Advertisement

Patient Heterogeneity in Health Economic Decision Models for Chronic Obstructive Pulmonary Disease: Are Current Models Suitable to Evaluate Personalized Medicine?

Open ArchivePublished:May 20, 2016DOI:https://doi.org/10.1016/j.jval.2016.04.002

      Abstract

      Objectives

      To assess how suitable current chronic obstructive pulmonary disease (COPD) cost-effectiveness models are to evaluate personalized treatment options for COPD by exploring the type of heterogeneity included in current models and by validating outcomes for subgroups of patients.

      Methods

      A consortium of COPD modeling groups completed three tasks. First, they reported all patient characteristics included in the model and provided the level of detail in which the input parameters were specified. Second, groups simulated disease progression, mortality, quality-adjusted life-years (QALYs), and costs for hypothetical subgroups of patients that differed in terms of sex, age, smoking status, and lung function (forced expiratory volume in 1 second [FEV1] % predicted). Finally, model outcomes for exacerbations and mortality for subgroups of patients were validated against published subgroup results of two large COPD trials.

      Results

      Nine COPD modeling groups participated. Most models included sex (seven), age (nine), smoking status (six), and FEV1% predicted (nine), mainly to specify disease progression and mortality. Trial results showed higher exacerbation rates for women (found in one model), higher mortality rates for men (two models), lower mortality for younger patients (four models), and higher exacerbation and mortality rates in patients with severe COPD (four models).

      Conclusions

      Most currently available COPD cost-effectiveness models are able to evaluate the cost-effectiveness of personalized treatment on the basis of sex, age, smoking, and FEV1% predicted. Treatment in COPD is, however, more likely to be personalized on the basis of clinical parameters. Two models include several clinical patient characteristics and are therefore most suitable to evaluate personalized treatment, although some important clinical parameters are still missing.

      Keywords

      Introduction

      Personalized medicine has the potential to improve the (cost-)effectiveness of treatments and contribute to health care cost containment. Hence, interest in personalized medicine has increased exponentially in the past decade [
      • O’Donnell J.C.
      Personalized medicine and the role of health economics and outcomes research: issues, applications, emerging trends, and future research.
      ]. The concept of personalized medicine has different definitions [
      • Redekop W.K.
      • Mladsi D.
      The faces of personalized medicine: a framework for understanding its meaning and scope.
      ,

      Personalized Medicine. U.S. Food and Drug Administration. Available from: http://www.fda.gov/scienceresearch/specialtopics/personalizedmedicine/default.htm. [Accessed June 17, 2015].

      ]. Some definitions focus on the use of genetics, proteomic, cytomic, and/or metabolic biomarkers to define patient genotypes and phenotypes most likely to benefit most from a certain treatment. Other definitions are broader and refer to customizing treatment to the individual characteristics, needs, and preferences of a patient during all stages of care, including prevention, diagnosis, treatment, and follow-up [

      Personalized Medicine. U.S. Food and Drug Administration. Available from: http://www.fda.gov/scienceresearch/specialtopics/personalizedmedicine/default.htm. [Accessed June 17, 2015].

      ]. The expectations of personalized medicine are high, but so far, the cost-effectiveness of such technologies on average has not been found to be any better than that of nonpersonalized interventions [
      • Hatz M.H.
      • Schremser K.
      • Rogowski W.H.
      Is individualized medicine more cost-effective? A systematic review.
      ,
      • Phillips K.A.
      • Ann Sakowski J.
      • Trosman J.
      • et al.
      The economic value of personalized medicine tests: what we know and what we need to know.
      ]. Thus, economic evaluation studies are needed to show whether the personalized approach indeed leads to greater efficiency in health care compared with the one-size-fits-all approach.
      If based on health economic models, such economic evaluations of personalized treatment strategies based on demographic and clinical patient characteristics require models that are able to address the heterogeneity due to those patient characteristics appropriately. However, most currently available cost-effectiveness models are cohort Markov or state-transition models that have only limited capability for addressing patient heterogeneity. Most published cost-effectiveness analyses similarly focused on results for the average patient. Current guidelines for health economic modeling recognize the importance of patient heterogeneity, but there is a lack of consensus on how to best address this in a modeling framework [
      • Ramaekers B.L.
      • Joore M.A.
      • Grutters J.P.
      How should we deal with patient heterogeneity in economic evaluation: a systematic review of national pharmacoeconomic guidelines.
      ].
      One of the major chronic diseases for which treatment is increasingly personalized is chronic obstructive pulmonary disease (COPD). In 2011, the new Global Initiative for chronic Obstructive Lung Disease guidelines proposed a new classification system for patients with COPD based on lung function, symptoms, and exacerbations. Although this new classification is a step forward to personalized treatment for COPD, proposed treatment guidelines for the subgroups of patients still need to be validated [

      Global Initiative for Chronic Obstructive Lung Disease. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease. Available from: http://www.goldcopd.com. [Accessed June 17, 2015].

      ]. Manufacturers of drugs and devices are increasingly focused on specific phenotypes of patients, such as patients with frequent exacerbations, a rapid decline in lung function, persistent systemic inflammation, raised eosinophil concentrations, and bacterial colonization [
      • Han M.K.
      • Agusti A.
      • Calverley P.M.
      • et al.
      Chronic obstructive pulmonary disease phenotypes: the future of COPD.
      ,
      • Woodruff P.G.
      • Agusti A.
      • Roche N.
      • et al.
      Current concepts in targeting chronic obstructive pulmonary disease pharmacotherapy: making progress towards personalised management.
      ]. In addition, caregivers providing multidisciplinary integrated care programs develop different treatment modules for subgroups of patients who need physical reactivation, smoking cessation support, nutritional interventions, treatment of depression, and so forth. In contrast, most decisions about the reimbursement of new treatments are still made for large groups of patients and on the basis of evaluations that do not consider patient heterogeneity. This is expected to change rapidly because biologics for COPD are under development [

      The Pharmaceutical Research and Manufacturers of America. Medicines in development for chronic obstructive pulmonary disease (COPD). 2012. Available from: http://www.phrma.org/sites/default/files/pdf/copd2012.pdf. [Accessed June17, 2015].

      ] and payers are likely to limit the reimbursement of these expensive drugs to specified subgroups. Moreover, new drugs for COPD that have recently been launched or are still under development are often quite similar to the currently marketed drugs, which makes it increasingly difficult to demonstrate their value for money unless one can specify the subgroup for which they are particularly beneficial.
      In the past years, several health economic decision models for COPD have been published. In 2011, our modeling group took the initiative to establish a network of people involved in COPD modeling around the world (COPD modeling teams, pharmaceutical companies interested in COPD modeling, epidemiologists, clinicians, etc.). Up to now, we have organized three 1-day meetings in 2011, 2012, and 2014 with the aim to compare the different available models with respect to model structure and input parameters and to cross-validate the models against each other [
      • Hoogendoorn M.
      • Feenstra T.L.
      • Asukai Y.
      • et al.
      Cost-effectiveness models for chronic obstructive pulmonary disease: cross-model comparison of hypothetical treatment scenarios.
      ]. Proceedings of the third meeting regarding patient heterogeneity are described in the present article.
      The aim of this article was to assess how suitable current models are to evaluate the cost-effectiveness of personalized treatment options for COPD. First, we explored which type of patient heterogeneity is included in currently available COPD cost-effectiveness models to see whether the models are able to evaluate subgroups of patients with COPD that are considered clinically relevant. Second, we investigated the impact of patient characteristics on the outcomes. Finally, we validated the outcomes of specific subgroup analyses with the models against subgroup results of clinical trials to assess whether the models are suitable for performing subgroup analyses. These questions are relevant because most of the models are initially built to evaluate treatment options for a large group of patients with COPD.

      Methods

      Procedure

      In March 2014, the COPD modeling groups that participated in previous meetings as well as new groups identified through publications or other participants within the network were contacted to explore their interest in participation in the third COPD modeling meeting and in running the model simulations required for this heterogeneity analysis. Modeling groups were first asked to specify which patient characteristics are currently included in their model and which input parameters are specified by subgroup. Second, they were asked to run their model for hypothetical patients who differed in terms of patient and disease characteristics to explore the impact of these characteristics on the outcomes of the models. Third, they were asked to simulate the outcomes for subgroups of patients that had similar characteristics as the subgroups of two large randomized controlled clinical trials to validate the model outcomes. Results were returned to the organizers of the meeting in a structured format in Microsoft Excel 2 weeks before the meeting. The combined results of the models were circulated to all participants of the COPD modeling meeting 1 week before the meeting to give participants time to reflect on the outcomes. The results were presented during the meeting and explanations for the differences in outcomes between the models were discussed.

