Avai lable onl ine at www.sc iencedirect .com

V A L U E I N H E A L T H ] ( 2 0 1 4 ) ] ] ] – ] ] ]

1098-3015$36.00 – s

Published by Elsevie

http://dx.doi.org/10.

Conflicts of inteconnection with thcompanies as consu

E-mail: Gisela.Ko

* Address correspFrance.

journal homepage: www.elsevier .com/ locate / jva l

Treating to Target with Etanercept in Rheumatoid Arthritis:Cost-Effectiveness of Dose Reductions When RemissionIs AchievedGisela Kobelt, PhD*

University of Lund, Lund, Sweden; European Health Economics, Mulhouse, France

A B S T R A C T

Background: Current management of rheumatoid arthritis (RA)focuses on inducing remission as early as possible to avoid lastingjoint damage, and maintenance of remission has become important.A 12-month clinical trial in 834 patients with moderate RAinvestigated whether etanercept 50 mg/wk could be reduced to halfdose or discontinued in patients who achieved low disease activityafter 36 weeks. Objective: The objective of this study was to estimatethe cost-effectiveness of the three maintenance strategies. Methods:A Markov model integrated the three strategies from the clinical trialand extrapolated to 10 years using data from the Swedish RA registry.Assumed treatment strategies after the trial were similar in allthree arms, with patients failing to maintain remission onhalf-dose etanercept or methotrexate alone switching to the full doseof etanercept and patients maintaining remission on full-dose eta-nercept allowed switching to half dose. Resource use and utilitieswere taken from an observational study. Results are presented as

ee front matter Copyright & 2014, International S

r Inc.

1016/j.jval.2014.04.005

rest: Gisela Kobelt is the president of Europeane development of this manuscript. Gisela Kobelltant in the field of economic evaluation.

belt@he-europe.com.

ondence to: Gisela Kobelt, European Health Econom

cost/quality-adjusted life-year (QALY) (both discounted 3%) in thesocietal perspective. Results: The cost/QALY gained with half-doseetanercept versus methotrexate ranged from €14,000 to €29,000:Longer simulations result in a higher cost/QALY, as the acquisitioncost of etanercept increases. Half-dose etanercept technically domi-nates the full dose (lower costs [€�3000 to 6300] and similar effective-ness [0.007–0.011]). Conclusions: Although ultimately all threestrategies explored achieve a similar outcome as all three continu-ously manage patients to maintain remission, it appears that a dosereduction is the most advantageous strategy in patients with moder-ate disease activity.Keywords: cost-effectiveness, etanercept, modeling, remission,rheumatoid arthritis.

Copyright & 2014, International Society for Pharmacoeconomics andOutcomes Research (ISPOR). Published by Elsevier Inc.

Introduction

Biologic treatment of active rheumatoid arthritis (RA) has becomestandard, when classical disease-modifying arthritis drugs(DMARDs) including methotrexate (MTX) fail. International guide-lines have defined general treatment approaches including adiscussion of costs and cost-effectiveness [1,2], but also fortreatment of early RA [3] and individual treatment to targetremission [4].

The use of biologic agents very early in the disease has,however, been limited by their cost, despite a growing body ofevidence in clinical trials that disease remission can be achievedand maintained better in early than in later disease [5–9]. Thesame findings have been confirmed in clinical practice whereearly intervention (o3 months) was the strongest predictor ofsuccessful remission [10]. Despite the restricted use in an esti-mated 10% to 20% of the patients in Europe [11], these treatmentsare among the most expensive items on many national drug

budgets (e.g., Sweden www.lif.se). Treating earlier thus raisesfears of a further budget increase because a larger proportion ofpatients may use the agents for a longer time. It is thereforeimportant to evaluate the benefit of starting early in order toachieve remission rapidly, with a possibility to modulate treat-ment for patients in remission, compared with the cost of such astrategy.

An early modeling study based on data from the SouthernSwedish RA registry indicated that early treatment was lesscostly and provided more benefits than a late treatment start[12]. We have also proposed an exploratory cost-effectivenessmodel for Sweden, using data from registries, clinical practicecohorts, and a clinical trial with etanercept (ETA) in patients withactive moderate disease activity (DA) (disease activity score[DAS28] 3.2 r 5.1) despite MTX therapy [13]. The two strategiescompared were combination treatment with ETA plus MTX andMTX alone. In the model, ETA was reduced to half dose in 50% ofthe patients who achieved remission while all other patients

ociety for Pharmacoeconomics and Outcomes Research (ISPOR).

Health Economics, SAS, who were paid consultants to Pfizer int has also worked and is working with several pharmaceutical

ics, Parc des Collines 15 rue Victor Schoelcher, Mulhouse 68200,

V A L U E I N H E A L T H ] ( 2 0 1 4 ) ] ] ] – ] ] ]2

remained on full dose. Patients who discontinued ETA/MTX wereswitched to a different biologic. Patients on MTX only continuedon MTX regardless of whether they reached remission or not butwere switched to a biologic if they discontinued. Over 10 years,patients in the ETA/MTX arm gained 1.15 quality-adjusted life-years (QALYs) than did patients treated with MTX, at an incre-mental cost of €15,000 (societal perspective, i.e., all costs regard-less of who pays).

