if chosen

+OBR

Fig. 2. Model structure, including primary and switch comparisons. Note: The OBR is selected based on resistance testing. In the model, theOBR selected included nucleotide reverse transcriptase inhibitors and the fusion inhibitor, enfuvirtide. Treatment failure was defined as eitherdetectible HIV-1-RNA levels or a decline in the CD4 cell count or both; switch may occur before or several time periods after the start of CD4 cellcount decline. OBR =optimized background regimen; PIs =protease inhibitors.



Input Parameter Values

Baseline Population Characteristics

The population entering the model was as-sumed to be equivalent to that enrolling in thePOWER 1 and 2 clinical trials in terms of gender,age distribution and CD4 cell count distribution.The values used in the model are from the pooledpopulation and are shown in table I.[15,27]

Transition Probabilities

The transition probabilities across the CD4cell count range health states were estimated se-parately for the first treatment regimens and theswitch regimen as well as by virological responseat 24 weeks. For each first regimen and for theswitch regimen, clinical trial data were used toestimate the proportion of trial participants ineach virological response category and the dura-tion and magnitude of the CD4 cell count chan-ges for each virological response category. Thesedata then were used to calculate the 3-month tran-sition probabilities used in the Markov model.Each of these steps is described in the subsectionsbelow.

Virological Response Rates: First Regimens

Intention-to-treat data from the two random-ized, open-label, controlled, clinical trials (POWER1and 2) were used for the base-case analysis for theDRV/r and the control first treatment regimens. Toallow for the transition rates to vary according to theearly virological response to treatment, the clinicaltrial populations were subdivided by their virologicalresponse to treatment at 24 weeks as follows:� Response to less than 50 copies/mL;� Response of 50 copies to 1 log10 reduction or

greater; and� Response of less than 1 log10 reduction inHIV-1

RNA.Virological response categories were estimated

for the DRV/r and control regimens from thepooled POWER 1 and 2 trial data for individualstaking enfuvirtide (45.8% use forDRV/r and 41.9%use for control) and those not taking enfuvirtide(table II).[28-31]

For the TPV/r switch regimen, published datawere used from the RESIST 1 and 2 (Random-ized Evaluation of Strategic Intervention in

multi-drug reSistant patients with Tipranavir)trials to estimate the HIV-1-RNA outcomes at24 weeks for TPV/r and, separately, for individ-uals taking enfuvirtide (27.1%) and those nottaking enfuvirtide[29-31] (table II).

Duration of Immunological Response: First andSwitch Regimens

Three stages of CD4 cell count change for eachindividual after starting a new treatment werealso assumed in the model:1. Initial, rapidly increasing CD4 cell count;2. Stable or slowly increasing CD4 cell count;and3. Declining CD4 cell count.

Table III presents the durations used in thebase-case analysis for each stage of CD4 cell countchange for the first treatment regimens:1. The durations for the initial rapid rate of in-crease were based on the observed changes inCD4 cell count during the first 48 weeks of theclinical trials;2. The durations for the stable or slower rate ofincrease were based on long-term studies that in-dicated continuing increases in the CD4 cell countover several years for individuals who achievedundetectable HIV-1-RNA levels compared withthose who did not;[32-36]

Table I. Population characteristics for all participants in the

POWER 1 and 2 clinical trials[15,27]

Input parameter Pooled POWER 1 and 2 data (%)

Gender

Male 88.6

Female 11.4

Age (years)

20–39 27.5

40–64 70.6

65+ 2.0

CD4 cell count range

(cells/mm3)

0–50 23.1

51–100 15.3

101–200 24.7

201–350 18.8

351–500 9.8

>500 8.2

POWER =Performance Of TMC114/r When evaluated in treatment-

Experienced patients with protease inhibitor Resistance.

88 Mauskopf et al.


3. The duration for the decline in CD4 cell countbefore switching was assumed to be 0.5 years inthe base-case analysis. It was assumed that a switchdid not occur until after the CD4 cell count startedto decline because of the limited availability ofalternative treatments.

Table III also includes the assumptions aboutthe durations for the switch TPV/r regimen. Thedurations for the switch TPV/r regimen were as-sumed to be the same as the durations for the firstcontrol regimen for all virological response groupsfor both the initial and the stable or slowly in-creasingCD4 cell count change periods. The dura-tion of the declining CD4 cell count period for the

switch regimen was assumed to equal the individ-ual’s remaining lifetime, because there was assumedto be no alternative regimen.

Magnitude of Immunological Response:First and Switch Regimens

The magnitudes and variabilities of the initialrapid increase and the subsequent slower increasein CD4 cell counts for the first regimens, by virol-ogical response category and attributable to eachtreatment regimen, were based on the meansand standard deviations for the CD4 cell countincreases at 24 weeks and 48 weeks observedin the pooled POWER 1 and 2 clinical trials’

Table II. Virological response rates at 24 weeks for first regimens from pooled POWER 1 and 2, and RESIST 1 and 2 clinical trials with and

without enfuvirtide[28-31]

Virological response categories DRV/r(n =131; %)

Control

(n =124; %)

TPV/r(n = 584; %)

Proportion of patients using enfuvirtide 45.8 41.9 27.1

‡1 log10 reduction 75.0 25.0 58.2

<50 copies/mL 43.3 13.5 23.9

50 copies to ‡1 log10 reduction 31.7 11.5 34.3

<1 log10 reduction 25.0 75.0 41.8

Proportion of patients not using enfuvirtide 54.2 58.1 72.9

‡1 log10 reduction 66.2 18.0 34.9

<50 copies/mL 46.5 11.1 23.9

50 copies to ‡1 log10 reduction 19.7 6.9 11.0

<1 log10 reduction 33.8 81.9 65.1

DRV/r =darunavir/ritonavir; POWER =Performance Of TMC114/r When evaluated in treatment-Experienced patients with protease inhibitor

Resistance; RESIST =Randomized Evaluation of Strategic Intervention in multi-drug resiStant patients with Tipranavir; TPV/r = tipranavir/ritonavir.

Table III. Durations of CD4 cell count changes by 24-week virological response for the first and switch treatment regimens[28,32-36]

<50Copies/mL 50Copies/mL to ‡1 log10reduction

<1 Log10 reduction

1. Initial CD4 cell count increase

DRV/r 0.5 years 0.5 years 0.5 years

Control 1 year 0.5 years 0.5 years

TPV/r 0.5 years 0.5 years 0.5 years

2. Stable or slowly increasing CD4 cell count

DRV/r 2 years 0.5 years 0 years

Control 1.5 years 0.5 years 0 years

TPV/r 2.0 years 0.5 years 0 years

3. Declining CD4 cell count before switching from first treatment regimen

DRV/r 0.5 years 0.5 years 0.5 years

Control 0.5 years 0.5 years 0.5 years

TPV/r Remaining lifetime Remaining lifetime Remaining lifetime

DRV/r =darunavir/ritonavir; TPV/r = tipranavir/ritonavir.



intention-to-treat population, adapted to 3-monthvalues. The magnitude of the CD4 cell count res-ponse was assumed to depend on the virologicalresponse category and the regimen, regardless ofwhether or not enfuvirtide was used.

The CD4 cell count changes by virological re-sponse category were not published for the TPV/rswitch regimen. The values for tipranavir were es-timated based on the published trial data from theRESIST 1 and 2 tipranavir trials assuming differ-ential response among virological response cate-gories similar to that observed for DRV/r.[30,31,37]

The rate of decline of CD4 cell count was esti-mated based on data on rates of CD4 cell countdecline in those who are untreated[38] and from astudy presenting estimates of the rate of CD4 cellcount decline for those taking HAART while ex-periencing virological failure.[24]

Table IV presents the estimated 3-month ratesof change of CD4 cell counts for the initial periodof increase, the stable or slowly increasing periodand the period of decline. These rates of changewereused to compute the transition probabilities betweenthe CD4 cell count ranges.

Mortality Rates

Mortality was explicitly included in the eco-nomic model and was estimated based on datafrom large observational database studies andfrom US life tables. People with HIV infection

may experience death due to causes related orunrelated to HIV. For HIV-related causes, themodel used the mortality rates for causes relatedto HIV taken from Mocroft et al.,[21] who esti-mated mortality rates using the EuroSIDAdatabase (table V).