      Participating Models

      Nine COPD models participated in the model simulations. Six of these nine models also participated in the previous meeting in 2012 [
      • Price D.
      • Gray A.
      • Gale R.
      • et al.
      Cost-utility analysis of indacaterol in Germany: a once-daily maintenance bronchodilator for patients with COPD.
      ,
      • Borg S.
      • Ericsson A.
      • Wedzicha J.
      • et al.
      A computer simulation model of the natural history and economic impact of chronic obstructive pulmonary disease.
      ,
      • Hansen R.N.
      • Xu X.
      • Sullivan S.D.
      A dynamic cohort model of chronic obstructive pulmonary disease and its treatments.
      ,
      • Hoogendoorn M.
      • Rutten-van Molken M.P.
      • Hoogenveen R.T.
      • et al.
      Developing and applying a stochastic dynamic population model for chronic obstructive pulmonary disease.
      ,
      • Samyshkin Y.
      • Kotchie R.W.
      • Mork A.C.
      • et al.
      Cost-effectiveness of roflumilast as an add-on treatment to long-acting bronchodilators in the treatment of COPD associated with chronic bronchitis in the United Kingdom.
      ,
      • Menn P.
      • Leidl R.
      • Holle R.
      A lifetime Markov model for the economic evaluation of chronic obstructive pulmonary disease.
      ]. A short description of these six models can be found in the publication about the proceedings of this second COPD modeling meeting [
      • Hoogendoorn M.
      • Feenstra T.L.
      • Asukai Y.
      • et al.
      Cost-effectiveness models for chronic obstructive pulmonary disease: cross-model comparison of hypothetical treatment scenarios.
      ]. Three new models also participated. The simulation model of Asukai et al. [
      • Asukai Y.
      • Baldwin M.
      • Fonseca T.
      • et al.
      Improving clinical reality in chronic obstructive pulmonary disease economic modelling: development and validation of a micro-simulation approach.
      ] was published in 2013. The other two models have not yet been published, but have been presented at international conferences. The model of Briggs et al. (GALAXY COPD model) was presented at the ISPOR Annual International meeting 2013 and the ISPOR 17th Annual European Congress 2014 [
      • Briggs A.
      • Lomas D.
      • Rutten van-Molken M.
      • et al.
      Developing a new model of COPD: from conceptualization to implementation to validation.
      ,
      • Risebrough N.A.
      • Briggs A.
      • Baker T.M.
      • et al.
      Validating a model to predict disease progression outcomes in patients with COPD.
      ]. The model of Dal Negro [
      • Dal Negro R.W.
      Current annual cost calculation is the best predictor of mortality at three years in COPD.
      ] was also presented at the ISPOR 17th Annual European Congress 2014 [
      • Dal Negro R.W.
      Current annual cost calculation is the best predictor of mortality at three years in COPD.
      ].

      Content of the Modeling Challenge

      For the first part of the modeling challenge, groups reported all patient characteristics that are currently included in their models. Furthermore, groups provided the level of detail in which the following input parameters had been specified in the models: disease progression, exacerbation frequency, mortality, case-fatality of an exacerbation, utilities during stable disease, utilities during exacerbations, maintenance costs, and exacerbation-related costs.
      In part two, modeling groups simulated the outcomes for hypothetical subgroups of patients that differ in terms of sex, age, smoking status, and level of forced expiratory volume in 1 second (FEV1) % predicted to see how patient heterogeneity affected effects and costs within one model and between models. Sex, age, smoking status, and FEV1% predicted were chosen because these are the factors that are included in most models. In the first simulation, outcomes for a 65-year-old, ex-smoking, male patient with severe COPD were calculated. In each of the following four simulations, one of the patient characteristics was changed: female instead of male patient, 75-year-old instead of 65-year-old patient, smoking instead of ex-smoking patient, and patient with moderate COPD instead of severe COPD. The outcomes of each of these four simulations were compared with the results of the first simulation.
      In part three, the validity of the subgroup analyses with the models was investigated by comparing the model outcomes against published subgroup results of two large randomized controlled clinical trials, the 4-year Understanding Potential Long-Term Impacts on Function with Tiotropium (UPLIFT) trial and the 3-year TOwards a Revolution in COPD Health (TORCH) trial [
      • Tashkin D.P.
      • Celli B.
      • Senn S.
      • et al.
      UPLIFT Study Investigators
      A 4-year trial of tiotropium in chronic obstructive pulmonary disease.
      ,
      • Calverley P.M.
      • Anderson J.A.
      • Celli B.
      • et al.
      TORCH Investigators
      Salmeterol and fluticasone propionate and survival in chronic obstructive pulmonary disease.
      ]. In the UPLIFT trial, almost 6000 patients with moderate-to-very severe COPD were randomized to placebo, defined as all regular respiratory medications except inhaled anticholinergics, or tiotropium 18 µg plus all regular respiratory medications except other inhaled anticholinergics [
      • Tashkin D.P.
      • Celli B.
      • Senn S.
      • et al.
      UPLIFT Study Investigators
      A 4-year trial of tiotropium in chronic obstructive pulmonary disease.
      ]. In the TORCH trial, slightly more than 6000 patients with moderate-to-very severe COPD were randomized to placebo, defined as all COPD medications except long-acting bronchodilators and inhaled corticosteroids, salmeterol 50 µg, fluticasone 500 µg, or the combination of salmeterol 50 µg and fluticasone 500 µg [
      • Calverley P.M.
      • Anderson J.A.
      • Celli B.
      • et al.
      TORCH Investigators
      Salmeterol and fluticasone propionate and survival in chronic obstructive pulmonary disease.
      ]. For both the UPLIFT trial and the TORCH trial, several subgroup analyses have been published mainly by sex, age, smoking status, and level of FEV1% predicted [
      • Jenkins C.R.
      • Jones P.W.
      • Calverley P.M.
      • et al.
      Efficacy of salmeterol/fluticasone propionate by GOLD stage of chronic obstructive pulmonary disease: analysis from the randomised, placebo-controlled TORCH study.
      ,
      • Tashkin D.
      • Celli B.
      • Kesten S.
      • et al.
      Effect of tiotropium in men and women with COPD: results of the 4-year UPLIFT trial.
      ,
      • Morice A.H.
      • Celli B.
      • Kesten S.
      • et al.
      COPD in young patients: a pre-specified analysis of the four-year trial of tiotropium (UPLIFT).
      ,
      • Tashkin D.P.
      • Celli B.
      • Kesten S.
      • et al.
      Long-term efficacy of tiotropium in relation to smoking status in the UPLIFT trial.
      ]. The modeling groups were asked to adjust the starting population of the model to the baseline characteristics of the specific subgroups within the placebo group of the trial in terms of percentage of males, mean age, percentage of current smokers, and mean FEV1% predicted (or distribution over the GOLD severity stages moderate, severe, or very severe COPD). The baseline characteristics of the subgroups in the trial are presented in Table 1. For comparison with the UPLIFT trial placebo arm, groups were asked to run the simulations assuming that patients received all regular respiratory medications except inhaled anticholinergics. Simulation of the TORCH trial placebo arm was done assuming that patients did not use long-acting bronchodilators and inhaled corticosteroids.
      Table 1Baseline characteristics of subgroups within the placebo group of the UPLIFT and TORCH trials used as starting population of the model simulations
      • Jenkins C.R.
      • Jones P.W.
      • Calverley P.M.
      • et al.
      Efficacy of salmeterol/fluticasone propionate by GOLD stage of chronic obstructive pulmonary disease: analysis from the randomised, placebo-controlled TORCH study.
      ,
      • Tashkin D.
      • Celli B.
      • Kesten S.
      • et al.
      Effect of tiotropium in men and women with COPD: results of the 4-year UPLIFT trial.
      ,
      • Morice A.H.
      • Celli B.
      • Kesten S.
      • et al.
      COPD in young patients: a pre-specified analysis of the four-year trial of tiotropium (UPLIFT).
      ,
      • Tashkin D.P.
      • Celli B.
      • Kesten S.
      • et al.
      Long-term efficacy of tiotropium in relation to smoking status in the UPLIFT trial.
      CharacteristicUPLIFT trialTORCH trial
      By sexBy ageBy smoking statusBy COPD severity
      MalesFemalesAge < 50 ySmokersEx-smokersModerate COPDSevere COPD
      N22227841724321791535775
      Sex: males (%)10005862787276
      Age (y), mean ± SD65 ± 862 ± 9<5061 ± 966 ± 965 ± 965 ± 8
      Current smokers (%)26405710004743
      Postbronchodilator FEV1% predicted, mean ± SD47 ± 1349 ± 1347 ± 1550 ± 1546 ± 1559 ± 740 ± 6
      GOLD II: moderate COPD (%)44484854411000
      GOLD III: severe COPD (%)45433638480100
      GOLD IV: very severe COPD (%)1071571000
      COPD, chronic obstructive pulmonary disease; GOLD, Global Initiative for Chronic Obstructive Lung Disease; TORCH, TOwards a Revolution in COPD Health; UPLIFT, Understanding Potential Long-Term Impacts on Function with Tiotropium.

