Modelling the cost-effectiveness of biologic treatments for psoriatic arthritis

Modelling the cost-effectiveness of biologic treatments for psoriatic arthritis
of 9
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  Modelling the cost-effectiveness of biologictreatments for psoriatic arthritis Laura Bojke 1 , David Epstein 1 , Dawn Craig 2 , Mark Rodgers 2 , Nerys Woolacott 2 ,Huiqin Yang 2 and Mark Sculpher 1  Abstract Objectives.  A probabilistic model was developed to determine the cost-effectiveness of three biologics,etanercept, infliximab and adalimumab, compared with palliative care for the treatment of active andprogressive PsA in patients who have an inadequate response to standard treatment (including DMARDs). Methods.  A previous model was revised to evaluate the impact of biologics on both skin and joint diseaseand to include new evidence from the clinical review and evidence synthesis. Initial response to biologicswas determined using the PsA response criteria. The impact of biologics on the arthritis component of thedisease is then modelled via a change in the HAQ and the impact of the psoriasis component measuredusing the Psoriasis Area and Severity Index. Results.  For PsA patients with mild to moderate skin disease, the incremental cost-effectiveness ratio(ICER) for etanercept  vs  palliative care is around £18000, and the ICER for infliximab  vs  etanercept isaround £44000 per quality-adjusted life year (QALY). Adalimumab is extendedly dominated. The prob-ability that etanercept is cost effective is 0.436 at a threshold of £20000 per QALY. Etanercept is alsolikely to be cost effective for patients with moderate to severe psoriasis or negligible skin involvement. Conclusions.  Further investigation is required to reduce uncertainties around a number of model param-eters, in particular the length of time over which biologics are assumed to be effective and the progressionof HAQ on and off treatment. Key words:  Cost-effectiveness, Biologics, Psoriatic arthritis, Decision model. Introduction PsA is defined as a unique inflammatory arthritis affectingthe joints and connective tissue and is associated withpsoriasis of the skin or nails [1]. The prevalence of psor-iasis in the general population has been estimated to bebetween 2 and 3% [1], and the prevalence of inflammatoryarthritis in patients with psoriasis has been estimated tobe up to 30% [2]. In the UK, DMARDs are often used totreat active and progressive PsA. For patients that have aninadequate response to DMARDs, biologic agents thattarget tumour necrosis factor (TNF) are available.The economic costs of PsA are likely to be substantial.In the USA, the mean annual direct (health and social care)cost per patient with PsA was $3638 according to datafrom Medstat MarketScan in 1999  2000 [3], but similardata are not yet available for the UK. Expenditures onbiologic therapies have been increasing rapidly in recentyears [4], although this might be at least partly offset bycost savings elsewhere [5].The cost-effectiveness of the biologics etanercept andinfliximab has been previously assessed by the NationalInstitute for Health and Clinical Excellence (NICE) [6].However, this did not include the impact of biologics onthe psoriasis component of PsA. In addition, since thisappraisal was published, another biologic, adalimumab,is also available for the treatment of PsA. To addressthese issues, a new decision model was constructed todetermine the cost-effectiveness of the biologics etaner-cept, infliximab and adalimumab for the treatment of active and progressive PsA in patients who have an inad-equate response to standard treatment (including DMARDtherapy). The model was developed alongside a system-atic review and synthesis of clinical evidence and a 1 Centre for Health Economics and  2 Centre for Reviews andDissemination, University of York, Heslington, York, UK.Correspondence to: Laura Bojke, Centre for Health Economics, Alcuin A Block, University of York, Heslington, York YO10 5DD, UK.E-mail: 2 February 2011; revised version accepted 7 June 2011. ! The Author 2011. Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved. For Permissions, please email: RHEUMATOLOGY   Rheumatology 2011;50:iv39  iv47doi:10.1093/rheumatology/ker245   a  t   J  BM or r  e l  l  L i   b r  a r  y , Uni   v e r  s i   t   y of  Y or k  on S  e  p t   e m b  e r 1 2  ,2  0 1  3 h  t   t   p :  /   /  r h  e  um a  t   ol   o g y . oxf   or  d  j   o ur n a l   s  . or  g /  D o wnl   o a  d  e  d f  r  om  systematic review of the cost-effectiveness literature,which was undertaken to inform a NICE appraisal. Thisarticle reports the methods and results of the  de novo economic model, which updates a previous model [7] byincorporating the biologic adalimumab, incorporating theeffect of biologics on the psoriasis aspect of the disease[measured using the Psoriasis Area Severity Index (PASI)],incorporating any additional trials and updating of modelparameters, including withdrawal rates, elicited evidenceon rebound after treatment withdrawal, costs of diseaseand natural history data. Methods Overview  A probabilistic cohort model tracks patients with PsA overa lifetime (40 years). Patients either receive a biologictreatment (etanercept, infliximab or adalimumab) or pallia-tive care (no biologic therapy). The model is implementedin the R programming software. The structure and eventsin the model aimed to be consistent with licensed indica-tions and current British Society of Rheumatologists (BSR)[8] and British Association of Dermatologists (BAD) [9]guidelines for the use of biologics in PsA [7]. The modeladopts the perspective of the UK National Health Service(NHS) and personal social services to quantify costs andhealth outcomes are measured as quality-adjusted lifeyears (QALYs). The price year assumed for costs is2008/2009 and the annual discount rate is 3.5% [10] forboth costs and QALYs.Initial response to biologic treatment is determinedusing the PsA response criteria (PsARC). The impact of biologics on the functional status (arthritis) component of the disease is then modelled via a change in the HAQ,which ranges from 0 to 3, with 3 being the most severestate. The impact of the psoriasis component of the dis-ease is measured using the PASI.The mean age of the population is assumed to be47 years, with at least 7 years since diagnosis of PsA,based on the average characteristics of participants inthe randomized controlled trials (RCTs) [7]. Patientsfulfil the BSR criteria in terms of previous therapies [7].In the base case, the HAQ at the start of the model is1.05, based on the average in the RCTs [7]. We assumepatients in the base case have mild to moderate psoriasiswith a PASI score of 7.5 (Ian Bruce, Arthritis ResearchUK Epidemiology Unit, School of Translational Medicine,personal communication, 20 November 2009). Subgroupsare also presented by varying the baseline HAQ and PASIscores to reflect more or less psoriasis involvement. Model structure The model structure is shown in Fig. 1. The base casemodel assumes that those with a PsARC response at3 months will continue on biologic treatment. The re-sponse time is consistent with the trials and the accepted F IG . 1  Model structure. Figure reproduced with permission from Rodgers M, Epstein D, Bojke L  et al  . Etanercept,infliximab and adalimumab for the treatment of psoriatic arthritis: a systematic review and economic evaluation. HealthTechnol Assess 2011;15:1  329. No PsARC ACycle 1 Months 1 – 3 Cycle 2 Months 3 – 6 Biologic  j No treatment N NPASI 75 only:withdrawn frombiologic  j Response of neither:withdrawn frombiologic  j PASI 75 No PASI 75 PASI 75 No PASI 75  p.m. BCAResponse of both:continue onbiologic  j  p.w.  p.m. BA  p.m.  p.w. A  p.m. PsARC  p.m. CAAResponse of PsARConly: continue onbiologic  j iv40 Laura Bojke  et al.   a  t   J  BM or r  e l  l  L i   b r  a r  y , Uni   v e r  s i   t   y of  Y or k  on S  e  p t   e m b  e r 1 2  ,2  0 1  3 h  t   t   p :  /   /  r h  e  um a  t   ol   o g y . oxf   or  d  j   o ur n a l   s  . or  g /  D o wnl   o a  d  e  d f  r  om  initiation period according to BSR guidelines [11]. Thosewho do not respond according to PsARC will discontinuebiologic treatment and switch to palliative care (with asso-ciated natural history progression—see Fig. 2). For thoseinitially receiving palliative care and those failing to re-spond according to PsARC, the arthritis componentsteadily deteriorates over time. The psoriasis componentof the disease is assumed to remain constant.Following a PsARC response, or no response to bio-logics according to PsARC, the impact of treatment onthe two aspects of PsA are modelled: the arthritis andthe