Systematic review of models used in economic analyses in moderate-to-severe asthma and COPD

Abstract

Background:

Respiratory diseases exert a substantial burden on society, with newer drugs increasingly adding to the burden. Economic models are often used, but seldom reviewed.

Purpose:

To summarize economic models used in economic analyses of drugs treating moderate-to-severe/very severe asthma or chronic obstructive pulmonary disease (COPD).

Methods:

This study searched Medline and Embase from inception to the end of February 2015 for cost-effectiveness/utility analyses that examined at least one drug against placebo, another drug, or other standard therapy in asthma or COPD. Two reviewers independently searched and extracted data with differences adjudicated via consensus discussion. Data extracted included model used and its qualities, validation methods, treatments compared, disease severity, analytic perspective, time horizon, data collection (pro- or retrospective), input rates and sources, costs and sources, planned sensitivity analyses, criteria for cost-effectiveness, reported outcomes, and sponsor.

Results:

This study analyzed 53 articles; 14 (25%) on asthma and 39 (75%) COPD. Markov models were commonly used for both asthma and COPD-related economic evaluations. Relatively few studies validated their model. For asthma-related studies, 10 examined inhaled corticosteroids and nine studied omalizumab. Placebo or standard therapy was the comparison in 11 studies and active drugs in the remainder.

Conclusions:

Few studies include validation of their models. Furthermore, controversy concerning some results was uncovered in this study, which needs to be avoided in the future.

Keywords::

• Asthma
• COPD
• Economic evaluations
• Validation

Introduction

Respiratory diseases such as chronic obstructive pulmonary disease (COPD) and asthma are persistent, with little chance of complete remission. As well, both diseases are widespread, occurring throughout the world. The 2010 Global Burden of Disease study undertaken by the World Health Organization estimated that more than 9.6% of the world’s population have either COPD or asthma, affecting about 329 million and 334 million people, respectively1.

Both COPD and asthma negatively impact patients in terms of health status, duration, and quality-of-life, as well as costs2. These ailments also substantially impact society with respect to resource utilization (e.g., medical care, hospitalization, drugs and other treatments), and associated costs. A recent study from France estimated it cost €9382 (57% direct costs) for each patient with COPD3. In the US, Ford et al.4 predicted that the burden of COPD in that country would rise to $49 billion by 2020. O’Neill et al.5 estimated that the cost of treating refractory asthma in the UK was between £2912 and £4217 per annum per patient. Thus, the costs are very large and increasing, which causes concern amongst payers and formulary managers.

Novel therapies are costly to develop. Estimates of the cost for the development and launching of a new drug run as high as USD $1.8 billion6^,7. At the same time, success rates for new drugs have been declining and, thereby, returns on investment have been negatively impacted6^,8. In order to recoup development expenses, the costs of developing novel therapeutics (both successful and unsuccessful) are usually passed onto the consumer. At the point of purchase, the consumer determines whether there is value for money in purchasing the novel therapeutics.

Consequently, there has been an emphasis on economics, which assesses value for money9. Pharmacoeconomic analyses apply the concepts of economics to the comparison of drugs10. Interest in this area is high, as evidenced by recent reviews of the pharmacoeconomics of both asthma11^,12 and COPD13. However, these reviews have focused mainly on either the results of the analyses11 or the quality of the reporting12^,14. A review focusing on the models used to arrive at the pharmacoeconomic outcomes has not been undertaken to date.

There are three basic models for assessing the pharmacoeconomics of drugs; prospective, predictive and retrospective15. Prospective models are often associated with clinical trials comparing drugs or treatments, which allows the prospective gathering of data. The predictive model is usually referred to as a decision analytic analysis, but also includes Markov models and discrete event simulations. These types predict the average outcome, given a set of inputs. The retrospective model uses patient charts or databases, using data that have already been collected. In general, prospective models focus on short-term acute treatments and usually cannot be extrapolated to long-term therapy15.

Despite reviews of the economics of drugs in asthma or COPD, the models used have seldom been their focus, as evidenced by a search of the literature. While there was one recent review identified by Kirsch13 that examined only Markov models in COPD, that study reviewed only smoking cessation interventions, rather than drugs to treat asthma or COPD. We, therefore, undertook the present research to review the types of models that have been utilized in cost-effectiveness analyses of drugs used to treat asthma and COPD since the introduction of biologic treatment.

Methods

This research was guided by the PRISMA statement for reviews16. Criteria for acceptable studies were developed a priori, as was the search strategy. All steps of the process were done in duplicate by independent reviewers.

Eligibility criteria

Studies eligible for inclusion in this study were economic evaluations related to pharmacotherapy interventions for at least moderate-to-severe asthma or chronic obstructive pulmonary disease or both. The target population consisted of patients having severe disease (or worse); however, patient samples often presented mixed severities. We, therefore, allowed patients having moderate-to-severe disease or perhaps a small proportion with milder forms. For this analysis, only the severity of the disease itself was considered; treatments exclusively used for acute exacerbations, whether severe or not, were not accepted. Asthma was considered sufficiently severe if identified as such by the authors or if patients were symptomatic on >800 µg/day of corticosteroids or if their average peak expiratory volume (PEV) was <70% (i.e., below the midpoint of the range for the moderate disease). COPD was considered sufficiently severe if the majority of patients were classified as having GOLD-3 or GOLD-4 disease or if the average forced expiratory volume (FEV) was <65% (i.e., below the midpoint of the range for moderate disease). We accepted studies dealing with either adults or pediatrics. Patients could not be immunocompromised or have any other concomitant respiratory diseases. Either inpatient or outpatient populations or both could be examined. We also accepted dynamic progression models of COPD that included disease severity that ranged from mild-to-severe or very severe disease as well as death, such as Markov models or discrete event simulations.

There must have been two or more therapeutic options compared, at least one of which was a drug. Acceptable types of analyses were those that combined both costs and outcomes into an incremental cost-effectiveness ratio (ICER). That is, we included cost-effectiveness and cost-utility analyses, but excluded cost-minimization and cost-consequence analyses.

The first biologic to be licensed (for any indication) was infliximab, in 199817. Its original indication was Crohn’s disease, but the potential for using biologics in treating respiratory diseases was quickly recognized18. Trials were soon undertaken in patients with COPD19. It may be postulated that the era of biologic treatment for asthma began with the Australian approval of omalizumab in 20021 7. However, the process began in the year 200020. The authors assumed that economic analyses of biologic treatment related to asthma would not have been published prior to the year 2000. As such, only full text, peer-reviewed articles published between 2000–2015 were considered.

Information sources

The search engines used to identify relevant articles were: Embase 1980–2015 week 7, and Medline 1946 to February week 2, 2015. Within these databases, searches were limited to the above-mentioned dates. Retrieved articles were hand searched for relevant studies not identified by search engines. Furthermore, PubMed was searched for more recent studies not yet included in the other search engines used.

Search and study selection

Two reviewers independently searched the databases using key words ‘cost effectiveness.mp OR exp cost-benefit analysis’, ‘costs and cost analysis OR cost utility.mp’, ‘willingness to pay.mp’, ‘asthma OR COPD’, and ‘full text’. Results were compared first after examining the titles and abstracts (first filtration) and again after retrieving full text articles (second filtration). At each stage, discrepancies were adjudicated via consensus discussion.

Data items

Data collected included author names, year of publication, country, sponsor, model type and structure, validation of the model, disease and its severity, patients examined, data collection (i.e., prospective vs retrospective), treatments compared, time horizon(s), analytic perspective(s), clinical rates used and their sources, resources costed and sources of utilization rates, currency and year of costing, discount rate for costs and outcomes, criterion used to determine cost-effectiveness, outcomes examined, sensitivity analyses planned, reported costs for each treatment, reported outcomes, incremental cost-effectiveness ratios (ICERs), sensitivity analyses reported, and conclusions made by the authors. Items were recorded in a spreadsheet and results were then compared between data extractors. All cases of difference were settled though discussion.

Data analysis

Models were defined in three broad categories: trial based (prospective), decision models (predictive), and models based on patient data records (retrospective). Three types of trial models were possible, including (1) the dedicated economic RCT which was designed specifically for an economic analysis, (2) the piggyback trial, where an economic arm was added to the clinical protocol of the RCT and all data were collected prospectively, and (3) the retrospective RCT, where an existing RCT was used and economic data were fitted post-hoc. Four types of decision models were considered. They included (1) the classical decision tree, (2) the Markov model (and all of its variants), (3) patient level simulations (e.g., discrete event simulations), and (4) spreadsheet models where calculations were done on a spreadsheet as opposed to dedicated software-based models. Finally, patient data based models comprised three types: (1) administrative databases, which were designed for other purposes such as the payment of claims for healthcare services, (2) patient registries designed to prospectively collect patient data for specific diseases, and (3) patient charts, which could be either paper or electronic. Data were tabulated and analyzed descriptively.