      Outcomes

      For part two of the challenge, the following outcomes were reported: disease progression defined as decline in lung function, mortality, quality-adjusted life-years (QALYs) per patient, and total costs per patient. These outcomes were calculated for both 1-year and lifetime horizons. For part three of the challenge, the total number of exacerbations per patient-year and all-cause mortality was calculated including the uncertainty around the outcomes if possible. The time horizon of the model simulations was similar to the duration of the clinical trials: 3 years for the TORCH trial and 4 years for the UPLIFT trial.

      Results

      Table 2 presents the patient characteristics included in the nine participating cost-effectiveness models. Most models include sex, age, smoking status, and FEV1% predicted. Newer models also included patient characteristics such as previous exacerbations, body mass index (BMI), and comorbidities (Asukai_simulation model and Briggs). Table 3 presents the level of specification for the most important input parameters. The level of detail in which the input parameters are specified varies greatly between the models. Table 4 presents the results for the model simulations of hypothetical patients who differ in terms of sex, age, smoking status, and FEV1% predicted. The validation of subgroup analysis in the models against the outcomes of the subgroup analyses in the clinical trials is presented in Fig. 1, Fig. 2, Fig. 3, Fig. 4. In subsequent sections, the results will be discussed separately for the four patient characteristics that most of the models have in common: sex, age, smoking status, and FEV1% predicted. For each characteristic we first present the results for the simulations of hypothetical patients followed by the simulations of the subgroups in the clinical trials.
      Table 2Patient characteristics included in nine COPD cost-effectiveness models
      CharacteristicAsukai_Markov modelAsukai_simulation modelBorgBriggsDal NegroHansenHoogendoornSamyshkinWacker
      SexXXXXXXX
      AgeXXXXXXXXX
      SmokingXXXXXX
      FEV1 % predictedXXXXXXXXX
      HeightX
      Previous exacerbationsXXX
      BMIXX
      ComorbiditiesXXX
      No. of ER visits/hospitalizationsX
      Rapid decline in FEV1XX
      Dyspnea (mMRC)X
      Disease-specific quality of life (SGRQ)X
      Utility (EQ-5D)X
      Activity (6 MWD)X
      Residual volumeX
      DLCOX
      Lung transplantationX
      BMI, body mass index; COPD, chronic obstructive pulmonary disease; DLCO, diffusion lung capacity for carbon monoxide; ER, emergency room; EQ-5D, EuroQol five-dimensional questionnaire; FEV1, forced expiratory volume in 1 s; mMRC, modified Medical Research Council dysnpnea scale; RV, residual volume; 6MWD, six-minute walking distance; SGRQ, St. George’s Respiratory Questionnaire.
      Table 3Level of specification of important input parameters
      ModelType of input parameter
      Disease progression specified byExacerbation frequency specified byMortality specified byUtilities during stable disease specified byUtilities during exacerbations specified byMaintenance costs specified byExacerbation-related costs specified by
      Asukai_Markov modelSex, age, smoking (indirectly)FEV1% pred.Sex, age, FEV1% pred.FEV1% pred.Exacerbation severityFEV1% pred.Exacerbation severity
      Asukai_simulation modelSex, age, smoking, FEV1% pred.FEV1% pred.Sex, age, FEV1% pred.Sex, age, FEV1%pred., BMI, comorbidities, no. or ER visits/hospitalizationsExacerbation severityFEV1% pred.Exacerbation severity
      BorgAge, FEV1% pred., rapid declineFEV1% pred., frequent exacerbationsAge, FEV1% pred.FEV1% pred.Exacerbation severity, FEV1% pred
      A disutility that depends on exacerbation severity is applied to the utility of stable disease.
      FEV1% pred.Exacerbation severity
      BriggsSex, age, smoking, FEV1% pred. + all other listed in Table 2Sex, age, smoking, FEV1% pred. + all other listed in Table 2Sex, age, smoking, FEV1% pred. + all other listed in Table 2Sex, age, smoking, FEV1% pred. + all other listed in Table 2Sex, age, smoking, FEV1% pred. + all other listed in Table 2Sex, age, smoking, FEV1% pred. + all other listed in Table 2Sex, age, smoking, FEV1% pred. + all other listed in Table 2
      Dal NegroSex, age, RV, DLCO, BODE, % FEV decline, comorbiditiesAge, RV, DLCO, BODE, % FEV decline, comorbiditiesAge, RV, DLCO, BODE, % FEV decline, comorbiditiesAge, RV, DLCO, BODE, % FEV decline, comorbiditiesAge, FEV1% pred., RV, DLCO, BODE, % FEV decline, comorbiditiesAge, FEV1% pred., RV, DLCO, BODE, % FEV decline, comorbiditiesAge, FEV1% pred., RV, DLCO, BODE, % FEV decline, comorbidities
      HansenSmoking, FEV1% pred.FEV1% pred.Age, FEV1% pred.FEV1% pred.FEV1% pred.FEV1% pred.FEV1% pred.
      HoogendoornSex, age, smoking, FEV1% pred.FEV1% pred.Sex, age, smoking, FEV1% pred.FEV1% pred.Exacerbation severitySex, age, FEV1% pred.Exacerbation severity
      SamyshkinSex, age, FEV1% pred.FEV1% pred.Sex, age, FEV1% pred.FEV1% pred.Exacerbation severityFEV1% pred.Exacerbation severity
      WackerSmoking, FEV1% pred.FEV1% pred, lung transplantSex, age, smoking, FEV1% pred., lung transplantFEV1% pred., lung transplantFEV1% pred., exacerbation severity, lung transplantAge, FEV1% pred., lung transplantAge, FEV1% pred., exacerbation severity, lung transplant
      BODE, Body-mass index, airflow Obstruction, Dyspnea and Exercise index; DLCO, diffusion lung capacity for carbon monoxide; FEV1, forced expiratory volume in 1 s; pred., predicted; RV, residual volume.
      low asterisk A disutility that depends on exacerbation severity is applied to the utility of stable disease.
      Table 4One-year model outcomes for a hypothetical reference patient (analysis 1) and for patients who differ on one patient characteristic compared with the reference patient (analyses 2–5)
      AnalysisModel
      Asukai_MarkovAsukai_simulationBorg
      The runs from this model reflect a prevalent mix of 52% male patients and 48% female patients, of which 41% are current smokers.
      BriggsDal NegroHansenHoogendoornSamyshkinWacker
      1. Male patient, 65 y, ex-smoking with severe COPD (=comparator)
       Disease progression (%)
      Disease progression was defined as the percentage of patients with severe COPD moving to very severe COPD except for the model of Briggs et al. in which disease progression was defined as the decline in FEV1% predicted.
      12.80.45.5–0.51 pred.10.05.22.811.92.5
       Mortality (%)4.04.77.22.810.03.17.14.912.4
       QALYs0.7520.6270.6880.581NA0.6580.7230.7240.520
       Costs (2014 $)19501560235038302680–40202260168022702480
      2. Female patient, 65 y, ex-smoking with severe COPD
       Disease progression (%)
      Disease progression was defined as the percentage of patients with severe COPD moving to very severe COPD except for the model of Briggs et al. in which disease progression was defined as the decline in FEV1% predicted.
      12.80.9
      Subgroup analysis not possible for this model.
      –0.81% pred.10.0
      Subgroup analysis not possible for this model.
      2.816.72.5
       Mortality (%)3.63.3
      Subgroup analysis not possible for this model.
      2.610.0
      Subgroup analysis not possible for this model.
      4.74.97.9
       QALYs0.7540.406
      Subgroup analysis not possible for this model.
      0.623NA
      Subgroup analysis not possible for this model.
      0.7230.7210.530
       Costs (2014 $)19501590
      Subgroup analysis not possible for this model.
      30902680–4020
      Subgroup analysis not possible for this model.
      205023202540
      3. Male patient, 75 y, ex-smoking with severe COPD
       Disease progression (%)
      Disease progression was defined as the percentage of patients with severe COPD moving to very severe COPD except for the model of Briggs et al. in which disease progression was defined as the decline in FEV1% predicted.
      11.40.55.6–0.57 pred.20.05.23.611.62.1
       Mortality (%)6.113.824.15.520.09.313.212.227.2
       QALYs0.7420.5980.6240.593NA0.6430.7230.6960.470
       Costs (2014 $)19001460209040104020–53602200209021802290
      4. Male patient, 65 y, smoking with severe COPD
       Disease progression (%)
      Disease progression was defined as the percentage of patients with severe COPD moving to very severe COPD except for the model of Briggs et al. in which disease progression was defined as the decline in FEV1% predicted.
      12.82.3
      Subgroup analysis not possible for this model.
      –1.27% pred.
      Subgroup analysis not possible for this model.
      8.63.6
      Subgroup analysis not possible for this model.
      7.6
       Mortality (%)4.04.8
      Subgroup analysis not possible for this model.
      3.6
      Subgroup analysis not possible for this model.
      3.18.1
      Subgroup analysis not possible for this model.
      17.7
       QALYs0.7520.627
      Subgroup analysis not possible for this model.
      0.548
      Subgroup analysis not possible for this model.
      0.6570.723
      Subgroup analysis not possible for this model.
      0.500
       Costs (2014 $)19501600
      Subgroup analysis not possible for this model.
      3840
      Subgroup analysis not possible for this model.
      22701680
      Subgroup analysis not possible for this model.
      2450
      5. Male patient, 65 y, smoking with moderate COPD
      For this subgroup, disease progression is not a relevant outcome because the percentage of patients with moderate COPD moving to severe COPD is not comparable with the first subgroup analysis that serves as comparator.
       Mortality (%)2.53.35.31.78.02.55.03.06.3
       QALYs0.7810.6530.7410.623NA0.7020.7370.7720.570
       Costs (2014 $)330450106039102010–26801580127015401460
      COPD, chronic obstructive pulmonary disease; NA, not available for this model; pred., predicted; QALY, quality-adjusted life-year.
      low asterisk The runs from this model reflect a prevalent mix of 52% male patients and 48% female patients, of which 41% are current smokers.
      Disease progression was defined as the percentage of patients with severe COPD moving to very severe COPD except for the model of Briggs et al. in which disease progression was defined as the decline in FEV1% predicted.
      Subgroup analysis not possible for this model.
      § For this subgroup, disease progression is not a relevant outcome because the percentage of patients with moderate COPD moving to severe COPD is not comparable with the first subgroup analysis that serves as comparator.
      Fig. 1
      Fig. 1Comparison of model results for the subgroup sex with empirical results of the 4-year UPLIFT trial for (A) exacerbations and (B) all-cause mortality. *Difference is statistically significant. Difference in the simulation model of Asukai could not be tested because no uncertainty was available around the estimates. UPLIFT, Understanding Potential Long-Term Impacts on Function with Tiotropium.
      Fig. 2
      Fig. 2Comparison of model results for the subgroup age with empirical results of the 4-year UPLIFT trial for (A) exacerbations and (B) all-cause mortality. *Difference is statistically significant. Difference in the simulation model of Asukai could not be tested because no uncertainty was available around the estimates. UPLIFT, Understanding Potential Long-Term Impacts on Function with Tiotropium.
      Fig. 3
      Fig. 3Comparison of model results for the subgroup smoking status with empirical results of the 4-year UPLIFT trial for (A) exacerbations and (B) all-cause mortality. *Difference is statistically significant. Difference in the simulation model of Asukai and the trial (exacerbations) could not be tested because no uncertainty was available around the estimates. Uplift, Understanding Potential Long-Term Impacts on Function with Tiotropium.
      Fig. 4
      Fig. 4Comparison of model results for the subgroup FEV1% predicted with empirical results of the 3-year TORCH trial for (A) exacerbations and (B) all-cause mortality. *Difference between moderate and severe COPD is statistically significant. Difference is most likely statistically significant, but standard errors were missing. Difference in the simulation model of Asukai and the model of Dal Negro could not be tested because no uncertainty was available around the estimates. COPD, chronic obstructive pulmonary disease; TORCH, TOwards a Revolution in COPD Health.