Annual all-cause mortality rates, measuredper 100 000 people, were taken from US 2004 lifetables.[39] These mortality rates were for all causesand thus double-counted the very small numberof annual HIV-related deaths in the total US pop-ulation. The US all-cause mortality rates wereadjusted upwards by a factor of 3.6 based on acomparison in Jensen-Fangel et al.[22] of mortalityrates in people with HIV infection to a matchedpopulation sample without HIV infection.[22]

Using the US all-cause mortality data and theJansen-Fangel relative-risk factor, the weightedaverage non-HIV-related death rates for each ofthe three age ranges included in the model, 20–39,

Table IV. Estimated 3-month mean rate of CD4 cell count change (cells/mm3) by 24-week virological response for the first and switch

regimens[24,28,30,31,37,38]

Treatment regimen <50Copies/mL 50Copies/mL to ‡1 log10reduction

<1 Log10 reduction

1. Initial CD4 cell count increase (cells/mm3)

DRV/r 54 74 24

Control 27 32 4

TPV/r 38 52 17

2. Stable or slowly increasing CD4 cell count (cells/mm3)

DRV/r 6.5 17 Not applicable

Control 6.5 12 Not applicable

TPV/r 0 0 Not applicable

3. Declining CD4 cell count (cells/mm3)

DRV/r 4.6 4.6 4.6

Control 4.6 4.6 4.6

TPV/r 4.6 4.6 4.6

DRV/r =darunavir/ritonavir; TPV/r = tipranavir/ritonavir.

Table V. HIV-related annual mortality rates by CD4 cell count

range[21]

CD4 cell count range (cells/mm3) Mocroft (EuroSIDA) (%)

0–50 17.6

51–100 5.5

101–200 2.2

201–350 0.8

351–500 0.4

>500 0.4

90 Mauskopf et al.


40–64 and 65+ years, were computed for men andwomen using the 5-year age distribution from thePOWER 1 and 2 clinical trials (table VI).

Annual mortality rates were transformed into3-month mortality rates, and a 3-month mortal-ity rate adjustment factor, estimated from an ex-ponential curve fitted to the all-cause mortalitydata, was applied to the non-HIV-related mortal-ity rates in each cycle to account for populationaging. In each cycle in the Markov model, the3-month probability of dying from either HIV-related causes or non-HIV-related causes wascomputed first, before the transitions between theCD4 cell counts were estimated.

Utilities

In the economic evaluation, both life-yearsand QALYs were measured to reflect the long-term health outcome benefits associatedwith treat-ment of different durations. Life-years were esti-mated using HIV-related and non-HIV-relatedmortality rate data described above. QALYs are acommon outcome reported in economic modelsfor HIV treatment, capturing both morbidity andmortality changes with new treatments. An addi-tional outcome measure estimated by the modelwas the length of time spent in each CD4 cell countrange over an individual’s remaining lifetime. Thisdirectly captures the impact of DRV/r on the timespent in less severe health states and also providesestimates of the natural units used to compute thelifetime costs in the model.

Utility weight estimates were taken fromSimpson et al.[40] The weights were derived from21 000 participants in HIV clinical trials, includ-ing participants in the USA,[41] whose wellbeingwas assessed using the EuroQol quality of lifeinstrument. These values were used for the base-case analysis because they were presented for theCD4 cell count ranges used in the economicmodel, were the most recent published values,

and included data from people in the USA.Simpson et al.[40] transformed the EuroQol qual-ity of life instrument scores into utility weightsusing the preference weight modelling transfor-mation developed by Dolan.[42] Simpson et al.[40]

presented the utility weights by CD4 cell countrange as well as by HIV-1-RNA level. In order toobtain utility values for the health states in ourmodel, which are based on CD4 cell count ranges,we calculated the mean values across all HIV-1-RNA subgroups included in Simpson et al.[40] foreach CD4 cell count range. The resulting utilityweights used in our model are shown in table VII.

Resource Use and Costs

All costs included in the model are presented in2008US dollars. The costs for antiretroviral drugswere derived using a microcosting approach. Theproportion of people using each drug [includingPIs and OBR (NRTIs and enfuvirtide)] in eachtreatment arm was taken from the POWER 1 and2 clinical trials. The mean daily cost for each ofthese drugs was estimated from the recommendeddoses in the Department of Health and HumanServices (DHHS)[9] guidelines and unit costs wereobtained fromMedispan’s Price-Chek PC[43] whole-sale acquisition costs. The daily cost for darunavir(excluding the ritonavir boosting)was set atUS$27.40for a daily dose of 1200mg (600mg bid). Average3-month costs for each treatment regimen includedin the model then were computed. These costs are

Table VI. Annual non-HIV-related mortality rates in the USA for people with HIV infection by age group[21,39]

Age group (years) Men annual deaths/100 000 Women annual deaths/100000

20–39 643.7 350.6

40–64 1741.2 1031.4

65+ 9312.3 6138.4

Table VII. Estimates of utility weights by CD4 cell count range[40]

CD4 cell count range (cells/mm3) Utility values

0–50 0.781

51–100 0.853

101–200 0.853

201–350 0.931

351–500 0.933

>500 0.946



shown in table VIII for regimens with and withoutenfuvirtide. The estimated drug costs did not includemark-up or dispensing fees.

Other costs for treating people with HIV infec-tion were estimated using published estimates ofall other costs for each CD4 cell count range.[13]

These costs included inpatient care, outpatient care,emergency department visits and other outpatientdrug costs and are presented in table IX. The costsare inflated from 2003 to 2008 dollars using themedical care component of the US consumer priceindex.[44]

The costs reported by Gebo and colleagues[13]

probably did not include costs for end-of-lifecare, because individuals very close to death maynot have been interviewed for the study. A sepa-rate cost for end-of-life care is thus added in themodel. In the base-case analysis, costs for careduring the final 3 months of life are obtainedfrom Yang et al.,[45] and the model assumes thatthese costs are similar regardless of the causeof death. The base-case analysis end-of-life carecost, inflated to 2008 dollars using the medicalcare component of the consumer price index,[44] isUS$27 098. Time costs, community care costs andproductivity losses were not included.

Discount Rate

The discount rate used for the costs and bene-fits was 3%, as preferred byUS decision-makers.[20]

The model was also run using undiscounted valuesas part of the variability analysis.

Uncertainty and Variability Analyses

To determine the impact of uncertainty aboutthe parameter estimates on the results of the cost-utility analysis, both extensive one-way sensitivityanalysis and PSA were performed using realisticranges and probability distributions for each para-meter. Parameters included in the analyses werethose that involved sampling uncertainty, such asefficacy, drug use, utility, cost and mortality. Inthe one-way analysis, each parameter was variedindividually across a realistic range, and model re-sults were recorded separately for each parameter.For the PSA, Monte Carlo simulations were per-formed. For each simulation, random values fromprobability distributions were drawn for each pa-rameter simultaneously, and model results wererecorded for the set of randomly drawn values.Table X displays the ranges used in the one-waysensitivity analysis and the distributions used inthe PSA. These values were derived from pub-lished data sources.[46] The results of the one-waysensitivity analysis are presented in the form ofa tornado diagram for the incremental cost perQALY gained. The PSA results are presented ingraphs showing the 1000Monte Carlo simulationson the cost-effectiveness plane and the cost-effec-tiveness acceptability curve.

To determine the impact of variability in mod-elling assumptions, the population that mightbe treated with DRV/r and the differences in

Table VIII. Three-monthly drug regimen costs (in 2008 US$)[9,28,43]

Drug regimen

components

DRV/r Control TPV/r

Boosted PIs 2814 2786 3346

NRTIs 2948 3222 3083

Enfuvirtide 6527 6527 6527

Total with enfuvirtide 12 290 12 535 12 955

Total without enfuvirtide 5763 6008 6428

DRV/r =darunavir/ritonavir; NRTIs = nucleoside reverse transcrip-

tase inhibitors; PIs =protease inhibitors; TPV/r = tipranavir/ritonavir.

