Pharmacoeconomic Analysis of 3 Treatment Strategies for Cytomegalovirus Retinitis in Patients with AIDS

Pharmacoeconomic Analysis of 3Treatment Strategies forCytomegalovirus Retinitis inPatients with AIDSRobert I. Griffiths,1,2 Giselle C. Bleecker,1 Douglas A. Jabs,2 Douglas T. Dieterich,3Laura Coleson,4 Debbie Winters,5 Richard Wolitz,6 and Earl P. Steinberg1,2,4

1 Covance Health Economics and Outcomes Services Inc., Washington, D.C., USA2 Johns Hopkins University School of Medicine, Baltimore, Maryland, USA3 New York University School of Medicine, New York, New York, USA4 Johns Hopkins University School of Hygiene and Public Health, Baltimore, Maryland, USA5 Infectious Diseases Associates of Central New Jersey, Somerville, New Jersey, USA6 Kaiser Permanente, San Francisco, California, USA

Summary A decision-analytical simulation model was constructed to perform a phar-macoeconomic analysis of the following 3 treatment strategies for previouslyuntreated cytomegalovirus (CMV) retinitis in patients with AIDS: (i) intravenousfoscarnet (IVF) for induction and maintenance therapy; (ii) intravenousganciclovir (IVG) for induction and maintenance therapy; and (iii) intravenousganciclovir for induction therapy, followed by oral ganciclovir for maintenancetherapy (IVG-ORG). Patients who experienced significant adverse effects during,or who failed, initial therapy were switched once to one of the other 2 treatments.The model was used to estimate the direct medical cost (from the perspective ofa public payer), survival, and survival adjusted for disutility because of lost vi-sion, for each strategy in the first year following treatment initiation.

The expected first-year costs of treatment initiated with IVF, IVG and IVG-ORG were $US47 918, $US38 817 and $US32 036 (1994 values), respectively,while expected first-year survival was 41 weeks, 35 weeks and 35 weeks, respec-tively. The incremental cost per incremental year of survival using IVF was$US78 000 versus IVG and $US138 000 versus IVG-ORG before adjustment forlost vision, and $US93 000 versus IVG and $US166 000 versus IVG-ORG afteradjustment for lost vision. About 23% of the cost of the IVG treatment strategywas attributable to treatment-related adverse events, compared with 14% of thecost of IVF and 16% of the cost of IVG-ORG. Because of the high failure ratewith IVG-ORG, initial treatment with IVG-ORG frequently led to switching toanother treatment. Only 27% of the costs associated with the IVG-ORG treatmentstrategy were in fact attributable to the cost of induction and maintenance therapyprior to a switch to alternative treatment.

In this analysis, initial treatment with IVG-ORG was the least costly approachfor treating CMV retinitis in patients with AIDS. Initial treatment with IVF

ORIGINAL RESEARCH ARTICLE Pharmacoeconomics 1998 Apr; 13 (4): 461-4741170-7690/98/0005-0461/$07.00/0

© Adis International Limited. All rights reserved.

resulted in slightly longer survival adjusted for vision-related quality of life. Newtreatments for AIDS may reduce the survival benefit of initial treatment with IVF.

Cytomegalovirus (CMV) retinitis, an intraocu-lar infection that can cause retinal destruction, de-velops in up to 40% of patients with AIDS.[1-5]

Therapies for CMV retinitis that are approved bythe US Food and Drug Administration (FDA) in-clude intravenous ganciclovir (IVG), intravenousfoscarnet (IVF), oral ganciclovir, an intraocularganciclovir implant and intravenous cidofovir.[6-14]

While IVG and IVF have been used for severalyears, oral ganciclovir, intraocular ganciclovir im-plants and cidofovir have only recently becomeavailable. Although all of these therapies slow theprogression of CMV infection, none of them eradi-cates it. Consequently, patients with CMV retinitisare at continuous risk of partial or total loss of vi-sion.

Available treatments for CMV retinitis vary intheir effectiveness, adverse effects and cost. IVGand IVF are particularly costly, and are inconve-nient because of the need for 2 or 3 daily infusionsduring the induction phase, daily infusions duringthe maintenance phase and frequent monitoring fortoxicity. Because these treatments require frequentintravenous infusions, they also necessitate a cen-tral venous catheter, which creates a risk of cathe-ter-related infections. In addition, IVF can be neph-rotoxic, while IVG is myelosuppressive.[12] Oralganciclovir has lower bioavailability than IVG.Therefore, although oral ganciclovir causes less neu-tropenia than IVG, patients treated with oral gan-ciclovir have experienced a shorter time to progres-sion of disease than those treated with IVG.[14,15]

Although the efficacy and safety of these treat-ments have been examined in several randomisedclinical trials,[11-15] and the cost of intravenous ver-sus oral ganciclovir has been evaluated,[16] the rela-tive cost effectiveness of these therapies, account-ing for patient utility related to vision outcomes,has not been addressed. We developed a computer-simulation model to assess the clinical, quality-of-life (QOL) and economic impacts of 3 differentinitial treatment regimens for previously untreated

CMV retinitis: (i) IVF for induction and mainte-nance therapy; (ii) IVG for induction and mainte-nance therapy; and (iii) intravenous ganciclovir in-duction therapy, followed by oral ganciclovirmaintenance therapy (IVG-ORG). We did not in-clude the intraocular ganciclovir implant,[17] com-bination IVG plus IVF, or intravenous cidofovir[18]

in our analysis because adequate clinical trial datafor estimating model parameters were not avail-able, and because none of these regimens had beenapproved by the FDA at the time our study wasperformed.

Methods

Decision-Analytical Model

Using the SMLTREE version 2.99 computerprogram (Jim Hollenberg, Rosslyn, New York,USA), we constructed a computerised decision-analytical simulation model of alternative initialtreatment regimens for CMV retinitis. The modelincludes several potential outcomes, namely: pa-tient survival; other clinical events related to theprogression of CMV retinitis; potential adverseevents associated with each of the 3 treatment regi-mens; and the values that patients place on differentlevels of visual acuity (utilities) [fig. 1].