      Sex

      Seven models included sex as patient characteristic (Asukai_Markov, Asukai_simulation, Briggs, Dal Negro, Hoogendoorn, Samyshkin, and Wacker). Sex was mainly included to specify disease progression (six models) and mortality (six models) (Asukai_Markov, Asukai_simulation, Briggs, Dal Negro, Hoogendoorn, and Samyshkin) (Table 3). Results of the comparison of hypothetical subgroups of patients showed, however, that three of the models that included sex reported a disease progression rate that was similar for men and women (Asukai_Markov, Dal Negro, and Hoogendoorn) (Table 4), whereas three models had faster disease progression rates for women (Asukai_simulation, Briggs, and Samyshkin). Five models had a higher 1-year mortality rate for male patients than for female patients (Asukai_Markov, Asukai_simulation, Brigs, Hoogendoorn, and Wacker), whereas in one model mortality rate in men and women was comparable (Samyshkin). A few models specified utilities during stable disease and/or COPD-related maintenance costs by sex (Table 3) (Asukai_simulation, Briggs, and Hoogendoorn). The simulation model of Asukai et al. specifying utility weights by sex reported a lower number of QALYs for a female patient than for a male patient, despite a lower female mortality, whereas the model of Briggs et al. had a slightly higher number of QALYs for female patients. Of the two models specifying maintenance costs by sex, one used higher costs for women (Hoogendoorn), whereas the other model (Briggs) included lower costs for women than for men. Validation of the model results with respect to sex against empirical data showed that in the UPLIFT study male patients had fewer exacerbations (rate ratio [RR] = 0.89; 95% confidence interval [CI] 0.81–0.98) compared with female patients, whereas this was found only in the model of Briggs (Fig. 1A). The other models reported equal rates for women and men. Four-year all-cause mortality in the UPLIFT trial was reported to be higher for men (RR = 1.49; 95% CI 1.21–1.84)), which was found in two models (Hoogendoorn and Wacker) (Fig. 1B).

      Age

      All nine participating models included age as characteristic mainly to specify disease progression (seven models) and mortality (nine models). In one model, disease progression was lower for a 75-year-old patient than for a 65-year-old patient (Asukai_Markov). In the models of Dal Negro and Hoogendoorn, disease progression increased with increasing age, whereas in the remaining models disease progression was about equal for both ages (Table 4). In all models, 1-year mortality was higher for a 75-year-old patient than for a 65-year-old patient (1.5-3.3 times higher). Only one model reported higher QALYs for an older patient (Briggs). Three of the four models that specified costs by age (Briggs, Dal Negro, Hoogendoorn, and Wacker) reported higher costs for an older patient (Briggs, Dal Negro, and Hoogendoorn). Validation of the model results with respect to age against the outcomes of the UPLIFT study showed that in all models except one, the exacerbation rate in the subgroup of patients younger than 50 years was equal to the exacerbation rate in the total population (mean age 65 years), which was in accordance with the trial results (Fig. 2A). Only in the model of Briggs, the subgroup of patients younger than 50 years had a significantly lower exacerbation rate than did the total population. In the subgroup of younger patients in the UPLIFT study, 4-year all-cause mortality was significantly lower compared with that of the total population (RR = 0.43; 95% CI 0.25–0.74), which was also found in four models (Briggs, Hansen, Hoogendoorn, and Wacker) (Fig. 2B).

      Smoking Status

      Six of the nine models included smoking status as patient characteristic (Asukai_Markov, Asukai_simulation, Briggs, Hansen, Hoogendoorn, and Wacker). In all six models, disease progression was specified by smoking status (Table 3) as smokers having a higher disease progression compared with ex-smokers (Table 4). Three models specified mortality by smoking status (Briggs, Hoogendoorn, and Wacker). One-year mortality in these models was higher for smokers (Table 4). Only in the model of Briggs et al., smoking status was modeled to have an association with other input parameters such as utilities and costs (Table 3). As a result, the 1-year QALYs and costs were (almost) equal for a patient who smoked compared with an ex-smoking patient in the models (Table 4). The UPLIFT study showed no difference in exacerbations between smokers and ex-smokers, which was in line with the results of most models, except for two (Briggs and Wacker) that reported a higher exacerbation rate for ex-smokers. Four-year mortality in the UPLIFT trial was also not found to be different between smokers and ex-smokers. Three models, however, reported a significantly higher mortality rate for ex-smokers than for smokers (Hansen, Hoogendoorn, and Wacker), most likely because ex-smokers in the UPLIFT trial were older than the smokers and not because of their difference in smoking status. This is confirmed by Table 4, which isolates the impact of smoking status from the impact of age.