Table IX. Three-month cost by CD4 cell count range in the USA (in 2008 US$)[13,44]

CD4 cell count range (cells/mm3) Other drug costs (US$) Medical care service costs (US$)

0–50 2409 10033

51–100 1582 4719

101–200 1582 4719

201–350 1010 3396

351–500 1010 3396

>500 822 2292

92 Mauskopf et al.


Table X. List of parameters tested in the sensitivity analyses

Variable Base-case analysis Distribution for probabilistic

sensitivity analysis

Range for one-way sensitivity analyses

Gender distribution POWER trials values Beta 95% CI for men

Age distribution POWER trials values Dirichlet (multivariate

generalization of beta)

95% CI for 65+ years age group

Starting CD4 cell count

distribution

POWER trials values Dirichlet Included in probabilistic sensitivity analysis

and variability analysis, but not one-way


Response rate

(<50 copies/mL):

DRV/r and control

regimens

POWER trial values with and

without enfuvirtide for DRV/r andcontrol

Dirichlet 95% CI

Response rate

(<50 copies/mL): TPV/rRESIST trial values with and

without enfuvirtide for TPV/rDirichlet 95% CI

CD4 cell count

increase: DRV/r andcontrol regimens, initial

rapid increase

POWER trial values by HIV-1

RNA response

Normal 95% CI

CD4 cell count

increase: TPV/rregimen, initial rapid

increase

RESIST trial values imputed by

HIV-1 RNA response

Normal 95% CI

CD4 cell count

increase: slower

increase

Trial values by virological

response

Normal 95% CI

Rate of CD4 cell count

decline

Calculated using PLATO

collaboration data: -18.48 cells

per year

Triangle with a minimum of

-73.28 and a maximum of zero

cells per year

Zero cells per year or untreated value of

-73.28 cells per year

Duration of rapidly

increasing CD4 cell

count

For those with <50 copies/mL:

0.5 years for DRV/r and TPV/r1 year for control

0.5 years for those with

50 copies/mL to ‡1 log10reduction

0.5 years for those with <1 log10reduction

Discrete distribution:

Base-case analysis plus

3 months (20%)

Base-case analysis (60%)

Base-case analysis less

3 months (20%)

Base-case value –3 months

Duration of stable or

slowly increasing CD4

cell count

For those with <50 copies/mL:

2 years for DRV/r and TPV/r1.5 years for control

0.5 years for those with

50 copies/mL to ‡1 log10reduction

Zero years for those with <1 log10reduction

Poisson with base-case

analysis as mean value

From zero to 5 years for those with

<50 copies/mL

From zero to 2 years for those with

50 copies/mL to ‡1 log10 reduction

From zero to 1 year for those with <1 log10reduction

Use of enfuvirtide POWER and RESIST trial values Beta 95% CI

Cost of control regimen Use of antiviral drugs based on

POWER trial data

Triangle with a minimum and a

maximum equal to one-way

sensitivity analysis upper and

lower limits

Lower limit: control arm PI is 100% LPV/rUpper limit: all base-case analysis LPV/ruse distributed among other base-case

analysis PIs proportionate to their use in the

POWER trials

HIV-related mortality

by CD4 cell count

Mocroft rates[21] Normal 95% CI

Increased risk: normal Varied relative-risk factor using 95% CI

Continued next page



treatment patterns, a variety of scenario analyseswere performed. In the base-case analysis, modeltime horizons of 5 years and lifetime were con-sidered, and discount rates were set to 3% peryear for costs and health outcomes. Scenarioswere tested for different model time horizons anddifferent discount rates.

In the base-case analysis, the starting CD4 cellcount distribution was set equal to that observedin the POWER 1 and 2 clinical trials. The trialpopulations included people with relatively highstarting CD4 cell counts as well as those with verylow starting CD4 cell counts. The variability anal-ysis tested scenarios in which the starting CD4 cellcount distributions were changed.

In order to account for possible treatment pat-terns that might be observed in clinics or in differ-ent regions of the USA, the variability associatedwith different drug treatment patterns was tested.For example, scenarios were tested in whichTPV/r was included as part of the first treatmentregimen control group. In this case, the switchtherapy was assumed to be similar in cost andefficacy to the POWER 1 and 2 trials controlgroup, and the results from the RESIST 1 and 2trials were combined directly with those from thecontrol group in the POWER 1 and 2 trials forthe first treatment regimen and compared withthe DRV/r group, despite possible differences intrial design and trial population characteristics.

Table XI presents a complete listing of the sce-nario analyses run in order to estimate the changesin economic outcomes associated with alternative

population characteristics, treatment patterns andmodelling assumptions.

Results

Base-Case Analysis Results

The results of the 5-year analysis comparing aregimen containing DRV/r with a control PI regi-men are presented (table XII). The results showthat the inclusion of DRV/r in the first treatmentregimen is associated with an increased discountedlife expectancy of 0.15 years and an increaseddiscounted quality-adjusted life expectancy of0.20 QALYs.

Table XII also presents the discounted 5-yearcost estimates for each treatment arm subdividedby the type of cost: antiretroviral costs, otherdrug costs, other medical care costs and end-of-life care costs. The total discounted 5-year costsare lower for the DRV/r regimen than for thecontrol PI regimen because increased antiretro-viral drug costs for DRV/r are more than offsetby lower costs for other medications, other medi-cal care and end-of-life costs.

The discounted incremental cost per life-yeargained and perQALYgained for aDRV/r regimencomparedwith the control regimenwas -US$11248and -US$8186, respectively (table XII). In bothcases, the DRV/r regimen represents a dominantstrategy compared with the control PI regimen, withboth better patient outcomes and lower costs.

Table X. Contd

Variable Base-case analysis Distribution for probabilistic


Range for one-way sensitivity analyses

Non-HIV-related

mortality by age

US life-table general population

mortality rates multiplied by

Jensen-Fangel et al.[22] relative-

risk factor (3.6)

Costs by CD4 cell

count

Gebo et al.[13] values inflated to

2007 dollars

Triangle with a minimum and

maximum equal to –20% of

mean values

–20% of mean values

End-of-life care costs Yang et al.[45] values inflated to

2008 dollars

Normal 95% CI

Utility weights Simpson et al.[40] values Normal 95% CI based on Schackman et al.[46]

CI = confidence interval; DRV/r = darunavir/ritonavir; PLATO =Pursuing Later Treatment Options collaboration; POWER =Performance Of

TMC114/r When evaluated in treatment-Experienced patients with protease inhibitor Resistance; RESIST =Randomized Evaluation of

Strategic Intervention in multi-drug resiStant patients with Tipranavir; TPV/r = tipranavir/ritonavir.

94 Mauskopf et al.


The results of the lifetime analysis show thatincluding DRV/r in the first treatment regimen isassociated with an increased discounted life ex-pectancy of 1.25 years and an increased dis-counted quality-adjusted life expectancy of 1.27QALYs (table XIII). The health benefits in nat-ural units of time spent in the different CD4 cellcount ranges over a person’s remaining lifetime

are also shown (table XIII). Compared with con-trol PIs, treatment with the DRV/r regimen isassociated with increased time in the higher CD4cell count ranges and decreased time in the lowerCD4 cell count ranges.

The total discounted lifetime costs are higherfor the DRV/r regimen than for the control regi-men because of the increased number of yearsalive (11.12 vs 9.86), during which treatment isneeded (table XIII). In fact, the discounted costper year alive is lower for the DRV/r regimenthan for the control PI regimen. The lifetime non-antiretroviral drug costs, other medical care costsand end-of-life care costs are very similar for theDRV/r and the control regimens.

The discounted incremental cost per life-yeargained and per QALY gained for a DRV/r regi-men compared with the control PI regimen wasUS$30436 andUS$30046, respectively (tableXIII).

Results of the Uncertainty Analysis

One-Way Sensitivity Analysis

The cost-utility ratios were most sensitive tothe following variables for the 5-year model timehorizon, in descending order of sensitivity: the

Table XI. List of scenarios tested in the variability analysis

Scenarios

Base-case analysisa

Time horizon

2 Years

10 Years

Discount rates (%)

0

5

6

Baseline CD4 cell count distribution

Everyone <350 cell/mm3

Everyone <200 cells/mm3

Tipranavir use in first control regimen (switch regimen then is

assumed to be POWER control regimen) (%)

0

5

10

20

50

75

100

Duration of declining CD4 cell count before regimen switch

0 Years for all virological response groups

3 Years for ‡1 log10 reduction group, 1 year for <1 log10reduction group

Time to enfuvirtide discontinuation after CD4 cell count decline

begins

3 Months

12 Months

a The base-case values for the variables changed in the scenario

analysis were as follows: time horizon of 5 years and lifetime; 3%discount rate; starting CD4 cell count distribution: 0–50, 23.1%;

51–100, 15.3%; 101–200, 24.7%; 201–350, 18.8%; 351–500,

9.8%; >500, 8.2%; tipranavir not used in first control regimen mix

and everyone switched to a tipranavir-containing regimen after

first treatment regimen fails; 6 months of CD4 cell count decline

before regimen switch; and enfuvirtide discontinued 6 months

after CD4 cell count decline begins in switch regimen.