In this article, we propose a confirmatory analysis based on adouble-blind clinical trial (Prospective Randomized EtanerceptStudy to Evaluate Reduced dose Etanercept Versus full doseEtanercept [PRESERVE]) that investigated the use of half-doseETA (25 mg/wk) for patients achieving remission [14]. The resultsare presented as cost per QALYs for Sweden, in the societalperspective, in 2010 euro (costs and effects discounted 3%).

Methods

Patients

The clinical trial [14] was initiated with an open phase in which834 patients with moderate RA (mean DAS28 4.4; mean functionalscore [health assessment questionnaire, HAQ] 1.1; mean diseaseduration 6.9 years) were enrolled and treated with the combina-tion of ETA and MTX for 36 weeks. Low disease activity (LDA;defined as DAS28 r 3.2) and remission (defined as DAS28 o 2.6)were assessed throughout the 36-week period. Patients whoachieved sustained LDA (mean DAS28 r 3 �2 from weeks 12 to36 and DAS28 r 3 �2 at week 36) were randomized to receive ETA50 mg or ETA 25 mg weekly both with MTX background therapy,or MTX alone for 52 weeks. Patients not fulfilling the random-ization criteria left the study.

At the end of the open-label phase, 80% (604) of these patientsfulfilled the randomization criteria. After 1 year, 82.6%, 79.1%, and42.6% had maintained LDA on full ETA, half ETA and MTX,respectively. Corresponding proportions for patients in remissionwere 66.7%, 60.2%, and 29.4%.

The economic analysis compares the three treatment strat-egies, but the induction phase in the trial and the patients notfulfilling the inclusion criteria into the double-blind phase at theend of this phase (nonrandomized) are incorporated into eachstrategy for completeness. This does not influence differencesbetween the three strategies in the current analysis because theopen-label phase is identical, but this would allow analyzing theeffect of different proportions of patients achieving LDA orremission with different induction therapies in the first phase.

The first cycle in the model represents the 88-week PRESERVEtrial, while subsequent (annual) cycles use data from two other2-year clinical trials with ETA [6,15] to estimate underlyingprogression while on treatment as well as discontinuation rates.Progression while on standard treatment with nonbiologicDMARDs was estimated from literature data [16].

The Model

Model structureThe model is developed as a microsimulation model pro-grammed in TreeAge, with a basic Markov model structure asthe vast majority of models in RA. This also allowed using datasets from previous analyses. The model has six basic states: fourtreatment states (full ETA plus MTX, half ETA plus MTX, MTXalone, and standard treatment); one state for patients who do notreach LDA in the initial 36-week period (“nonrandomized”); andone for dead. Transitions between these states are driven bypatients’ response to treatment arms: maintaining LDA (notransition), failing to maintain LDA (triggering a treatment

change), discontinuing treatment (leading to mixed standardtherapy as observed in the Swedish RA registry [11,12]), or death(according to normal mortality for women in Sweden).

Within each of these treatment states, patients are furtherdistributed into five health states defined by functional score(HAQ score o 0.6; 0.6–1.1; 1.1–1.6; 1.6–2.1; 42.1). Disease pro-gression is expressed as transitions between these HAQ states.Similarly, treatment response affects HAQ transitions, in additionto DA. HAQ has also been shown to correlate well with costs andutilities [12,13,17]. Costs in the model are thus driven by HAQ, byDA, and by what treatment patients are on. Similarly, utility isdriven by HAQ and DA. At each cycle, patients in the threestrategies thus move between treatment options defined by thelevel of DA as well as between HAQ levels. Patients who do notachieve LDA in the first treatment period will remain on treat-ment with full ETA plus MTX unless they discontinue (withtransitions between the HAQ state) or die. A simplified structureof the model is presented in Figure 1, and a summary of thetransitions and assumptions is given in Table 1.

Patient flowAll patients in the simulation start with LDA. Patients in the MTXor half-ETA strategies will stay on their treatment until they failto maintain LDA. Patients in the full-ETA strategy who maintainLDA throughout the 12 months of the clinical trial will switch tohalf ETA at the end of the trial and continue in that state as longas they maintain LDA. Patients with LDA are assumed not todiscontinue treatment.

Patients who fail to maintain LDA on half ETA in the two ETAstrategies are assumed to return to full ETA at the next cycle andreach LDA again, with the rationale (and in the absence of anyobservational data) that because they had reached LDA on fullETA previously they can be expected to reach it again. Sensitivityanalysis is performed for only half or none of these patientsreaching LDA again. Subsequently, they can fail to maintain LDAand either continue on full ETA but with high DA, or discontinueand receive standard treatment.