      FEV1% predicted

      All nine models included FEV1% predicted. In all models except the Markov model of Asukai, disease progression was specified by FEV1% predicted. All models used FEV1% predicted to specify exacerbation frequency, mortality, utilities, and maintenance costs. In all nine models, 1-year mortality and costs were lower for a patient with moderate COPD (FEV1% predicted about 65%) than for a patient with severe COPD (FEV1% predicted about 40%), while the number of QALYs was higher in all models (Table 4). Five models also specified utilities during exacerbations by FEV1% predicted (Briggs, Borg, Dal Negro, Hansen, and Wacker). Exacerbation costs were specified by FEV1% predicted by four models (Briggs, Dal Negro, Hansen, and Wacker) (Table 3).
      Validation of the model results with respect to FEV1% predicted against the outcomes of the 3-year TORCH study showed that in three models (Borg, Briggs, and Wacker) the exacerbation rate in patients with severe COPD was significantly higher than in patients with moderate COPD (Fig. 4A), which was in line with the trial results. In the TORCH trial, mortality in patients with severe COPD was higher than in patients with moderate COPD (RR = 1.33; 95% CI 1.09–1.63). Four models reported a higher mortality rate in patients with severe COPD (Borg, Hansen, Hoogendoorn, and Wacker).

      Discussion

      By comparing patient heterogeneity in currently available COPD cost-effectiveness models, this study aimed to assess how suitable current models are to evaluate the cost-effectiveness of personalized treatment options for COPD. The patient characteristics that almost all nine models included were sex, age, smoking status, and FEV1% predicted.
      Results with respect to sex showed that the empirical data supported a difference in exacerbations between female and male patients. Several studies in the literature confirm that female patients have a higher exacerbation rate than do male patients [
      • McGarvey L.
      • Lee A.J.
      • Roberts J.
      • et al.
      Characterisation of the frequent exacerbator phenotype in COPD patients in a large UK primary care population.
      ,
      • Müllerová H.
      • Shukla A.
      • Hawkins A.
      • Quint J.
      Risk factors for acute exacerbations of COPD in a primary care population: a retrospective observational cohort study.
      ,
      • Husebø G.R.
      • Bakke P.S.
      • Aanerud M.
      • et al.
      Predictors of exacerbations in chronic obstructive pulmonary disease--results from the Bergen COPD cohort study.
      ,
      • Beeh K.M.
      • Glaab T.
      • Stowasser S.
      • et al.
      Characterisation of exacerbation risk and exacerbator phenotypes in the POET-COPD trial.
      ]. Only one of the current COPD models included this relationship. Two models reported a significantly higher mortality rate in male patients, which was also found in the UPLIFT trial. However, it should be noted that in the trial female patients were on average 3 years younger and had a slightly higher FEV1% predicted, but they were also more likely to smoke. The difference in mortality is less clear from the literature. Some studies reported lower mortality rates for women [
      • Montserrat-Capdevila J.
      • Godoy P.
      • Marsal J.R.
      • Barbé-Illa F.
      Risk factors for mortality in patients with chronic obstructive pulmonary disease [in Spanish].
      ,
      • Schembri S.
      • Anderson W.
      • Morant S.
      • et al.
      A predictive model of hospitalisation and death from chronic obstructive pulmonary disease.
      ,
      • de Torres J.P.
      • Cote C.G.
      • López M.V.
      • et al.
      Sex differences in mortality in patients with COPD.
      ], whereas others found no difference [
      • Patel M.S.
      • Natanek S.A.
      • Stratakos G.
      • et al.
      Vastus lateralis fiber shift is an independent predictor of mortality in chronic obstructive pulmonary disease.
      ,
      • Stolz D.
      • Meyer A.
      • Rakic J.
      • et al.
      Mortality risk prediction in COPD by a prognostic biomarker panel.
      ,
      • de Voogd J.N.
      • Wempe J.B.
      • Postema K.
      • et al.
      More evidence that depressive symptoms predict mortality in COPD patients: is type D personality an alternative explanation?.
      ,
      • Briggs A.
      • Spencer M.
      • Wang H.
      • et al.
      Development and validation of a prognostic index for health outcomes in chronic obstructive pulmonary disease.
      ]. Only two models specified quality of life by sex, whereas several studies found a lower quality of life for female patients [
      • Tsiligianni I.
      • Kocks J.
      • Tzanakis N.
      • et al.
      Factors that influence disease-specific quality of life or health status in patients with COPD: a review and meta-analysis of Pearson correlations.
      ,
      • Fishwick D.
      • Lewis L.
      • Darby A.
      • et al.
      Determinants of health-related quality of life among residents with and without COPD in a historically industrialised area.
      ,
      • Rutten-van Mölken M.P.
      • Oostenbrink J.B.
      • Tashkin D.P.
      • et al.
      Does quality of life of COPD patients as measured by the generic EuroQol five-dimension questionnaire differentiate between COPD severity stages?.
      ].
      All models included a strong association between age and mortality, which was in line with studies found in the literature [
      • Montserrat-Capdevila J.
      • Godoy P.
      • Marsal J.R.
      • Barbé-Illa F.
      Risk factors for mortality in patients with chronic obstructive pulmonary disease [in Spanish].
      ,
      • Schembri S.
      • Anderson W.
      • Morant S.
      • et al.
      A predictive model of hospitalisation and death from chronic obstructive pulmonary disease.
      ,
      • Patel M.S.
      • Natanek S.A.
      • Stratakos G.
      • et al.
      Vastus lateralis fiber shift is an independent predictor of mortality in chronic obstructive pulmonary disease.
      ,
      • Stolz D.
      • Meyer A.
      • Rakic J.
      • et al.
      Mortality risk prediction in COPD by a prognostic biomarker panel.
      ,
      • de Voogd J.N.
      • Wempe J.B.
      • Postema K.
      • et al.
      More evidence that depressive symptoms predict mortality in COPD patients: is type D personality an alternative explanation?.
      ,
      • Briggs A.
      • Spencer M.
      • Wang H.
      • et al.
      Development and validation of a prognostic index for health outcomes in chronic obstructive pulmonary disease.
      ]. Validation against trial results also showed a difference in mortality between young and old patients for the trial as well as most models. Most models did not specify quality of life by age, whereas several studies have shown that older patients with COPD have a lower quality of life [
      • Tsiligianni I.
      • Kocks J.
      • Tzanakis N.
      • et al.
      Factors that influence disease-specific quality of life or health status in patients with COPD: a review and meta-analysis of Pearson correlations.
      ,
      • Punekar Y.S.
      • Rodriguez-Roisin R.
      • Sculpher M.
      • et al.
      Implications of chronic obstructive pulmonary disease (COPD) on patients’ health status: a western view.
      ]. Three models specified COPD-related costs by age. Several studies investigated the association between age and COPD-related costs and found no significant impact of age on costs [
      • Darnell K.
      • Dwivedi A.K.
      • Weng Z.
      • Panos R.J.
      Disproportionate utilization of healthcare resources among veterans with COPD: a retrospective analysis of factors associated with COPD healthcare cost.
      ,
      • Mapel D.W.
      • McMillan G.P.
      • Frost F.J.
      • et al.
      Predicting the costs of managing patients with chronic obstructive pulmonary disease.
      ,
      • Miravitlles M.
      • Murio C.
      • Guerrero T.
      • Gisbert R.
      Costs of chronic bronchitis and COPD: a 1-year follow-up study.
      ,
      • Jansson S.A.
      • Andersson F.
      • Borg S.
      • et al.
      Costs of COPD in Sweden according to disease severity.
      ]. This may be because age and FEV1% predicted are correlated and in three of these studies FEV1% was also included in the multivariate model and found to be a significant predictor.
      The impact of smoking on disease progression was included in six of the models. Smokers were modeled to have a higher disease progression compared with ex-smokers, which was in line with the literature [
      • Vestbo J.
      • Edwards L.D.
      • Scanlon P.D.
      • et al.
      ECLIPSE Investigators. Changes in forced expiratory volume in 1 second over time in COPD.
      ,
      • Celli B.R.
      • Thomas N.E.
      • Anderson J.A.
      • et al.
      Effect of pharmacotherapy on rate of decline of lung function in chronic obstructive pulmonary disease: results from the TORCH study.
      ,
      • Scanlon P.D.
      • Connett J.E.
      • Waller L.A.
      • et al.
      Lung Health Study Research Group
      Smoking cessation and lung function in mild-to-moderate chronic obstructive pulmonary disease. The Lung Health Study.
      ]. Three models included a higher all-cause mortality rate for smokers than for ex-smokers in their input data (Table 4). No difference in mortality rate between smokers and ex-smokers was found in the UPLIFT trial. Validation of model outcomes against trial results showed that three models reported a higher mortality rate for ex-smokers. This was probably because ex-smokers in the trial were on average 5 years older than smokers and had a lower FEV1% predicted. Of 17 studies on predictors of mortality found in the literature, 4 studies found a higher mortality rate for smokers, 6 studies found no association, and 10 studies did not investigate this association.
      With respect to FEV1% predicted, the models are in line with the empirical data and the literature. All models include an association between FEV1% predicted and exacerbation frequency, quality of life, mortality, and costs. In some models, the differences in exacerbations and mortality are however very small and not significant due to the large uncertainty around the estimates. The literature also showed that disease progression in terms of lung function decline seems to decrease when the FEV1 decreases [
      • Vestbo J.
      • Edwards L.D.
      • Scanlon P.D.
      • et al.
      ECLIPSE Investigators. Changes in forced expiratory volume in 1 second over time in COPD.
      ,
      • Celli B.R.
      • Thomas N.E.
      • Anderson J.A.
      • et al.
      Effect of pharmacotherapy on rate of decline of lung function in chronic obstructive pulmonary disease: results from the TORCH study.
      ,
      • Casanova C.
      • de Torres J.P.
      • Aguirre-Jaíme A.
      • et al.
      The progression of chronic obstructive pulmonary disease is heterogeneous: the experience of the BODE cohort.
      ]. This association was not included in all models, because five of the models used the Lung Health Study as (one of the) data source (s) for disease progression, which is a study in patients with mild-to-moderate COPD showing the opposite relation (decline increases when FEV1 decreases) [
      • Scanlon P.D.
      • Connett J.E.
      • Waller L.A.
      • et al.
      Lung Health Study Research Group
      Smoking cessation and lung function in mild-to-moderate chronic obstructive pulmonary disease. The Lung Health Study.
      ]. It is not surprising that the model performance with respect to patient heterogeneity was best for FEV1% predicted, because in almost all models the FEV1% predicted is the key parameter. Five of the nine models are cohort Markov models with Markov states defined according to the 2007 GOLD classification; that is, the severity of COPD was defined by the FEV1% predicted.