psoriasis components. Change in HAQ score is as-sumed to be relative to baseline HAQ. For PsARC re-sponders, there will be an HAQ gain that corresponds toa drop in HAQ score. There is very little data to informwhether patients maintained on biologic therapy willmaintain the initial improvement in HAQ over the longterm. Previous models have assumed that the initial gainis maintained while the patient continues with biologictherapy, and this is the base case in the current model(Fig. 2). Opinions were elicited from rheumatologists toinform this assumption [12] (see Rodgers  et al  . [7] for fur-ther details). For PsARC non-responders, a slight HAQgain is estimated in the initial (3-month) treatment assess-ment period, after which they will no longer receive bio-logics and will switch to palliative care and the associatednatural history progression of HAQ.The psoriasis component of the disease is modelled viathe PASI. A patient who achieves a PASI of 75 has gainedat least a 75% improvement in psoriasis compared withbaseline PASI. Patients who do not achieve a PASI75 response will nevertheless have some proportionategain in PASI while they continue taking a biologic,though this will be  < 75% improvement (see Rodgers et al  . [7] for further detail). QALYs associated with eachtreatment strategy are calculated by mapping the calcu-lated HAQ and PASI scores onto utilities (see below) andare then transformed into QALYs over a lifetime.The base case model assumes a constant rate of with-drawal from biologics after 12 weeks (for those with an ini-tial PsARC response). Withdrawal might occur for lack of continuing efficacy (secondary non-response), adverseevents or other reasons. In this model, patients are as-sumed to return to palliative care after withdrawal frombiologic therapy. There is little evidence about patient out-comes (known as rebound) after withdrawal from biologictherapy. Other models have assumed either that patientsreturn to their baseline HAQ (rebound equal to initial gain)or that patients rebound to the HAQ score that wouldhave been expected had they never taken biologic ther-apy (rebound equal to natural history). In the base case, itis assumed that rebound is equal to initial gain in terms of HAQ and PASI scores (Fig. 2). This assumption was in-formed by elicitation from experts (see Rodgers  et al  . [7]).Patients with PsA have been found to have lower lifeexpectancy than the general population [13]. However,there is no evidence that biologics affect mortality. Thedecision model assumes no difference in mortality ratesbetween treatments, or between biologic treatments andno treatment. All-cause mortality was estimated from UKlife tables. A Gompertz function was fitted to these data(Table 1). F IG . 2  Illustration of the progression of arthritis according to HAQ scores. Figure reproduced with permission fromRodgers M, Epstein D, Bojke L  et al  . Etanercept, infliximab and adalimumab for the treatment of psoriatic arthritis:a systematic review and economic evaluation. Health Technol Assess 2011;15:1  329. 00.511.5     I 22.5301020304050 Year HAQ established on drug (zero progression)HAQ natural historyHAQ if withdrawn at Year 5  iv41 Cost-effectiveness of biologic treatments   a  t   J  BM or r  e l  l  L i   b r  a r  y , Uni   v e r  s i   t   y of  Y or k  on S  e  p t   e m b  e r 1 2  ,2  0 1  3 h  t   t   p :  /   /  r h  e  um a  t   ol   o g y . oxf   or  d  j   o ur n a l   s  . or  g /  D o wnl   o a  d  e  d f  r  om  Estimating QALYs The effectiveness parameters of the model were obtainedfrom published literature, manufacturers’ parameter esti-mates, evidence synthesis and a structured elicitation of expert opinion. Clinical effectiveness  A Bayesian mixed treatment comparison [14] was usedto generate the following parameters for the decisionmodel (for each biologic and for palliative care): the prob-ability of PsARC response at 12 weeks, probability of achieving PASI 50, 75 and 90 response at 12 weeks andthe associated HAQ for PsARC responder/non responder(see Rodgers  et al  . [7] for further detail). Data were usedfrom six RCTs comparing one of the three biologics toplacebo ([15  21], IMPACT [randomized period] [22  33],IMPACT 2 [randomized] [34  41], ADEPT [42  50],Genovese 2007 [51]). The analysis was undertaken usingWinBUGS version 1.4.2 [52]. The