Patients in the MTX monotherapy strategy who do not main-tain LDA will also switch to full ETA at the end of the clinical trialand reach LDA again. If LDA is maintained over the followingcycle, they will then also have the possibility to switch to halfETA. In the absence of any data, the proportion is arbitrarily set to20%, with the rationale that few clinicians would want to lowerthe ETA dose in patients who had shown to flare when switchedto MTX from full ETA. Sensitivity analysis is performed for 50%going on half ETA. Thus, in this strategy, there will be patients onfull ETA with both LDA and high DA; as above, only patients withhigh DA may discontinue.

TransitionsMortality. The probability of dying is based on populationmortality (Statistiska Centralbyrån www.scb.se/befolkningssta-tistik) and thus depends only on age. Although patients withsevere and continuously active RA have a higher mortality risk,previous models have shown that the effect is limited, partic-ularly in this moderate group of early patients managed towardLDA in all three groups [11]. Incorporating RA-specific mortalitywould thus unnecessarily complicate the model.

Discontinuation. Only patients on full ETA with high DA areassumed to discontinue, as patients on half ETA or MTX areswitched to full ETA. During the first cycle, the probabilities fordiscontinuation are taken directly from the PRESERVE clinicaltrial [14]. For subsequent years, discontinuation rates areadjusted using matched patients from a 2-year clinical trial in

Table 1 – Summary of transitions and assumptions.

Treatment changesMaintain LDA Stay on the same treatment

No discontinuationSwitch from full to half ETA

(LDA maintained for 12 mo)Only once in each strategy

- Half-ETA strategy: 100% (0%, 50%)- Full-ETA strategy: 100% (0%, 50%)- MTX strategy: 20% (50%)

Return to full ETA upon failure to maintain 100%

LDAReach LDA again on full ETA 100%Fail to maintain LDA on full ETA Stay on full ETA or discontinueDiscontinuation (on full ETA and high DA) As observed

- First year: PRESERVE trial [13]- Subsequent years: matched patients from COMET trial [14]

Switch to mixed standard therapy [11,12]

HAQ changesPatients included in the double-blind

phase of PRESERVEFirst year:

- Annual change as in PRESERVE [13] for all three strategies (full and half ETA and MTX)

Subsequent years:

- High DA: estimated from a matched sample of patients with high DA from twotrials at 0.010/y (COMET and TEMPO [14,15])

- LDA: 50% of high DA rate, 0.005/y

Patients not included in the double-blindphase of PRESERVE

On full ETA, 0.0102/y (as above)

Patients on standard therapy Average progression rate from the literature, 0.031/y [16]

COMET, combination of methotrexate and etanercept in active early rheumatoid arthritis [6]; DA, disease activity; full ETA, etanercept 2 � 25mg/wk; half ETA, etanercept 25 mg/wk; HAQ, health assessment questionnaire (functional score); LDA, low disease activity; MTX,methotrexate; PRESERVE, Prospective Randomized Etanercept Study to Evaluate Reduced dose Etanercept Versus full dose Etanercept [14];TEMPO, Trial of Etanercept and Methotrexate with Radiographic Patient Outcomes [15].

Fig. 1 – Schematic illustration of the model structure. Patients achieving remission in the open-label phase with full-dose ETAare randomized to continue on full-dose ETA or move to half-dose ETA or discontinue ETA, all with background MTX. Patientswho discontinue and flare return to full-dose ETA and switch to half-dose ETA if they achieve remission again. Patients onhalf-dose ETA who fail to maintain remission return to full-dose ETA. Patients who continue on full-dose ETA switch to half-dose ETA if they maintain remission for an additional cycle. Patients who fail to maintain remission in any of the armscontinue on full-dose ETA or are allowed to discontinue. ETA, etanercept; MTX, methotrexate.

V A L U E I N H E A L T H ] ( 2 0 1 4 ) ] ] ] – ] ] ] 3

Table 2 – Transition probabilities.

Discontinuation MTX strategy Half-ETA strategy Full-ETA strategy Source

to full ETA to full ETA to standard therapyFirst year 0.295 0.134 0.104 PRESERVE first-year discontinuation [14]Subsequent years 0.133 0.060 0.047 45% of first year according to COMET [6]Maintaining LDA 0.426 0.791 0.826 Based on PRESERVE first year [14],

extrapolated to following cyclesMaintaining remission 0.294 0.602 0.667 Patients not maintaining LDA go to full ETA

COMET, combination of methotrexate and etanercept in active early rheumatoid arthritis [6]; DA, disease activity; full ETA, etanercept 2 � 25mg/wk; half ETA, etanercept 25 mg/wk; MTX, methotrexate; PRESERVE, Prospective Randomized Etanercept Study to Evaluate Reduced doseEtanercept Versus full dose Etanercept [14].

V A L U E I N H E A L T H ] ( 2 0 1 4 ) ] ] ] – ] ] ]4

which second-year discontinuation was 45% of the first year [6](Table 2).

Maintenance of LDA. Probabilities to maintain LDA for the firstyear are calculated directly from the clinical trial, and the samerates are used in subsequent years, which may result in con-servative estimates. Indeed, data from the only cohort of similarpatients treated with half ETA followed over several yearsindicate that failure rates are highest in the first 6 months andthen decrease substantially [8].