      Outcomes of the different models were compared with published subgroup results of two large trials: the UPLIFT and TORCH trials. Some of the COPD models were built to extrapolate the results of clinical trials (Asukai_Markov and Samyshkin) and were populated with input data from these trials, whereas other models are population models using a wide range of different data sources as input (Hoogendoorn, Hansen, and Wacker). For the latter models, comparison with trial results may therefore be less valid because it is well known that patients included in clinical trials are a subgroup of the total COPD population. This was, for example, shown in a study of Kruis et al. [
      • Kruis A.L.
      • Ställberg B.
      • Jones R.C.
      • et al.
      Primary care COPD patients compared with large pharmaceutically-sponsored COPD studies: an UNLOCK validation study.
      ] that found that patients participating in large clinical trials sponsored by pharmaceutical companies are on average younger, more likely to be male, have a lower FEV1% predicted, and tend to have more exacerbations than do patients in primary care. In the current analysis, this selection bias issue was most likely the explanation that no difference in mortality was found between smokers and ex-smokers in the UPLIFT trial. Patients who continue to smoke are probably less likely to be included in clinical studies because they progress faster to more severe stages, have more comorbidities, and are more likely to die because of other smoking-related diseases. Nevertheless, among the patients who continue to smoke despite having COPD there may be relatively more patients who are less susceptible to smoking-related diseases. The true impact of trial populations being a selective group of patients on the outcomes of this study is difficult to assess. In the present study, we mainly focused on differences in outcomes between subgroups within one model or trial and therefore the impact might be limited.
      The results of this study showed that all currently available models are capable of running simulations for different age- and COPD severity classes. Most models also have the ability to run analyses separately for men and women and for smokers and ex-smokers. The validity of these subgroup analyses within the models is questionable, because important input parameters have not been specified by sex, age, or smoking status. For example, the specification of higher exacerbations for women or lower quality of life with older age is not included in most models. Most models are developed to evaluate treatment options for the total general COPD population or the average COPD trial population.
      Although it is not unlikely that future treatment will be increasingly tailored to age, sex, and smoking status, treatment is more likely to be personalized on the basis of clinical parameters especially when considering ethical debates and societal preferences. Exacerbations, for example, are highly associated with the number of previous exacerbations [
      • Müllerová H.
      • Shukla A.
      • Hawkins A.
      • Quint J.
      Risk factors for acute exacerbations of COPD in a primary care population: a retrospective observational cohort study.
      ,
      • Husebø G.R.
      • Bakke P.S.
      • Aanerud M.
      • et al.
      Predictors of exacerbations in chronic obstructive pulmonary disease--results from the Bergen COPD cohort study.
      ,
      • Beeh K.M.
      • Glaab T.
      • Stowasser S.
      • et al.
      Characterisation of exacerbation risk and exacerbator phenotypes in the POET-COPD trial.
      ,
      • Make B.J.
      • Eriksson G.
      • Calverley P.M.
      • et al.
      A score to predict short-term risk of COPD exacerbations (SCOPEX).
      ]. Mortality in COPD is associated not only with age and FEV1 but also with comorbidities, BMI, dyspnea, and several other clinical parameters [
      • Montserrat-Capdevila J.
      • Godoy P.
      • Marsal J.R.
      • Barbé-Illa F.
      Risk factors for mortality in patients with chronic obstructive pulmonary disease [in Spanish].
      ,
      • de Torres J.P.
      • Cote C.G.
      • López M.V.
      • et al.
      Sex differences in mortality in patients with COPD.
      ,
      • Stolz D.
      • Meyer A.
      • Rakic J.
      • et al.
      Mortality risk prediction in COPD by a prognostic biomarker panel.
      ,
      • Briggs A.
      • Spencer M.
      • Wang H.
      • et al.
      Development and validation of a prognostic index for health outcomes in chronic obstructive pulmonary disease.
      ,
      • Schols A.M.
      • Slangen J.
      • Volovics L.
      • Wouters E.F.
      Weight loss is a reversible factor in the prognosis of chronic obstructive pulmonary disease.
      ]. BMI also seems to be associated with quality of life in COPD [
      • Tsiligianni I.
      • Kocks J.
      • Tzanakis N.
      • et al.
      Factors that influence disease-specific quality of life or health status in patients with COPD: a review and meta-analysis of Pearson correlations.
      ,
      • Villar Balboa I.
      • Carrillo Muñoz R.
      • Regí Bosque M.
      • et al.
      Factors associated with the quality of life in patients with chronic obstructive pulmonary disease [in Spanish].
      ] and comorbidities may affect COPD-related costs [
      • Mapel D.W.
      • McMillan G.P.
      • Frost F.J.
      • et al.
      Predicting the costs of managing patients with chronic obstructive pulmonary disease.
      ,
      • Miravitlles M.
      • Murio C.
      • Guerrero T.
      • Gisbert R.
      Costs of chronic bronchitis and COPD: a 1-year follow-up study.
      ,
      • Huber M.
      • Wacker M.
      • Vogelmeier C.F.
      • Leidl R.
      Excess costs of comorbidities in chronic obstructive pulmonary disease: a systematic review.
      ,
      • Huber M.
      • Wacker M.
      • Vogelmeier C.F.
      • Leidl R.
      Comorbid influences on generic health-related quality of life in COPD: a systematic review.
      ]. Only the more recently developed models of Asukai (simulation) and Briggs included these types of clinical parameters and are therefore more suitable than the other models to evaluate personalized treatment. Recent reviews on phenotyping in COPD showed that currently about four phenotypes have been defined in COPD that may require a different treatment strategy: emphysema, COPD with chronic bronchitis, COPD combined with asthma, and COPD with frequent exacerbations [
      • Miravitlles M.
      • Soler-Cataluña J.J.
      • Calle M.
      • et al.
      Spanish guideline for COPD (GesEPOC): update 2014.
      ,
      • Segreti A.
      • Stirpe E.
      • Rogliani P.
      • Cazzola M.
      Defining phenotypes in COPD: an aid to personalized healthcare.
      ,
      • Rothe T.
      Phenotype specific therapy of COPD [in German].
      ,
      • Vestbo J.
      COPD: definition and phenotypes.
      ]. Information on effectiveness and cost-effectiveness of treatment options for these subgroups is needed to guide clinical guideline development and decisions for reimbursement. Three of the nine models included information on one of the phenotypes, that is, COPD with frequent exacerbations, but none of the current models is able to evaluate treatment options for the other three phenotypes. Future models should include all clinical patient characteristics currently considered to influence disease severity, prognosis, and treatment response in COPD.
      The present study had some limitations. First, not all models could perform uncertainty analysis around the results for exacerbations and all-cause mortality. Therefore, it was not always possible to determine whether predicted differences between subgroups are important within the range of uncertainty estimated. The information in the published articles of the two clinical trials was also sometimes not sufficient to calculate whether the difference between subgroups was statistically significant or not. Results of the trials by smoking status and by COPD severity did not present SDs or standard errors around the mean exacerbation rate [
      • Jenkins C.R.
      • Jones P.W.
      • Calverley P.M.
      • et al.
      Efficacy of salmeterol/fluticasone propionate by GOLD stage of chronic obstructive pulmonary disease: analysis from the randomised, placebo-controlled TORCH study.
      ,
      • Tashkin D.P.
      • Celli B.
      • Kesten S.
      • et al.
      Long-term efficacy of tiotropium in relation to smoking status in the UPLIFT trial.
      ]. The exacerbation rates in the UPLIFT trial in smokers and ex-smokers were, however, almost equal, 0.77 versus 0.83, a difference that is most likely not significant [
      • Tashkin D.P.
      • Celli B.
      • Kesten S.
      • et al.
      Long-term efficacy of tiotropium in relation to smoking status in the UPLIFT trial.
      ]. In contrast, the difference in exacerbation rate in the TORCH trial between people with severe and moderate COPD was large, 1.24 versus 0.82, and therefore this difference seems significant given the population size [
      • Jenkins C.R.
      • Jones P.W.
      • Calverley P.M.
      • et al.
      Efficacy of salmeterol/fluticasone propionate by GOLD stage of chronic obstructive pulmonary disease: analysis from the randomised, placebo-controlled TORCH study.
      ].
      A second limitation was that the starting population of the models was adjusted only with respect to sex distribution, age, smoking distribution, and FEV1%predicted to mirror the population of either the TORCH trial or the UPLIFT trial. The most recent models include much more characteristics (e.g., BMI and comorbidities) that could have been adjusted to create model populations that were even more similar to the trial populations. However, this would have made a comparison with the other models impossible.
      A third limitation of the present study was that we only assessed whether there were significant differences in outcomes between the subgroups simulated with the models and whether that result was in line with the trial results and the literature. We did not focus on the results of the models in absolute terms. However, the results of the different models showed substantial variation, especially for mortality. A more detailed external validation of model results against the outcomes of the trials for the total population will be a topic of another article.
      Finally, although this study aimed to compare cost-effectiveness models, it mainly focused on the comparison of clinical outcomes, not on costs. Only Table 4 presents information on costs. Validation of costs was not possible because the trials did not present costs specified by subgroups. Clinical outcomes are important outcomes for cost-effectiveness models as well because they have a high impact on quality of life and costs.
      In conclusion, this study showed that most currently available COPD cost-effectiveness models include the relevant patient characteristics sex, age, smoking status, and FEV1% predicted and are therefore able to evaluate the cost-effectiveness of personalized treatment based on these parameters. Most models, however, do not include all important associations between these characteristics and input parameters. Furthermore, treatment in COPD is more likely to be personalized on the basis of clinical parameters. Two models also included several clinical patient characteristics, such as previous exacerbations, BMI, and comorbidities, and seem to be more suitable to evaluate personalized treatment. Inclusion of other clinical parameters, such as emphysema, chronic bronchitis, and coexistence of asthma, is relevant to make the models suitable to evaluate the cost-effectiveness of treatment options for currently defined phenotypes in COPD.
      Source of financial support: This study was financially supported by Boehringer Ingelheim International, GlaxoSmithKline The Netherlands, Novartis International, and Takeda Pharmaceuticals International.