probabilities of PsARCand PASI 75 responses may not be independent, and themodel incorporates an estimate of this correlation basedon data provided to the authors by one of the manufac-turers (Abbott Laboratories) [42]. Impact of withdrawals from biologics For patients who respond to biologics according toPsARC, there is an ongoing risk of withdrawal from bio-logic therapy at some point in the future (post 12-weekassessment period). The rate of withdrawal is esti-mated from a meta-analysis of registry data from severalcountries [7] and is assumed to be constant after the first12 weeks. As these data were not randomized, the choiceof therapy may be subject to selection bias, and thereforethese data were not considered reliable enough to esti-mate whether withdrawal rates differed betweenbiologics. Natural history  HAQ progression while on palliative care (natural history)was estimated from patients described in the Norfolk Arthritis Register (NOAR) who, as far as was possible todetermine, matched the licensing indications for a biologicdrug, but were not using one [7]. HAQ was estimated toincrease (worsen) over time, with a 12-week HAQ increaseof 0.018. Utilities HAQ and PASI scores are mapped onto utilities from dataprovided by one of the manufacturers (Wyeth), who car-ried out linear regressions of the EQ-5D utility (EuroQolGroup)  vs  HAQ and PASI in participants in key RCTs. Aninteraction term (HAQ*PASI) was also included in the re-gression, but was dropped from the final analysis becausethe coefficient was small and insignificant [7]. The basecase utility function was: Expected utility ¼ 0 : 897  0 : 298  HAQ  0 : 004  PASI ð SE Þ ð 0 : 006 Þ ð 0 : 006 Þ ð 0 : 0003 Þ Table 1 shows a summary of parameter values for themodel, including their sources. For further details, seeRodgers  et al  . [7]. T  ABLE  1  Model data Description Mean SE Distribution Source Change in cost for 1 U change in HAQ 103 67 Normal Kobelt  et al.  [54]Three-month cost for mild-to-moderatepsoriasis if uncontrolled by biologics198 9 Normal Ref costs [56]Three-month cost for psoriasis in remission 16 1 Normal Hartman  et al.  [58]Change in HAQ while on treatment per 3-monthperiod0 0.02 Normal Expert elicitationChange in HAQ while not on treatment per3-month period0.018 0.007 Gamma NOAR dataLog withdrawal rate from biologics per year   1.823 0.2044 Normal RegistersProbability of PsARC response on placebo 0.249 0.0384 Beta EvidencesynthesisChange in HAQ given a PsARC response onplacebo  0.2436 0.04746 NormalProbability of PASI 75 response on placebo 0.044 0.009 Beta Etan Inflix Adal Cost of drugs (first 3 months) 2495 5523 2495 Normal BSR guidelinesand BNFCost of drugs for Months 4  6 2443 2965 2443 NormalCost of drugs, subsequent 3 months 2385 2965 2385 NormalProbability of PsARC response on biologic 0.713 0.795 0.587 Beta EvidencesynthesisChange in HAQ in first 3 months given noPsARC response on biologic  0.190   0.194   0.130 NormalChange in HAQ in first 3 months givenPsARC response on biologic  0.630   0.657   0.477 NormalProbability of PASI 75 response on biologic 0.1768 0.7687 0.4772 Beta iv42 Laura Bojke  et al.   a  t   J  BM or r  e l  l  L i   b r  a r  y , Uni   v e r  s i   t   y of  Y or k  on S  e  p t   e m b  e r 1 2  ,2  0 1  3 h  t   t   p :  /   /  r h  e  um a  t   ol   o g y . oxf   or  d  j   o ur n a l   s  . or  g /  D o wnl   o a  d  e  d f  r  om  Estimating costs Drug acquisition costs and costs of administration andmonitoring were obtained from BSR recommendations andpharmaceutical list prices [53]. The health service costs of treating arthritis were measured from a UK-based studythat estimated the effect of HAQ on costs in patientswith RA [54, 55]. The NHS costs used for treating mildto moderate psoriasis in patients who do not use biologicsor who do not respond to biologics were obtained fromNHS unit costs of phototherapy [56] and a UK RCT [57].  