POWER =Performance Of TMC114/r When evaluated in treatment-

Experienced patients with protease inhibitor Resistance.

Table XII. Five-year cost-utility analysis of DRV/r compared with

control: base-case analysis

Outcome

measuresaDRV/r Control Difference

Life-years (years) 4.17 4.02 0.15

QALYs 3.80 3.60 0.20

5-Year costs

ART US$131 591 US$122 052 US$9538

Other drugs US$19 261 US$21 011 -US$1749

Other medical

care

US$62 219 US$70 200 -US$7981

End-of-life care US$4216 US$5698 -US$1482

Total US$217 288 US$218 962 -US$1674

Cost per life-year

gained

-US$11 248

Cost per QALY

gained

-US$8186

a Discounted at 3% per year. Note: negative incremental cost-

effectiveness ratios indicate that DRV/r dominates control.

ART = antiretroviral therapy; DRV/r = darunavir/ritonavir; QALYs =quality-adjusted life-years.



rate of CD4 cell count change during the periodof stable or slowly increasing CD4 cell count; theduration of the period of stable or slowly increas-ing CD4 cell count; HIV-related mortality rates;and costs for medical resources other than antire-troviral drugs (figure 3a). For the lifetime modeltime horizon (figure 3b), the cost-utility ratioswere most sensitive to the following variables indescending order of sensitivity: the rate of CD4cell count decline after virological failure, butwhile still on multidrug therapy; the rate of CD4cell count change during the period of stable orslowly increasing CD4 cell count; HIV-relatedmortality rates; and the duration of the period ofrapidly increasing CD4 cell count. Nevertheless,for all ranges tested, the incremental cost perQALY gained remained below US$50 000.

Probabilistic Sensitivity Analysis

The results of probabilistic sensitivity analysisfor a 5-year model time horizon are presented in

figures 4a and 5a, while figures 4b and 5b presentsimilar results for a lifetime model time horizon.The probabilistic sensitivity analysis shows thatfor a 5-year model time horizon the cost-utilityratio will be below a threshold of US$50 000 perQALY, with a probability of 0.921 (figure 5a).For a lifetime model time horizon, the cost-utilityratio will be below a threshold of US$50 000 perQALY, with a probability of 0.950 (figure 5b).

Results of the Variability Analysis

A series of scenario analyses that representdifferent model structure assumptions, treatmentpractices or population characteristics were per-formed, and the results show that cost-utility ra-tios were most sensitive to changes in the modeltime horizon (-US$67 434 for a 2-year time hori-zon compared with US$30 046 for a lifetime timehorizon) (table XIV). Also, cost-utility ratios weremore sensitive to alternative scenarios when the

Table XIII. Lifetime cost-utility analysis of DRV/r compared with control: base-case analysis

Outcome measuresa DRV/r Control Difference

Life-years (years) 11.12 9.86 1.25

QALYs 10.03 8.76 1.27

Mean time in each CD4 cell count range over remaining lifetime (years; cells/mm3)

0–50 1.03 1.49 -0.46

51–100 0.72 0.88 -0.16

101–200 1.84 1.93 -0.09

201–350 3.02 2.60 0.41

351–500 2.33 1.64 0.68

>500 2.18 1.31 0.86

Lifetime costs

ART US$310 508 US$272 335 US$38 173

Other drugs US$54 895 US$53 638 US$1257

Other medical care US$182 304 US$182 657 -US$353

End-of-life care US$17 651 US$18 657 -US$1006

Total US$565 358 US$527 287 US$38 071

Cost per life-year gained US$30 436

Cost per QALY gained US$30 046

a Discounted at 3% per year.

ART = antiretroviral therapy; DRV/r = darunavir/ritonavir; QALYs = quality-adjusted life-years.

Fig. 3. One-way sensitivity analysis: tornado diagram for (a) 5-year time horizon and (b) lifetime time horizon. Note A: duration of stable orslowly increasing CD4 cell count: run 1 =0 years; run 2= 5 years for less than 50 copies/mL group, 2 years for 50 copies/mL to 1 log10 or greaterreduction group and 1 year for less than 1 log10 reduction group. Note B: cost of initial control regimen: run 1 = based on the use of LPV/r;run 2= based on no use of LPV/r, with the percentage of individuals who had been receiving LPV/r in the base-case analysis spread to the otherprotease inhibitors proportionate to their use in the trial. Note: negative incremental cost-effectiveness ratios indicate that darunavir/ritonavirdominates control. ART = antiretroviral therapy; CI = confidence interval; LPV/r = lopinavir/ritonavir; QALY =quality-adjusted life-year.

96 Mauskopf et al.


−$40 000 −$20 000 $0 $20 000

Rate of slow CD4 increase (95% CI)

a

b

Duration of slow CD4 increase (Note A)

HIV-related mortality (95% CI)

Non-ART medical costs (± 20%)

Cost of control regimen (Note B)

Rate of CD4 decline (−73.28, 0 cells/yr)

Duration of rapid CD4 increase (± 3 months)

Darunavir/r virologic response (95% CI for <50 copies)

Control virologic response (95% CI for <50 copies)

Rate of rapid CD4 increase for switch regimen (95% CI)

Rate of rapid CD4 increase for initial regimens (95% CI)

End-of-life costs (95% CI)

Use of enfuvirtide (95% CI)

Relative risk of non-HIV-related mortality (95% CI)

Utility values (95% CI)

Age distribution (95% CI for 65+)

Gender distribution (95% CI for males)

Switch virologic response (95% CI for <50 copies)

Par

amet

er

Incremental cost per QALY gained

Sensitivity Run 1 Sensitivity Run 2

$20 000 $25 000 $30 000 $35 000 $40 000

Rate of CD4 decline (−73.28, 0 cells/yr)

Rate of slow CD4 increase (95% CI)

HIV-related mortality (95% CI)

Duration of rapid CD4 increase (± 3 months)

Relative risk of non-HIV-related mortality (95% CI)

Duration of slow CD4 increase (Note A)

Cost of control regimen (Note B)

Darunavir/r virologic response (95% CI for <50 copies)

Rate of rapid CD4 increase for initial regimens (95% CI)

Rate of rapid CD4 increase for switch regimen (95% CI)

Utility values (95% CI)

Control virologic response (95% CI for <50 copies)

Non-ART medical costs (± 20%)

Use of enfuvirtide (95% CI)

Age distribution (95% CI for 65+)

Switch virologic response (95% CI for <50 copies)

Gender distribution (95% CI for males)

End-of-life costs (95% CI)

Par

amet

er




model time horizon was 5 years than when thetime horizon was lifetime. For all scenarios tested,the incremental cost per QALY gained remainedbelow US$50 000.

Discussion

The results of this cost-effectiveness analysisindicate that the inclusion of DRV/r in an ARTregimen for patients who have previously failedto respond toARTs is cost effective, under a broadrange of assumptions. In addition, it is likely thatfewer years of life will be spent in severe diseasestates with treatment with a DRV/r-containingregimen in this population. Importantly, the re-sults of this analysis are robust to changes in mod-elling assumptions and input parameter values. ThePSA demonstrates that, with a threshold value ofUS$50000perQALYgained, treatmentwithDRV/rrather than the control PI regimen is a more favour-able choice in 92.1% of the simulations for the 5-yeartime horizon and in 95.0% of the simulations for thelifetime time horizon.

The results for DRV/r are consistent with thosereported in studies of the cost effectiveness ofmultidrug ART compared with monotherapy ordual therapy for initial treatment. These resultsinclude US$13 000–23 000 per QALY in the

USA,[47] Can$971–806 per life-year gained inCanada[48] and d17 698 per QALY gained in theUK.[49] This suggests that the continuing gains inhealth outcomes associated with the use of thenewer drugs for people with HIV infection arepossible, with cost-effectiveness ratios similar inorder of magnitude to those of earlier therapeuticadvances.