      References

        • O’Donnell J.C.
        Personalized medicine and the role of health economics and outcomes research: issues, applications, emerging trends, and future research.
        Value Health. 2013; 16: S1-S3
        • Redekop W.K.
        • Mladsi D.
        The faces of personalized medicine: a framework for understanding its meaning and scope.
        Value Health. 2013; 16: S4-S9
      1. Personalized Medicine. U.S. Food and Drug Administration. Available from: http://www.fda.gov/scienceresearch/specialtopics/personalizedmedicine/default.htm. [Accessed June 17, 2015].

        • Hatz M.H.
        • Schremser K.
        • Rogowski W.H.
        Is individualized medicine more cost-effective? A systematic review.
        Pharmacoeconomics. 2014; 32: 443-455
        • Phillips K.A.
        • Ann Sakowski J.
        • Trosman J.
        • et al.
        The economic value of personalized medicine tests: what we know and what we need to know.
        Genet Med. 2014; 16: 251-257
        • Ramaekers B.L.
        • Joore M.A.
        • Grutters J.P.
        How should we deal with patient heterogeneity in economic evaluation: a systematic review of national pharmacoeconomic guidelines.
        Value Health. 2013; 16: 855-862
      2. Global Initiative for Chronic Obstructive Lung Disease. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease. Available from: http://www.goldcopd.com. [Accessed June 17, 2015].

        • Han M.K.
        • Agusti A.
        • Calverley P.M.
        • et al.
        Chronic obstructive pulmonary disease phenotypes: the future of COPD.
        Am J Respir Crit Care Med. 2010; 182: 598-604
        • Woodruff P.G.
        • Agusti A.
        • Roche N.
        • et al.
        Current concepts in targeting chronic obstructive pulmonary disease pharmacotherapy: making progress towards personalised management.
        Lancet. 2015; 385: 1789-1798
      3. The Pharmaceutical Research and Manufacturers of America. Medicines in development for chronic obstructive pulmonary disease (COPD). 2012. Available from: http://www.phrma.org/sites/default/files/pdf/copd2012.pdf. [Accessed June17, 2015].