In an alternative analysis, the more stringent criterion of fullremission is applied; that is, only those patients who achieve aDAS28 value of less than 2.6 in the open period (63 � 6%) areincluded in the analysis and patients not maintaining fullremission are switched back to full ETA. All other variables inthe simulation remain the same (Table 2).

HAQ progression. HAQ is modeled as changes from the baselinevalues in the clinical trial, by HAQ state (Table 3). For patientsqualifying for the double-blind period of the trial, changes duringthe open 36-week phase and the subsequent double-blind periodof 52 weeks are taken directly from the trial, as shown in Table 2.No difference was found between the three strategies because theindividual groups were too small to yield reliable estimates; also,in the setting of this model, differences between the groups areessentially driven by the maintenance of LDA or discontinuation.

Patients who fail to maintain LDA after the trial and are thustreated with full-dose ETA progress at 0.0102 HAQ points per year.This was estimated from a sample of patients extracted from theCOMET and the TEMPO trials [6–15], with the same baselinecharacteristics as patients in PRESERVE [14] and who had not reachedLDA after 6 months despite treatment with ETA. The estimate uses

Table 3 – HAQ changes (first year).

States Patients not achieving LDA in theperiod

Open period(PRESERVE)

First yearTEM

1 (HAQ score o0.6) �0.208 �0.12 (HAQ score 0.6 o1.1) �0.131 �0.23 (HAQ score 1.1 o1.6) �0.132 �0.34 (HAQ score 1.6 o2.1) �0.391 �0.25 (HAQ score r2.1) �0.634 �0.2

COMET, combination of methotrexate and etanercept in active early rheu[23]; PRESERVE, Prospective Randomized Etanercept Study to Evaluate Reof Etanercept and Methotrexate with Radiographic Patient Outcomes [15* No patients were identified in these groups and progression in state 3

the period between 6 and 18 months to exclude the initial treatmenteffect in the two trials. Patients who do not reach LDA in the firstperiod and continue on full ETA progress at this rate as well.

For patients who maintain LDA, an arbitrary annual progressionrate of 0.005 HAQ points per year (half the rate of patients with highDA) is used as in a previous model [13], based on the argument thatdespite LDA there was still detectable synovitis. In view of thesevery low rates, no distinction by HAQ state is applied.

Patients in all groups who discontinue ETA return to theirbaseline HAQ at the next cycle and will then progress at theannual rate of 0.031 HAQ [16].

Costs and effectsCosts and effects (utilities) depend on HAQ (five states) and DA(high or low) and are based on a regular cross-sectional popula-tion-based observational study in Southern Sweden as in pre-vious models, adjusted to 2010 [13,18,19] (Table 4). The surveycovers an estimated 90% of all patients in the area (N ¼ 1027) andincludes all types of resource consumption (health care, com-munity services, patient costs, and production losses) as well as autility measure (EuroQol five-dimensional questionnaire). Regres-sion analyses had shown that all costs were driven by function(HAQ), except sick leave, which was dependent on DA and ageand was hence calculated separately for patients with high or lowDA. For the current microsimulation technique, the original costsper HAQ state were transformed into continuous variables byextrapolation between the states.

The cost of ETA is €14,341 and the cost of MTX is €47, based onSwedish list prices [20].The initial cost (representing the first 36weeks) is the same for all patients, because all had the sametreatment, and is calculated as 36/52 of the annual cost corre-sponding to the HAQ and DAS levels of the total group. Patients

first Patients included in the double-blind period

(COMET/PO)

Open period(PRESERVE)

First year (trial)(PRESERVE)

18 �0.155 �0.11348 �0.466 �0.10225 �0.795 �0.24128 �0.957 �0.241*

58 �1.458 �0.241*

matoid arthritis [6]; HAQ, health assessment questionnaire (function)duced dose Etanercept Versus full dose Etanercept [14]; TEMPO, Trial].was carried forward.

Table 4 – Costs and utilities (€2010, 1€ ¼ 9.2 SEK).

States Utilities Costs (€2010) *

Low diseaseactivity

High diseaseactivity

Healthcare costs

Patientcosts

Long-term work absence/early retirement†

1 (HAQ score o0.6) 0.78 0.709 915 313 2,5342 (HAQ score 0.6 o1.1) 0.704 0.568 1,719 455 5,1733 (HAQ score 1.1 o1.6) 0.676 0.441 2,654 821 8,1054 (HAQ score 1.6 o2.1) 0.562 0.446 3,778 1,501 9,4055 (HAQ score r2.1) 0.408 0.213 9,607 1,440 12,671

DA, disease activity; HAQ, health assessment questionnaire (function) [23].* Costs in the model are continuous; numbers presented here for illustrative purposes are costs at the midpoint in the states, calculated usingbootstrapping [12].

† Sick leave was calculated separately with regression analysis and added to the total cost of individual patients because it depended on high/low disease activity rather than function (SEK 12,313 � [186 � age] þ [2,689 � high DA]) [13,19].