Direct comparisons between the DRV/r resultsof this analysis and those from economic studies ofother treatment options for treatment-experiencedpatients were difficult. The most comparable ana-lyses would be analyses for TPV/r and enfuvirtide,because their trials enrolled patient populationswith a similar degree of treatment experience andcompared the new antiretroviral drugs with an op-timized, investigator-selected combination ther-apy, containing currently available PIs.[37,50] TheUS cost-effectiveness estimate for TPV/r com-pared with currently available PIs was US$56 315per QALY,[51] a higher value than the DRV/r es-timate. Published US results for QALY gains forboth enfuvirtide and tipranavir were similar to theDRV/r results in order of magnitude. In particu-lar, the estimated discounted QALYs gained were0.64 QALYs for tipranavir[51] and 0.7 QALYs or1.5 QALYs for enfuvirtide.[52,53] A 1.27 QALYgain was estimated for DRV/r.

−$15 000

−$10 000

−$5000

$0

$5000

$10 000

$15 000

$20 000a b

−0.2 −0.1 0 0.1 0.2 0.3 0.4 0.5

Difference in QALYs

Diff

eren

ce in

tota

l cos

ts

−$20 000

$0

$20 000

$40 000

$60 000

$80 000

−1 −0.5 0 0.5 1 1.5 2 2.5 3

Difference in QALYsD

iffer

ence

in to

tal c

osts

Individual simulationsBase caseUS$50 000 per QALY gained

Fig. 4. Probabilistic sensitivity analysis: cost-effectiveness plane for darunavir/lopinavir compared with control for (a) 5-year model timehorizon and (b) lifetime model time horizon. QALYs = quality-adjusted life-years.

98 Mauskopf et al.


An important result that emerged from thisanalysis is that the cost-effectiveness ratio is verysensitive to the time horizon chosen for the analy-sis. For example, treatment with the DRV/r regi-men is less costly andmore effective than treatmentwith the control PI regimen (dominates) using a5-year time horizon, whereas DRV/r is more costlyand more effective using a lifetime time horizon,with a cost-effectiveness ratio of US$30046 perQALY. This result explains the large differencebetween the model results of Sax and colleagues[52]

and Hornberger and colleagues[53] for the cost ef-fectiveness of enfuvirtide. The study by Sax et al.[52]

uses a lifetime horizon and estimates a cost-effectiveness ratio ofUS$69500 perQALY,whereasthe study by Hornberger et al.[53] uses a 10-yeartime horizon and estimates a cost-effectivenessratio of US$24604 per QALY. There are otherdifferences between the assumptions of the twostudies but, based on our sensitivity analysis, themodel results are not very sensitive to the other fac-tors. For comparison purposes, using ourmodel, theestimated cost-effectiveness ratios for a DRV/r re-gimen using 10-year and lifetime time horizons areUS$7290 per QALY and US$30046 per QALY,respectively. The estimates of Simpson and collea-gues[51] for the cost effectiveness of TPV/r using the

results of theRESIST trials, however, use a differenttime horizon. Themodel runs for each regimen until90% of the modelled population is dead. Thisis closer to the lifetime horizon of the study bySax et al.,[52] and the results are similar, with a cost-effectiveness ratio compared with control PIs ofUS$56315 per QALY.

It is interesting to note that in addition tolonger life expectancy, the time spent in the lowerCD4 cell count ranges was predicted to be lesswith the DRV/r regimen than with the control PIregimen. This is important not only for individualwellbeing, but also for the many other costs asso-ciated with HIV infection that are not included inthis analysis. For example, productivity losses aregreatest once a person starts to develop oppor-tunistic infections.[54,55] This generally happensonce CD4 cell counts fall below 200 cells/mm3.These costs were not included in this analysis, butthey could further increase the cost effectivenessof DRV/r.

As mentioned earlier, as tipranavir was stillinvestigational when the POWER 1 and 2 trialswere initiated, it was not included in the controlPI group of the POWER trials. Therefore, thebase-case analysis of this economic evaluationdid not compare DRV/r with TPV/r directly

0

0.2

0.4

0.6

0.8

1.0

a b

$0 $20 000 $40 000 $60 000 $80 000 $100 000


Pro

babi

lity

0

0.2

0.4

0.6

0.8

1.0

$0 $20 000 $40 000 $60 000 $80 000 $100 000


Pro

babi

lity

CEACUS$50 000 per QALY gained

Fig. 5. Probabilistic sensitivity analysis: cost-effectiveness acceptability curve for darunavir/lopinavir compared with control for (a) 5-yearmodel time horizon and (b) lifetime model time horizon. CEAC = cost-effectiveness acceptability curve; QALY = quality-adjusted life-year.



because head-to-head trial data were not avail-able, and published information on the RESIST 1and 2 trials did not provide detailed informationon the CD4 cell count increases by virologicalresponse category needed in the economic model.

In recognition of the fact that TPV/r is cur-rently being used in the USA and that partici-pants in the tipranavir RESIST trials and thePOWER trials have a broadly comparable pro-file, an indirect comparison was performed. Thisindirect comparison was performed by modelling

a scenario in which the currently recommendeddose of TPV/r plus OBR was the initial com-parator for DRV/r, with the switch regimen beingthe control PI used in the POWER studies (plusOBR). This scenario analysis showed that DRV/rwas a cost effective first treatment compared withTPV/r (lifetime incremental cost per QALYgained of US$32 709). This result must, however,be interpreted with caution because of the as-sumptions made to map the published tipranavirdata into the virological response categories.

Table XIV. Results of variability analyses for 5-year and lifetime model time horizons

Scenarios Incremental cost per QALY gained

5-Year model time horizon (US$) Lifetime model time horizon (US$)

Base-case analysis -8186 30 046

Time horizon

2 Years -67434 -67 434

10 Years 7290 7290

Discount rates (%)

0 -6757 35 228

5 -9157 26 757

6 -9648 25 167

Baseline CD4 cell count distribution (cells/mm3)

Everyone <350 -9983 30 518

Everyone <200 -9806 31 902

Tipranavir use in first control regimen (switch regimen is POWER 1 and 2 control regimen) (%)

0 -8104 29 880

5 -7524 29 982

10 -6888 30 088

20 -5559 30 308

50 -1032 31 057

75 -3508 31 807

100 -8974 32 709

Duration of declining CD4 cell count before regimen switch

0 Months -8301 29 912

3 Years for ‡1 log10 reduction group,

1 year for <1 log10 reduction group

11593 35 738

Time to enfuvirtide discontinuation after CD4 cell count decline begins in switch regimen

3 Months -7638 30 057

12 Months -9201 30 032

a The base-case values for the variables changed in the scenario analysis were as follows: time horizon of 5 years and lifetime; 3% discount

rate; starting CD4 cell count distribution: 0–50, 23.1%; 51–100, 15.3%; 101–200, 24.7%; 201–350, 18.8%; 351–500, 9.8%; >500, 8.2%;

tipranavir not used in first control regimen mix and everyone switched to a tipranavir-containing regimen after first treatment regimen fails;

6 months of CD4 cell count decline before regimen switch; and enfuvirtide discontinued 6 months after CD4 cell count decline begins in

switch regimen.

Note: negative incremental cost-effectiveness ratios indicate that DRV/r dominates control.

DRV/r =darunavir/ritonavir; POWER =Performance Of TMC114/r When evaluated in treatment-Experienced patients with protease inhibitor

Resistance; QALYs =quality-adjusted life-years.

100 Mauskopf et al.


This model includes many parameter valuesand model assumptions that are specific to theUSA, including drug costs, costs by CD4 cellcount ranges, utility values andmortality rates, aswell as model assumptions about treatmentpatterns and population characteristics. Never-theless, these input parameter values and modelassumptions can be readily changed to generateestimates of the cost effectiveness of a DRV/rregimen in other countries. The results of suchcountry-specific analyses for the UK, Sweden,Italy and Belgium are presented elsewhere in thissupplement.[56] More generally, a major strengthof this model is that it is designed to be flexibleand allows very extensive sensitivity and scenarioanalyses to be performed that allow the impor-tance of uncertainties related to modelled para-meters to be interpreted as well as the impact ofvarying model assumptions, characteristics of thetarget population and treatment patterns on thecost-effectiveness results. For example, cost effec-tiveness can be estimated for DRV/r at many timehorizons. In addition, the assumptions about theduration of efficacy or the rate of increase or de-crease of the CD4 cell count by virological res-ponse and treatment arm can be varied by theuser to test the impact of using different datasources or assumptions about efficacy.

