Economic Modeling in Chronic Obstructive Pulmonary Disease

Economic Modeling in Chronic Obstructive Pulmonary Disease Maureen Rutten-van Mölken and Todd A. Lee Erasmus Medical Center, Institute for Medical Technology Assessment, Rotterdam, The Netherlands; and Hines Veterans Affairs Hospital and Northwestern University, Chicago, Illinois Calculating the cost-effectiveness of interventions is an important step in accurately assessing the health and financial burdens of a disease. Although clinical trials that include cost data can be used to compare the cost-effectiveness of specific interventions, they only deal with outcomes within the time frame of the trial. Health economic models can synthesize epidemiologic, clinical, economic, and quality-of-life data from many different sources and extrapolate results to a point many years in the future. The models generally compare interventions with respect to the costs per life-year gained or per quality-adjusted life-year gained. The use of health economic models to assess the economic burden of chronic obstructive pulmonary disease (COPD) and the value of interventions is growing, and will continue to do so as the burden of the disease is better appreciated. Several COPD disease-state models have been described; each uses a consistent definition of COPD severity that is based on FEV 1 % predicted, but the models differ in the allowed transitions, disease progression estimates, utility weights, and costs. This article reviews COPD health economic models and discusses the importance of survival benefits and utilities (health state valuations) for COPD in economic models. Keywords: chronic obstructive pulmonary disease; health economic model; cost-effectiveness Chronic obstructive pulmonary disease (COPD) is a major cause of chronic morbidity and mortality. Its high prevalence and related mortality are increasing at a time when death from cardiovascular events is decreasing, giving a clear signal that improved intervention strategies through early diagnosis and treatment are required to reduce the economic and social costs of this disease. Data from the United States and Europe offer insights into the enormous economic and social burden of COPD for societies and insurance payers. When considering the economics of COPD, the arguments must stretch beyond simply the efficacy, safety, and cost of an intervention to include broader aspects of improved health, such as survival and impact on quality of life, within a standardized framework. In general, economic evaluations, or cost-effectiveness analyses, are employed because they allow valued judgments on the efficiency of treatments within the context of survival and quality of life. Parameters such as cost per life-year gained (LYG) and per quality-adjusted life year (QALY) combine costs with measures of quality of life and survival time to help policy makers determine priorities and strategies for health care. This framework allows comparison of the value of healthcare interventions both within and across different diseases and an assessment of the impact of investment now on future expenditure and health outcomes. Thus, under- (Received in original form March 27, 2006; accepted in final form April 18, 2006 ) Correspondence and requests for reprints should be addressed to Dr. Maureen Rutten-van Mölken, Institute for Medical Technology Assessment, Erasmus MC, P.O. Box 1738, 3000 DR Rotterdam, The Netherlands. E-mail: m.rutten-vanmolken@ erasmusmc.nl Proc Am Thorac Soc Vol 3. pp 630 634, 2006 DOI: 10.1513/pats.200603-095SS Internet address: www.atsjournals.org standing the health economics of interventions for COPD will be an important component in caring for patients with COPD. A clearer quantitative economic assessment of COPD and its interventions is required to more effectively inform health care decision making. Although clinical trial data are frequently used to compare the cost-effectiveness of specific interventions, they have limitations. These include the choice of outcome and when the outcome is measured. Frequently, surrogate endpoints (e.g., lung function) are measured as opposed to measuring final outcomes, such as quality of life and death. The endpoints are usually recorded at a discreet time point, and don t go beyond the end of the trial, so no ongoing effect over a subject s lifetime is measured. Survival data often tend to relate only to the followup period of the trial, and are not usually extended long-term, which offers insufficient information to accurately estimate any potential gain in life expectancy. Given these limitations, increasingly more sophisticated