V A L U E I N H E A L T H ] ( 2 0 1 4 ) ] ] ] – ] ] ] 5

who discontinued before the end of the 36-week period areassumed to do so on average in the middle of the period andare assigned half the cost.

Effects were modeled in the same way as costs, with an initialchange in utility based on HAQ changes during the 36 weeks addedat the start of the simulation, and annual HAQ changes thereafter.

Analyses

The simulations are run for 2, 5, and 10 years. The 2-year timeframe should be interpreted as representing the clinical trial effectonly because the changes achieved during the trial that lead todose adjustments or discontinuation are implemented in themodel at the end of the trial (first cycle) and thus affect resultsin the second cycle. Five- and 10-year horizons not only allow thelong-term effect to be incorporated but also half-dose ETA to beused in the strategies starting with full ETA or MTX alone.

Results are presented both for patients achieving LDA and forpatients achieving complete remission (DAS o 2.6). An additionalanalysis explores the effect on the results if half ETA is allowed onlyat the start of the simulation in the half-ETA strategy, and switch-ing to half ETA is disallowed in the full-ETA and the MTX strategies.

Analyses are performed with 30,000 runs where results wereshown to be stable. Results are presented as costs (2010 euro;1€ ¼ 9.2 SEK [12–month average]) per QALY gained, and bothcosts and effects are discounted with 3%.

Results

Main Analyses

For all analyses, we first present a comparison of each of the twoETA strategies with MTX and then the two ETA strategies witheach other (Table 5).

The cost per QALY for the half-ETA strategy versus MTX variesbetween €14,000 and €29,000, depending on the time frame:Longer durations of the simulations increase the incrementalcost-effectiveness ratio (ICER), as incremental costs of the ETAstrategies versus MTX become higher. Half ETA technicallydominates full ETA (i.e., it has lower costs and slightly bettereffectiveness) although differences are small.

When using full remission as the criteria for randomizationand switching, ICERs are higher (€25,000 and €56,000 dependingon the time frame) because fewer patients are allowed to switchto half ETA, thus increasing costs in the ETA strategies (Table 5).

If no assumptions about dose reductions are implemented afterthe trial in the MTX and full-ETA arm, costs in both arms increase.

In the MTX arm, the increase is limited and outcome is not affectedbecause switches would happen only toward the end of the 5-yearsimulation. In the full-ETA arm, costs increase significantlybecause all patients stay on full ETA regardless of whether theyachieve remission or not. The half-ETA arm is not affected becausehalf dose is introduced only at the start of the simulation. As aconsequence, the ICER for half ETA compared with MTX decreases,while the ICER for full ETA compared with MTX increases,reinforcing the dominance of the half-ETA strategy (Table 5).

Total costs over 5 years are €100,500 in the MTX arm and€103,200 in the half-ETA arm. Treatment costs were €49,700 and€56,800, respectively, but direct health care costs decreased from€13,300 to €8,500 with half ETA.

After 5 years, a substantial proportion of patients are on ETAin all three strategies, with most of them in remission at the endof the last cycle (63% in the MTX arm, 75% in the half-ETA arm,and 67% in the full-ETA arm) (Table 6).

Sensitivity Analysis

Because in all three strategies patients are managed toward LDAover time, results are similar and we present sensitivity analysisonly for the key assumptions made in terms of cost and QALYchanges because in such a situation ICERs become unstable andessentially meaningless (Table 7). Varying the assumption aboutproportions of patients switching to half ETA after randomization tofull ETA substantially changes the results. Fewer patients going onhalf ETA increases costs, but also slightly reduces effects as aconsequence of our assumption that only patients failing on fullETA discontinue. Thus, patients failing tomaintain LDA on half ETAwill not discontinue but can achieve it again upon return to fullETA. Increasing the proportion of patients switching to half ETA inthe MTX arm decreases costs but does not affect the outcomebecause switches happen at the end of the 5-year simulation.Varying the proportions of patients achieving remission again whenrestarting full ETA has a limited effect in the two ETA arms butincreases costs and reduces outcome in the MTX strategy, decreas-ing thereby the ICER, particularly in the half-ETA arm. If the higherdiscontinuation rates in the first year in the MTX strategy are set tobe equal as in the two ETA strategies, half ETA also dominates MTXbecause with a lower discontinuation rate with MTX more patientswould switch to full ETA and thus increase the cost.

Discussion

After more than a decade of using TNF-α inhibitors, research iscontinuing on when and how best to use these agents. Theiradverse-event profile has been well characterized, making them

Table 6 – Cohort distribution at the end of 5 y in thethree strategies.

Status Proportion of patients

MTXstrategy

Half-ETAstrategy

Full-ETAstrategy

Full ETA 35.0 31.5 31.1Half ETA 6.1 25.9 26.6Only MTX 2.6 0.0 0.0Dead 0.5 0.5 0.5Discontinued 55.9 42.2 41.8Start last cycle in

remission32.3 46.6 41.0

Remission at end of lastcycle

27.5 43.1 38.9

Full ETA, etanercept 2 � 25 mg/wk; half ETA, etanercept 25 mg/wk;MTX, methotrexate.