Results

The search process is depicted in Figure 1. We were able to use data from 53 articles, including 14 (25%) in patients with asthma12^,21–33 and 39 (75%) in patients with COPD33–71. No papers were found that examined patients having both diseases concomitantly.

Figure 1. Literature search tree.

Among the 14 papers on asthma, eight were from Europe, five (36%) from North America, and one (7%) from South America. Nine (64%) were sponsored by the pharmaceutical industry, two (14%) by non-government foundations, and three (21%) did not mention funding. Table 1 summarizes the models that have been used. In asthma patients, one was trial based (retrospectively modeled an existing RCT), 11 were decision models (10 Markov and one spreadsheet decision analysis), and two were based on electronic patient data records.

Table 1. Models used in economic analyses of severe asthma and COPD.

Model type	Sub-type	Disease examined
		Asthma	COPD	Total
Trial based	Prospective economic trial	0	0	0
	RCT with economic data collection (piggyback)	0	6	6
	Existing RCT	1	1	2
	Sub-total	1	7	8
Decision model	Markov	10	10	22
	Decision tree	0	1	1
	Patient level simulation	0	1	1
	Spreadsheet	1	2	3
	Sub-total	11	16	27
Patient data records	Administrative database	0	0	0
	Patient/disease registry	0	0	0
	Chart review/electronic medical records	2	1	3
	Sub-total	2	0	2
Total		14	39	53

Table 2 provides details of the models, describing their structures and characteristics. Among the Markov models used in asthma, only three reported validation of some sort, but most often providing few details. All of the studies collected data retrospectively. Ten of the studies examined inhaled corticosteroids and nine studies omalizumab. Placebo or standard therapy was the comparison in 11 studies and active drugs in the remainder. Six focused on adults only, seven included adolescents as well as adults, and one was restricted to children <18. The perspective of the analysis was societal in three, the healthcare system in nine, and third party payer in two. Two studies used a time horizon of 12 weeks, seven used 1 year, one used 10 years and six examined over a lifetime. Two studies employed two different time horizons.

Table 2. Model structure and characteristics in economic analyses of severe asthma.

Reference (Year)	Country	Sponsor	Model structure	Validation	Patients/severity	Data collection	Treatments compared	Time horizon	Perspective
Brown et al.21 2007	Canada	Novartis	Markov model25 with five states: (1) day-to-day asthma (no exacerbations), (2) clinically significant non-severe exacerbations, (3) clinically significant severe exacerbations, (4) all cause death, (5) asthma related death Cycle length = 2 weeks	not reported	adolescents and adults (12–73) severe persistent allergic asthma despite ICS + LABA	retrospective	omalizumab add-on for 5 years standard therapy	Lifetime	Canadian healthcare system
Campbell et al.22 2010	USA	Novartis	Markov model with 6 states: (1) chronic asthma, (2) exacerbation (oral steroids), (3) exacerbation (emergency room), (4) exacerbation (hospitalization), (5) asthma mortality, (6) other mortality 2 week transitions	not reported	average age = 40, 60% females moderate-to-severe persistent asthma inadequate control with ICS	retrospective	omalizumab add-on for 5 years standard therapy	1 year lifetime	US payer
Dal Negro et al.23 2011	Italy	not stated	Pre–post comparison using data from a patient database	not reported	23 patients aged 26–45 with adequate data persistent atopic asthma resistant to common therapies	retrospective	12 months pre-omalizumab 12 months with omalizumab	1 year	Regional Health System
Dal Negro et al.24 2012	Italy	Novartis	Pre–post comparison using data from a patient database	not reported	16 patients >12 years old, 50% females persistent atopic asthma severe, difficult to treat	retrospective	12 months pre-omalizumab 36 months with omalizumab	1 year	Regional Health System
Dewilde et al.25 2006	Sweden	Novartis	Markov model with five states: (1) daily symptoms, (2) clinically significant non-severe exacerbations, (3) clinically significant severe exacerbations, (4) all cause death, (5) asthma related death Cycle length = 2 weeks	not reported	average age = 40, 67% female severe persistent asthma	retrospective	omalizumab add-on for 5 years standard therapy	lifetime	Societal (Sweden)
Gerzeli et al.26 2012	Italy	Chiesi Farmaceutici	Markov model with five states: (1) successful control, (2) sub-optimal control, (3) outpatient managed exacerbation, (4) inpatient managed exacerbation, (5) death Cycle length = 1 week	not reported	Adults aged 16–65 moderate-to-severe persistent asthma	retrospective	betamethasone/formoterol fluticasone propionate/salmeterol	12 weeks 1 year	Italian National Health Service
Miller et al.27 2007	Canada	AstraZeneca	Spreadsheet model	not reported	patients aged 45 (range 12–84), 59% female moderate-to-severe asthma	retrospective	betamethasone/formoterol fluticasone propionate/salmeterol	1 year	Health Care System societal (Canada)
Norman et al.12 2013	UK	NICE	Markov model with five states: (1) day-to-day symptoms, (2) clinically significant non-severe exacerbations, (3) clinically significant severe exacerbations, (4) all cause death, (5) asthma related death Cycle length: first = 16 weeks, others = 3 months	developed by one analyst checked by another internal validation	adults + adolescents ≥12 and children 6–11 years old Severe exacerbations despite daily ICS + LABA Step 4 or 5 treatment	retrospective	omalizumab add-on for 5 years standard therapy	lifetime	UK National Health Service
Novartis28 2007	UK	Novartis	Markov model with five states: (1) day-to-day symptoms, (2) clinically significant non-severe exacerbations, (3) clinically significant severe exacerbations, (4) all cause death, (5) asthma related death Cycle length: first = 16 weeks, others = 3 months	not reported	adults + adolescents ≥12 years old severe persistent IgE mediated allergic asthma	retrospective	omalizumab add-on for 5 years standard therapy	lifetime (40 years)	UK National Health Service
Obaand Salzman29 2004	USA	none stated	modeled existing RCTs	not reported	males or females 12–75 years old patients with allergic asthma symptomatic despite inhaled corticosteroids	retrospective	omalizumab SC for 7 months placebo	52 weeks	Third party payer
Price and Briggs31 2002	UK	GlaxoSmithKline	Markov model with five states: (1) successful control, (2) hospital-managed exacerbation, (3) primary-care managed exacerbation, (4) sub-optimal control, and (5) treatment failure Cycle length: 1 week	not reported	adults and children >12 years old moderate-to-severe asthma treated with inhaled drugs	retrospective	fluticasone propionate/salmeterol fluticasone propionate	12 weeks	UK National Health Service
Rodríguez-Martínez et al.30 2013	Colombia	none stated	Markov model with three states: (1) no symptoms, (2) sub-optimal control, (3) asthma exacerbation Cycle length: 1 week	not reported	children <18 years old persistent asthma	retrospective	beclomethasone budesonide fluticasone ciclesonide	12 months	Compulsory Insurance Plan of Colombia
van Nooten et al.32 2013	Netherlands	Novartis	Markov model with five states: (1) daily symptoms, (2) clinically significant non-severe exacerbations, (3) clinically significant severe exacerbations, (4) all cause death, (5) asthma related death Cycle length = 2 weeks	validated by Brown et al.21 and Dewilde et al.25	adults aged 40 uncontrolled allergic asthma	retrospective	omalizumab add-on for up to 5 years standard therapy	lifetime	societal
Wu et al.33 2007	USA	National Heart, Blood and Lung Institute	Markov model with six states: (1) chronic, (2) acute, (3) urgent care visits, (4) emergency room visits, (5) hospitalization, and (6) death Cycle length = monthly	used in at least two previous analyses	adults >18 severe persistent asthma	retrospective	omalizumab add-on standard therapy	10 years	Societal (all direct costs)

ACT, Asthma Control Test; ICS, inhaled corticosteroids; LABA, long-acting beta-agonists.

Table 3 summarizes the inputs into the models and their sources for asthma-related pharmacoeconomic studies. Eleven studies used RCTs as their primary source for clinical data and three used patient data. In seven analyses, discounting was not done because the time horizon did not exceed 1 year. In all of the others, discounting was done for both costs and outcomes. Cost were discounted from 3–5% and outcomes from 1–3.5%. Ten included only direct costs and four included both direct and indirect costs. Ten out of 12 reported sensitivity analyses, and two did not perform any.