Hypothetical cohorts of 100 000 patients wererun through each of the treatment arms, one at atime, and outcomes were assessed at 2-week inter-vals for 1 year or until death, whichever came first.For each patient who was started on one of the 3treatment regimens that we examined, the costs oftreatment for CMV retinitis and of managing re-lated clinical events, duration of survival, and du-ration of survival adjusted for lost visual acuitywere tracked, taking account of switches to alter-native therapies in the event of treatment failure orsignificant adverse effects. All events, whether re-lated to initial treatment or to the alternative ther-apies to which a patient may subsequently haveswitched, were attributed to the initial treatment

462 Griffiths et al.

� Adis International Limited. All rights reserved. Pharmacoeconomics 1998 Apr; 13 (4)

strategy. Events related to the treatment or out-comes of AIDS, other than survival, but not alsorelated to CMV retinitis or CMV retinitis therapy,were not included in the model.

Treatment Regimens

The dosage schedules for the treatment re-gimens included in our model were based on thoseused in previous clinical trials.[11,15] We assumedthat patients assigned to IVF received 14 days ofinduction therapy at a dosage of 90 mg/kg twicedaily, followed by 90 mg/kg once daily as mainte-nance therapy. Patients assigned to IVG were as-sumed to receive 5 mg/kg twice daily during 14days of induction, followed by daily maintenancetherapy with 5 mg/kg. We assumed that patientsassigned to IVG-ORG received the same inductionregimen as IVG recipients, before switching tomaintenance therapy with oral ganciclovir 1000mg3 times daily.

AssumptionsWe assumed that 10% of patients in each treat-

ment arm would be hospitalised during the firstweek of induction therapy and then receive 2home-care nursing visits during the second weekof induction to facilitate the transition to self-ad-ministered therapy and to monitor for treatment-related toxicity. The remaining 90% of patientswere assumed to receive induction therapy at homewith 2 home-care visits in each of the first 2 weeks.

We assumed that all patients would receive acomplete blood count (CBC) with white cell countdifferential and blood chemistry panel (SMA-18)twice weekly, and would have 1 ophthalmologistand 1 infectious disease specialist outpatient visitduring induction.

We further assumed that patients receiving IVFor IVG would have a central venous catheter in-serted in an ambulatory surgical centre for induc-tion, while patients receiving oral ganciclovirmaintenance therapy would receive intravenousinduction therapy through a peripheral line.

We also assumed that, during maintenance ther-apy, IVF and IVG patients would receive: 1 home-care nursing visit per week to flush the catheter,

Figure 1 goes as a landscape figure here

Treatment Strategies for CMV Retinitis 463


Treatment Strategies for C

MV

Retinitis

Author

© A

dis In

tern

atio

na

l Limite

d. A

ll righ

ts rese

rved

.

Fig. 1. Decision-analytical model. In the model, localised and bloodstream catheter infections are tracked separately. Abbreviation and symbol: IV = intravenous; [+] =model compressed at this stage for purposes of illustration.

Yes

No

Alive atbeginingof next2-weekinterval

[+]

[+]

Yes

No

Switch toalternativetherapyin next2-weekintervalYes

No

Retinaldetachment incontralateral eyeYes

No

ContralateraleyeinvolvementYes

No

[+]

[+]

Retinaldetachment

Treat withIV ganciclovirinduction andmaintenance

Treat withIV foscarnetinduction andmaintenance

Treat withIV ganciclovirinduction andoral ganciclovirmaintenance

CatheterinfectionYes

No

Neutropenia

Catheterinfection

Yes

No[+]

[+]

Yes

No[+]

[+]

Design component Treatment-induced adverse events Events related to disease progression

change the dressing at the catheter site and monitorfor treatment adverse effects; 1 outpatient visit toan ophthalmologist, and 1 to an infectious diseasespecialist, every 4 weeks; and fundus photographyevery 8 weeks to monitor for progression of CMVretinitis.

Another assumption was that patients receivingoral ganciclovir maintenance would receive 1 CBCwith white cell count differential and 1 bloodchemistry panel per week, and would have 1 out-patient visit to an ophthalmologist and 1 to an in-fectious disease specialist every 4 weeks, and fun-dus photography every 8 weeks.

We assumed a constant number of visits for oph-thalmology care for all patients, independent oftheir degree of visual impairment, including blind-ness.

We further assumed that patients receiving IVFor IVG would require first reinduction 16 weeksfollowing first induction, second reinduction 24weeks after first induction, and subsequent rein-ductions at 4-week intervals thereafter. We as-sumed that patients receiving IVG-ORG would re-quire reinduction 8 weeks after first induction, andat 4-week intervals thereafter.[13,14] We also as-sumed that reinduction treatment regimens wereidentical to initial induction regimens, but that IVFand IVG patients would not receive a new cathetersolely for the purpose of reinduction.

Switching Between RegimensIn the model, patients could switch once from

IVF to IVG or vice versa, and from IVG-ORG toIVG alone. Patients who switched therapy begantheir new therapy after completing the current cy-cle of their initial therapy. For example, patientswho switched to IVG after initial induction and 8weeks of maintenance therapy with IVF weretreated in week 9 as if they were receiving a firstreinduction with IVG. Patients who switched fromIVG-ORG to IVG alone incurred the cost of acatheter insertion in conjunction with their next in-duction. Once a patient switched to a different ther-apy, they remained on the new therapy for the du-ration of the simulation. We assumed that all

patients who died remained on therapy until theydied.

Toxicity Related to Therapy

Neutropenia, defined as an absolute neutrophilcount <500 cells/mm3, could occur during anytreatment regimen as a result of toxicity of any ofthe CMV retinitis therapies or of other AIDS-relatedtherapies. We assumed that patients who developedneutropenia would receive granulocyte-colonystimulating factor (G-CSF) 300�g 3 times perweek for the remainder of their time on that ther-apy. In addition, patients who developed neu-tropenia while receiving IVF remained on G-CSFtherapy when they switched treatments. We as-sumed that neither a hospitalisation nor additionaloffice visits were required to manage neutropenia.