Table 5 – Main results (€2010).

Cost (€) Incremental cost (€) QALY Incremental QALY ICER (€)

2 yLDAMTX 50,389 1.723Half ETA vs MTX 50,904 515 1.760 0.037 14,000Full ETA vs MTX 54,867 4,478 1.753 0.030 148,288Full ETA vs half ETA 3,963 �0.007 Half ETA dominates

RemissionMTX 52,685 1.714Half ETA vs MTX 53,385 700 1.742 0.028 25,180Full ETA vs MTX 56,574 3,889 1.729 0.015 262,770Full ETA vs half ETA 3,189 �0.013 Half ETA dominates

5 yLDAMTX 100,479 3.373Half ETA vs MTX 103,236 2,757 3.504 0.131 21,046Full ETA vs MTX 108,170 7,691 3.488 0.115 66,878Full ETA vs half ETA 4,934 �0.016 Half ETA dominates

RemissionMTX 103,970 3.333Half ETA vs MTX 109,198 5,228 3.445 0.112 46,679Full ETA vs MTX 113,517 9,547 3.421 0.088 108,489Full ETA vs half ETA 4,319 �0.024 Half ETA dominates

No switching to half ETA inMTX and full-ETA armsMTX 101,925 3.373Half ETA vs MTX 103,236 1,311 3.504 0.131 10,008Full ETA vs MTX 116,795 14,870 3.457 0.084 177,024Full ETA vs half ETA 13,559 �0.047 Half ETA dominates

10 yLDAMTX 168,808 5.575Half ETA vs MTX 178,510 9,702 5.906 0.331 29,311Full ETA vs MTX 185,129 16,321 5.895 0.320 51,003Full ETA vs half ETA 6,619 �0.011 Half ETA dominates

RemissionMTX 171,939 5.459Half ETA vs MTX 187,218 15,279 5.733 0.274 55,763Full ETA vs MTX 193,539 21,600 5.723 0.264 81,818Full ETA vs half ETA 6,321 �0.010 Half ETA dominates

Full ETA, etanercept 2 � 25 mg/wk; half ETA, etanercept 25 mg/wk; ICER, incremental cost-effectiveness ratio; LDA, low disease activity; MTX,methotrexate; QALY, quality-adjusted life-years.

V A L U E I N H E A L T H ] ( 2 0 1 4 ) ] ] ] – ] ] ]6

very safe therapies for a majority of patients. Their cost, however,limits their use to patients with active disease not responding toclassical DMARDs. Budget restrictions have induced clinicians toevaluate whether effectiveness can be maintained with lowerdoses or less frequent administration, or whether temporary orlong-term discontinuation is an option [6,8,21]. At the same time,the concept of “treat to target” has been widely accepted andguidelines established [4]. Thus, patients are managed inten-sively toward remission as early as possible in their diseasecourse to prevent erosions in the longer term. Earlier treatmentwould, however, lead to more patients being treated. Because it isnot yet possible to establish a definitive prognostic for severe andprogressive disease very early, treatment would have to be givento a relatively large population initially and thus represent anadditional financial burden. The potential to maintain remissionwith both less intensive management and dose reductions anddrug holidays, if proven and applied, will be able to limit andcontrol the budget impact of early treatment. Currently, fewstudies have explored these options, but results clearly show

Table 7 – Sensitivity analysis for key assumptions.

Scenario Costs in euros (€2010) QALYs

MTX Half ETA Full ETA MTX Half ETA Full ETA

Base case (5 y) 100,500 103,200 108,200 3.37 3.50 3.49Switching to half ETA in full-ETA arm (base case 100%)0% 116,800 3.4650% 112,400 3.47

Switching to half ETA in MTX arm (base case 20%)50% 98,400 3.37

Recovering remission on going back to full ETA (base case 100%)0% 103,500 104,700 109,100 3.27 3.44 3.4550% 102,000 103,900 108,600 3.32 3.47 3.47

Full ETA, etanercept 2 � 25 mg/wk; half ETA, etanercept 25 mg/wk; MTX, methotrexate; QALY, quality-adjusted life-year.

V A L U E I N H E A L T H ] ( 2 0 1 4 ) ] ] ] – ] ] ] 7

the potential of this approach in early disease: In the PRESERVEtrial, 79% of the patients maintained LDA after 12 months despitea 50% dose reduction of etanercept [6]; in the Optimal Protocol forTreatment Initiation with Methotrexate and Adalimumab trial,the proportion of patients with early RA maintaining LDA for 12months after discontinuation or continuation of the biologic wassimilar (www.eular.com); in the Behandel Strategieen trial, 52% of thepatients maintained LDA 7 years after discontinuation of infliximab[21].The common feature in these trials was the early introduction ofeffective treatment and close management of patients.