Table 3. Model inputs in economic analyses of severe asthma.

Reference (Year)	Costs	Currency and year	Discounting	Input rates	Data source	Items costed	Data sources	Criterion	Planned sensitivity analyses
Brown et al.21 2007	Direct	CAD converted to euros (year not stated)	Costs: 5% QALYs: 5%	exacerbation rates resource use utilities death rates	ETOPA39, INNOVATE40 ETOPA39 ETOPA39, Price et al.41 Lowhagen et al.42, Statistics Canada43	exacerbations drugs physician visits hospitalization omitted omalizumab administration costs	not stated Ontario Drug Benefit Ontario Schedule of Benefits CIHI	ICER = €35,000/QALY (CAD $50,000)	probabilistic, 1-way: (a) no discounting, (b) 3% discount on costs + outcomes, (c) 5-year time horizon, (d) including administration costs, (e) mortality 0%, (f) mortality 2.48%
Campbell et al.22 2010	Direct	USD, 2008	Costs: 3% QALYs: 3%	exacerbation rates utilities death rates	Bousquet et al.44, Humbert et al.40 Humbert et al.40, Tsuchiya et al.46, Price et al.41 Sullivan et al.47	care drugs hospitalization	Ayres et al.45 Bousquet et al.44 MarketScan database	not stated	probabilistic (5000 iterations), 1-way on all inputs using 95% confidence limits 1-way analyses on costs, all rates, response definition
Dal Negro et al.23 2011	Direct	euros, 2008	not applicable	hospitalizations drug use physician visits	patients charts patients charts patients charts	hospitalizations drugs physician visits	DRG (Regional Tariff), National Hospital Discharge Database Italian Drugs Formulary, Off-Patent Drugs Reference Price System Regional Tariffs, DYSCO48	$27,000–$165,000/QALY	not stated
Dal Negro et al.24 2012	Direct	euros, 2011	not applicable	hospitalizations drug use physician visits utilities	patients charts patients charts patients charts	hospitalizations drugs physician visits	DRG (Regional Tariff), National Hospital Discharge Database Italian Drugs Formulary, Off-Patent Drugs Reference Price System Regional Tariffs, DYSCO48	$27,000–$165,000/QALY	not stated
Dewilde et al.25 2006	Direct Indirect	Swedish kronor converted to euros, 2006	Costs: 3% Outcomes: 3%	exacerbation rates utilities death rates	INNOVATE40 INNOVATE40, Price et al.41 Lowhagen et al.42, Swedish life tables (No reference provided)	exacerbations hospital costs drugs time lost	Swedish Association of Local Authorities Swedish Association of Local Authorities Pharmaceutical Benefits Board Statistics Sweden	€53,384 (500,000 SEK)	Probabilistic sensitivity analyses (5000 simulations) 12 1-way sensitivity analyses varying rates for mortality, discount, utilities, time horizon, healthcare costs, costs for extra life years gained
Gerzeli et al.26 2012	Direct	euros, 2010	not applicable	transition rates resource use hospitalization mortality utilities	ICAT SE49 Priceand Briggs31, GOAL50, experts Ministry of Health de Marco et al.51 literature (average from 5 studies)	resource use hospitalization	Ministry of Health DRG (Regional Tariff)	not stated	Probabilistic sensitivity analyses (100,000 simulations) 1-way and bivariate on costs, transition probabilities (minimum/maximum values, threshold analysis)
Miller et al.27 2007	Direct Indirect	CAD, 2005	not applicable	clinical rates resource use	Vogelmeier et al.52 Vogelmeier et al.52, expert panel	drugs medical nursing care hospitalizations lost wages	Ontario Drug Benefit Ontario Schedule of Benefits Nursing Services Study Alberta Health Costing Statistics Canada	not stated	1-way on exacerbation rate, drugs costs, wages, hospitalization cost
Norman et al.12 2013	Direct	UK pounds, 2010	Costs: 3.5% Outcomes: 3.5%	exacerbation rates resource use utilities death rates	INNOVATE40, Lanier and Marshall53 INNOVATE40, Lanier and Marshall53 Bousquet et al.44, Lloyd et al.54 de Vries55, UK Life Tables	resource use	Humbert et al.40, Lanier and Marshall53, NHS Reference Costs	£30,000	3 scenarios: patients hospitalized, patients on maintenance oral steroids, patients with ≥3 exacerbations rates of exacerbation, effectiveness, mortality, utilities, horizon, dosing, costs
Novartis28 2007	Direct	UK pounds, 2006	Costs: 3.5% Outcomes: 3.5%	exacerbation rates resource use	INNOVATE40, ETOPA39 INNOVATE40, ETOPA39	resource use	Humbert et al.40, Lanier and Marshall53, NHS Reference Costs	£30,000	1-way on rates, time horizon, treatment duration
Oba and Salzman29 2004	Direct	USD, 2003	not applicable	clinical rates utilities	Busse et al.56, Soler et al.57 Busse et al.56, Soler et al.57	drugs hospitalization	Drug Topics published studies	not stated	best/worst case for costs, for each outcome
Price and Briggs31 2002	Direct	UK pounds, 2000	not applicable	clinical rates resource use	Kavuru et al.58 Kavuru et al.58	drugs attacks	MIMS Hoskins et al.59	not stated	Probabilistic (1000 simulations)
Rodríguez-Martínez et al.30 2013	Direct	UK pounds, 2007	not applicable	clinical rates resource use utilities	24 published papers asthma guidelines, local experts survey of patients’ parents	drugs hospital care	National manual of drug prices Hospital records	£9803.96/QALY (3× GDP in 2008)	Probabilistic (10,0000 simulations), 1-way, 2-way, and multi-way analyses
van Nooten et al.32 2013	Direct Indirect	euros, 2010	Costs: 4% Outcomes: 1.5%	clinical rates resource use	observational registry observational registry	care and drugs	Dutch cost lists, published studies	not stated	rates of exacerbation, death; cost of drug, exacerbations time horizon, workdays lost, retaining non-responders
Wu et al.33 2007	Direct Indirect	USD, 2005	Costs: 3% Outcomes: 3%	clinical rates hospitalization utilities resource use	Paltiel et al.60, Walker et al.61 Busse et al.56, Corren et al.62 Niebauer et al.63 Eight published studies	drugs care productivity	Drug Topics Red Book eight published studies US Bureau of Labor	no consensus	1-way analyses on costs and rates

CAD, Canadian dollars; CIHI, Canadian Institute for Health Information; DRG, Diagnosis Related Group; SEK, Swedish kronor; USD, United States dollars.

Results of the models used in the economic analysis of drugs used for severe asthma appear in Table 4. Costs were reported in either US dollars, Canadian dollars, British pounds or Euros. Results varied widely; for example Brown et al.21 and Dal Negro et al.113 concluded that omalizumab compared to standard care was cost-effective, while Campbell et al.22 reported that omalizumab was not cost-effective compared to standard care.

Table 4. Model outputs from economic analyses of severe asthma.