Analytical methods  Any uncertain parameters in the model were representedby probability distributions. For probabilities (e.g. PsARCresponse), a beta distribution was assigned. For continu-ous measures such as the rate of change of HAQ withoutbiologic therapy, a gamma distribution was assigned. Thisensured that all values were strictly positive. All otheruncertain parameters were assigned normal distributions,with the mean and  S . E . shown in Table 1. Probabilisticsensitivity analysis was carried out using Monte Carlosimulation.The results of the model are presented by first reportingmean lifetime costs and QALYs for the four strategies andthen by comparing their cost-effectiveness, estimatingincremental cost-effectiveness ratios (ICERs) using stand-ard decision rules [59]. Although NICE does not specify aparticular cost-effectiveness threshold, a strategy is morelikely to be considered cost effective if the ICER is < £20000 per QALY, and less likely to be consideredcost-effective if the ICER is  > £30000 per QALY [10].The decision uncertainty is shown as the probability thateach intervention is the most cost effective for a givencost-effectiveness threshold. Results Base case results The expected lifetime costs without biologics (palliativecare only) are around £42000 in the base case for a pa-tient with PsA and mild to moderate psoriasis (Table 2).Total costs for all of the biologics are higher than this. Thelifetime discounted acquisition, administration and moni-toring cost of infliximab is around £52000, of etanerceptis about £33000 and of adalimumab is about £27000.These prescribing costs are much greater than any offsethealth care cost savings arising from improvements inarthritis and psoriasis.Expected QALYs are low in this model. The total life-time discounted health associated with palliative care is  5.24 QALYs (Table 2). This is because utility declinesfairly rapidly in patients with uncontrolled arthritis, and ispredicted to be  < 0 in later years. In the base case, utilitygains as a result of improvement in arthritis are predictedto be much greater than utility gains as a result of improve-ment in the psoriasis component of PsA. This situation arisesbecause of the relative importance of improvements inarthritis and psoriasis on health related quality of life(HRQoL) estimated by the utility function. For example, a75% improvement in PASI implies a reduction in PASIfrom 7.5 to 1.9, a 5.6-point reduction, resulting ina change in HRQoL (measured by EQ-5D) of 5.6  0.004=0.02. Etanercept is expected to reduce HAQby 0.63 points (Table 1), which would improve HRQoLby 0.63  0.298=0.188, showing that expected gainsfrom biologic therapy in arthritis have a much greaterimpact on HRQoL than psoriasis.The base case analysis (Table 2) suggests that inflixi-mab is the most effective treatment (in terms of expectedQALYs), followed by etanercept, then adalimumab.Infliximab is also the most costly treatment, followed byetanercept, then adalimumab. The results (Table 2) showthat for patients with PsA and mild to moderate skin dis-ease, etanercept is likely to be the most cost-effectiveoption, with an ICER  vs  palliative care of around £16000per QALY. Etanercept also has the highest probability of being cost effective at a threshold between £20000 and£30000 per QALY. The ICER of infliximab  vs  etanercept isaround £54000 per QALY, which is unlikely to be accept-able at conventional thresholds for the cost per QALY. Inthe base case, adalimumab is extendedly dominated.Therefore it would not be cost effective, on average, torecommend adalimumab because a greater QALY gaincan be achieved from etanercept within the threshold of £20000 per QALY. Subgroup analysis Subgroup analysis was conducted to explore the impactof differing severities of psoriasis on cost-effectiveness.For patients with PsA and moderate to severe skin dis-ease, the ICER of adalimumab  vs  palliative care is around£15000perQALY.TheICERofetanercept vs adalimumabis T  ABLE  2  Base case results Strategy QALYs Costs (£) ICER (£) P(c/e)_20k P(c/e)_30k  Palliative care 5.241 42205    0.414 0.282 Adalimumab 6.642 66408 Extendedlydominated0.044 0.020Etanercept 7.115 72178 15986 a 0.524 0.566Infliximab 7.430 89107 53750 b 0.018 0.132 a Compared with the next best strategy (excluding extendedly dominated option) palliative care.  b Compared with the next beststrategy (excluding extendedly dominated option) etanercept.  iv43 Cost-effectiveness of biologic treatments   a  t   J  BM or r  e l  l  L i   b r  a r  y , Uni   v e r  s i   t   y of  Y or k  on S  e  p t   e m b  e r 1 2  ,2  0 1  3 h  t   t   p :  /   /  r h  e  um a  t   ol   o g y . oxf   or  d  j   o ur n a l   s  . or  g /  D o wnl   o a  d  e  d f  r  om
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks