UBC Faculty Research and Publications

The impact of incorporating Bayesian network meta-analysis in cost-effectiveness analysis - a case study… Thorlund, Kristian; Zafari, Zafar; Druyts, Eric; Mills, Edward J; Sadatsafavi, Mohsen Mar 13, 2014

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-12962_2013_Article_209.pdf [ 681.78kB ]
JSON: 52383-1.0223446.json
JSON-LD: 52383-1.0223446-ld.json
RDF/XML (Pretty): 52383-1.0223446-rdf.xml
RDF/JSON: 52383-1.0223446-rdf.json
Turtle: 52383-1.0223446-turtle.txt
N-Triples: 52383-1.0223446-rdf-ntriples.txt
Original Record: 52383-1.0223446-source.json
Full Text

Full Text

METHODOLOGY Open AccessThe impact of incorporating Bayesian networkmeta-analysis in cost-effectiveness analysis - acase study of pharmacotherapies for moderateimportant role in providing estimates of comparative efficacy. Their use in the CEAs therefore results in methodologicalThorlund et al. Cost Effectiveness and Resource Allocation 2014, 12:8http://www.resource-allocation.com/content/12/1/8Hamilton, ON, CanadaFull list of author information is available at the end of the articleconsistency and reduced uncertainty.Keywords: Meta-analysis, Multiple treatment comparison, Bayesian analysis, Cost-effectiveness* Correspondence: thorluk@mcmaster.ca1Stanford Prevention Research Center, Stanford University, Stanford, CA, USA2Department of Clinical Epidemiology and Biostatistics, McMaster University,many jurisdictions requires estimates of comparative effto severe COPDKristian Thorlund1,2*, Zafar Zafari3, Eric Druyts4,5, Edward J Mills1,5 and Mohsen Sadatsafavi6AbstractObjective: To evaluate the impact of using network meta-analysis (NMA) versus pair wise meta-analyses (PMA) forevidence synthesis on key outputs of cost-effectiveness analysis (CEA).Methods: We conducted Bayesian NMA of randomized clinical trials providing head-to-head and placebo comparisonsof the effect of pharmacotherapies on the exacerbation rate in chronic obstructive pulmonary disease (COPD). Separately,the subset of placebo–comparison trials was used in a Bayesian PMA. The pooled rate ratios (RR) were used to populate adecision-analytic model of COPD treatment to predict 10-year outcomes.Results: Efficacy estimates from the NMA and PMA were similar, but the NMA provided estimates with higher precision.This resulted in similar incremental cost-effectiveness ratios (ICER). Probabilities of being cost-effective at willingness-to-paythresholds (WTPs) between $25,000 and $100,000 per quality adjusted life year (QALY) varied considerably between thePMA- and NMA-based approaches. The largest difference in the probabilities of being cost-effective was observed at aWTP of approximately $40,000/QALY. At this threshold, with the PMA-based analysis, ICS, LAMA and placebo had a 43%,30, and 18% probability of being the most cost-effective. By contrast, with the NMA based approach, ICS, LAMA, andplacebo had a 56%, 19%, and 21% probability of being cost-effective. For larger WTP thresholds the probability of LAMAbeing the most cost-effective became higher than that of ICS. Under the PMA-based analyses the cross-over occurred ata WTP threshold between $60,000/QALY-$65,000/QALY, whereas under the NMA-based approach, the cross-over occurredbetween $85,000/QALY-$90,000/QALY.Conclusion: Use of NMAs in CEAs is feasible and, as our case study showed, can decrease uncertainty aroundkey cost-effectiveness measures compared with the use of PMAs. The approval process of health technologies inicacy and cost-effectiveness. NMAs play an increasingly© 2014 Thorlund et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the CreativeCommons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, andreproduction in any medium, provided the original work is properly credited.a chronic disease of the airways that is responsible for ain patients with COPD [16]. In particular, five interven-Thorlund et al. Cost Effectiveness and Resource Allocation 2014, 12:8 Page 2 of 9http://www.resource-allocation.com/content/12/1/8IntroductionNetwork meta-analysis (NMA) (also