Patients who had a central venous catheter inplace were assumed to be at risk of 2 types of cathe-ter infections: bloodstream, or localised to the siteof catheter placement. We assumed that 75% ofpatients with bloodstream infection would be man-aged with vancomycin in an outpatient setting, andthat the remaining 25% would require hospitalisa-tion for 2 weeks. We assumed that localised infec-tion would be managed with a standard course oforal cefalexin. Finally, we assumed that all catheterinfections would be cured at the end of 2 weeks,and that a new infection could occur with the sameprobability as that for a patient with no prior infec-tion.

Events Related to CMV Retinitis

58% of patients were assumed to have unilateralCMV retinitis at the start of therapy and 42% wereassumed to have bilateral infection.[13] Patientswith unilateral disease were assumed to be at con-tinuous risk for contralateral eye involvement. Weassumed that contralateral eye involvement wouldnot result in additional costs, unless the patient de-veloped retinal detachment in the contralateral eye.Since estimates of the decline in visual acuity thatwere incorporated into our model were based onclinical trial data that reflected the impact of retinaldetachment on vision,[13] we made no separate es-



timate of the amount of vision loss that was speci-fically attributable to retinal detachment.

We did, however, estimate the resource costsassociated with management of retinal detach-ment. When retinal detachment occurred in eithereye, we assumed that it would be managed by over-night hospitalisation for vitrectomy, performedwith endolaser and silicone oil, and that no sub-sequent outpatient care would be specifically re-quired for management of the retinal detachment.We assumed that patients with prior retinal detach-ment were at the same risk of another detachmentas those with no prior detachment.

At the suggestion of our expert panel (see ‘DataSources’ section below), we did not include differ-ential rates of extraocular CMV infections andtheir costs in the model because these data were notavailable from the literature.

Visual Acuity and Utility

During each 2-week period, patients were as-sumed to have 1 of 3 levels of visual acuity in eacheye: 20/40 (patient activity not limited by visualimpairment) or better; worse than 20/40 to 20/100or better (patient typically no longer able to drive);worse than 20/100 (patient typically not able toperform regular activities without assistance). Theprobability that the visual acuity in a patient’s in-volved eye would be at each of these levels at thestart of therapy, and at 6 and 12 months after initi-ation of therapy, was drawn from the literature.[13]

We assumed that visual acuity in an uninvolved eyewas 20/40 or better. Visual acuity was tracked sep-arately in each eye.

We assigned a utility value ranging from 0 to 1,where 0 reflected death and 1 reflected perfect vi-sion with current health, to each of these levels ofvisual acuity, based on a previous study of the val-ues (utilities) that cataract patients placed on vari-ous degrees of lost vision.[19] We computed a utilityvalue for vision, based on the visual acuity in thebetter eye at baseline, at 6 months and at 12months, by multiplying the utility associated withthe visual acuity in the better eye by the probabilityof having that visual acuity, and then summing

these products. We further assumed that utility de-clined in a linear fashion between baseline and 6months, and between 6 months and 12 months.

At the end of each 2-week interval the modeldetermined whether a patient was still alive. If thepatient was alive, the utility value associated withthe visual acuity in their better eye was applied tothe corresponding 2-week interval and was used tocalculate survival adjusted for the disutility relatedto lost vision.

Data Sources

Clinical ProbabilitiesData used in the model were drawn from 4

sources: clinical trial data reported in the litera-ture;[11-15] primary data from the Studies of the Oc-ular Complications of AIDS (SOCA) Foscarnet-Ganciclovir Cytomegalovirus Retinitis Trial(FGCRT), which were obtained from the NationalTechnical Information Service database (USNationalTechnical Information Service, Spring-field, Virginia, USA); package inserts for each ofthe 3 therapies we examined; and a panel of 5 cli-nicians with expertise in treating CMV retinitis (DrDouglas A. Jabs, Dr Douglas T. Dieterich, DebbieWinters, Laura Coleson, Dr Richard Wolitz).

Using SOCA FGCRT mortality data,[11-13] weperformed a lifetable analysis using the SAS® lan-guage and procedures computer program version6.03 (SAS Institute Inc., Cary, North Carolina,USA), to calculate the risk of death during each2-week interval among SOCA patients assigned toreceive IVF or IVG. We assumed that mortalityamong patients treated with oral ganciclovir wasthe same as that for patients treated with IVG(SMLTREE).

Cumulative probabilities of clinical eventsother than mortality were obtained from the litera-ture when available, and from the expert panel inother instances. We assumed that all cases of neu-tropenia occurred within the first month of therapyand that 50% of these occurred within the first 2weeks. For the remaining events, we plottedWeibull cumulative probability curves[20] throughdata points reported in the literature, selecting the



shape and scale parameters of each Weibull curveto obtain the best fit to the data. We then derivedthe hazard function for each of these cumulativeprobability curves and used the function to assigna hazard for each event to each 2-week interval inthe model. All probabilities used in the model,along with the source of each, are shown in table I.

CostsOur expert panel identified medical services

that are typically provided as part of CMV retinitistherapy, including those used to treat adverseevents related to therapy or to progression of CMVretinitis. We estimated cost of care from the pers-pective of a government payer in 1994. For the costof the facility component of each hospitalisation(for induction of CMV retinitis therapy, manage-ment of catheter-related bloodstream infection, andrepair of retinal detachment), we determined theappropriate Medicare diagnosis-related group(DRG) and the national average payment for thatDRG.[22]

We assigned a common procedure terminology(CPT) code and then a Medicare resource-basedrelative value scale payment[23] to each physicianservice provided in an inpatient or outpatient set-ting, such as vitrectomy with endolaser for repairof retinal detachment, central venous catheter in-sertion, and ophthalmologist and infectious diseaseoutpatient visits.