The modeling study presented here explores the economicconsequences of treating to target with ETA, followed by a 50%dose reduction or discontinuation for patients with LDA.Although it might appear obvious that reducing ETA will lead tolower costs, we argue that this is the case only initially, in a purebudget perspective. In the longer term, when differences inoutcome, discontinuation, and treatment switches are taken intoaccount, the result is less predictable, as can be seen from thesmall differences in costs of the three strategies explored here.The model adopts a longer time horizon than the clinical trial toallow all strategies to treat toward maintenance of remission.Thus, patients who fail to maintain LDA on half ETA or MTX areallowed to return to full ETA and patients on full ETA and withLDA for two cycles are allowed to reduce the dose. This repre-sents a number of issues that require discussion.

First, in the absence of data regarding the maintenance of LDAbeyond the trial, this is based on assumptions, albeit supportedby data from other cohorts [8] and sensitivity analysis is pre-sented. Second, we have not included switching to differentbiologic treatments when patients fail to maintain LDA on ETA;rather patients discontinue and receive standard DMARD therapyas documented for Sweden. Although some data on therapychanges are available from the Swedish biologics registry[12,22], these concern the earlier years of biologic therapy andmainly TNF-α inhibitors and are thus not representative ofcurrently available alternatives and clinical practice. Also, themodel would have been unnecessarily complicated because ourobjective was not to assess current practice but to estimate thepotential future impact of dose reductions. Finally, with thesimilarity of the treatment options in all three strategies, differ-ences become very small, and cost-effectiveness ratios thereforenot very meaningful. We therefore prefer to discuss the results ina more general manner.

The strategy with half ETA after the 36-week induction periodprovides a better outcome, at only a slightly increased cost,compared with MTX. The proportion of patients discontinuingtreatment in the MTX strategy, however, was higher than in theETA arms and one could assume that these patients would havebeen switched back to a different biologic. We have not

incorporated this in the base case because the cost increasedue to more biologic treatment could be considered biasingagainst MTX. Rather, we show a sensitivity analysis with identi-cal dropout rates in all three strategies, which results in half ETAdominating not only full ETA but MTX as well. This seems toindicate that lowering the dose of biologics is a better option than“biologic holidays” when remission is achieved.

In all simulations, the half-ETA strategy technically dominatesthe full-ETA strategy. This may be surprising because one wouldindeed expect not only higher costs but also better outcome withfull ETA. Although this is indeed the case in the PRESERVE trial [14],the difference is very small. It is also somewhat further reducedbecause of the model structure. Patients on full ETA who do notmaintain remission can discontinue biologic therapy and receivestandard DMARD therapy with the corresponding effectiveness.Patients on half ETA, however, will move to full ETA when they donot maintain remission and achieve remission again. In thisscenario, we would argue that the two strategies provide the sameoutcome, but logically the costs are higher with full ETA. If wecompare scenarios in which half dose is introduced only atrandomization and not after the end of the trial, however, thestrategy with half ETA is clearly dominating full ETA, and providesbetter outcome at a limited cost-increase compared with MTX.

In conclusion, in this model we explored the costs and out-comes when patients with moderate RA are treated to target(remission) initially and then managed to maintain remission.The model is based on a number of data sets, but neverthelessrepresents a hypothetical scenario. Although ultimately all threestrategies explored—maintaining patients on full ETA, reducingthe dose to half ETA, or discontinuing ETA—give rather similarresults because all three continuously manage patients to main-tain remission, it appears that a dose reduction is the mostadvantageous strategy in patients with moderate RA. Orexpressed differently, maintaining biologics at the full dose inall patients in the case of remission is an unnecessary invest-ment in treating patients with moderate RA, but discontinuingbiologics altogether may not provide the expected advantage interms of costs and jeopardize outcomes. It must, however, bepointed out that at the time of this publication, the product labelfor ETA is only for 50 mg weekly for moderate to severe RA anddoes not include a lower dose.

Since the advent of the TNF inhibitors, treatment approachesand outcome in RA have been intensely explored, including long-term follow-up in registries. Thus, large amounts of data areavailable that allow building hypotheses about treatmentchanges, including dose modifications as proposed in this study.The importance of dose modifications and treatment holidays isobviously not limited to RA and can be explored in other diseases,provided enough high-quality data are available.

V A L U E I N H E A L T H ] ( 2 0 1 4 ) ] ] ] – ] ] ]8

Acknowledgments

We are grateful for modeling support from Ingrid Lekander andErik Landtfeldt at Optum Insight, Sweden.

Source of financial support: The study was sponsored byPfizer, Inc., to European Health Economics.

R E F E R E N C E S

[1] Smolen JS, Landewe R, Breedveld FC, et al. EULAR recommendations forthe management of rheumatoid arthritis with synthetic and biologicaldisease-modifying antirheumatic drugs. Ann Rheum Dis 2010;69:964–75.

[2] Furst DE, Keystone EC, Braun J, et al. Updated consensus statement onbiological agents for the treatment of rheumatic diseases, 2011. AnnRheum Dis 2012;71(Suppl. 2): i2–45.