known as multipleor mixed treatment comparisons) are becoming widelyaccepted for establishing comparative efficacy betweencompeting health technologies [1-4]. In contrast withconventional pair wise meta-analysis (PMA), NMAsallow for comparisons between interventions that havenot been compared head-to-head in randomized clin-ical trials (RCTs), and offer additional precision by‘borrowing strength’ from indirect evidence [1,2,5-7].In medical decision-making, NMAs are commonlyused in health technology assessments produced bygovernment agencies or pharmaceutical companiesin connection with technology approval submissions[8-10]. In this context, NMAs can provide reliable andconsistent evidence on the efficacy and safety of theconsidered interventions. The contemporary technol-ogy approval process in many jurisdictions is informedby evaluating comparative efficacy as well as cost-effectiveness analysis (CEA) comparing the new technol-ogy with the alternative choices. NMAs are increasinglypopular frameworks for synthesizing evidence on com-parative efficacy [2,3]. Despite the merits of NMAs, it isstill common that evidence synthesis for the CEA is basedon conventional PMA meta-analysis. While some integra-tion of NMAs and CEAs are beginning to take placein commercially prepared health technology assessment(HTA) reports, we are not aware of any published applica-tions intended to inform decision-making.In addition, it is well accepted that CEAs should becomprehensive [11]. That is, the analysis should includeall available treatment options; and the evidence synthe-sis should be based on all the available evidence [12]. ACEA based on PMA meta-analyses may however fallshort in these two aims. First, evidence on comparativeefficacy and safety may not be available for all treatmentsvia PMA meta-analysis because not all options havebeen compared head-to-head or with a common controlintervention. Second, when more than two options arecompared, the evidence synthesis for a PMA is oftenbased on taking one technology as the ‘reference’ andlooking for comparative studies of other technologieswith that reference. In this vein, head-to-head compari-sons between the considered interventions, as well asrelevant comparisons with older interventions might bediscarded, and so the full evidence-base is not utilized inthe CEA. NMAs on the other hand can produce esti-mates of comparative efficacy for all considered options,and allow for inclusion of all relevant randomized evi-dence (i.e., both direct and indirect evidence). ThereforeNMAs are likely to more optimally and rationally utilizethe available evidence, and the resulting added precisionand accuracy may translate into a more confident adop-tion decision.tions were considered: no treatment (placebo), inhaledcorticosteroids (ICS), long-acting beta-agonists (LABA),long-acting muscarinic agents (LAMA), and the combin-ation of ICS and LABA (ICS + LABA). Several agents areavailable within each of these three drug classes (e.g.,salmeterol, formoterol, and indacaterol are all LABAs)but they were considered equally effective in this ana-lysis. While some may challenge this assumption, thereare a number of reasons for employing this assumptionin our study. First, our study is predominantly of an edu-cational nature, and thus, simplicity in assumptions iskey. Second, the NMA on which this study is based alsoassumed class-effects [16]. Third, other NMA that havedistinguished between therapies within classes havesubstantial economic and humanistic burden [13]. Exac-erbations (lung attacks) are hallmarks of COPD, and areassociated with significant costs, impaired quality of life,and risk of mortality [14]. There are multiple pharmaco-therapies available for COPD and there is considerabledebate on which pharmacotherapy should be used asfirst line treatment in COPD [15]. There is inconsistentevidence as to whether pharmacotherapies can changethe course of COPD. Nevertheless, pharmacotherapieshave a proven impact on reducing the exacerbation ratein COPD [16]. There are several RCTs comparing suchtherapies with placebo (i.e., no treatment), as well as alarge number of RCTs providing head-to-head compari-sons between such therapies [16].NMA model and dataEfficacy data was taken from a recent NMA on the effectof pharmacotherapies in reducing the exacerbation ratesThe use of PMAs rather than NMAs for evidence syn-thesis in economic evaluations therefore represents amissed opportunity for optimizing decision-making [5].To provide insights on the benefit of using NMAs, ra-ther than PMAs, in CEAs we use an illustrative case ofpharmacotherapies for chronic obstructive pulmonarydisease (COPD). We demonstrate how the precisiongained on efficacy estimated via the NMA, as opposed toPMA, can reduce the uncertainty around CEA outputs andcan result in more confident adoption decisions. We alsoprovide practical guidance on the step-wise processesneeded to incorporate the NMA analysis into the CEAprocess.Methods and materialWe use a motivating example of pharmacotherapies forthe treatment of moderate to severe of COPD. COPD isfailed to demonstrate statistically significant differenceswithin classes [17]. Lastly, the assumption of ‘class effect’Abbreviations: ICS (Inhaled Corticosteroids); LABA (Long-acting Beta Agonists); LAMA (Long-acting Muscarinic Agents). A BFigure 1 Treatment networks constituting the evidence-base used in the PMA (left) and NMA (right) analyses.Thorlund et al. Cost Effectiveness and Resource Allocation 2014, 12:8 Page 3 of 9http://www.resource-allocation.com/content/12/1/8for medications within the same class has long been anaccepted paradigm in COPD [18].Details of the NMA are provided elsewhere [16]. Theoutcome (effect) of interest in synthesizing such evi-dence was the impact of the intervention on the yearlyrate of COPD exacerbation. A total of 19 trials (14 two-arm trials, 1 three-arm trial, and 4 four-arm trials)including a total of 28,172 patients informed theevidence-base. Most interventions had been comparedhead-to-head in at least one RCT. The effect measure ofthe NMA was the rate ratio (RR) comparing each treat-ment versus no treatment (i.e., placebo) for yearly inci-dence rates of exacerbations (an RR less than one meansthe treatment reduced the exacerbation rate, comparedwith no treatment). One-year RR estimates were ob-tained using a Bayesian Poisson regression NMA model[10]. Separately, Bayesian Poisson regression PMAs wereused to obtain conventional pair wise RRs for each ofthe considered interventions versus no treatment, fromFigure 2 Markov model used for the performed cost-effectiveness anthe placebo-based RCTs. Figure 1(A) presents the treat-ment network of available comparisons, and Figure 1(B)presents the full treatment network.Economic model and dataA decision-analytic model of COPD was created thattranslated the measures of treatment effect [16], com-bined with parameters representing the epidemiology[13,19] and natural history [20,21] of COPD, into thecosts [22,23], exacerbation rates and quality-adjusted lifeyears (QALYs) associated with each treatment [16,20,21].The time-horizon was 10 years with one-year time cy-cles. A constant yearly rate of exacerbations was as-sumed, thus allowing for the NMA RR estimate to beemployed for determining transition probabilities foreach of the ten cycles. Yearly mortality rates were takenfrom American life Tables [24]. The yearly discount ratewas set to 3% for both health and cost outcomes. Theanalysis adopted a third-party payer perspective. Allalysis.Table 1 Parameter estimates and their probability distributioParameter Assumed input value at GIIAnnual COPD mortality [20] 0.00393 0Utility and disutilities [19]Baseline 0.72Minor exacerbation 0.658Major exacerbation 0.447Exacerbations rates and probabilities [19]Frequency 1.22Minor (%) 0.93Thorlund et al. Cost Effectiveness and Resource Allocation 2014, 12:8 Page 4 of 9http://www.resource-allocation.com/content/12/1/8Major (%) 0.07Minor exacerbation 80$Major exacerbation 3250$Indirect maintenance cost [20] 215$Direct exacerbations costs ($) [19]Minor exacerbations 161$Major exacerbations 6501$General practitioner visit 70$costs were converted and presented as annual costs inyear 2011 US dollars ($).Figure 2 demonstrates the structure of the model. Inmodeling the natural history of patients with moderate tosevere COPD, we used the Global Burden of Lung Disease(GOLD) criteria to classify COPD into mild, moderate, andsevere. However, as the RCTs informing the evidence baseevaluated the impact of treatments in patients with moder-ate/severe COPD, we excluded the state of mild COPD. InSpecialist visit 90$Direct medication costs ($) [22]Inhaled corticosteroids (ICS) 450$Long-acting beta-agonists (LABA) 500$ICS + LABA 1000$Long-acting muscarinic agents 750$Table 2 Incidence rate ratio estimates for the consideredinterventions based on pair-wise meta-analysis (PMA)and network meta-analysis (NMA)Intervention Rate ratios (95% CrI)PMA NMAPlaceboICS 0.81 (0.68-0.95) 0.81 (0.72-0.91)LABA 0.87 (0.75-1.01) 0.87 (0.78-0.96)ICS + LABA 0.71 (0.60-0.88) 0.70 (0.62-0.79)LAMA 0.73 (0.59-0.91) 0.74 (0.67-0.82)PTC: pairwise treatment comparison, MTC: multiple treatment comparison, CrI:credible interval.ns used to populate the modelOLD stages Assumed probability distribution at GOLD stagesIII II III.006762 – –0.67 B(160, 62) B(59, 29)0.475 B(164, 85) B(47, 52)0.408 B(22, 27) B(39, 57)1.47 Γ(14884, 12200) Γ(21609, 14700)0.90 Γ(8649, 9300) Γ(8100, 9000)0.10 Γ(12.25, 175) Γ(25, 250)134$ Γ(320, 4) Γ(536, 4)5417$ Γ(13000, 4) Γ(21668, 4)524$ Γ(860, 4) Γ(2096, 4)Γ(644, 4)Γ(26004, 4)Γ(280, 4)addition to COPD states, individuals in the model couldalso independently move through the states representingbeing a current smoker, ex-smoker, and never-smoker. Indi-viduals could not revert from a worse COPD state to a bet-ter COPD state.Each state of COPD was associated with an annual ex-acerbation rate for each treatment, which was calculated asthe product of a baseline (no treatment) rate multiplied bythe RR of the treatment versus no treatment. Exacerbationswere categorized as either minor or major. The impact oftreatment was assumed to be independent of the severity ofthe exacerbation.Table 1 provides the parameter estimates and theirprobability distributions used to populate the model.Estimates in original reports for the majority of the pa-rameters were accompanied by confidence intervals orstandard errors. As such, each parameter was modeledas a probability distribution to match the reportedlevel of uncertainty. On the other hand, cost compo-nents often were not accompanied by uncertainty, andwe a priori decided to model costs to have a gammadistribution with a coefficient of variation of 0.25. Costof medications were assumed fixed at their knownvalue in 2013.Γ(360, 4)––––AnalysisThe Bayesian NMA model was run in WinBUGS v.1.4.3[25], and the economic model was run in R v2.14 [26].WinBUGS and R code is available from the authors uponrequest. The step-wise implementation of the PMA andNMA analyses and the CEA is described further in theAdditional file 1. A total of 10,000 posterior distributionsamples were used for the CEA, separately for the NMAand PMA meta-analyses. The model outputs on costs andQALYs were used to calculate the ICERs and incrementalnet monetary benefits (INMB), with no treatment asthe reference group, and to draw the cost-effectivenessplanes and cost-effectiveness acceptability curves (CEACs).thresholds between $0/QALY and $100,000/QALY. TheICERs from the PMA- and NMA-based analyses weresimilar, but the probabilities of being cost-effective at theexplored WTP thresholds varied considerably. Thelargest difference in the probabilities of being cost-effective was observed at a WTP of approximately$40,000/QALY. At this threshold, with the PMA-basedanalysis, ICS, LAMA and placebo had a 43%, 30%, and18% probability of being the most cost-effective. By con-trast, with the NMA based approach, ICS, LAMA, andplacebo had a 56%, 19%, and 21% probability of beingcost-effective. As illustrated in both Table 4 and Figure 4,the differences between the two approaches were alsod qysiThorlund et