approaches to economic assessment, which incorporate disease modeling, are employed to provide survival and cost per LYG estimates beyond the clinical trial and to explore uncertainties and drivers of cost-effectiveness. In an economic assessment of a chronic disease such as COPD, decision analytical modeling, such as Markov modeling, is now the favored approach. In a Markov model, the progress of a notional group of patients through a finite number of health states is modeled over time, with patients initially placed in one health state; the probability of transition to another state in the model is defined within a given time period, known as a Markov cycle. Models are created through incorporation of clinical data, especially from epidemiologic and population databases that map the natural history of the disease in terms of survival, allowing extrapolation to a point in the future for cost-effectiveness assessment. A limitation of Markov modeling is that previous disease events experienced by a patient do not influence the probability of a transition from one state to another (known as the Markovian assumption) (1). In this review, we offer a brief overview of published Markov models developed to assess the cost-effectiveness of COPD interventions, highlighting some of the differences between these models and reviewing ongoing research to improve utility estimates for COPD severity stages and exacerbation status. Utility is a measure of the preference for or desirability of a specific level of health status or specific health outcome, and is used to weigh improvements in life expectancy according to the quality of life experienced within the calculation of QALYs. CURRENT COPD-RELATED ECONOMIC MODELS Five COPD progression models, all of which are Markov models (2 6), and which have evaluated different interventions (summarized in Table 1) are described below. The most recent model published, developed by the Institute for Medical Technology Assessment in Rotterdam and the National Institute of Public Health and the Environment in Bilthoren, NL (6), is a uniquely dynamic population model. A model in development as part of the Burden of Obstructive Lung Disease (BOLD) initiative is also described (7).

Rutten-van Mölken and Lee: Economic Models in COPD 631 TABLE 1. MODEL-BASED COST-EFFECTIVENESS STUDIES OF CHRONIC OBSTRUCTIVE LUNG DISEASE INTERVENTIONS Author (Reference) Interventions Time Horizon Conclusions of the Base Case Analyses Sin and colleagues (2) Inhaled CS in various COPD severity stages 3 yr Can $17,000 when treatment is restricted to patients with ATS stages II and III COPD Borg and colleagues (3) Two hypothetical interventions, one halving lung Lifetime 30 yr were needed for the first intervention function decline and one reducing to become cost-effective compared to exacerbations by 25% the latter intervention Spencer and colleagues (4) Salmeterol/fluticasone vs. usual care 25 yr Can $75,000 per QALY when there is an effect on exacerbations only; Can $11,000 per QALY when there is a survival effect Oostenbrink and colleagues (5) Ipratropium, tiotropium, salmeterol 1 yr In Canada, tiotropium highest probability of being cost-effective for ceiling ratios above 160 per exacerbation avoided Hoogendoorn and colleagues (6) Minimal counseling and intensive counseling Lifetime Minimal counseling cost saving and bupropion vs. current practice intensive counseling 15,000 per QALY Definition of abbreviations: ATS American Thoracic Society; CS corticosteroids; QALY quality-adjusted life-year. The first Markov model published assessed the cost-effectiveness of four different treatment strategies involving inhaled corticosteroids (ICS) with a time horizon of 3 yr (2). The strategies were: no ICS; ICS given to all patients; ICS given to patients with American Thoracic Society stage II or III disease (FEV 1 50% of predicted); or to stage III patients only (FEV 1 35% of predicted). The effect of corticosteroids was modeled as a 30% reduction in exacerbation rate and a 16% reduction in mortality. The model showed that ICS in stage II or III patients improved quality-adjusted life expectancy at a cost of $34,000 per QALY gained when no mortality benefit was assumed, and $17,000 per QALY gained when a mortality benefit was assumed. A different model (3), based on the Global Initiative for Chronic Lung Disease (GOLD) recommendations, was unique in that it included separate Markov chains for COPD severity and for COPD-related exacerbations. In the exacerbation component of the model, patients could be exacerbation free or have mild, moderate, or severe exacerbations. The type of