We used New Jersey Medicaid reimbursementamounts for home nursing visits and supplies re-lated to administration of CMV retinitis therapy.We used the Medicare fee schedule[23] to assignallowed charges to laboratory tests used to monitorCMV retinitis therapy, such as SMA-18 and CBCwith differential. Finally, we assigned the averagewholesale price (AWP)[24] to therapies for CMVretinitis, neutropenia and catheter-related infec-tion.

All costs and their sources, together with theutilities associated with visual acuity after 6months’ treatment are shown in table II.

Table I. Probability inputs for the decision-analytical model

Parameter Base-caseprobability

Source

Death:a

IVG 0.42 FGCRT; NTIS

IVF 0.33 FGCRT; NTIS

oral ganciclovir 0.42 Expert panel

Neutropenia (absolute neutrophil count <500 cells/mm3):a

IVG 0.34 SOCA[12]

IVF 0.14 SOCA[12]

oral ganciclovir 0.18 Package insert[21]

Localised catheter infection (per patient per year):

IVG 0.25 Expert panel

IVF 0.25 Expert panel


Bloodstream catheter infection (per patient per year):

IVG 0.79 Expert panel

IVF 0.79 Expert panel


Retinal detachment:b

IVG 0.21 SOCA[13]

IVF 0.19 SOCA[13]


Contralateral eye involvement:

IVG 0.17a SOCA[13]

IVF 0.16a SOCA[13]

oral ganciclovir 0.14c Drew et al.;[14]

OGEACSG[15]

Switch to alternative therapy:

IVG 0.11a SOCA[12]

IVF 0.46a SOCA[12]

oral ganciclovir 0.98d Expert panel

Bilateral disease at baseline 0.42 SOCA[13]

a Cumulative probability at 6 months after initiation of therapy.

b Cumulative probability in an involved eye at 6 months followinginvolvement.

c Weighted average of the percentage of patients in each clinicaltrial[14,15] who had a new lesion in a previously uninvolvedeye, based on photographic assessment.

d Expert panel estimated a cumulative probability of 0.9 at 4months; cumulative probability reported is for 6 months folow-ing initiation of therapy, based on extrapolation from the 4-month value using Weibull cumulative probability curve.

Abbreviations: FGCRT = Foscarnet-Ganciclovir CytomegalovirusRetinitis Trial database; IVF = intravenous foscarnet; IVG = intra-venous ganciclovir; NTIS = National Technical Information Service;OGEACSG = Oral Ganciclovir European and Australian Coopera-tive Study Group; SOCA = Studies of the Ocular Complications ofAIDS Research Group.



Base-Case Analysis

We performed Monte Carlo simulations inwhich 100 000 patients entered, one at a time, intoeach of the 3 treatment strategies. Patients werefollowed for 1 year. We calculated the mean cumu-lative cost and survival, adjusted and unadjustedfor disutility associated with lost visual acuity, thateach patient experienced in conjunction with theirinitial therapy and with any subsequent CMV reti-nitis therapy that they received after they switchedto another therapy.

We computed 95% confidence intervals forthese means, based on standard deviations gener-ated by the Monte Carlo simulations. We also com-puted the incremental cost per incremental life-yeargained associated with IVF versus IVG, and withIVF versus IVG-ORG. These cost-effectiveness ra-tios were computed by dividing the difference inmean cost by the difference in mean survival asso-ciated with the pair of treatment strategies beingcompared. Finally, we computed the proportions ofpatient management costs that were attributable toinduction and maintenance therapy, and to manag-ing the adverse effects of therapy, for each treat-ment strategy.

Sensitivity Analysis

In addition to the analysis performed with ourbest estimates for each parameter in the model, weperformed a series of sensitivity analyses, in whicheach parameter estimate was varied by 20% aboveand below the base-case value.

We also performed several sensitivity analysesto address the implications of variation in practicepatterns. For example, we explored the implica-tions of a 3-week, rather than a 2-week, inductionperiod. In addition, we assessed the implications ofperforming fundus photography every 4 weeks ins-tead of every 8 weeks during maintenance therapy,and of potential changes in patient outcomes, suchas equivalent survival across CMV retinitis treat-ment groups that could be associated with newmultidrug treatment regimens for AIDS.

Table II. Cost inputs and utility values for the decision-analyticalmodel. Where stated, the expert panel defined the treatmentregimen. Average wholesale prices of all drugs in this table wereobtained from the 1995 Drug Topics Red Book[24]

Parameter Base-case Source

Costs ($US; 1994 values)

Induction therapy for CMV retinitis (per 2-week treatmentinterval):

IVGa 2341 Expert panel

IVFa 2341 Expert panel

Maintenance therapy for CMV retinitis (per 2-week treatmentinterval):

IVGb 1231 Expert panel

IVFb 1727 Expert panel

oral ganciclovirc 739 Expert panel

Neutropenia (per 2-weektreatment interval)d

846 Expert panel

Catheter infection (per event):

localisede 108 Expert panel

bloodstreamf 1769 Expert panel

Retinal detachmentg 3290 Expert panel

Expected utility associated with visual acuity at 6 monthsfollowing initiation of therapy

IVG 0.68 SOCA;[13] SMLTREE

IVF 0.69 SOCA;[13] SMLTREE

Oral ganciclovir 0.68 SOCA;[13] SMLTREE

a Seven-day inpatient stay (DRG490); 7-day physician consult,daily (RBRVS 99262); Medicaid New Jersey (nursing visit,supplies); laboratory tests (Medicare fee schedule, HCPCS80018 and 85007); physician visits (RBRVS).

b Medicaid New Jersey (nursing visit, supplies); Medicare feeschedule; HCPCS 80018 and 85007; physician visits(RBRVS).

c Medicare fee schedule, HCPCS 80018 and 85007; physicianvisits (RBRVS).

d Granulocyte colony-stimulating factor (Red Book).

e Cefalexin (Red Book); Medicare fee schedule, HCPCS Codes87040 and 87070.

f Vancomycin (Red Book); HCPCS code 87040; 2-week inpatientstay (DRG 489); physician consult (RBRVS).

g Inpatient stay (DRG 36); professional fees (RBRVS 67108,vitrectomy with endolaser); anaesthesia (Kaiser reimburse-ment).