al. Cost Effectiveness and Resource Allocation 2014, 12:8 Page 5 of 9http://www.resource-allocation.com/content/12/1/8Treatments were also ranked according to their INMB atWTP of $50,000/QALY, separately for PMA- and NMA-based analyses.ResultsTable 2 presents the RRs and the associated credible in-tervals (CrI) for all treatment vs no treatment compari-sons based on the NMA and PMA meta-analyses. Thepooled RR estimates for all treatment vs no treatmentcomparisons were similar for the NMA and PMA meta-analyses, but the NMA results had higher precision,manifested in terms of tighter CrIs (Table 2).Table 3 presents the mean and 95% CrIs for costs, ex-acerbation rates, and QALYs. Figure 3 presents the un-certainty ellipses around the incremental cost and QALYestimates on the cost-effectiveness plane. Uncertaintyaround both costs and QALYs was reduced substantiallyin the NMA-based analysis. This reduction is visuallyapparent from the considerably smaller 95% credibleellipses NMA-based analysis compared with the PMA-based analysis in Figure 3.Table 4 presents the ICERs and probabilities of eachtreatment being cost-effective as WTP thresholds of$30,000, $50,000, $70,000, and $100,000. Figure 4presents the CEACs for all interventions from WTPTable 3 10-year average cost, number of exacerbations, anboth pairwise meta-analysis (PMA) and network meta-analIntervention Meta-analysis Costs ($)Placebo PMA 25 458 (19 927, 32 312)NMA 25 316 (19 950, 31 897)ICS PMA 27 116 (22 194, 33 307)NMA 26 979 (21 991, 33 420)LABA PMA 28 304 (23 081, 34 897)NMA 28 116 (23 002, 34 604)ICS + LABA PMA 30 849 (25 947, 36 741)NMA 30 496 (25 902, 36 133)LAMA PMA 28 840 (23 928, 35 073)NMA 28 816 (24 237, 34 548)notable for all WTP thresholds above approximately$25,000. In both analyses, LAMA were estimated morelikely to be cost-effective than ICS for high WTP thresh-old, but the point where these probabilities crossed weredifferent between the PMA- and NMA-based analyses.In particular, with the PMA-based approach the point ofprobabilities crossing was between $60,000/QALY and$65,000/QALY, whereas the point of crossing with theNMA-based approach was between $85,000/QALY and$90,000/QALY.At WTP of $50,000/QALY, the ranking of the first threetreatments (ICS, LAMA, and no treatment) remained thesame between PMA- and NMA-based analyses. The onlydifference in the results was that the treatment with thelowest INMB for the PMA-based analysis was ICS + LABAwhereas for the NMA-based analysis it was LABA.DiscussionIn the present work we elaborated on the theoretical ad-vantages of using NMAs over PMAs in economic evalu-ations of health technologies, and used a case study todemonstrate the practical aspects of the use of NMAs aswell as the empirical differences in the outcomes of theeconomic evaluation when NMA instead of PMA is usedfor evidence synthesis. The results demonstrate how theuality adjusted life-years for each intervention usings (NMA)Number of exacerbations Quality adjusted life years (QALYs)12.6 (12.5, 12.8) 5.67 (5.36, 5.99)12.6 (12.5, 12.7) 5.67 (5.34, 5.99)10.3 (8.62, 11.9) 5.73 (5.40, 6.07)10.2 (9.10, 11.4) 5.73 (5.38, 6.06)11.0 (9.31, 13.2) 5.71 (5.39, 6.04)10.9 (9.86, 12.2) 5.71 (5.37, 6.05)9.13 (7.69, 11.2) 5.76 (5.42, 6.10)8.85 (7.83, 10.0) 5.75 (5.40, 6.11)9.33 (7.47, 11.4) 5.76 (5.42, 6.10)9.39 (8.37, 10.4) 5.77 (5.41, 6.12)Thorlund et al. Cost Effectiveness and Resource Allocation 2014, 12:8 Page 6 of 9http://www.resource-allocation.com/content/12/1/8ACEA can benefit from the gain in precision from usingthe entire network of evidence rather than the results ofpair wise comparisons alone. In our case study, whilethe added precision did not result in major changes inthe choice of the optimal treatment across a wide rangeBAbbreviations: QALY (Quality Adjusted LifLABA (Long-acting Beta Agonists); LAMAFigure 3 Cost-effectiveness plane illustrating the 95% credible ellipsemeta-analysis (A) and the network meta-analysis (B).of WTP, it prevented the counter-intuitive situation ofthe optimal treatment not having the maximum prob-ability of cost-effectiveness [27].The network of evidence underlying the case studywas a well-connected treatment network including largee years; ICS (Inhaled Corticosteroids);  (Long-acting Muscarinic Agents).  