		Treatments	Outcomes
		compared	Cost	Clinical	Amount	ICER	Sensitivity analyses	Conclusions
Brown et al.21 2007	lifetime	omalizumab* standard care	€61,257 €27,403	QALYs QALYs	7.57 6.49	€31,209	CI: €27,739–€40,840 69.7% of ICERS <€35,000 Range of 1-way results: €23,762–€66,443	omalizumab is cost-effective
Campbell et al.22 2010	1 year lifetime	omalizumab standard care omalizumab* standard care	USD $23,000 USD $3600 USD $174,500 USD $83,400	QALYs hospitalizations ER visits steroid exacerbations QALYs hospitalizations ER visits corticosteroid bursts QALYs QALYs	0.72 0.03 0.03 0.83 0.66 0.06 0.06 1.30 14.19 13.87	USD $306,200 $646,667 $646,667 $41,277 USD $287,200	1-way analyses: sensitive to changes in utilities, exacerbation mortality, omalizumab price, exacerbation rates, response definition none <$200,000/QALY	adding omalizumab to usual care improves QALYs at an increase in direct medical costs not cost-effective
Dal Negro et al.23 2011	1 year (pre/post)	omalizumab standard care	€586.71/month €241.49/month	QALYs FEV ACT QALYs FEV ACT	0.53 57.61 11.87 0.70 75.46 19.4	€25,983 €21.9/FEV% €57.3/ACT point	not done	cost-effective
Dal Negro et al.24 2012	1 year (pre/post) (averaged over 3 years post)	omalizumab standard care	€8038 €2869	QALYs hospitalizations ER visits QALYs hospitalizations ER visits	0.78 0 0.25 0.56 0.94 0.69	€23,880 €5499 €11,748	not done	cost-effective
Dewilde et al.25 2006	lifetime	omalizumab* standard care	€95,456 €52,702	QALYs QALYs	12.357 11.595	€56,091	Probabilistic ICER = €57,961 (€31,328–€120,552) 1-way: sensitive to exacerbation mortality rate, time horizon, discount rate 1-way ICERS: €39,675–€131,130	omalizumab provides cost offsets, improves quality-of-life and may have an attractive ICER in treating severe allergic asthma
Gerzeli et al.26 2012	12 weeks 1 year	B/F FP/S B/F FP/S	€175.03 €188.70 €745.48 €804.60	QALW WSC QALW WSC QALW WSC QALW WSC	9.737 6.381 9.696 6.015 42.448 42.279 31.152 29.628	dominant dominant dominated dominated dominant dominant dominated dominated	68% of simulations for QALW favoured B/F (dominated 63%) 64% of simulations for WSC favored B/F (dominated 59%) minor changes in sensitivity analyses not reported	cost-effective cost-effective
Miller et al.27 2007	1 year	B/F FP/S	HS: CAD $1315 Soc: CAD $1538 HS: CAD $1541 Soc: CAD $1854	serious exacerbations serious exacerbations serious exacerbations serious exacerbations	0.24 0.24 0.31 0.31	dominant dominant	B/F dominated in all but 1 sensitivity analysis where it was equivalent	cost-effective
Norman et al.12 2013	lifetime lifetime	omalizumab (≥12 years)* standard care omalizumab (6–11 years)* standard care	£72,938 £33,218 £92,497 £40,218	QALYs QALYs QALYs QALYs	14.13 13.66 17.39 16.72	£83,822 £78,009	Price = patient access scheme discount ICER = £83,822 with omalizumab list price aged ≥12 ICER = £78,009 with omalizumab list price aged 6–11 1-way ICERs: £46,029–£112,033 Hospital group ≥12 years: ICER = £31,782 Hospital group 6–11 years: ICER = £30,109 Oral steroid group ≥12 years: ICER = £34,386 ≥3 exacerbation group ≥12 years: ICER = £53,087 ≥3 exacerbation group 6–11 years: ICER = £48,537 1-way ICERs: £50,319– £98,688	not cost-effective not cost-effective
Novartis28 2007	lifetime (40 years)	omalizumab (≥12 years)* standard care				£30,600	£26,500 for the high-risk hospitalization group £21,700 using data from ETOPA study Sensitivity analyses: average ICER: £31,700 (£23,200–£48,200) Sensitivity using 0% mortality ICER: £73,200 NICE estimate: £38,900; 23.6% probability of being cost-effective	cost-effective not cost-effective
Oba and Salzman29 2004	52 weeks	omalizumab for 7 months placebo omalizumab for 7 months placebo omalizumab for 7 months placebo		patients with 0.5 point decrease in AQLQ patients with 1.5 point decrease in AQLQ SCD		$378 $517 $388–$523	CI: $146–$919 Best case: $280 ($107–$684) CI: $200–$1259 Best case: $383 ($146–$937) Sensitivity: $148–$1273	incremental cost far exceeds cost/SCD with other drugs Should restrict to very specific groups with severe asthma
Price and Briggs31 2002	12 weeks	FP fluticasone propionate	£15.77 £11.83	SCW	66% 47%	£20.83/SCW	95% CI for ICER: −£64.94 (dominant) to £112.66 Sensitivity: FP/S has 10–6% more SFWs	cost-effective at a willingness-to-pay of £45/SFW 80% of the time
Rodríguez-Martínez et al.30 2013	12 months	beclomethasone fluticasone ciclesonide budesonide	£106.16 £231.19 £270.15 £309.27	QALY QALY QALY QALY	0.9262 0.9325 0.9051 0.8999	£19,835.28 dominated dominated	exacerbation costs most influential	beclomethasone is cost-effective when WTP <£21,129
van Nooten et al.32 2013	lifetime	omalizumab* omalizumab (undiscounted) standard care standard care undiscounted	€151,619 €227,688 €95,754 €161,499	QALY QALY	25.74 22.93	€38,371 €38,528	Probabilistic: 75% of ICERs <€44,000/QALY 1-way sensitivity analyses (14 reported): all but one ICER €32,375–€80,332	cost-effective for this country
Wu et al.33 2007	10 years	omalizumab standard care	$131,000 $16,000	QALM	80.01 78.33	$821,000	Also reported an ICER of $120/SFD. Sensitivity: changing exacerbation rates, ICERS = $727,000–$853,000/QALY changing event rates, ICERs = $207,000–$1.42 million/QALY changing costs: ICER = $100,000/QALY	not cost-effective

B/F, betamethasone + formoterol; CI, 95% confidence interval; ER, emergency room; FP, fluticasone propionate; FP/S, fluticasone propionate + salmeterol; HS, Healthcare System perspective; QALW, quality adjusted life week; QALM, quality adjusted life month; QALY, quality adjusted life year; SCD, symptom controlled day; SCW, symptom controlled week; SFD, symptom free day; Soc, societal perspective; WSC, weeks with symptom control.

*Omalizumab given for the first 5 years only.

There were 39 studies that employed economic models to assess drugs used in severe COPD. Table 5 describes model characteristics in economic analysis of severe COPD. Economic analyses were carried out in Australia (1), Belgium (3), Canada (7), Denmark (1), Finland (1), Germany (1), Greece (1), Italy (1), the Netherlands (3), Norway (1), Singapore (1), Spain (1), Sweden (3), Switzerland (1), the UK (9), and the US (7). Most of the studies (30/39, 77%) were sponsored by pharmaceutical companies. The remainder of the studies did not explicitly state source of funding, but did not include authors who were affiliated with either a pharmaceutical or contract research firm. Approximately half of the articles (19/39, 49%) did not mention any validation process. The shortest time period used to analyze data was 16 weeks and the longest was a lifetime. Thirteen of the 39 studies used a time horizon of between 3–5 years, and 10 used a time horizon of 1 year. Few studies (five) approached their analysis from a societal perspective, while most used a third party payer perspective. Most studies (67%) collected data retrospectively, with 13 of the 39 (33%) studies collecting data prospectively. A Markov structured model was used in 19 studies. There were eight studies that were based on randomized controlled trials.

Table 5. Model structure and characteristics in economic analyses of severe COPD.

Reference (Year)	Country	Sponsor	Model structure	Validation	Patients/severity	Data collection	Treatments compared	Time horizon	Perspective
Ayres et al.45 2003	UK	GlaxoSmithKline	economic RCT based on ISOLDE64	not needed	adults aged 49–75 moderate-to-severe COPD	prospective	(1) FP (2) placebo	6 months	UK National Health Service Societal
Briggs et al.65 2006	UK	GlaxoSmithKline	Modeled ISOLDE Trial64	not needed	adults aged 64, 75% males moderate-to-severe COPD	prospective	(1) FP (2) placebo	3 years	UK National Health Service
Briggs et al.66 2010	UK	GlaxoSmithKline	economic RCT based on TORCH67 Regression model using TORCH67 data	not needed	adults average age 65, 76% males moderate-to-severe COPD FEV1 <60%	prospective	(1) FP/S (2) FP (3) S (4) placebo	3 years	Health Service/Third party payer
Chuck et al.68 2008	Canada	GlaxoSmithKline	Markov model with 3 states: Stage 1 (FEV ≥50%), Stage 2 (FEV 35–49.9%, Stage 3(FEV <35%); also exacerbations (mild, moderate, severe) Cycle length: 3 months	not stated	mostly severe COPD (FEV1 <60%)	retrospective	(1) LABA + IC (all stages) (2) LABA + IC (Stages 2–3) (3) LABA + IC (Stage 3 only) (4) IC only	3 years lifetime	Health System
Collet et al.69 2001	Canada	Jouveinal Inc.	economic RCT (piggyback)	peer reviewed by FRSQ (a Government Agency)	adults aged 70, 66% males mostly severe COPD	prospective	(1) OM-85 BV (immunostimu-lating agent) (2) placebo	6 months	Health System
Dalal et al.70 2010	USA	GlaxoSmithKline	spreadsheet model based on pooled data from 2 RCTs	not stated	adults >40, FEV <50% severe COPD	retrospective	(1) FP/S (2) S	1 year	Third party payer
Dal Negro et al.35 2007	Italy	GlaxoSmithKline	Markov chain with 5 states: COPD Stages 1–4 and death Cycle length: 1 year	comparison with another similar study in Italy	adults severe (Stage 3) and very severe (Stage 4)	retrospective	(1) FP/S (2) B/F (3) FP (4) S (5) control (standard therapy)	lifetime	National Health Service
Earnshaw et al.36 2009	USA	GlaxoSmithKline	Markov model with 4 states: (1) moderate, (2) severe, (3) very severe, (4) death Also mild, moderate, severe exacerbations within states.		adults 40–80, FEV <60%	retrospective		lifetime	Third party payer
Gani et al.71 2010	UK	Boehringer Ingelheim & Pfizer	Markov model 3 states: (1) Mild, (2) moderate, (3) severe	Oostenbrink et al.72	Age not specified: Mild-to-severe COPD	retrospective	(1) tiotropium (2) ipratropium (3) salmeterol (4) placebo	1 year	National Health Service
Hettle et al.73 2012	UK and Belgium	Boehringer Ingelheim	Markov model (1) moderate, (2) severe, (3) very severe, (4) death	Oostenbrink et al.74	75.4% male, 64.5 age moderate-to-very severe	prospective	(1) tiotropium + usual care (2) tiotropium	4 years	Health care payer
Hogan et al.75 2010	USA	Novartis	Not stated	Not specified	Not stated	Prospective	(1) formaterol (2) ipratropium (3) placebo	12 weeks	Not stated
Hoogendorn et al.76 2013	The Netherlands	Boehringer Ingelheim	Markov model (1) moderate, (2) severe, (3) very severe, (4) death	Oostenbrink et al.74	74% male; mean age 62.9	Prospective	(1) tiotropium (2) salmeterol	1 and 5 years	German statutory health insurance Societal
Jones et al.77 2003	UK	GlaxoSmithKline	Not specified	Not specified	83% male: mean age 62	Prospective	(1) salmeterol 50 mg (2) salmeterol 100 mg	16 weeks	Healthcare provider
Karabis et al.78 2014	USA	Mapi HEOR and Forest research	Markov model (1) moderate, (2) severe, (3) very severe, (4) death	Not specified	53% male Mean age 64.3%	Retrospective	(1) aclidimium (2) tiotropium	5 years	Healthcare payer
Lee et al.79 2006	Singapore	Department of Medicine, National University Hospital	Not stated	Not specified	Not specified	Retrospective	Addition of tiotropium to usual care	Not specified	Not specified
Lofdahl et al.80 2005	Sweden	AstraZeneca	Not stated	Not specified	Patients at least 40 years of age	Prospective	B/F combined inhaler B/F each as monocomponent	Not specified	Healthcare provider
Maniadakis et al.81 2006	Greece	Boehringer Ingelheim	Markov 3 states: Moderate, severe, very severe	Oostenbrink et al.72	Not specified	Retrospective	Tiotropium salmeterol	1 year	Greek National Health Service
Mittmann et al.82 2011	Australia, Canada Sweden	AstraZeneca	Not specified	CLIMB82	Not specified	Retrospective	(1) B/F + tiotropium (2) placebo + tiotropium	12 weeks	Healthcare provider
Naik et al.83 2010	US	Duquesne University School of Pharmacy, Pittsburgh, PA, USA	Markov: 2 states on treatment and maintenance therapy	Not specified	Not specified	Retrospective	tiotropium salmeterol	1 year	Third party payer
Najafzadeh et al.84 2008	Canada	Faculties of medicine and pharmacy across Canada	Decision analytic	Not specified	Not specified	Prospective	S, F/S, tiotropium	5 year	Canadian healthcare system
Neyt et al.85 2010	Belgium	Belgian Health Care Knowledge Centre Academic Centre for Primary Care,	Decision analytic	Not specified	Not specified	Prospective	tiotropium No ttiotropium	1 year	Healthcare payer
Nielsen et al.86 2013	Denmark Finland, Norway, Sweden	AstraZeneca	modelled CLIMB trial82	not stated	adults (average age = 62) moderate-to-severe COPD FEV1 = 38%)	Retrospective	(1) B/F tiotropium (2) placebo + tiotropium	3 months	Societal Healthcare provider
Oba87 2007	USA	not stated	spreadsheet model	not stated	adults (average age = 66) moderate-to-severe COPD FEV1 = 44%)	Retrospective	(1) tiotropium placebo (2) S placebo (3) tiotropium salmeterol	1 year	Third party payer
Oba37 2009	USA	none stated	Markov model with 4 states: (1) stable, (2) exacerbation requiring a physician visit, (3) exacerbation requiring hospitalization, (4) death Cycle length: 3 months	not stated	adults (average age = 65)	retrospective	(1) FP/S (2) FP (3) S (4) placebo	3 years	Third party payer
Onukwugha etnbsp;al.89 2008	USA	PhRMA Foundation	medical records	A pulmonologist validated diagnosis and severity levels	adults (average age = 68.9) >92% moderate, severe, or very severe COPD average FEV1 = 51%	retrospective	(1) tiotropium (2) ipratropium	1 year	Veterans Affairs Health Care System
Oostenbrink et al.90 2004	Netherlands Belgium	Boehringer Ingelheim GmBH	RCT (piggyback) by Vincken et al.91	not applicable	Adults ≥40 years old moderate-to-severe COPD (FEV1 ≤65%)	prospective	(1) tiotropium (2) ipratropium	1 year	Societal (direct costs only)
Oostenbrink et al.92 2005	Netherlands Canada	Boehringer Ingelheim GmBH	Markov model with three states: moderate, severe, very severe. Each state can have non-severe or severe exacerbations	Model outputs compared with 2 clinical trials	adults (average age 64) moderate-to-severe COPD	retrospective	(1) tiotropium (2) salmeterol (3) ipratropium	1 year
Price et al.93 2011	Germany	Novartis	Markov model with four health states (mild, moderate, severe, very severe) and death. States also had 2 sub-states with exacerbations: (1) severe (managed as outpatients) and (2) very severe (managed in in hospital) Cycle length: 3 months using a cohort of 1000 patients	Clinical rates and resource use validated by a clinician Model structure validated by an external economist Model outputs compared with other models and published clinical rates/mortality	adults (average age 64) moderate-to-severe COPD	retrospective	(1) indacaterol (2) tiotropium	3 years	German Healthcare provider
Price et al.94 2013	UK	Novartis UK	Markov model with four health states (mild, moderate, severe, very severe) and death. States also had 2 sub-states with exacerbations: 1) severe (managed as outpatients) and 2) very severe (managed in in hospital) Cycle length: 3 months	previously validated by Price et al.93 Experts validated inputs Data outputs validated against epidemiologic studies and other economic analyses	Adults (average age = 63.6) moderate-to-severe COPD	retrospective	(1) indacaterol (2) tiotropium (3) salmeterol	3 years	UK National Health Service
Price et al.95 2014	Sweden	Novartis	patient level simulation model Cycle length: 6 months	previously done by Asukai et al.96	adults (average age = 63) moderate-to-severe COPD with a low risk of exacerbations	retrospective	(1) indacaterol/glycopyrronium (2) indacaterol + glycopyrronium (3) FP/S	1, 3, 5, 10 years; lifetime	Swedish societal
Rutten-van Mölken et al.97 2007	Spain	Boehringer Ingelheim Pfizer Global	Markov model with 4 states: moderate, severe, very severe, death Cycle length: 8 days for the first; others = 1 month	not stated	moderate, severe, or very severe COPD	retrospective	(1) tiotropium (2) salmeterol (3) ipratropium	5 years	Societal (Spain) Spanish National Healthcare System
Rutten-van Mölken et al.98 2007	UK	ALTANA Pharma	RCT piggyback trial	not applicable	Adults ≥40 years old severe-to-very severe COPD	prospective	roflumilast placebo	1 year	UK society UK National Health Service
Samyshkin et al.99 2013	Switzerland	Takeda Pharma	Markov model with 3 states: severe, very severe, and dead Cycle length: 1 month	not stated	severe-to-very severe COPD associated with bronchitis; aged >40 years old	retrospective	(1) roflumilast + LAMA LAMA (2) roflumilast + LABA/ICS LABA/ICS (3) roflumilast + LABA/ICS + LAMA LABA/ICS + LAMA	lifetime	Swiss payer
Samyshkin et al.50 2014	UK	Takeda Pharmaceuticals	Markov model with 3 states: severe, very severe, and dead Cycle length: 1 month	not stated	severe-to-very severe COPD associated with bronchitis; aged 64 years old	retrospective	roflumilast + LABA LABA	30 years	UK National Health Service
Sin et al.100 2004	Canada	Institute of Health Economics	Markov model; states included three disease stages as well as death Cycle length: 3 months	not stated	Stage 1: FEV1 ≥50% Stage 2: FEV1 35% <50% Stage 3: FEV1 <35%	retrospective	ICs in all patients ICs for Stage 2 + 3 ICs for Stage 3 only none	3 years, lifetime	Societal
Spencer et al.101 2005	Canada	GlaxoSmithKline	Markov model with 4 states: (1) mild, (2) moderate, (3) severe, (4) death Cycle length: 3 months	One author provided ‘validation of the model assumptions and costing estimates’	Adults aged 63 ± 8.6 years old FEV = 25–70%	retrospective	FP/S usual care	25 years	Canadian healthcare payer
van der Palen et al.102 2006	Netherlands	Netherlands Asthma Foundation Amicon Insurance GlaxoSmithKline Boehringer Ingelheim		not stated	adults aged 40–75 moderate-to-severe (FEV 20–80%)	retrospective	FP placebo (FP withdrawn)	6 months	Healthcare system

CAD, Canadian dollars; CHF, Swiss francs; DKK, Danish kroner; NOK, Norwegian kroner; OECD, Organization for Economic Cooperation and Development; RAMQ, Régie de l’Assurance Maladie du Québec; RCT, randomized controlled trial; SEK, Swedish kronor; USD, United States dollars.