The model presented in this paper also has sev-eral other strengths. First, the clinical outcomesare derived directly from DRV/r and TPV/r clini-cal trial data. Second, the impact of alternativetreatment regimens on disease progression ismodelled as a function of the virological responseat 24 weeks. The duration of efficacy is thus al-lowed to be longer for individuals who achieveHIV-RNA levels less than 50 copies/mL than forthose who do not, as has been shown in manypublished studies.[32-36] In addition, the CD4 cellcount change after starting a new treatment isexplicitly modelled as having three sequentialphases, an initial rapidly increasing phase, fol-lowed by a second stable or slowly increasingphase, which is followed by a third decreasingphase. This follows the patterns of CD4 cell countchange that have been observed in clinical trialsof antiretroviral drugs and cohort studies.[17,23,25,33]

The magnitude of CD4 cell count change in each

phase is also allowed to vary based on the virol-ogical response at 24 weeks.

There are also, however, several limitations orareas of uncertainty that could impact the modelresults. First, the assumption was made in themodel that there are no important differences be-tween DRV/r and comparator regimens with re-gard to AEs. Therefore, the model did not takeinto account drug-specific costs or health effectsfrom AEs other than those included in the CD4cell count range costs. It should be noted, how-ever, that the safety and tolerability data gath-ered in the POWER trials suggest that DRV/ris well tolerated and has a tolerability profilecomparable to that of the control PIs used in thePOWER trials with a lower incidence of diar-rhoea,[15] an important and troublesome side-effect of currently available PIs.[57,58] Therefore,the inclusion of AE-related components specifi-cally in the model should probably not negativelyimpact the cost effectiveness of DRV/r.

Also, in the model we assumed that the effi-cacy of a regimen used after failure on either theDRV/r or the control PI initial regimen would bethe same as that observed in clinical trials of thatregimen. It is possible that the switch regimenwould be less effective, but this possibility wasnot included in the model. In addition, the modeldid not allow the CD4 cell count change to varyby the starting CD4 cell count.[59,60] We per-formed a simple analysis of the DRV/r clinicaltrial data to test for this effect. As the results didnot show such an effect, we did not include this inthe model.

The inclusion of only one switch regimen thatis assumed to be continued indefinitely may alsobe regarded as a limitation in the model structure.This decision, however, was made because of thelimited options currently available for treatment-experienced HIV-1-infected individuals, and inaddition, modelling different subsequent therapyswitches from the pool of currently available treat-ments is likely to have very limited impact on thecosts and health outcomes of alternative first re-gimens. As a number of new drugs, some belong-ing to novel drug classes, have very recently beenapproved or are expected to be licensed during thenext 12–18 months, future economic models may



fortunately, however, have the possibility to in-clude multiple, successive, options for follow-uptreatment, even for highly treatment-experiencedpatients.[61-64]

A final limitation of the model, which is inher-ent to almost all economic assessments of HIV-1therapy in the HAART era, is the issue of extrap-olation of the clinical trial efficacy data to routinepractice and the lack of long-term data with re-gard to the treatment impact on clinical effec-tiveness. This model, like many other recentlydeveloped economic models of HAART, projectsthe efficacy findings of randomized trials beyondthe time horizon of these trials to predict the ef-fectiveness of those treatments over a longer timeperiod. This approach is, however, justified by thelong-term efficacy results of the POWER trialsshown in follow-up open-label studies[65] and,more importantly, by the strong, confirmed linkbetween the level of viral suppression and theCD4 cell count change under treatment, which iscentral to the model, and by the established re-lationship between the CD4 cell count and theactual short-term risk of clinical disease progres-sion.[17,24] Furthermore, extensive sensitivity ana-lysis to assess the impact of changing assumptionsregarding the virological and immunological effi-cacy parameters used in the model did not affectthe conclusions regarding the overall cost effec-tiveness of DRV/r in the target population.

The results of this study have three importantimplications for decision-makers. First, whateverthe time horizon of interest, a DRV/r regimen iscost effective and may even be cost saving in theshort run. HIV-1 treatment patterns change rap-idly as new drugs and management strategies aredeveloped. Second, a DRV/r regimen results inincreased life expectancy and QALYs for peoplewith HIV-1 infection, and this will increase thenumber of people living with HIV/AIDS in lesssevere disease states. Third, the model resultsclearly show that the lifetime costs of treatingHIV-1 infection will be substantially higher thanin the past because of new treatment options. Thelifetime costs estimated in this analysis are higherthan those in the study by Schackman et al.[4]

largely because of the increased life expectancy aswell as the assumption that the final regimen that

is used until death is the more costly TPV/r regi-men rather than the older ‘salvage’ regimens.

It should be noted that, in addition to thecurrently approved 600/100mg bid dose, thephase III development programme of DRV/r iscurrently also comprising an evaluation of the ef-ficacy, safety and tolerability of a lower, once-daily dose of 800mg with 100mg ritonavir intreatment-experienced patients without darunavirresistance-associated mutations (RAMs). Thesepatients accounted for 82% of the overall popu-lation included in the TITAN (TMC114-C214;TMC114/r In Treatment-experienced pAtientsNaive to lopinavir) trial, which included treatment-experienced patients with a profile similar to thatusually encountered in the clinical setting.[66,67]

This lower DRV/r dose has shown impressive ef-ficacy results and a favourable tolerability profilein antiretroviral-naive patients[68] and in the sub-group of POWER patients with no darunavirRAMs, in which it generated a virological responsethat was comparable to that observed with 600/100mg bid (62% vs 67%; difference 5; 95% con-fidence interval -30, 23), with both regimens beingsuperior to control PIs (11%; p< 0.0001).[27] Dos-ing frequency – in particular once-daily dosing –and pill burden have been shown to impactadherence to ART positively and thereby improvethe long-term treatment outcome, compared withantiretroviral drugs that have more complex dos-ing schedules.[69-71] Therefore, the future approveduse of this lower, once-daily dose of DRV/r in themajority of treatment experienced, HIV-1-infectedpatients in routine clinical practice may positivelyinfluence the cost effectiveness of DRV/r-basedHAART in this population in the near future. Fur-ther research is, however, warranted to confirm this.

Conclusions

Darunavir/ritonavir is estimated to be cost ef-fective in highly treatment-experienced patientsfrom a US societal perspective, compared withother currently available PIs, and is expected toyield an average of 0.20 additional QALYs pertreatment-experienced patient over 5 years and1.27 additional QALYs over a lifetime.

102 Mauskopf et al.


Acknowledgements

The authors wish to thank the investigators and thepatients and their families for their participation and supportduring the POWER studies. The authors are also especiallygrateful to Tony Vangeneugden and Ben Van Baelen foranalysing and providing the POWER trial data in line with themodel structure and required inputs, and Eric Lefebvre,Martine De Pauw, Frederic Godderis, Piet De Doncker andthe rest of the darunavir study team for their contributions.The authors also wish to acknowledge Catherine McCarthyBragg (medical writer, Gardiner-Caldwell Communications,Macclesfield, UK) for her assistance in editing the manuscriptand collating author contributions. This project was finan-cially supported by Johnson & Johnson Pharmaceutical Ser-vices, Beerse, Belgium.

JM and AB are employees of RTI Health Solutions andhave received grant support from Janssen Cilag, the manufac-turer of darunavir, to assist with the preparation of this manu-script. They were not restricted by Janssen Cilag in theiranalysis or in their interpretation of the final results. SM is anemployee of Johnson & Johnson Pharmaceutical Services,New Jersey, USA. ES is an employee of Johnson & JohnsonPharmaceutical Services, Beerse, Belgium, and owns stockoptions and shares in this company.