s for each intervention versus placebo based on the pair wiseababis rrefThorlund et al. Cost Effectiveness and Resource Allocation 2014, 12:8 Page 7 of 9http://www.resource-allocation.com/content/12/1/8Table 4 Incremental cost-effectiveness ratios (ICERs) and probvarious willingness-to-pay thresholdsIntervention Meta-analysisICER Prob$30,000/QALYPlacebo aPlacebo PTC Reference 35%MTC Reference 37%ICS PTC 27044 40%MTC 27614 49%LABA PTC 65509 1%MTC 64339 1%ICS + LABA PTC 57933 1%MTC 52116 0%LAMA PTC 38427 21%MTC 41203 13%ICS asLABA PTC Dominated –MTC Dominated –ICS + LABA PTC 89843 –MTC 96749 –LAMA PTC 25930 –MTC 57854 –studies and head-to-head RCTs for almost all compari-sons. As such, the use of the entire available evidencebase synthesized through NMA resulted in similar pointestimate for the effect size but with an increased preci-sion. However, situations may occur where NMA esti-mates are not close to their PMA counterparts; andwhere the combination of indirect and direct evidencedoes little to increase the precision [7]. However, thetheoretical justifications underpinning the use of NMAinstead of PMA are unrelated to the empirical gains incertainty and stand valid regardless of any particularresults.The performed analyses come with some limitations.We used a simple decision-analytic model of COPD forthe case study, mainly based on the modeling assump-tions used by previous authors [20,21]. The simplicity ofthis model allowed us to focus on the practical aspectsand illustration of the results; but we acknowledge thatto inform policy, a deeper analysis, including a detailedset of sensitivity and alternative analyses will be required.For example, our model did not account for the poten-tial impact of treatments on disease progression [20], acontroversial aspect of the treatment that needs to beconsidered in a sensitivity analysis. Our model also didnot account for potential long-term adverse events asso-ciated with corticosteroid treatment and their associatedcosts. However, the complexity of building a decision-ilities of each intervention being the most cost effective atlity of being cost-effective by willingness-to-pay threshold$50,000/QALY $70,000/QALY $100,000/QALYeference10% 3% 0%12% 3% 2%42% 38% 31%55% 44% 31%6% 6% 5%2% 1% 0%6% 11% 18%7% 1% 29%34% 43% 46%24% 35% 38%erence– – –– – –– – –– – –– – –– – –model is not intensified by the use of NMA versus PMAfor evidence synthesis.The implications of the results are rather straightfor-ward: the potential theoretical and practical gains inusing NMAs as opposed to PMAs in cost-effectivenessanalysis are too significant to be ignored. However, thisdoes not mean that CEAs should only ever rely onefficacy estimates from NMAs. NMA is a method of in-ference and as such is based on certain statistical as-sumptions that are generally more restrictive than theassumptions underlying PMA [2,3]. For example, thereare situations where NMA estimates may be morebiased than their PMA counterparts estimated only fromplacebo comparisons [28,29]. If in a particular contextwhere there are misgivings about the suitability of suchassumptions, the investigator might deliberately choosePMA. Overall, a thorough assessment of the potentialbiases and confounders in both the NMA and the PMAis necessary before deciding which data and type of re-search synthesis method should be used for informingthe cost-effectiveness analysis.ConclusionIn summary, incorporating NMA in CEA offers con-sistency and added certainty in comparison with CEAinformed by conventional PMA. As the role of NMAs ininforming comparative efficacy in the evaluation of newAbbreviations: WTP (Willingness-to-pay); ICS (Inhaled Corticosteroids); LABA (Long-acting Beta Agonists); LAMA (Long-acting Muscarinic Agents). ABFigure 4 Cost-effectiveness acceptability curves when efficacy results are based on the pair