Table 6 details the inputs into economic models in severe COPD. Most (29/39, 74%) studies calculated direct costs only. Where applicable (i.e., in studies where the time horizon was more than 1 year), costs and outcomes were discounted between 3–5%, depending on local pharmacoeconomic requirements. Willingness-to-pay thresholds used for interpretation of results varied widely, between £20,000/QALY and $100,000/QALY.

Table 6. Model inputs in economic analyses of severe COPD.

Reference (Year)	Costs	Currency and year	Discounting	Input rates	Data source	Items costed	Data sources	Criterion	Planned sensitivity analyses
Ayres et al.45 2003	Direct, indirect (days lost)	UK pounds, 1998	not applicable	clinical resources	ISOLDE64 ISOLDE64	drugs hospitalization medical care days lost	Published costs (University of Kent) Published costs (University of Kent) Published costs (University of Kent) UK Department of Transport	not stated (willingness-to-pay)	Rates of FEV, exacerbations with/without indirect costs
Briggs et al.65 2006	Direct	UK pounds 1998–1999	Costs: 3.5% Outcomes: 3.5%	clinical resources utilities	ISOLDE64 ISOLDE64 ISOLDE64	drugs hospitalization medical care	British National Formulary Published costs (University of Kent) Published costs (University of Kent)	£20,000/QALY	Best/worst case on imputed data
Briggs et al.66 2010	Direct only	USD, 2007	Costs: 3% QALYs: 3%	clinical resources utilities	TORCH67 TORCH67 TORCH67, Dolan103	drugs	IMS	$100,000/QALY	not done
Chuck et al.68 2008	Direct only	CAD, 2006	Costs: 3% and 5%	clinical resources utilities	10 published studies 4 published RCTs Oostenbrink et al.90, Borg et al.104	drugs hospitalization medical care	Alberta Health Costing Alberta Health Costing Alberta Health Costing	$50,000/QALY	Probabilistic (100,000 simulations) 1-way sensitivity analyses: Rates of disease progression, exacerbation, mortality, discount rate, QALY value, costs Also, excluded TORCH Trial (milder disease) {Calverley}
Collet et al.69 2001	(days lost, need for help, caregiver burden)	CAD, 1995	not applicable	clinical resources	Collet et al.105 Collet et al.105	drugs medical care days lost	RAMQ RAMQ Average wage in Quebec	not stated	Probabilistic (1000 simulations), cost and clinical rates
Dalal et al.70 2010	Direct only	USD, 2007	not applicable	clinical	Anzueto et al.106, Ferguson et al.88	drugs hospital costs medical care	Drug Topics Redbook Stanford et al.107 Nurmagambetov et al.108	not stated	not stated
Dal Negro et al.35 2007	Direct only	euros, year not stated	not stated	clinical exacerbations	ISOLDE64; Calverley et al.109^,110, Szafranski et al.111 ISOLDE64, Price et al.41	drugs Care	National Health Service Luccioni et al.112, Dal Negro et al.113		1-way analyses on input rates, costs
Earnshaw et al.36 2009		USD, 2006	Costs: 3% Outcomes: 3%		TORCH67			$50,000/QALY	Probabilistic (10,000 simulations) 1-way on rates of exacerbation, mortality, duration of exacerbations, utility weights, transition probabilities
Gani et al.71 2010	Direct	£, 2009	Not applicable	Clinical resources	Brusasco et al.114, Vincken et al.91, Casaburi et al.115	Drugs hospitalization	Britton116, Department of Health NHS reference costs, Griffiths et al.117, Curtis, MIMS online, personal communication	£20,000/QALY	Probabilistic sensitivity analysis
Hertel et al.118 2012	Direct	£, otherwise not stated	3.5% (costs and outcomes)	Clinical Utilities Resources	Mills et al.119, Calverley et al.67, British National Formulary	Drugs Maintenance Community treated exacerbation	British National Formulary	£20,000–30,000/QALY	Probabilistic sensitivity analysis
Hettle et al.73 2012	Drug costs	£, euro 2011	UK Costs/Benefits 3.5% Belgium costs 3% and befits 1.5%	Utilities resource	Rutten van Mölken et al.97	Drugs Maintenance Hospitalization	Cleemput et al.120 2008	£30,000/QALY	Deterministic and probabilistic sensitivity analysis
Hogan et al.75 2010	Direct	Not stated	Not applicable	Clinical Resources Utilities	Dahl et al.121	Study drugs; rescue medication	Dahl et al.121 2001	Not stated	1-way sensitivity analysis
Hoogendorn et al.76 2013	Direct, Out of pocket, Productivity loss	€, 2010	3%	Clinical	POET-COPD122	hospitalizations, days at intensive care unit, ambulance transportations, emergency room visits and contact with five types of healthcare providers	Koopmanschap et al.123	Not stated	One way and probabilistic sensitivity analysis
Jones et al.77 2003	Direct costs	Not stated	Not applicable	Clinical Resources Utilities	Boyd et al.124 Boyd et al.124 patient diary	ER visits Outpatient hospital visits Inpatient hospital visits GP visits study drugs rescue medication concurrent medication	Patient diary cards	Not stated	Carried out but not reported - Not specified
Karabis et al.78 2014	Direct costs	USD 2012	3% Costs and benefits	Clinical Resources Utilities	Kerwin et al.125 Quanjer et al.126 Rutten-van Moülken and Lee127 Yu et al.128	Drug COPD management Exacerbations	Not specified Yu et al.128	$50,000/QALY	Probabilistic
Lee et al.79 2006	Direct costs	Singapore dollars	Not specified	Clinical Resources	Casaburi et al.115, Brusasco et al.114, Niewoehner et al.129	Drug exacerbation hospitalization	Not specified Not specified Ministry of Health; national healthcare group data	Not specified	1-way sensitivity
Lofdahl et al.80 2005	Direct costs	Swedish kroner (converted to euros)	Not applicable	Clinical Resources	Calverley et al.67	Total healthcare costs	Calverley et al.67	Not specified	1 way sensitivity
Maniadakis et al.81 2006	Direct costs	2005 euros	Not applicable	Resources utility	Not specified Oostenbrink et al.72	Study medication Hospital and other exacerbation	General hospital Heraklion	Not specified	1-way sensitivity
Mittmann et al.82 2011	Direct costs	2009, Canadian, Australian dollars; Swedish kroner	Not applicable	Resources Clinical	CLIMB82	Study drugs Healthcare utilization	CLIMB82	Not specified	1 way sensitivity analysis
Naik et al.83 2010	Direct costs	Not stated	Not applicable	Clinical Resources	Casaburi et al.115 Vincken et al.91 Donohue et al.130	Drug costs, hospitalization costs, monitoring costs, and physician visit costs	Red book Medical Insurance Reimbursement-based Maximum Allowable Charges	Not specified	1 way sensitivity analysis
Najafzadeh et al.84 2008	Direct costs	2006 Canadian dollars	Not applicable	Clinical Resources Utility	OPTIMAL131^,132	Drug Medical resource utilization	Pharmanet Drug Master List Medical Services Commission. Payment Schedule.	Not specified	1 way sensitivity analysis