exacerbation influenced the probability of moving to either milder or worse disease, and of death. Patients can improve to the next milder COPD state, after which disease progression continues. The model used data from a variety of sources on disease progression, expected exacerbation frequency/duration, mortality, costs, burden of illness, and the relationships between variables. The costeffectiveness of two hypothetical interventions, one that would halve lung function decline, and one that reduced exacerbations by 25%, was calculated. The findings were important, as they showed that an intervention that slows disease progression as assessed by lung function decline only becomes cost-effective after many years (about 30 years) when compared with an intervention that reduces exacerbation rates. In 2005, a further Markov model (4) was described using the same four mutually exclusive disease states (mild, moderate, and severe disease, and death) as those of a previous study (2) to model costs, exacerbations, survival, QALYs, and cost effectiveness (Figure 1). Spencer and colleagues compared a long-acting 2 -agonist/inhaled corticosteroid combination (salmeterol/fluticasone propionate) to usual care over a 25-yr time horizon (4). The transition probabilities of disease progression (for smokers and ex-smokers), death, and the exacerbation probabilities were taken from the published medical literature. Efficacy data for the combination therapy were obtained from the TRISTAN (Trial of Inhaled Steroids and Long-Acting -Agonists) study (8) that involved patients with a documented history of frequent COPD exacerbations. The model demonstrated that, when combination therapy impacted exacerbation rates only, the cost per QALY was Can $74,887 (95% confidence interval 21,985 128,671), but if the therapy impacted mortality, the survival effect resulted in a huge reduction; the incremental cost per QALY was Can $11,125 (95% confidence interval 8,710, dominated) compared with usual care. Sensitivity analyses showed that the cost per QALY was insensitive to the size of the mortality benefit, which supports the importance of this endpoint in determining the cost-effectiveness. To assess the cost-effectiveness of bronchodilator therapy in patients with COPD, Oostenbrink and colleagues developed a stochastic Markov model with a time horizon of 1 yr to compare the cost-effectiveness of tiotropium, ipratropium, and salmeterol in The Netherlands and Canada (5). Differences in effectiveness of the three inhaled bronchodilators were modeled as differences in transition and exacerbation rates, which were derived from six randomized studies of tiotropium. The model allowed for patients moving to less severe COPD severity stages. The number of quality-adjusted life-months was not different between the treatment groups. The cost-effectiveness acceptability frontier of exacerbations showed that tiotropium had the highest expected net benefit for ceiling ratios above 10 per exacerbation avoided in Canada and above 0 per exacerbation avoided in The Netherlands. In The Netherlands, the ceiling ratios for costeffectiveness were better because of the higher treatment costs and increased length of hospital stay in case of an exacerbation compared with Canada. A DYNAMIC, POPULATION-BASED MODEL Recently, a Dutch model (6) was used to determine the future burden of COPD in The Netherlands and to assess two smoking Figure 1. Simplified diagrammatic representation of the Markov model for chronic obstructive lung disease (COPD). The four disease states are represented by ovals and the transitions between them by arrows (4).

632 PROCEEDINGS OF THE AMERICAN THORACIC SOCIETY VOL 3 2006 cessation interventions. The model is both population-based and dynamic (Figure 2). It starts from the age-, sex-, and smokingclass distribution in the entire Dutch population and models the incidence of COPD depending on this distribution. The model follows patients with COPD over their course of disease, distinguishing four COPD severity stages according to GOLD, and never, current, and former smokers. Progression of COPD is defined as the annual decline in FEV 1% predicted, and depends on age, sex, smoking, and FEV 1% predicted. Estimates of progression were based on data from the Lung Health Study (9). Changes in demography and smoking influence the incidence, prevalence, progression, mortality, and costs of COPD. As a result, the model can estimate the impact of intervening at any stage in the course of COPD, from primary prevention to endstage COPD. This cannot be done with the COPD models described previously because they all