wise meta-analysis (A) and the networkmeta-analysis (B).Thorlund et al. Cost Effectiveness and Resource Allocation 2014, 12:8 Page 8 of 9http://www.resource-allocation.com/content/12/1/8report of the ISPOR Task Force on Indirect Treatment Comparisons GoodWelton NJ, Abrams KR, Bujkiewicz S, Spiegelhalter D, Sutton AJ: HowThorlund et al. Cost Effectiveness and Resource Allocation 2014, 12:8 Page 9 of 9http://www.resource-allocation.com/content/12/1/8valuable are multiple treatment comparison methods in evidence-basedhealth-care evaluation? Value Health 2011, 14(2):371–80.6. Higgins JP, Whitehead A: Borrowing strength from external trials in aResearch Practices: part 1. Value Health 2011, 14(4):417–28.4. Mills EJ, Bansback N, Ghement I, Thorlund K, Kelly S, Puhan MA, Wright J:Multiple treatment comparison meta-analyses: a step forward intocomplexity. Clin Epidemiol 2011, 3:193–202.5. Cooper NJ, Peters J, Lai MC, Juni P, Wandel S, Palmer S, Paulden M, Conti S,health technologies is growing, NMAs could and shouldbe considered for informing the evidence used in CEA.Additional fileAdditional file 1: Implemention of WinBUGS and R.Competing interestKristian Thorlund and Edward Mills are founding partners of RedwoodOutcomes Inc. Redwood Outcomes consults to a number pharmaceuticalcompanies, of which several are manufacturers of at least one brandbelonging to the classes of COPD drugs considered for this article.Authors’ contributionsKT conceived the idea of the study, contributed to the design of the study,wrote up the first manuscript, conducted the pair wise meta-analysis andnetwork meta-analysis, and contributed to the interpretation of findings. ZZcontributed to the design of the study, programmed the Markov model andran all cost-effectiveness analyses, contributed to the writing of the manuscriptand contributed to the interpretation of findings. ED contributed to the design ofthe study, contributed to the identification of literature to inform model parametervalues, contributed to the writing of the manuscript and contributed to theinterpretation of findings. EM contributed to the design of the study, the writingof the manuscript, and the interpretation of findings. MS contributed to thedesign of the study, contributed to the identification of literature to inform modelparameter values, supervised the development of the cost-effectiveness Markovmodel, contributed to the writing of the manuscript and contributed to theinterpretation of findings. All authors read and approve the final manuscript.Author details1Stanford Prevention Research Center, Stanford University, Stanford, CA, USA.2Department of Clinical Epidemiology and Biostatistics, McMaster University,Hamilton, ON, Canada. 3Collaboration for Outcome Research and Evaluations,Faculty of Pharmaceutical Sciences, University of British Columbia, Vancouver,Canada. 4Department of Experimental Medicine, University of BritishColumbia, Vancouver, BC, Canada. 5Faculty of Medicine, University of Ottawa,Ottawa, ON, Canada. 6Faculty of Medicine, University of British Columbia,Vancouver, BC, Canada.Received: 13 July 2013 Accepted: 26 February 2014Published: 13 March 2014References1. . Hoaglin DC, Hawkins N, Jansen JP, Scott DA, Itzler R, Cappelleri JC,Boersma C, Thompson D, Larholt KM, Diaz M, Barrett A: Conductingindirect-treatment-comparison and network-meta-analysis studies: reportof the ISPOR Task Force on Indirect Treatment Comparisons Good ResearchPractices: part 2. Value Health 2011, 14(4):429–37.2. Mills EJ, Ioannidis JP, Thorlund K, Schunemann HJ, Puhan MA, Guyatt GH:How to use an article reporting a multiple treatment comparisonmeta-analysis. JAMA 2012, 308(12):1246–53.3. Jansen JP, Fleurence R, Devine B, Itzler R, Barrett A, Hawkins N, Lee K,Boersma C, Annemans L, Cappelleri J: Interpreting indirect treatmentcomparisons and network meta-analysis for health-care decision making:meta-analysis. Stat Med 1996, 15(24):2733–49.7. Thorlund K, Mills EJ: Sample size and power considerations in networkmeta-analysis. Syst Rev 2012, 1(1):41.8. Sutton AJ, Ades AE, Cooper NJ, Abrams KR: Use of indirect and mixedtreatment comparisons for technology assessment. Pharmacoecon 2008,26(9):15.9. Wells GA, Sultan SA, Chen L, Khan M, Coyle D: Indirect evidence: indirecttreatment comparison in meta-analysis. Ottawaa: Canadian Agencies forDrugs and Technology in Health (CADTH); 2009.10. Dias S, Welton N, Sutton A, Ades A: NICE DSU Technical Support Document2 - A generalised linear modelling framework fro pairwise and network meta-analysisof randomised controlled trial. 