[3] Combe B, Landewe R, Lukas C, et al. EULAR recommendations for themanagement of early arthritis: report of a task force of the EuropeanStanding Committee for International Clinical Studies IncludingTherapeutics (ESCISIT). Ann Rheum Dis 2007;66:34–45.

[4] Smolen JS, Aletaha D, Bijlsma JW, et al. Treating rheumatoid arthritis totarget: recommendations of an international task force. Ann RheumDis 2010;69:631–7.

[5] Breedveld FC, Weisman MH, Kavanaugh AF, et al. The PREMIER study: amulticenter, randomized, double-blind clinical trial of combinationtherapy with adalimumab plus methotrexate versus methotrexatealone or adalimumab alone in patients with early, aggressiverheumatoid arthritis who had not had previous methotrexatetreatment. Arthritis Rheum 2006;54:26–37.

[6] Emery P, Breedveld F, van der Heijde D, et al. Two-year clinical andradiographic results with combination etanercept-methotrexatetherapy versus monotherapy in early rheumatoid arthritis: a two-year,double-blind, randomized study. Arthritis Rheum 2010;62:674–82.

[7] Brocq O, Millasseau E, Albert C, et al. Effect of discontinuing TNFalphaantagonist therapy in patients with remission of rheumatoid arthritis.Joint Bone Spine 2009;76:350–5.

[8] Botsios C, Furlan A, Ostuni P, et al. Effects of low-dose etanercept inmaintaining DAS-remission previously achieved with standard dose inpatients with rheumatoid arthritis. Ann Rheum Dis 2009;66:54.

[9] Emery P, Kvien TK, Combe B, et al. Combination etanercept andmethotrexate provides better disease control in very early (o¼4 months)

versus early rheumatoid arthritis (44 months and o2 years): posthoc analyses from the COMET study. Ann Rheum Dis 2012;71:989–92.

[10] Gremese E, Salaffi F, Bosello SL, et al. Very early rheumatoid arthritisas a predictor of remission: a multicentre real life prospective study.Ann Rheum Dis 2013;72(6):858–62.

[11] Kobelt G, Kasteng F. The burden of rheumatoid arthritis and access tonew treatments. Available from: www.comparatorreports.se. [AccessedJanuary 13, 2014].

[12] Kobelt G, Lindgren P, Geborek P. Costs and outcomes for patients withrheumatoid arthritis treated with biological drugs in Sweden: a modelbased on registry data. Scand J Rheumatol 2009;38:409–18.

[13] Kobelt G, Lekander I, Lang A, et al. Cost-effectiveness of etanercepttreatment in active early RA followed by dose adjustment. Int J TechnolAssess Health Care 2011;27:193–200.

[14] Smolen J, Nash P, Durez P, et al. Maintenance, reduction, or withdrawalof etanercept after treatment with etanercept and methotrexate inpatients with moderate rheumatoid arthritis (PRESERVE): a randomisedcontrolled trial. Lancet 2013;381:918–29.

[15] Klareskog L, van der Heijde D, de Jager JP, et al. Therapeutic effect of thecombination of etanercept and methotrexate compared with eachtreatment alone in patients with rheumatoid arthritis: double-blindrandomised controlled trial. Lancet 2004;363:675–81.

[16] Scott D, Pugner K, Kaarela K, et al. The links between joint damageand disability in rheumatoid arthritis. Rheumatology 2000122–32.

[17] Kobelt G, Lindgren P, Singh A, Klareskog L. Cost effectiveness ofetanercept (Enbrel) in combination with methotrexate in the treatmentof active rheumatoid arthritis based on the TEMPO trial. Ann RheumDis 2005;64:1174–9.

[18] Jacobsson LT, Lindroth Y, Marsal L, et al. Rheumatoid arthritis: whatdoes it cost and what factors are driving those costs? Results of asurvey in a community-derived population in Malmo, Sweden. Scand JRheumatol 2007;36:179–83.

[19] Kobelt G, Lindgren P, Lindroth Y, et al. Modelling the effect of functionand disease activity on costs and quality of life in rheumatoid arthritis.Rheumatology (Oxford) 2005;44:1169–75.

[20] FASS. Läkemedel i Sverige (Pharmaceutical Lexicon). Stockholm,Sweden: Lakemedelsinformation AB, 2011.

[21] Van den Broek M, Klarenbeek NB, Dirven L, et al. Discontinuation ofinfliximab and potential predictors of persistent low disease activity inpatients with early rheumatoid arthritis and disease activity score-steeredtherapy: subanalysis of the BeSt study. Ann Rheum Dis 2011;70(8):1389–94.

[22] Lekander I, Borgstrom F, Svarvar P, et al. Cost-effectiveness of real-world infliximab use in patients with rheumatoid arthritis in Sweden.Int J Technol Assess Health Care 2010;26:54–61.

[23] Fries J, Spitz P, Kraines R, Holman H. Measurement of patient outcomein arthritis. Arthritis Rheum 1980;23:137–45.