follow a cohort of patients with COPD over time, mostly until they have all reached the death state. New cases of COPD do not enter these models. Hence, they are not able to estimate how COPD epidemiology and costs change at a country level. Over the 25-yr time frame, the model predicted an increase of 6, 3, 0.9, and 0.8 patients with mild, moderate, severe, and very severe COPD, respectively, per 1,000 inhabitants. When costs per patient were kept constant, Dutch COPD-related healthcare expenditures were predicted to increase from 280 million in 2000 to 495 million in 2025. The model was used to estimate the cost-effectiveness of minimal counseling by the general practitioner and intensive counseling plus bupropion compared to usual care, with the assumption that these interventions would be implemented for a period of 25 yr and taken up by 25% of all diagnosed smoking patients with COPD each year. When patients stop smoking, the model allows for a one-time increase in lung function and an ongoing slower decline in lung function, resulting in reduced mortality and better severity distribution. Overall, smoking cessation intervention resulted in some increase in prevalence because of decreased severity, and a gain in life-years, a gain in QALYs and cost savings, because of reduced movement to the more severe COPD states. The model demonstrated that minimal counseling is cost-saving and that intensive counseling plus bubropion is cost-effective at 14,000 per QALY. This Dutch COPD model is part of a large dynamic multistate life table model, the Chronic Disease Model, developed by the National Institute for Public Health and the Environment, which models the prevalence, incidence, and mortality of a large num- ber of different diseases in the Dutch population (10, 11). Because of this, the COPD model accounts for competing death risks. A disadvantage of the model is that it does not include COPD exacerbations. AN ECONOMIC MODEL BASED ON THE BOLD INITIATIVE The objective of the BOLD economic model, as part of the BOLD initiative (7), is to develop a health policy tool that can be used to calculate site-specific estimates of the current and future economic impact of COPD. It is an interactive, web-based model using information from the BOLD prevalence survey, local cost and population estimates, and disease progression estimates from the Framingham Heart Study. At present, 11 countries (China, Turkey, Austria, Poland, South Africa, Iceland, Norway, Germany, The United States, Canada, and The Philippines) have completed the prevalence portion of the BOLD project or are in the process of completing the prevalence survey. In addition, BOLD is associated with a similar project, PLATINO (Latin American Project for the Investigation of Obstructive Lung Disease), which was conducted in countries in Central and South America (12, 13). The structure of the BOLD model is similar to that of the COPD models described here, and includes 10 health states (Figure 3). The model does not allow transitions to less severe GOLD stages in patients with COPD. Both cohort and offspring data from the Framingham Heart Study are used to follow patients with COPD over a nearly 20-yr period to determine the rates of disease progression in males and females by smoking status. To demonstrate an application of the BOLD model, and to estimate the cost-effectiveness of ICS, data were used from the Inhaled Steroid Effects Evaluation in COPD study (14), an individual patient data meta-analysis of seven large randomized placebo-controlled trials investigating the effects of ICS in COPD. These data were applied to the BOLD model from the perspective of a US population. Three scenarios were assessed: (1) ICS has an equal benefit on mortality across both GOLD stages II and III COPD; (2) there is an ICS-related differential mortality benefit between GOLD stages II and III; (3) there is an ICS-related differential mortality benefit between GOLD stages II and III, and a benefit of reduced exacerbations. The simulations compared no ICS versus ICS in GOLD stages II and III for life, and the cost-effectiveness ratio for ICS was evaluated at Years 5 and 20 of follow-up. The results suggest that ICS treatment is a cost-effective alternative in patients with stages II and III COPD. In the first scenario, cost-effectiveness per LYG was $29,064 at 5 yr and Figure 2. Dynamic population-based model of COPD incidence and progression (6). FEV 1 % pred percentage of predicted FEV 1 ;RR relative risk. Decline depends on age, sex, smoking, and FEV 1 % pred. Figure 3. Structure of the Burden of Obstructive Lung Disease (BOLD) model.