Neyt et al.85 2010	Direct costs	Not stated	Not applicable	Clinical Resources Utility	O’Reilly et al.133 Roede et al.134 Belgian database	Drug Hospital related exacerbation Other exacerbations	O’Reilly et al.133 Roede et al.134 Belgian database	Not specified	1 way sensitivity analysis
Nielsen et al.86 2013	Direct and indirect	DKK, 2011 euros, 2011 (Finland) NOK, 2011 SEK, 2011	not applicable	clinical resources	CLIMB82 CLIMB82	drugs hospital care medical care sick leave	national medicines agencies national Ministry of Health national Ministry of Health OECD Labour costs
Oba87 2007	Direct only	USD, 2005	not applicable	clinical utilities	6 published trials (meta-analyzed) Niewoehner et al.129	drugs Hospital medical	Drug Topics Red Book Niewoehner et al.129, Pennachio135 DRG Code 99214 for COPD	not stated	1-way on costs, unscheduled visits, incremental QALYs
Oba37 2009	Direct only	USD, 2006	Costs: 3% Outcomes: 3%	clinical utility	TORCH67 TORCH67	drugs medical hospital	Drug Topics Red Book Pennachio135, Thompson Healthcare 5 published studies, Thompson Healthcare	not stated	Probabilistic (10,000 simulations) 1-way on clinical rates, discount rate, costs, hospitalizations, mortality and utilities
Onukwugha et al.89 2008	Direct only	USD, 2004	not applicable	clinical medical hospital	patient records patient records patient records	drugs hospitalization emergency room	Veterans Affairs Veterans Affairs Maryland Health Consortium list	cost savings was preferred outcome	Probabilistic (1000 draws) 1-way on rates of exacerbation and adherence, costs
Oostenbrink et al.90 2004	Direct only	euros, 2001	not applicable	clinical rates resources quality-of-life	Vincken et al.91 Vincken et al.91 Vincken et al.91	drugs laboratory medical care hospital care	List prices List prices average physician fees in the trial Oostenbrink et al.136	willingness-to-pay	1-way on costs, resource use, threshold values for improvement and quality of life
Oostenbrink et al.92 2005	Direct only	euros, 2001	not applicable	clinical utilities	Brusasco et al.114, Casaburi et al.115, Vincken et al.91 Borg et al.104, Paterson et al.138, Spencer and Jones139	drugs clinical	Oostenbrink et al.90^,137{2 studies in 2004},Ontario Drug Benefit Formulary Oostenbrink et al.90^,137, Ontario Schedule of Benefits	willingness-to-pay	Probabilistic (5000 simulations) 1-way on rates of transition, exacerbation, baseline disease severity, utility weights, costs
Price et al.93 2011	Direct only	euros, 2010	Costs: 3% Outcomes: 3%	clinical utilities mortality	INHANCE140, INLIGHT-2141, UPLIFT142 Ståhl et al.143 Rutten-van Mölken et al.97	drugs clinical	German Drug Index (Lauer Taxe) Uniform Value Scale	not stated	Probabilistic (1000 simulations) Univariate on clinical rates, time horizon, mortality and costs
Price et al.94 2013	Direct only	UK pounds, 2011	Costs: 3.5% Outcomes: 3.5%	clinical utility values mortality	Donohue et al.140, Kornmann et al.144, Price et al.93 Rutten-van Mölken et al.145 Rutten-van Mölken et al.97	clinical drugs	National Health Service British National Formulary	£20,000/QALY	Probabilistic (1000 simulations) 1-way on clinical rates, mortality, discount rate
Price et al.95 2014	Direct, indirect (productivity)	SEK, 2013	Costs: 3% Outcomes: 3%	clinical resources mortality utilities	Dahl et al.146, Vogelmeier et al.147, Mahler et al.148 Jansson et al.149, Jansson et al.150 Lindberg et al.151 Rutten-van Mölken et al.152	drug costs clinical care productivity	FASS 2013 drug prices Jansson et al.149, Jansson et al.150 Jansson et al.150	not stated	Probabilistic (1000 cohorts with 10,000 patients) 1-way on rates of improvement and exacerbation, baseline severity
Rutten-van Mölken et al.97 2007	Direct and indirect (sick leave)	euros, 2005	Costs: 6% Outcomes: 6%	clinical resources mortality utilities	Oostenbrink et al.72, Miravitlles et al.153 Miravitlles et al.153 Miravitlles et al.154 Badia et al.155, Spencer et al.101	clinical hospital drugs sick leave	SOIKOS database SOIKOS database SOIKOS database National Health Service	willingness-to-pay	Probabilistic (5000 simulations) 1-way on disease severity, Discount rate
Rutten-van Mölken et al.98 2007	Direct, indirect (work/leisure days lost, impairment)	euros, 2004	not applicable	clinical hospital work lost/impairment	directly observed directly observed patient questionnaire	drugs, care workdays lost	National Health Service, Published costs (University of Kent) UK Office of National Statistics	willingness-to-pay	Imputation method, method of calculating productivity costs, unit costs for hospitalization
Samyshkin et al.99 2013	Direct only	CHF, 2011	Costs: 2.5% Outcomes: 2.5%	clinical resources utilities mortality	Bateman et al.156, Mills et al.119 Calverley et al.157, Oostenbrink et al.72 Rutten-van Mölken et al.158 Spencer et al.101	drugs clinical	Swiss Medicines Compendium Schramm{2005}, TARMED Suisse	CHF 72,000/QALY (€60,000/QALY)	Probabilistic 1-way on clinical rates. costs, discount rate
Samyshkin et al.50 2014	Direct only	UK pounds, year not stated	Costs: 3.5% Outcomes: 3.5%	clinical resources utilities mortality	Rabe159, Bateman et al.156, data on file Calverley et al.157 Rutten-van Mölken et al.158 Ekberg-Aronsson et al.160	drugs exacerbations	British National Formulary NHS lists	£20,000–£30,000/QALY	Probabilistic (2000 simulations) 1-way on clinical rates, discount rate, patient age, disease severity, utility values, costs
Sin et al.100 2004	Direct + indirect (lost wages)	CAD, 2000	Costs: 5%	clinical resources utilities	4 published trials 4 published trials several published articles	exacerbations medical Care hospital care lost wages	Alberta Health Care Plan Alberta Health Care Plan Alberta Health Costing Canadian National Population Health Survey	$50,000/QALY	Probabilistic (100,000 simulations) 1-way on time horizon, response rates
Spencer et al.101 2005	Direct only	CAD, 2002	Costs: 5% Outcomes: 5%	clinical resources utilities	TRISTAN110, 7 RCTs TRISTAN110, 7 RCTs Health Survey for England161, Brazier et al.161, Spencerand Jones139	maintenance therapy exacerbations Hospitalization	Ontario Case Costing Program Ontario Case Costing Program London Health Science Centre	$100,000/QALY	Probabilistic analyses 1-way on clinical rates, discount rates
van der Palen et al.102 2006	Direct only	euros, 2002	not applicable	clinical resources	COPE102 COPE102, patient charts	drugs medical care hospital care	Local price list Oostenbrink{2002} Oostenbrink{2002}	not stated	Probabilistic (1000 simulations) 1-way on costs, probabilities