Abbreviations: AWP = average wholesale price; DRG = diagnosis-related group; HCPCS = Health Care Financing AdministrationCommon Procedure Coding System; IVF = intravenous foscarnet;IVG = intravenous ganciclovir; RBRVS = Resource Based RelativeValue Scale; SOCA = Studies of the Ocular Complications of AIDSResearch Group; SMLTREE = SMLTREE version 2.99 computerprogram (Jim Hollenberg, Rosslyn, New York, USA).



Results

Base-Case Analysis

The cost, survival and utility-adjusted survivalresults of the baseline analysis are presented intable III. A strategy involving treatment with IVFwas estimated to cost $US9101 (23%) more thana strategy involving treatment with IVG, and$US15 882 (50%) more than a treatment strategyconsisting of IVG-ORG.

However, treatment with IVF was associatedwith longer unadjusted and adjusted survival thantreatment with either IVG or IVG-ORG. The re-sulting incremental cost per incremental year ofsurvival was approximately $US78 000 for treat-ment with IVF rather than IVG, and approximately$US138 000 per patient for treatment with IVFrather than IVG-ORG. The corresponding ratios ofincremental cost per incremental year of utility-adjusted survival were about 20% higher.

23% of the cost associated with the IVG treat-ment strategy was attributable to management ofadverse effects of therapy (neutropenia and cathe-ter infections), whereas 14% of the cost of the IVFtreatment strategy and 16% of the cost of the IVG-ORG treatment strategy was related to manage-ment of treatment adverse effects.

Induction and maintenance therapy prior toswitching to an alternative therapy accounted for60% of the cost of the IVF treatment strategy and53% of the cost of the IVG treatment strategy, butonly 27% of the cost of the IVG-ORG treatment

strategy. In contrast, induction and maintenancetherapy after switching to the alternative therapyaccounted for 46% of the cost of the IVG-ORGtreatment strategy, compared with 17% of the costof the IVF treatment strategy and 15% of the costof the IVG treatment strategy, reflecting the factthat patients treated in the IVG-ORG strategyswitched to an alternative therapy much more fre-quently than patients treated with one of the otherstrategies.

Sensitivity Analyses

In one-way sensitivity analyses, the incrementalcost per incremental year of utility-adjusted sur-vival for treatment with IVF versus treatment withIVG was moderately sensitive (>10% increase ordecrease in the cost-effectiveness ratio) to the costof induction and maintenance therapy with IVF andthe cost of maintenance therapy with IVG (fig. 2).

However, the incremental cost per incrementalyear of utility-adjusted survival for treatment withthe IVF strategy versus treatment with the IVGstrategy was not sensitive to the probability ofdeath, neutropenia, contralateral eye involvementor switching to an alternative therapy, within therange of parameter estimates we examined for eachtreatment strategy (fig. 3).

The cost effectiveness of treatment with the IVFversus the IVG-ORG strategy was moderately sen-sitive to the cost of induction and maintenancetherapy with IVF, and to the cost of induction withIVG (fig. 4).

Table III. Costs (1994 values) and outcomes associated with the 3 treatment regimens for CMV retinitis. All costs and outcomes relate to apatient’s expected experience over a 12-month period, after receiving initial treatment with the treatment listed. Since the simulation modeldid not track patients for longer than 12 months, the expected survival over 12 months that is reported here is less than a patient’s actualexpected survival

Treatmentstrategy

Expected direct medical costper patient (95% CI) [$US]

Expected survival(95% CI) [weeks]

Expected utility-adjustedsurvival (95% CI) [weeks]

Incremental cost ($US) of using IVFper incremental year of

survival utility-adjustedsurvival

IVF 47 918 (47 783-48 053) 40.78 (40.68-40.88) 31.95 (31.97-32.03) NA NA

IVG 38 817 (38 664-38 970) 34.70 (34.60-34.80) 26.86 (26.78-26.94) 77 838 92 795

IVG-ORG 32 036 (31 915-32 157) 34.78 (34.68-34.88) 26.98 (26.90-27.06) 137 644 165 836

Abbreviations: CI = confidence interval; IVF = intravenous foscarnet; IVG = intravenous ganciclovir; IVG-ORG = intravenous ganciclovirinduction then oral ganciclovir maintenance; NA = not applicable.



In contrast to the situation for the IVF treatmentstrategy versus the IVG strategy, the cost effective-ness of treatment with IVF relative to treatmentwith IVG-ORG was sensitive to the probability ofdeath for IVF and for IVG (fig. 5).

Because of recent trends toward using longerinduction periods for IVG, we also assessed theimpact on our findings of increasing the durationof the IVG induction period from 2 to 3 weeks.With this longer induction period, the expected di-rect costs of a treatment strategy with IVF, IVG andIVG-ORG were $US3000 to $US5000 higher($US52 777, $US42 215 and $US35 725 per pa-tient, respectively), and the incremental cost perincremental year of utility-adjusted survival for IVFrelative to IVG increased by 15% to $US106 439,while the incremental cost per incremental year ofutility-adjusted survival for IVF relative to IVG-ORG increased by 7% to $US177 341. However,increasing the frequency of fundus photographyduring maintenance therapy, from once every 8weeks to once every 4 weeks, increased the incre-mental cost per incremental year of utility-adjusted

survival by less than 1% for both comparisons($US928 for IVF vs IVG, $US1658 for IVF vsIVG-ORG).

Finally, we explored the implications of an as-sumption that new multidrug treatments for AIDSwill result in equivalent survival for patients withCMV retinitis, regardless of the specific CMV reti-nitis treatment that is used. Under this assumption(that patients treated with IVG or IVG-ORG havethe same survival as those treated with IVF), theestimated cost per patient of the IVG strategy in-creased by $US8741 (23%) to $US47 558 and thecost of the IVG-ORG strategy increased by$US6961 (22%) to $US38 997. Thus, under an as-sumption of equivalent survival, treatment with ei-ther IVF or IVG resulted in similar costs and effec-tiveness, while treatment with IVG-ORG wasabout $US9000 less expensive, with similar effec-tiveness to that of the other 2 regimens.