        • Hoogendoorn M.
        • Feenstra T.L.
        • Asukai Y.
        • et al.
        Cost-effectiveness models for chronic obstructive pulmonary disease: cross-model comparison of hypothetical treatment scenarios.
        Value Health. 2014; 17: 525-536
        • Price D.
        • Gray A.
        • Gale R.
        • et al.
        Cost-utility analysis of indacaterol in Germany: a once-daily maintenance bronchodilator for patients with COPD.
        Respir Med. 2011; 105: 1635-1647
        • Borg S.
        • Ericsson A.
        • Wedzicha J.
        • et al.
        A computer simulation model of the natural history and economic impact of chronic obstructive pulmonary disease.
        Value Health. 2004; 7: 153-167
        • Hansen R.N.
        • Xu X.
        • Sullivan S.D.
        A dynamic cohort model of chronic obstructive pulmonary disease and its treatments.
        Value Health. 2012; 15: A54
        • Hoogendoorn M.
        • Rutten-van Molken M.P.
        • Hoogenveen R.T.
        • et al.
        Developing and applying a stochastic dynamic population model for chronic obstructive pulmonary disease.
        Value Health. 2011; 14: 1039-1047
        • Samyshkin Y.
        • Kotchie R.W.
        • Mork A.C.
        • et al.
        Cost-effectiveness of roflumilast as an add-on treatment to long-acting bronchodilators in the treatment of COPD associated with chronic bronchitis in the United Kingdom.
        Eur J Health Econ. 2014; 15: 69-82
        • Menn P.
        • Leidl R.
        • Holle R.
        A lifetime Markov model for the economic evaluation of chronic obstructive pulmonary disease.
        Pharmacoeconomics. 2012; 30: 825-840
        • Asukai Y.
        • Baldwin M.
        • Fonseca T.
        • et al.
        Improving clinical reality in chronic obstructive pulmonary disease economic modelling: development and validation of a micro-simulation approach.
        Pharmacoeconomics. 2013; 31: 151-161
        • Briggs A.
        • Lomas D.
        • Rutten van-Molken M.
        • et al.
        Developing a new model of COPD: from conceptualization to implementation to validation.
        Value Health. 2013; 16: A234
        • Risebrough N.A.
        • Briggs A.
        • Baker T.M.
        • et al.
        Validating a model to predict disease progression outcomes in patients with COPD.
        Value Health. 2014; 17: A560-A561
        • Dal Negro R.W.
        Current annual cost calculation is the best predictor of mortality at three years in COPD.
        Value Health. 2014; 17: A590-A591
        • Tashkin D.P.
        • Celli B.
        • Senn S.
        • et al.
        • UPLIFT Study Investigators
        A 4-year trial of tiotropium in chronic obstructive pulmonary disease.
        N Engl J Med. 2008; 359: 1543-1554
        • Calverley P.M.
        • Anderson J.A.
        • Celli B.
        • et al.
        • TORCH Investigators
        Salmeterol and fluticasone propionate and survival in chronic obstructive pulmonary disease.
        N Engl J Med. 2007; 356: 775-789
        • Jenkins C.R.
        • Jones P.W.
        • Calverley P.M.
        • et al.
        Efficacy of salmeterol/fluticasone propionate by GOLD stage of chronic obstructive pulmonary disease: analysis from the randomised, placebo-controlled TORCH study.
        Respir Res. 2009; 10: 59
        • Tashkin D.
        • Celli B.
        • Kesten S.
        • et al.
        Effect of tiotropium in men and women with COPD: results of the 4-year UPLIFT trial.
        Respir Med. 2010; 104: 1495-1504
        • Morice A.H.
        • Celli B.
        • Kesten S.
        • et al.
        COPD in young patients: a pre-specified analysis of the four-year trial of tiotropium (UPLIFT).
        Respir Med. 2010; 104: 1659-1667
        • Tashkin D.P.
        • Celli B.
        • Kesten S.
        • et al.
        Long-term efficacy of tiotropium in relation to smoking status in the UPLIFT trial.
        Eur Respir J. 2010; 35: 287-294
        • McGarvey L.
        • Lee A.J.
        • Roberts J.
        • et al.
        Characterisation of the frequent exacerbator phenotype in COPD patients in a large UK primary care population.
        Respir Med. 2015; 109: 228-237
        • Müllerová H.
        • Shukla A.
        • Hawkins A.
        • Quint J.
        Risk factors for acute exacerbations of COPD in a primary care population: a retrospective observational cohort study.
        BMJ Open. 2014; 4: e006171
        • Husebø G.R.
        • Bakke P.S.
        • Aanerud M.
        • et al.
        Predictors of exacerbations in chronic obstructive pulmonary disease--results from the Bergen COPD cohort study.
        PLoS One. 2014; 9: e109721
        • Beeh K.M.
        • Glaab T.
        • Stowasser S.
        • et al.
        Characterisation of exacerbation risk and exacerbator phenotypes in the POET-COPD trial.
        Respir Res. 2013; 14: 116
        • Montserrat-Capdevila J.
        • Godoy P.
        • Marsal J.R.
        • Barbé-Illa F.
        Risk factors for mortality in patients with chronic obstructive pulmonary disease [in Spanish].
        Aten Primaria. 2015; 47: 498-504
        • Schembri S.
        • Anderson W.
        • Morant S.
        • et al.
        A predictive model of hospitalisation and death from chronic obstructive pulmonary disease.
        Respir Med. 2009; 103: 1461-1467
        • de Torres J.P.
        • Cote C.G.
        • López M.V.
        • et al.
        Sex differences in mortality in patients with COPD.
        Eur Respir J. 2009; 33: 528-535
        • Patel M.S.
        • Natanek S.A.
        • Stratakos G.
        • et al.
        Vastus lateralis fiber shift is an independent predictor of mortality in chronic obstructive pulmonary disease.
        Am J Respir Crit Care Med. 2014; 190: 350-352
        • Stolz D.
        • Meyer A.
        • Rakic J.
        • et al.
        Mortality risk prediction in COPD by a prognostic biomarker panel.
        Eur Respir J. 2014; 44: 1557-1570
        • de Voogd J.N.
        • Wempe J.B.
        • Postema K.
        • et al.
        More evidence that depressive symptoms predict mortality in COPD patients: is type D personality an alternative explanation?.
        Ann Behav Med. 2009; 38: 86-93
        • Briggs A.
        • Spencer M.
        • Wang H.
        • et al.
        Development and validation of a prognostic index for health outcomes in chronic obstructive pulmonary disease.
        Arch Intern Med. 2008; 168: 71-79
        • Tsiligianni I.
        • Kocks J.
        • Tzanakis N.
        • et al.
        Factors that influence disease-specific quality of life or health status in patients with COPD: a review and meta-analysis of Pearson correlations.
        Prim Care Respir J. 2011; 20: 257-268
        • Fishwick D.
        • Lewis L.
        • Darby A.
        • et al.
        Determinants of health-related quality of life among residents with and without COPD in a historically industrialised area.
        Int Arch Occup Environ Health. 2015; 88: 799-805
        • Rutten-van Mölken M.P.
        • Oostenbrink J.B.
        • Tashkin D.P.
        • et al.
        Does quality of life of COPD patients as measured by the generic EuroQol five-dimension questionnaire differentiate between COPD severity stages?.
        Chest. 2006; 130: 1117-1128
        • Punekar Y.S.
        • Rodriguez-Roisin R.
        • Sculpher M.
        • et al.
        Implications of chronic obstructive pulmonary disease (COPD) on patients’ health status: a western view.
        Respir Med. 2007; 101: 661-669
        • Darnell K.
        • Dwivedi A.K.
        • Weng Z.
        • Panos R.J.
        Disproportionate utilization of healthcare resources among veterans with COPD: a retrospective analysis of factors associated with COPD healthcare cost.
        Cost Eff Resour Alloc. 2013; 11: 13
        • Mapel D.W.
        • McMillan G.P.
        • Frost F.J.
        • et al.
        Predicting the costs of managing patients with chronic obstructive pulmonary disease.
        Respir Med. 2005; 99: 1325-1333
        • Miravitlles M.
        • Murio C.
        • Guerrero T.
        • Gisbert R.
        Costs of chronic bronchitis and COPD: a 1-year follow-up study.
        Chest. 2003; 123: 784-791
        • Jansson S.A.
        • Andersson F.
        • Borg S.
        • et al.
        Costs of COPD in Sweden according to disease severity.
        Chest. 2002; 122: 1994-2002
        • Vestbo J.
        • Edwards L.D.
        • Scanlon P.D.
        • et al.
        ECLIPSE Investigators. Changes in forced expiratory volume in 1 second over time in COPD.
        N Engl J Med. 2011; 365: 1184-1192
        • Celli B.R.
        • Thomas N.E.
        • Anderson J.A.
        • et al.
        Effect of pharmacotherapy on rate of decline of lung function in chronic obstructive pulmonary disease: results from the TORCH study.
        Am J Respir Crit Care Med. 2008; 178: 332-338
        • Scanlon P.D.
        • Connett J.E.
        • Waller L.A.
        • et al.
        • Lung Health Study Research Group
        Smoking cessation and lung function in mild-to-moderate chronic obstructive pulmonary disease. The Lung Health Study.
        Am J Respir Crit Care Med. 2000; 161: 381-390
        • Casanova C.
        • de Torres J.P.
        • Aguirre-Jaíme A.
        • et al.
        The progression of chronic obstructive pulmonary disease is heterogeneous: the experience of the BODE cohort.
        Am J Respir Crit Care Med. 2011; 184: 1015-1021
        • Kruis A.L.
        • Ställberg B.
        • Jones R.C.
        • et al.
        Primary care COPD patients compared with large pharmaceutically-sponsored COPD studies: an UNLOCK validation study.
        PLoS One. 2014; 9: e90145
        • Make B.J.
        • Eriksson G.
        • Calverley P.M.
        • et al.
        A score to predict short-term risk of COPD exacerbations (SCOPEX).
        Int J Chron Obstruct Pulmon Dis. 2015; 10: 201-209
        • Schols A.M.
        • Slangen J.
        • Volovics L.
        • Wouters E.F.
        Weight loss is a reversible factor in the prognosis of chronic obstructive pulmonary disease.
        Am J Respir Crit Care Med. 1998; 157: 1791-1797
        • Villar Balboa I.
        • Carrillo Muñoz R.
        • Regí Bosque M.
        • et al.
        Factors associated with the quality of life in patients with chronic obstructive pulmonary disease [in Spanish].
        Aten Primaria. 2014; 46: 179-187
        • Huber M.
        • Wacker M.
        • Vogelmeier C.F.
        • Leidl R.
        Excess costs of comorbidities in chronic obstructive pulmonary disease: a systematic review.
        PLoS One. 2015; 10: e0123292
        • Huber M.
        • Wacker M.
        • Vogelmeier C.F.
        • Leidl R.
        Comorbid influences on generic health-related quality of life in COPD: a systematic review.
        PLoS One. 2015; 10: e0132670
        • Miravitlles M.
        • Soler-Cataluña J.J.
        • Calle M.
        • et al.
        Spanish guideline for COPD (GesEPOC): update 2014.
        Arch Bronconeumol. 2014; 50: 1-16
        • Segreti A.
        • Stirpe E.
        • Rogliani P.
        • Cazzola M.
        Defining phenotypes in COPD: an aid to personalized healthcare.
        Mol Diagn Ther. 2014; 18: 381-388
        • Rothe T.
        Phenotype specific therapy of COPD [in German].
        Praxis (Bern 1994). 2014; 103: 1509-1514
        • Vestbo J.
        COPD: definition and phenotypes.
        Clin Chest Med. 2014; 35: 1-6