References1. Centers for Disease Control and Prevention (CDC). HIV/

AIDS surveillance report, 2005. Vol 17. Atlanta: US De-partment of Health and Human Services, Centers for Dis-ease Control and Prevention; 2007 Jun [online]. Availablefrom URL: http://www.cdc.gov/hiv/topics/surveillance/resources/reports/2005report/pdf/2005surveillancereport.pdf[Accessed 2009 Dec 10]

2. Braithwaite RS, Justice AC, Chang CC, et al. Estimating theproportion of patients infected with HIV who will die ofcomorbid diseases. Am J Med 2005; 118: 890-8

3. Walensky RP, Paltiel AD, Losina E, et al. The survivalbenefits of AIDS treatment in the United States. J InfectDis 2006; 194: 11-19

4. Schackman BR, Gebo KA, Walensky RP, et al. The lifetimecost of current human immunodeficiency virus care in theUnited States. Med Care 2006; 44: 990-7

5. Hutchinson AB, Farnham PG, Dean HD, et al. The eco-nomic burden of HIV in the United States in the era ofhighly active antiretroviral therapy: evidence of continuingracial and ethnic differences. J Acquir Immune Defic Syndr2006; 43: 451-7

6. Clavel F, Hance AJ. HIV Drug Resistance. N Engl J Med2004; 350: 1023-35

7. Mocroft A, Ledergerber B, Viard JP, et al., EuroSIDAStudy Group. Time to virological failure of 3 classes ofanti-retrovirals after initiation of highly active antiretrovi-ral therapy: results from the EuroSIDA study group. J In-fect Dis 2004; 190: 1947-56

8. Sabin CA, Scullard G, Easterbrook P, et al. Treatment ex-haustion of highly active antiretroviral therapy (HAART)among individuals infected with HIV in the UnitedKingdom: multicentre cohort study. BMJ 2005; 330: 695-9

9. Department of Health and Human Services (DHHS). Panelon antiretroviral guidelines for adults and adolescents.Guidelines for the use of antiretroviral agents in HIV-1-infected adults and adolescents; [online]. Available fromURL: http://AIDSinfo.nih.gov [Accessed 2009 Dec 10]

10. BHIVAWriting Committee. British HIVAssociation (BHIVA)guidelines for the treatment of HIV-infected adults with anti-retroviral therapy (2006). HIVMedicine 2006; 7: 487-503

11. International AIDS Society – USA Panel. Treatment foradult HIV infection: 2006 recommendations of the Inter-nationalAIDSSociety –USAPanel. JAMA2006; 296: 827-43

12. Chen RY, Accortt NA, Westfall AO, et al. Distribution ofhealthcare expenditures for HIV-infected patients. ClinInfect Dis 2006; 42: 1003-10

13. GeboK, Fleishman J, Conviser R, et al. Contemporary costsof HIV healthcare in the HAART era [abstract 537]. 13thConference on Retroviruses and Opportunistic Infections;2006 Feb 5–9; Denver, Colorado, USA

14. Tibotec Inc. Prezista (darunavir). Full prescribing informa-tion; [online]. Available from URL: http://www.prezista.com/prezista/documents/us_package_insert.pdf [Accessed2009 Dec 10]

15. Clotet B, Bellos M, Molina JM, et al. Efficacy and safety ofdarunavir–ritonavir at week 48 in treatment-experiencedpatients with HIV-1 infection in POWER 1 and 2: a pooledsubgroup analysis of data from two randomised trials.Lancet 2007; 369: 1169-78

16. EACSEuroguidelinesGroup. SummaryEuropean guidelinesfor the clinical management and treatment of HIV infectedadults in Europe [summary, pocket version]. 10th EuropeanAIDS Conference; 2005 Nov 10–17; Dublin, Ireland

17. Phillips AN, Lundgren JD. The CD4 lymphocyte count andrisk of clinical progression: the T cell in HIV infection anddisease. Curr Opin HIV AIDS 2006; 1: 43-9

18. Phillips AN, Dunn D, Sabin C, et al. Long term probabilityof detection of HIV-1 drug resistance after starting antire-troviral therapy in routine clinical practice. AIDS 2005; 19:487-94

19. BHIVA Writing Committee. British HIV Association(BHIVA) guidelines for the treatment of HIV-infectedadults with antiretroviral therapy (2005). HIV Med 2005;6 Suppl. 2: 1-61

20. Gold MR, Siegel JE, Russell LB, et al. Cost-effectiveness inhealth and medicine. New York: Oxford University Press,1996

21. Mocroft A, Ledergerber B, Katlama C, et al. Decline in theAIDS and death rates in the EuroSIDA Study: an ob-servational study. Lancet 2003; 362: 22-9

22. Jensen-Fangel S, Pedersen L, Pedersen C, et al. Lowmortality in HIV-infected patients starting highly activeantiretroviral therapy: a comparison with the general po-pulation. AIDS 2004; 18: 89-97

23. Deeks SG, Barbour JD, Grant RM, et al. Duration andpredictors of CD4 T-cell gains in patients who continuecombination therapy despite detectable plasma viremia.AIDS 2002; 16: 201-7

24. Ledergerber B, Lundgren JD,Walker AS, et al. Predictors oftrend in CD4-positive T-cell count and mortality amongHIV-1-infected individuals with virological failure to allthree antiretroviral-drug classes. Lancet 2004; 364: 51-62



25. Kaufmann GR, Furrer H, Ledergerber B, et al. Character-istics, determinants, and clinical relevance of CD4 T cellrecovery to <500 cells/ml in HIV type 1-infected individualsreceiving potent antiretroviral therapy. Clin Infect Dis2005; 41: 361-72

26. International AIDS Society USA (IAS–USA). Treatment foradult HIV infection 2004. Recommendations of the Interna-tional AIDS Society – USA Panel. JAMA 2004; 292: 251-65

27. Tibotec Pharmaceuticals. Clinical data of TMC114-C213(POWER 1) and TMC114-C202 (POWER 2) Phase IIbclinical trial for darunavir/r. Data on file. Dec 2005

28. Johnson & Johnson. Results of POWER 1, POWER 2, andPOWER 3 Phase II clinical trials for darunavir/r. Data onfile, 2006

29. Cooper D, Hicks C, Cahn P, et al. 24 Week study analyses:the efficacy of tipranavir/ritonavir (TPV/r) is superior toLPV/r), and the tipranavir/r treatment is enhanced by in-clusion of genotypically active antiretrovirals in the opti-mized background regimen (OBR) [abstract 561]. 12thConference on Retroviruses and Opportunistic Infections2005; Feb 22–25; Boston, Massachusetts, USA

30. Hicks C and the RESIST-1 study team. A phase 3 random-ized, controlled, open-label multicenter trial comparing ti-pranavir/ritonavir (TPV/r) to an optimized comparatorprotease inhibitor/r (CPI/r) regimen in antiretroviral (ARV)experienced patients: 24-week data [abstract 3726]. 44th In-terscience Conference on Antimicrobial Agents and Che-motherapy; 2004 Oct 31–Nov 2; Washington, DC, USA

31. Cahn P and the RESIST 2 Study Team. 24-Week data fromRESIST 2: phase 3 study of the efficacy and safety ofbackground therapy plus tipranavir/ritonavir (TPV/r) oroptimized ritonavir-boosted standard-of-care (SOC) com-parator PI (CPI) in a large randomized multicenter trial intreatment-experienced HIV+ patients [abstract PL14.3].7th International Congress on Drug Therapy in HIV In-fection; 2004 Nov 14–18; Glasgow, UK

32. Tarwater PM, Margolick JB, Jin J, et al. Increase and pla-teau of CD4 T-cell counts in the 3(1/2) years after initiationof potent antiretroviral therapy. J Acquir Immune DeficSyndr 2001; 27: 168-75

33. Kaufmann GR, Perrin L, Pantaleo G, et al. CD4 T-lym-phocyte recovery in individuals with advanced HIV-1 in-fection receiving potent antiretroviral therapy for 4 years:the Swiss HIV Cohort Study. Arch Intern Med 2003; 163:2187-95

34. Hunt PW, Deeks SG, Rodriguez B, et al. Continued CD4cell count increases in HIV-infected adults experiencing4 years of viral suppression on antiretroviral therapy.AIDS 2003; 17: 1907-15

35. Smith CJ, Sabin CA, Lampe FC, et al. The potential forCD4 cell increases in HIV-positive individuals who controlviraemia with highly active antiretroviral therapy. AIDS2003; 17: 963-9

36. Garcia F, de Lazzari E, Plana M, et al. Long-term CD4+

T-cell response to highly active antiretroviral therapy ac-cording to baseline CD4+ T-cell count. J Acquir ImmuneDefic Syndr 2004; 36: 702-13

37. Hicks CB, Cahn P, Cooper DA, et al. Durable efficacy oftipranavir–ritonavir in combination with an optimisedbackground regimen of antiretroviral drug for treatment-experienced HIV-1-infected patients at 48 weeks in the

Randomized Evaluation of Strategic Intervention in multi-drug reSistant patients with Tipranavir (RESIST) studies:an analysis of combined data from two randomised open-label trials. Lancet 2006; 368: 466-75