Discussion

Several clinical trials have examined the rela-tive efficacy and adverse-effect profiles of IVF,

140 000

120 000

100 000

80 000

60 000Cos

t effe

ctiv

enes

s (in

crem

enta

l cos

tpe

r in

crem

enta

l util

ity-a

djus

ted

year

of

sur

viva

l) [$

US

]

IVG in

ducti

on th

erap

y

IVF in

ducti

on th

erap

y

IVG m

ainte

nanc

e th

erap

y

IVF m

ainte

nanc

e th

erap

y

Neutro

penia

Loca

lised

cath

eter

infe

ction

Cathe

r-rela

ted

blood

strea

m in

fecti

on

Baseline cost effectiveness = $US92 795

110% of baseline cost effectiveness = $US102 075


Cost variables

Fig. 2. One-way sensitivity analysis of the effect of varying the cost variables by �20% of the values in the base-case analysis onthe incremental cost effectiveness of intravenous foscarnet (IVF) over intravenous ganciclovir (IVG).



IVG and oral ganciclovir in the treatment of CMVretinitis in patients with AIDS.[11-15] While thecomparative clinical benefits of these therapieshave been well documented, the relative cost effec-tiveness of these alternative therapies has not beenassessed. Given the large number of patients withAIDS who receive Medicaid benefits and currentpressures on state Medicaid budgets, considerationof the costs and cost effectiveness of these treat-ments is quite relevant to public as well as privatepayers.

Selection of a treatment for CMV retinitis in apatient with AIDS is not straightforward. Eachtreatment alternative has advantages and draw-backs. For example, although a previous head-to-head comparative study suggested that survival ofpatients treated with IVF is longer than in patientstreated with IVG,[11] advances in managing neu-tropenia and treating AIDS, notably through prote-ase inhibitor therapy, are likely to affect both theabsolute and relative survival of patients withCMV retinitis treated with various therapies. Fur-thermore, 12% of patients treated with IVF experi-ence nephrotoxicity or other adverse effects that

necessitate a change to another therapy.[12] Simi-larly, although CMV retinitis progresses morequickly in patients treated with oral ganciclovirthan in those treated with IVF or IVG, patients re-ceiving oral therapy do not require the installationof an indwelling catheter. These trade-offs, as wellas considerations of the impact of partial or com-plete loss of vision, involve value judgments re-garding the quality, as well as duration, of life.

Findings From the Model

We developed a decision-analytical simulationmodel that facilitates comparison of alternativetreatment regimens for CMV retinitis on the basisof survival, vision-adjusted survival and the cost ofcare, from the perspective of a public payer such asMedicaid. Several of our findings are noteworthy.

First, treatment of CMV retinitis is expensive.The expected direct medical cost to a public payerfor the first year of management of CMV retinitisranges from approximately $US32 000, if initialmaintenance treatment is with oral ganciclovir (fol-lowing induction with IVG), to about $US48 000,

140 000

120 000

100 000

80 000

60 000Cos

t effe

ctiv

enes

s (in

crem

enta

l cos

tpe

r in

crem

enta

l util

ity-a

djus

ted

year

of

sur

viva

l) [$

US

]

IVG d

eath

IVF d

eath

IGV n

eutro

penia

IVF n

eutro

penia

IVG co

ntra

later

al ey

e inv

olvem

ent

IVG sw

itchin

g to

alte

rnat

ive th

erap

y




Probability variables

IVF co

ntra

later

al ey

e inv

olvem

ent

IVF sw

itchin

g to

alte

rnat

ive th

erap

y

Fig. 3. One-way sensitivity analysis of the effect of varying the probability variables by �20% of the values in the base-case analysison the incremental cost effectiveness of intravenous foscarnet (IVF) over intravenous ganciclovir (IVG).



if initial treatment is with IVF. These cost dif-ferences are substantial given the approximately30 000 to 60 000 patients with CMV retinitis andAIDS.

Second, our analysis suggests that a treatmentstrategy that begins with IVG-ORG is preferableto treatment with IVG for both induction andmaintenance since survival, whether or not ad-justed for quality of life, is similar for the 2 treat-ments, but treatment with IVG for induction andmaintenance is more expensive than inductionwith IVG-ORG.

Third, our analysis suggests that adjustment forvision-related quality of life results in about a 20%decrease in expected quality-adjusted survival foreach of the 3 therapies. Thus, although considera-tion of vision-related quality of life influences themagnitude of the cost effectiveness of these thera-pies, it does not influence considerations related towhich of these therapies is more cost effective thanthe others.

Fourth, relative to the cost effectiveness of manyother medical interventions, treatment with IVF inplace of IVG or IVG-ORG is fairly expensive. Rel-ative to induction and maintenance treatment withIVG, treatment with IVF costs $US78 000 per in-cremental year of survival ($US1500 per week),and $US93 000 per incremental year of survival($US1800 per week) when adjusted for disutilityassociated with lost vision; relative to induction withIVG-ORG, treatment with IVF costs $US138 000per incremental year of survival ($US2700 perweek) and $US167 000 per incremental year ofsurvival adjusted for disutility associated with lostvision ($US3200 per week).