Rutten-van Mölken and Lee: Economic Models in COPD 633 $13,043 at 20 yr. In the second scenario, these figures were $27,321 and $12,758, respectively. When ICS reduced the risk of mortality and the rate of exacerbations, as in the third scenario, the results were $1,581 per LYG at 5 yr and ICS was dominant (cheaper and more effective) at 20 yr. Some limitations of the analysis include a lack of quality-of-life data in the simulations, only a deterministic analysis was performed, and an inability to populate the model with estimates for each of the subpopulations modeled. UTILITY AND LIFE EXPECTANCY IN ESTIMATING QALYS IN COPD ECONOMIC MODELS With new treatment approaches suggesting opportunities to improve COPD management, one of the challenges that remains is improving COPD economic models to provide better QALY estimates. Currently, the COPD models lack good data on both the quality (defined by utility) and quantity (life expectancy) elements of QALYs. A limitation of current models lies in the accuracy of the projections of the impact of COPD interventions on life expectancy. Improved modeling of life expectancy will be particularly important when assessing new life-saving interventions. In addition, more sensitivity analyses around this parameter and validation of estimates against epidemiology evidence is required. However, at present, utility is the main driver in the calculation of QALYs in COPD economic models because interventions generally do not affect survival. Despite the importance of utility estimates for these models, robust estimates according to the severity of COPD remain undefined. In particular, reliable estimates for utility scores for exacerbations, which are especially important when assessing the effect of an intervention, are not available. This reflects the fact that, in clinical trials of COPD, assessments of disease status are usually made at scheduled clinic visits and, therefore, do not accurately capture the precise impact of exacerbations. Furthermore, the instruments that are currently employed to calculate utility (e.g., EQ-5D) are largely too insensitive. The lack of accepted utility values is reflected in the COPD models currently described. In the models, utility score for severity stages and exacerbation status vary. Three of the models (3, 5, 6) used an identical utility score for severity stages, but with respect to exacerbations, utility weights differed across the models with either a unique weight given for exacerbations, or a disutility factor associated with an exacerbation. The disutility value (the negative impact on quality of life associated with the exacerbation) was either a constant (2) or a function of the baseline utility weight for the patients (3, 5). Thus, a key area of ongoing research remains defining the extent and length of utility decrement during the disease course of COPD. CONCLUSIONS The number of COPD health economic models is growing. Comparing the five current models, all are Markov models that are generally similar with regard to the health states included. Disease progression, defined as decline in lung function, is the predominant movement through health states, although several models allow movement to less severe health states. Importantly, disease progression estimates used in each of the models are derived from various sources, with different populations and estimates of progression, which may subsequently impact the estimates of future costs associated with the disease. In addition, the model structures differ with respect to many elements, including the ability to model COPD incidence, the impact of smoking on disease progression, and the impact of exacerbation rates on disease progression. Ongoing research efforts in COPD economic models are directed toward constructing a model that is widely accepted and allows comparison of cost-effectiveness between treatments. The best available process to obtain crucial model parameters, such as disease progression and the impact of smoking, must be identified to provide some consensus on model structure and datasets. Because COPD is a systemic, multicomponent disease, future economic models will require incorporation of disease progression dependent upon more than just lung function. Value-ofinformation analysis of existing models, which involves estimating the value of collecting additional information on one or more crucial drivers of the cost-effectiveness in order to reduce or eliminate uncertainty, will also allow prioritization of future COPD research. Key priorities in future model development include better utility information and better definition of outcome measures. Questions that require better answers include: are exacerbations a good outcome measure in long-term analyses? Because interventions that increase survival will also accumulate more exacerbations, potentially offsetting reductions in the annual rate of exacerbation that the treatment may produce, are