38. Cook J, Dasbach E, Coplan P, et al. Modeling the long-termoutcomes and costs of HIV antiretroviral therapy usingHIV RNA levels: application to a clinical trial. AIDS ResHum Retroviruses 1999; 15: 499-508

39. Minino AM, Heron MP, Murphy SL, et al. Division of vitalstatistics. Deaths: final data for 2004. Natl Vital Stat Rep2007; 55: 1-119

40. Simpson KN, LuoMP, Chumney E, et al. Cost-effectivenessof lopinavir/ritonavir versus nelfinavir as the first-linehighly active antiretroviral therapy regimen for HIV infec-tion. HIV Clin Trials 2004; 5: 294-304

41. Walmsley S, Bernstein B, KingM, et al. Lopinavir–ritonavirversus nelfinavir for the initial treatment of HIV infection.N Engl J Med 2002; 346: 2039-46

42. Dolan P. Modeling valuations for EuroQoL health states.Medical Care 1997; 35: 1095-108

43. Medispan. Price-check PC; [online]. Available from URL:http://www.medispan.com/index.aspx [Accessed 2009Dec 10]

44. Department of Labor, Bureau of Labor Statistics. MedicalCare Consumer Price Index. US medical care, consumerprice index; [online]. Available from URL: http://da-ta.bls.gov/cgi-bin/surveymost?cu [Accessed 2009 Dec 10]

45. Yang Z, Norton EC, Stearns SC. Longevity and healthcare expenditures: the real reasons older people spendmore. J Gerontol 2003; 58: S2-10

46. Schackman BR, Goldie SJ, Freedberg KA, et al. Compar-ison of health state utilities using community and patientpreference weights derived from a survey of patients withHIV/AIDS. Med Decis Making 2002; 22: 27-38

47. Freedberg KA, Losina E, Weinstein MC, et al. The costeffectiveness of combination antiretroviral therapy forHIV disease. N Engl J Med 2001; 344: 824-31

48. Anis AH, Guh D, Hogg RS, et al. The cost effectiveness ofantiretroviral regimens for the treatment of HIV/AIDS.PharmacoEconomics 2000; 18: 393-404

49. Miners AH, Savin CA, Trueman P, et al. Assessing the cost-effectiveness of HAART for adults with HIV in England.HIV Med 2001; 2: 52-8

50. Nelson M, Arasteh K, Clotet B, et al. Durable efficacy ofenfuvirtide over 48 weeks in heavily treatment-experiencedHIV-1-infected patients in the T-20 versus optimizedbackground regimen only 1 and 2 clinical trials. J AcquirImmune Defic Syndr 2005; 40: 404-12

51. Simpson KN, Chumney ECG, Hicks CB, et al. Cost effec-tiveness of tipranavir in treatment-experienced HIV pa-tients in the USA [abstract PIN10]. 11th Annual Interna-tional Meeting of the International Society for Pharmaco-economics and Outcomes Research; 2006 May 20–24;Philadelphia, PA, USA

52. Sax PE, Losina E,WeinsteinMC, et al. Cost-effectiveness ofenfuvirtide in treatment-experienced patients with advancedHIV disease. J Acquir Immune Defic Syndr 2005; 39: 69-77

53. Hornberger J, Kilby JM,WintfeldN, et al. Cost-effectivenessof enfuvirtide in HIV therapy for treatment-experiencedpatients in the United States. AIDS Res Hum Retroviruses2006; 22: 240-7

104 Mauskopf et al.


54. Liu GG, Yin DD, Lyu R, et al. Economic costs of HIVinfection: an employer’s perspective. Eur J Health Econ2002; 3: 226-34

55. Sendi P, Schellenberg F, Unqsedhapand C, et al. Produc-tivity costs and determinants of productivity in HIV-infected patients. Clin Ther 2004; 26: 791-800

56. Moeremans K, Annemans L, Lothgren M, et al. Costeffectiveness of darunavir/ritonavir 600/100mg bid inprotease inhibitor-experienced, HIV-1-infected adults inBelgium, Italy, Sweden and the UK. PharmacoEconomics2010; 28 Suppl. 1: 107-28

57. Rockstroh J, Clumeck N, Spinosa-Guzman S, et al. Dar-unavir (TMC114)/r has tolerability and efficacy benefitsfor treatment-experienced patients compared with controlPIs; overview of the POWER trials [abstract P28]. 8th In-ternational Congress on Drug Therapy in HIV Infection;2006 Nov 12–16; Glasgow, UK

58. Sherman DS, Fish DN. Management of protease inhibitor-associated diarrhea. Clin Infect Dis 2000; 30: 908-14

59. Smith CJ, Sabin CA, Youle MS, et al. Factors influencingincreases in CD4 cell counts of HIV-positive personsreceiving long-term highly active antiretroviral therapy.J Infect Dis 2004; 190: 1860-8

60. Moore RD, Keruly JC. CD4+ cell count 6 years after com-mencement of highly active antiretroviral therapy in per-sons with sustained virologic suppression. Clin Infect Dis2007; 44: 441-6

61. Cooper D, Gatell J, Rockstroh J, et al. Results ofBENCHMRK-1, a phase III study evaluating the efficacyand safety of MK-0518, a novel HIV-1 integrase inhibitor,in patients with triple-class resistant virus [abstract105aLB]. 14th Conference on Retroviruses and Opportu-nistic Infections; 2007 Feb 25–28; Los Angeles, CA, USA

62. Lalezari J, Goodrich J, DeJesus E, et al. Efficacy and safetyof maraviroc plus optimized background therapy in vire-mic ART-experienced patients infected with CCR5-tropicHIV-1: 24-week results of a phase 2b/3 study in the US andCanada [abstract 104bLB]. 14th Conference on Retro-viruses and Opportunistic Infections; 2007 Feb 25–28; LosAngeles, CA, USA

63. Madruga JV, Cahn P, Grinsztejn B, et al. Efficacy and safetyof TMC125 (etravirine) in treatment-experienced HIV-1-infected patients in DUET-1: 24-week results from a ran-

domised, double-blind, placebo-controlled trial. Lancet2007; 370: 29-38

64. Lazzarin A, Campbell T, Clotet B. Efficacy and safety ofTMC125 (etravirine) in treatment-experienced HIV-1-infected patients in DUET-2: 24-week results from arandomised, double-blind, placebo-controlled trial. Lancet2007; 370: 39-48

65. Arasteh K, Grinsztejn B, de Bethune M-P, et al. Efficacyanalysis of darunavir/r in treatment-experienced POWER3 patients at week 96 [abstract P7.2/05]. 11th EuropeanAIDS Conference; 2007 Oct 24–27; Madrid, Spain

66. Madruga JV, Berger D, McMurchie M, et al. Efficacy andsafety of darunavir–ritonavir compared with that of lopi-navir–ritonavir at 48 weeks in treatment-experienced, HIV-infected patients in TITAN: a randomised controlledphase III trial. Lancet 2007; 370: 49-58

67. Berger DS, Northland R, Scribner A, et al. Effect of baselinefactors on virological response to darunavir/r and lopinavir/rat week 48 in TITAN [abstract P7.3/27]. 11th EuropeanAIDS Conference; 2007 Oct 24–27; Madrid, Spain

68. Ortiz R, DeJesus E, Khanlou H, et al. Efficacy and safety ofonce-daily darunavir/ritonavir versus lopinavir/ritonavir intreatment-naive HIV-1-infected patients at week 48. AIDS2008; 22: 1389-97

69. Ammassari A, Trotta MP, Murri R, et al. Correlates andpredictors of adherence to highly active antiretroviraltherapy: overview of published literature. J Acquir Im-mune Defic Syndr 2002; 31 Suppl. 3: S123-7

70. Maggiolo F, Ripamonti D, Gregis G, et al. Once-a-daytherapy for HIV infection: a controlled, randomized studyin antiretroviral-naive, HIV-1-infected patients. AntivirTher 2003; 8: 339-46

71. Negredo E, Bonjoch A, Clotet B. Benefits and concerns ofsimplification strategies in HIV-infected patients. J Anti-microb Chemother 2006; 58: 235-42

Correspondence: Dr Josephine Mauskopf, Vice President,Health Economics, RTI Health Solutions, 3040 CornwallisRoad, Research Triangle Park, NC 27709, USA.E-mail: [email protected]



cost effectiveness of darunavir/ritonavir in highly treatment-experienced, hiv-1-infected adults in...

Documents