Sensitivity Analyses

The sensitivity analyses we performed provide3 interesting insights. The first is that our conclu-sions are robust over a broad range of parameterestimates. For example, when we varied all parame-ter estimates by 20% above and below base-case

220 000

200 000

180 000

160 000

140 000

120 000

Cos

t effe

ctiv

enes

s (in

crem

enta

l cos

tpe

r in

crem

enta

l util

ity-a

djus

ted

year

of

sur

viva

l) [$

US

]

IVF in

ducti

on th

erap

y

IVG-O

RG indu

ction

ther

apy

IVF m

ainte

nanc

e th

erap

y

IVG-O

RG main

tena

nce

ther

apy

Neutro

penia

Loca

lised

cath

eter

infe

ction

Cathe

r-rela

ted

blood

strea

m in

fecti

on




Cost variables

Fig. 4. One-way sensitivity analysis of the effect of varying the cost variables by �20% of the values in the base-case analysis onthe incremental cost effectiveness of intravenous foscarnet (IVF) over the strategy of intravenous ganciclovir induction therapyfollowed by oral ganciclovir maintenance therapy (IVG-ORG).



values, only variation in our estimate of the cost ofIVF induction and maintenance therapy, IVGmaintenance therapy, and our estimate of survivalassociated with intravenous therapies had muchimpact. A 20% reduction in the assumed overallcost of IVF maintenance therapy correspondsroughly to a 30% reduction in the AWP of IVF. A30% reduction in the AWP of IVF is slightly belowthe best price paid by a Medicaid programme. Underthe assumption of a 30% price reduction for IVF,the cost-effectiveness ratio for treatment with IVFrelative to treatment with IVG-ORG, adjusted forlost vision, decreases by $US27 606 to $US138 230,meaning that treatment with IVF becomes morecost effective relative to treatment with IVG-ORG.

In a further sensitivity analysis, we explored theimplications of use of a 3-week, rather than a 2-week, period of induction with IVG. In this analy-sis, we assumed that although the direct costs oftreatment would change, the effectiveness of treat-

ment would not change. The effect of these as-sumptions was to increase the cost of treatmentwith IVF by $US4859, with IVG by $US3398 andwith IVG-ORG by $US3688. The cost of IVF in-creased more than the cost of IVG because of thelarge number of patients in the IVF strategy whoswitched to IVG induction and maintenance afterthe time to reinduction decreased to only 1 month.The cost of IVG-ORG increased more than the costof IVG because large numbers of IVG-ORG pa-tients switched to IVG induction and maintenancetherapy, while IVG patients switched to IVF induc-tion and maintenance. As a result, IVF became lesscost effective relative to IVG or IVG-ORG.

Finally, because of recent experience in a studyof treatment of relapsed CMV retinitis, which sug-gests that previously observed differences in survivalassociated with different CMV retinitis therapies nolonger apply,[20] and because of the availability ofseveral new treatments for AIDS including multi-

240 000

220 000

200 000

180 000

160 000

140 000

120 000

Cos

t effe

ctiv

enes

s (in

crem

enta

l cos

tpe

r in

crem

enta

l util

ity-a

djus

ted

year

of

sur

viva

l) [$

US

]

IVF d

eath

IVG-O

RG dea

th

IVF n

eutro

penia

IVG-O

RG neu

trope

nia

IVF co

ntra

later

al ey

e inv

olvem

ent

IVG-O

RG cont

ralat

eral

eye

involv

emen

t




Probability variables

IVF sw

itchin

g to

alte

rnat

ive th

erap

y

IVG-O

RG switc

hing

to a

ltern

ative

ther

apy

Fig. 5. One-way sensitivity analysis of the effect of varying the probability variables by �20% of the values in the base-case analysison the incremental cost effectiveness of intravenous foscarnet (IVF) over the strategy of intravenous ganciclovir induction therapyfollowed by oral ganciclovir maintenance therapy (IVG-ORG).



drug therapy, we performed an analysis to explorethe cost implications of comparable survival, re-gardless of specific treatment for CMV retinitis.Our findings suggest that if patients treated withIVG or IVG-ORG had the same survival as thosetreated with IVF, the cost of treatment with IVG-ORG would be approximately $US9000 less thanthe cost of treatment with IVF, while treatmentwith IVG would be approximately $US300 lessthan treatment with IVF.

Study Limitations

Several limitations of our analysis should beborne in mind when interpreting our findings.

First, because data on survival of patientstreated for CMV retinitis for more than 1 year areavailable from only a small number of patients,[11]

we limited our analysis to outcomes occurringwithin 1 year following initiation of treatment forCMV retinitis. Although by so doing we havegreater confidence in the estimates we report, wehave ignored possible differences in survival be-yond 1 year following initiation of different treat-ments. Consequently, it should be emphasised thatthe survival estimates reported in table II are meanestimates of expected survival over the year fol-lowing initiation of treatment for CMV retinitis. Ifit is assumed that, among patients alive at 12months, the hazard function for death over the suc-ceeding 6 months is similar to that over the preced-ing year, then our model suggests that expectedsurvival during the 18 months following treatmentis 53 weeks for IVF and 42 weeks for IVG andIVG-ORG. These latter estimates probably over-state expected survival, since the risk of deathamong patients between the ninth and twelfthmonths following initiation of treatment in theSOCA FGCRT was actually 2 to 3 times higherthan that during months 1 to 9.

Second, the structure of our analysis requiredthat patients remain on some type of anti-CMVretinitis therapy until their death or 12 months,whichever occurred first. This assumption, if in-accurate, results in an overestimate of the overallcosts for each of the treatment alternatives.

Third, we assumed that nephrotoxicity wouldhave no economic impact since nephrotoxic effectswere assumed to be reversible with a reduction indosage of antiviral treatment and, even when dos-ages are lowered, the same size vials are used re-sulting in the same cost to the third-party payer. Ifwrong, this assumption would bias the results infavour of IVF to some extent.

Fourth, the utility values we assigned to variouslevels of visual acuity were based on data collec-tion from patients with cataracts because there areno data on utilities associated with different visualacuities among HIV-infected individuals. Utilityvalues among patients with CMV retinitis (a rela-tively young population with a short life expec-tancy) may differ from those among cataract pa-tients (a predominantly elderly population).