2011.11. Cooper NJ, Sutton A, Abrams KR, Turner D, Wailoo A: Comprehensivedecision analytic modelling in economic evaluation: a Bayesianapproach. Health Econ 2003, 13(3):24.12. Sculpher MJ, Claxton K, Drummond M, McCabe C: Whither trial-basedeconomic evaluation for health care decision making? Health Econ 2006,15(7):677–87.13. Mannino DM, Duist AS: Global burden of COPD: risk factors, prevalence,and future trends. Lancet 2007, 370(9589):9.14. Connors AF Jr, Dawson NV, Thomas C, Harrell FE Jr, Desbiens N, Fulkerson WJ,Kussin P, Bellamy P, Goldman L, Knaus WA: Outcomes following acuteexacerbation of severe chronic obstructive lung disease. The SUPPORT Study.Respir and Crit Care Med 1996, 154(4):9.15. Celli BR, MacNee W; ATS/ERS Task Force: Standards for the diagnosis andtreatment of patients with COPD: a summary of the ATS/ERS positionpaper. Eur Respir J 2004, 23(6):932–946.16. Mills EJ, Druyts E, Ghement I, Puhan MA: Pharmacotherapies for chronicobstructive pulmonary disease: a multiple treatment comparison meta-analysis. Clin Epidemiol 2011, 3:107–29.17. Cope S, Capkun-Niggli G, Gale R, Jardim JR, Jansen JP: Comparative efficacyof indacaterol 150 mug and 300 mug versus fixed-dose combinations offormoterol + budesonide or salmeterol + fluticasone for the treatment ofchronic obstructive pulmonary disease–a network meta-analysi. Int JChron Obstruct Pulmon Dis 2011, 6:329–44.18. McAlister FA, Laupacis A, Wells GA, Sackett DL: Users’ Guide to the MedicalLiterature: XIX. Applying clinical trial results B. Guidelines for determiningwhether a drug is exerting (more than) a class effect. JAMA 1999, 282(14):7.19. Donaldson GC, Wedzicha JA: COPD exacerbations 1: epidemiology.Thorax 2011, 61(2):5.20. Spencer M, Briggs AH, Grossman RF, Rance L: Development of an economicmodel to assess the cost-effectiveness of treatment interventions for chronicobstructive pulmonary disease. Pharmacoeconomics 2005, 23:19.21. Buist AS, Vollmer WM, Sullivan SD, Weiss KB, Lee TA, Menezes AM,Crapo RO, Jensen RL, Burney PG: The Burden of Obstructive Lung DiseaseInitiative (BOLD): rationale and design. COPD 2005, 2:7.22. Dalai AA, Shah M, D’Souza AO, Rana P: Costs of COPD exacerbations inthe emergency department and inpatient setting. Respir Med 2011,105(3):7.23. MIMS: MIMS: prescription drug database and drug prescriptionguide. 2012, 59. [http://http://mims.co.uk/]24. Aries E: United States Life Tables. London, UK: CDC; 2007.25. Lunn D, Spiegelhalter D, Thomas A, Best N: The BUGS project: evolution,critique and future directions. Stat Med 2009, 28(25):3049–67.26. The R, Core T: R: A Language and Environment for Statistical Computing.Vienna, Austria: R Foundation for Statistical Computing; 2005.27. Sadatsafavi M, Najafzadeh M, Marra C: Technical note: acceptability curvescould be misleading when correlated strategies are compared. Med DecisMak 2008, 28(3):2.28. Song F, Harvey I, Lilford R: Adjusted indirect comparison may be lessbiased than direct comparison for evaluating new pharmaceuticalinterventions. J Clin Epidemiol 2008, 61(5):455–63.29. Song F, Clark A, Bachmann MO, Maas J: Simulation evaluation of statisticalproperties of methods for indirect and mixed treatment comparisons.BMC Med Res Methodol 2012, 12:138.doi:10.1186/1478-7547-12-8Cite this article as: Thorlund et al.: The impact of incorporating Bayesiannetwork meta-analysis in cost-effectiveness analysis - a case study ofpharmacotherapies for moderate to severe COPD. Cost Effectiveness andResource Allocation 2014 12:8.


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items