exacerbation-free months a relevant outcome parameter? In addition, when transition probabilities between COPD severity stages differ (even if only during the first year) and mortality rates depend on severity stage, can an effect on survival be modeled without any robust clinical evidence of a survival effect? Finally, should costs other than those that are COPD-related during the extended life be included, because QALYs are included? Through the BOLD initiative, the availability of site-specific epidemiologic and cost data will allow evaluation of the costeffectiveness for COPD interventions in a number of countryspecific settings. Ultimately, global epidemiologic and economic studies will allow for more uniform guidelines for the prevention and cost-effective treatment of patients with COPD (15). Conflict of Interest Statement : M.R.v.M. has received two research grants for modeling cost-effectiveness of COPD interventions from Boehringer Ingelheim. She has received a speakers fee from GlaxoSmithKline (UK 1,500) for presenting the current review during a COPD meeting in Rome on November 3 and 4, 2005. During 2005, the Institute for Medical Technology Assessment had a consultancy agreement with Boehringer Ingelheim and Altana Pharma (both for 50,000). T.A.L. participated as a speaker in the COPD meeting mentioned above, which was sponsored by GlaxoSmithKline, and received a US $116,646 grant for the BOLD project in 2005, which was sponsored by a consortium of pharmaceutical manufacturers. References 1. Briggs A, Sculpher M. An introduction to Markov modelling for economic evaluation. Pharmacoeconomics 1998;13:397 409. 2. Sin DD, Golmohammadi K, Jacobs P. Cost-effectiveness of inhaled corticosteroids for chronic obstructive pulmonary disease according to disease severity. Am J Med 2004;116:325 331. 3. Borg S, Ericsson A, Wedzicha J, Gulsvik A, Lundback B, Donaldson GC, Sullivan SD. A computer simulation model of the natural history and economic impact of chronic obstructive pulmonary disease. Value Health 2004;7:153 167. 4. Spencer M, Briggs AH, Grossman RF, Rance L. Development of an economic model to assess the cost effectiveness of treatment interventions for chronic obstructive pulmonary disease. Pharmacoeconomics 2005;23:619 637. 5. Oostenbrink JB, Rutten-van Molken MP, Monz BU, FitzGerald JM. Probabilistic Markov model to assess the cost-effectiveness of bronchodilator therapy in COPD patients in different countries. Value Health 2005;8:32 46. 6. Hoogendoorn M, Rutten-van Molken MP, Hoogenveen RT, van Genugten ML, Buist AS, Wouters EF, Feenstra TL. A dynamic population model of disease progression in COPD. Eur Respir J 2005; 26:223 233. 7. Buist AS, Vollmer WM, Sullivan SD, Weiss KB, Lee TA, Menezes AMB, Crapo RO, Jensen RL, Burney PGJ. The Burden of Obstructive Lung

634 PROCEEDINGS OF THE AMERICAN THORACIC SOCIETY VOL 3 2006 Disease Initiative (BOLD): rationale and design. COPD. Journal of Chronic Obstructive Pulmonary Disease 2005;2:277 283. 8. Calverley P, Pauwels R, Vestbo J, Jones P, Pride N, Gulsvik A, Anderson J, Maden C; Trial of Inhaled Steroids and Long-Acting 2 Agonists Study Group. Combined salmeterol and fluticasone in the treatment of chronic obstructive pulmonary disease: a randomised controlled trial. Lancet 2003;361:449 456. 9. Anthonisen NR, Connett JE, Kiley JP, Altose MD, Bailey WC, Buist AS, Conway WA Jr, Enright PL, Kanner RE, O Hara P. Effects of smoking intervention and the use of an inhaled anticholinergic bronchodilator on the rate of decline of FEV 1 : the Lung Health Study. JAMA 1994;272:1497 1505. 10. Rutten-van Molken MP, Postma MJ, Joore MA, van Genugten ML, Leidl R, Jager JC. Current and future medical costs of asthma and chronic obstructive pulmonary disease in The Netherlands. Respir Med 1999;93:779 787. 11. Feenstra TL, van Genugten ML, Hoogenveen RT, Wouters EF, Ruttenvan Molken MP. The impact of aging and smoking on the future burden of chronic obstructive pulmonary disease: a model analysis in The Netherlands. Am J Respir Crit Care Med 2001;164:590 596. 12. Menezes AM, Victora CG, Perez-Padilla R. The PLATINO project: methodology of a multicenter prevalence survey of chronic obstructive pulmonary disease in major Latin American cities. BMC Med Res Methodol 2004;4:15. 13. Menezes AM, Perez-Padilla R, Jardim JR, Muino A, Lopez MV, Valdivia G, Montes de Oca M, Talamo C, Hallal PC, Victora CG; the PLATINO Team. Chronic obstructive pulmonary disease in five Latin American cities (the PLATINO study): a prevalence study. Lancet 2005;366:1875 1881. 14. Sin DD, Wu L, Anderson JA, Anthonisen NR, Buist AS, Burge PS, Calverley PM, Connett JE, Lindmark B, Pauwels RA, et al. Inhaled corticosteroids and mortality in chronic obstructive pulmonary disease. Thorax 2005;60:992 997. 15. Nowak D, Berger K, Lippert B, Kilgert K, Caeser M, Sandtmann R. Epidemiology and health economics of COPD across Europe: a critical analysis. Treat Respir Med 2005;4:381 395.