Fifth, although our goal was to estimate costsfrom the perspective of a public payer in the US,such as Medicaid or Medicare, we were unable toobtain cost estimates for all resources from one orthe other of these programmes. Rather, in someinstances, we based cost estimates on rates paid bythe Medicare programme and, in other instances,we based cost estimates on Medicaid paymentrates. Moreover, since Medicaid payment ratesvary by state, our estimates of total direct costsassociated with each of the 3 treatment strategiesthat we have considered will not reflect those thatare actually incurred by each state Medicaid pro-gramme. These limitations, however, would notaffect our conclusions regarding the relative costeffectiveness of alternative initial treatment re-gimens for CMV retinitis and we believe that ourcost estimates are reasonable reflections of thoseincurred by public payers in the US. Furthermore,we recognise that there are differences between ourassumptions regarding the intensity and cost of re-sources, and those in other cost analyses of treat-ment for CMV retinitis.[16] These may result fromdifferences in practice patterns or insurance sys-tems between different countries in which theseanalyses have been conducted.

Sixth, we did not include the intraocularganciclovir implant, combination IVG plus IVF, or



intravenous cidofovir in our analysis, because ade-quate clinical trial data on these therapies were notavailable at the time our study was performed.

Finally, our analysis examined only direct medi-cal costs and did not consider indirect costs asso-ciated with these treatments.

Conclusion

Our results show that while all of the CMV reti-nitis therapies are costly, the use of IVG-ORG ther-apy is the least costly approach for treating CMVretinitis in patients with AIDS. Therapy with IVFis the most expensive alternative, but it results inslightly longer survival, adjusted for vision-relatedquality of life.

Acknowledgements

This research was performed with financial support fromGilead Sciences, Inc., Foster City, California, USA. By con-tractual agreement, however, Gilead Sciences had no controlover the design, interpretation or presentation of our analysis.

References1. Holland GN, Pepose JS, Pettit TH, et al. Acquired immune de-

ficiency syndrome: ocular manifestations. Ophthalmology1983; 90: 859-73

2. Jabs DA, Green WR, Fox R, et al. Ocular manifestations ofacquired immunodeficiency syndrome. Ophthalmology1989; 96: 1092-9

3. Jabs DA, Enger C, Bartlett JG. Cytomegalovirus retinitis andacquired immunodeficiency syndrome. Arch Ophthalmol1989; 107: 75-80

4. Drew WR. Cytomegalovirus infection in patients with AIDS. JInfect Dis 1988; 158: 449-56

5. Hoover DR, Saah AJ, Bacellar H, et al. Clinical manifestationsof AIDS in the era of pneumocystis prophylaxis. N Engl J Med1993; 329: 1922-6

6. Holland GN, Sidikaro Y, Kreiger AE, et al. Treatment of cyto-megalovirus retinopathy with ganciclovir. Ophthalmology1987; 94: 815-23

7. Jabs DA, Newman C, DeBustros S, et al. Treatment of cytomeg-alovirus retinitis with ganciclovir. Ophthalmology 1987; 94:824-30

8. Jacobson MA, O’Donnell JJ, Brodie HR, et al. Randomizedprospective trial of ganciclovir maintenance therapy for cyto-megalovirus retinitis. J Med Virol 1988; 25: 339-49

9. Jacobson MA, O’Donnell JJ, Mills J. Foscarnet treatment ofcytomegalovirus retinitis in patients with acquired immuno-deficiency syndrome. Antimicrob Agents Chemother 1989;33: 736-41

10. Lehoang P, Girard B, Robinet M, et al. Foscarnet in the treat-ment of cytomegalovirus retinitis in acquired immune defi-ciency syndrome. Ophthalmology 1989; 96: 865-74

11. Studies of Ocular Complications of AIDS Research Group.Mortality in patients with the acquired immunodeficiencysyndrome treated with either foscarnet or ganciclovir for cy-tomegalovirus retinitis. N Engl J Med 1992; 326: 213-20

12. Studies of Ocular Complications of AIDS Research Group.Morbidity and toxic effects associated with ganciclovir orfoscarnet therapy in a randomized cytomegalovirus retinitistrial. Arch Intern Med 1995; 155: 65-74

13. Studies of Ocular Complications of AIDS Research Group.Foscarnet-ganciclovir cytomegalovirus retinitis trial 4: visualoutcomes. Ophthalmology 1994; 101: 1250-61

14. Drew WL, Ives D, Lalezari JP, et al. Oral ganciclovir as main-tenance treatment for cytomegalovirus retinitis in patientswith AIDS. N Engl J Med 1995; 333: 615-20

15. The Oral Ganciclovir European and Australian CooperativeStudy Group. Intravenous versus oral ganciclovir: Eu-ropean/Australian comparative study of efficacy and safety inthe prevention of cytomegalovirus retinitis recurrence in pa-tients with AIDS. AIDS 1995; 9: 471-7

16. Graf von den Schulenburg JM, Wahling S, Stoll M. Germanhealth economic cost evaluation on oral ganciclovir in treat-ing cytomegalovirus retinitis. Pharmacoeconomics 1996; 10(5): 522-30

17. Martin DF, Parks DJ, Mellow SD, et al. Treatment of cytomeg-alovirus retinitis with an intraocular sustained-releaseganciclovir implant. Arch Ophthalmol 1994; 112: 1531-9

18. Lalezari JP, Stagg RJ, Kuppermann BD, et al. Intravenouscidofovir for peripheral cytomegalovirus retinitis in patientswith AIDS: a randomized, controlled trial. Ann Intern Med1997; 126: 257-63

19. Bass EB, Wills S, Scott IU, et al. Preference values for visualstates in patients planning to undergo cataract surgery. MedDecis Making 1997; 17 (3): 324-30

20. Lee ET. Statistical methods for survival data analysis. 2nd ed.New York: John Wiley & Sons, 1992

21. Cytovene package insert. Syntex Laboratories Inc., 199422. Fed Regist 1 Sep 1995; 60 (170)23. Fed Regist 8 Dec 1995; 60 (236)24. Drug Topics Red Book. Montvale (NJ): Medical Economics

Company Inc., 1995

Correspondence and reprints: Dr Robert I. Griffiths, CovanceHealth Economics and Outcomes Services Inc., 1100 NewYork Avenue, NW, Suite 200 East, Washington, DC 20005-3934, USA.E-mail: [email protected]



Documents

Pharmacoeconomic Analysis of 3 Treatment Strategies for Cytomegalovirus Retinitis in Patients with AIDS