Table 7 presents the results from the economic models used in severe COPD. Agents or comparators used in analysis include: aclidimium, betamethasone, fluticasone propionate, formeterol, indacaterol, indacaterol/glyopyrronium, ipratroprium, long-acting beta-agonists, OM-85 BV (immunostimulating agent), placebo, roflumilast, salmeterol, tiotropium, and usual care.

Table 7. Model outputs from economic analyses of severe COPD.

Reference (Year)	Time horizon	Sub-set	Treatment	Outcomes					Conclusions
				Cost	Clinical	Amount	ICER†	Sensitivity analyses
Ayres et al.45 2003	6 months	Societal viewpoint NHS viewpoint	FP* placebo* FP (direct cost)* placebo (direct cost)*	£3.65 £4.06 £2.75 £2.72	patient with 10% improvement in FEV1 exacerbation-free patient moderate-to-severe exacerbation-free patient patient with 10% improvement in FEV1 exacerbation-free patient moderate-to-severe exacerbation-free patient		−£3.39 −£7.19 −£3.28 £0.25 £0.54 £0.25	At a WTP of €15/day, FP would be cost-effective 90% of the time Sensitivity analyses: results were robust; ICERs varied between £0.22 and £0.25/day	cost-effective cost-effective
Briggs et al.65 2006	3 years	Observed cases Full analysis (imputed data)	FP placebo FP placebo	£2494 £1341 £2530 £1509	QALYs QALYs QALYs QALYs	1.88 1.74 1.86 1.74	£9500 £9282	ICERs: £5200–£13,200	cost-effective
Briggs et al.66 2010	3 years	Pooled USA Eastern Europe Western Europe Other	FP/S FP S placebo FP/S FP S placebo FP/S FP S placebo FP/S FP S placebo FP/S FP S placebo	$12,967 $11,486 $11,035 $9456 $24,830 $22,692 $21,717 $18,891 $7710 $6437 $5815 $4455 $7619 $6797 $6603 $5623 $6511 $5726 $5404 $4490	QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs	2.022 1.967 1.949 1.941 1.957 1.911 1.890 1.880 1.974 1.904 1.869 1.863 2.109 2.047 2.030 2.026 2.003 1.923 1.919 1.909	$43,346 dominated dominated $77,130 dominated dominated $2924 dominated dominated $24,048 dominated dominated $2100 dominated dominated	not done	FP/S is preferred as it had the lowest C/E ratio
Chuck et al.68 2008	3 years lifetime	Base case Base case	IC only LABA + IC (Stage 3 only) LABA + IC (Stages 2–3) LABA + IC (all stages) IC only LABA + IC (Stage 3 only) LABA + IC (Stages 2–3) LABA + IC (all stages)	$3719 $3758 $3853 $5204 $9636 $9788 $10,142 $13,282	QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs	2.405 2.406 2.408 2.411 6.763 6.769 6.776 6.783	$39,000 $47, 500 $450,333 $25,333 $50,571 $448,571	80% of ICERS were <$50,000 when treating Stage 3 only 55% of ICERS were <$50,000 when treating Stages 2 and 3 only none cost effective when all stages treated Without TORCH, ICER = $9669 when treating Stage 3 only $31,606 when treating Stages 2 and 3, $271,041 treating all stages Without TORCH, ICER = $7828 when treating Stage 3 only $26,356 when treating Stages 2 and 3, $235,827 treating all stages	FP/S is cost-effective for patients in Stages 2 and 3 FP/S is cost-effective for patients in Stages 2 and 3
Chuck et al.68 2008	6 months	Societal viewpoint	OM-85 BV placebo		day of hospitalization for respiratory disease		$45	CI: $18–$210 98.8% of simulations favored OM-85 BV	Cost-effective
Dalal et al.70 2010	1 year	Base case	FP/S S	$4842 $5066	Annualized exacerbation rate	4.76 5.67	dominant	CI for Cost: FP/S = $4731–$4952 CI for Cost: S = $4937–$5195	Cost-effective
Dal Negro et al.35 2007	lifetime	NHS only	S B/F FP/S controls F S B/F FP/S controls F	€33,369 €33,945 €34,038 €34,632 €34,754 €33,369 €33,945 €34,038 €34,632 €34,754	Average exacerbations/patient Average exacerbations/patient Average exacerbations/patient Average exacerbations/patient Average exacerbations/patient Average SFDs/patient Average SFDs/patient Average SFDs/patient Average SFDs/patient Average SFDs/patient	10.07 9.66 9.09 12.04 10.14 55 220 257 0 37		Only FP/S and S are considered in the final analysis Actual ICER = €682.07/exacerbation avoided (average cost/exacerbation = €1778) Actual ICER = €3.31/additional SFD	Reported: All but S dominated controls; S was slightly more costly ERROR = results not interpreted correctly FP/S is cost effective over salmeterol all others are dominated (B/F extended; F and C strongly dominated)
Earnshaw36 2009	Lifetime	none	Advair® 500/50 Fluticasone propionate Placebo	$42,019 $23,727 $18,986	QALY QALY QALY	7.42 6.58 6.74	33,865/QALY dominated –	77% of simulations resulted in an ICER of less than $50,000	Advair is cost effective compared to no maintenance treatment
Gani et al.71 2010	1 year	None	Tiotropium Salmeterol Ipratropium	£1439 £1565 £1631	QALY QALY QALY	0.744 0.73 0.723	dominated dominated	WTP threshold 97–98% probability of <£20,000/QALY	Tiotropium is cost effective compared to salmeterol or ipratropium
Hertel et al.118 2012	30 years	None	LABA LAMA LABA + roflumilast LAMA + roflumilast LAMA + LABA LAMA + LABA + roflumilast LABA/ICS LABA/ICS + roflumilast LABA + LAMA/ICS LABA + LAMA/ICS + roflumilast	£22,342 £22,370 £22,749 £22,779 £22,687 £23,100 £22,468 £22,878 £22,816 £23,230	QALY QALY QALY QALY QALY QALY QALY QALY QALY QALY	5.39 5.42 5.43 5.46 5.45 5.48 5.43 5.46 5.48 5.51	LAMA + LABA + roflumilast versus LAMA + LABA, was £13,764/QALY	Results robust	addition of roflumilast to the standard of care regimens is cost-effective
Hettle et al.73 2012	4 year		Tiotropium Usual care	£6475 £5679	QALY QALY	2.645 2.594	£15,567/QALY	The probability of cost-effectiveness at a willingness-to-pay threshold of £30,000 was over 60%.	Adding tiotropium to usual care is cost-effective.
Hogan et al.75 2010	12 weeks		Placebo Ipratropium Formoterol 12 Formoterol 24	$38.93 $76.34 $214.91 $418.92	QALY QALY QALY QALY	1.50 1.1 6.60 4.8	Dominated 25.20/QoL Dominated		Formoterol 12 provided greater QoL outcome than ipratropium bromide at an additional cost of $554.28 per year.
Hoogendorn et al.76 2013	1 and 5 year		Tiotroprium – 1 year Salmeterol – 1 year Tiotroprium – 5 year Salmeterol – 5 year	1570 1380 7520 6896	QALY QALY QALY QALY	0.746 0.734 3.26 3.18	Euro8141/QALY	Results robust	Tiotropium is cost-effective
Jones et al.77 2003	16 weeks		Salmeterol Placebo	£192.37 £102.26	Symptom free night	60% 48%	£5.67 per symptom free night	Reduction in hospital costs offsets cost of salmeterol
Karabis et al.78 2014	5 years		Aclidimium tiotropium	126,274 USD 128,591 USD	QALY QALY	3.50 3.49	dominated
Lee et al.79 2006	1 year		Tiotroprium + usual care Usual care
Lofdahl et al.80 2005	1 year		B/F Single inhaler B monocomponent F monocomponent Placebo	2518 3194 3653 3213				Results robust	Single inhaler cost-effective compared to F or placebo
Maniadakis et al.81 2006	1 year		Tiotroprium Salmeterol	€2504 €2655	Exacerbation free months	11.15 10.98	dominated	The probability that tiotropium is cost-effective was 65% at a ceiling value of €0 and reached 77% at a ceiling ratio of €1000
Mittmann et al.82 2011
Naik et al.83 2010	1 year		Tiotroprium Salmeterol No treatment	1408.59 1268.66 392.1	Exacerbation avoided	1.12855 1.05156 0.69443	1817.37 2454.48		tiotropium is more cost-effective than salmeterol and no treatment
Najafzadeh et al.84 2008	1 year		Tiotroprium + placebo Tiotropium + F/S Tiotroprium + S	C$2678 C$4042 C$2801	QALY QALY QALY	0.7092 1.35 1.69
Neyt et al.85 2010	1 year		tiotropium real-world tiotropium theoretical salmeterol real world	427.71 621 225.46	Incremental QALY	0.013			Tiotroprium is not cost-effective
Nielsen et al.86 2013	3 months 3 months	Denmark: Societal Finland: Societal Norway: Societal Sweden: Societal Denmark: Health system Finland: Health system Norway: Health system Sweden: Health system	B/F + tiotropium placebo + tiotropium B/F + tiotropium placebo + tiotropium B/F + tiotropium placebo + tiotropium B/F + tiotropium placebo + tiotropium B/F + tiotropium placebo + tiotropium B/F + tiotropium placebo + tiotropium B/F + tiotropium placebo + tiotropium B/F + tiotropium placebo + tiotropium	DKK 3713 DKK 3756 467€ 436€ NOK 3313 NOK 3827 SEK 4494 SEK 4575 DKK 3713 DKK 3756 467€ 436€ NOK 3313 NOK 3827 SEK 4494 SEK 4575	exacerbations/patient exacerbations/patient	0.11 0.29 0.11 0.29 0.11 0.29 0.11 0.29 0.11 0.29 0.11 0.29 0.11 0.29 0.11 0.29	dominant dominated 174 € dominant dominated dominant dominated DKK 1580 307 € dominant dominated SEK 1573	Dominant in 50% of simulations no change if antibiotics are included Dominant in 32% of simulations 149€ if antibiotics are included Dominant in 74% of simulations no change if antibiotics are included Dominant in 52% of simulations no change if antibiotics are included DKK 1449 if antibiotics are included 281€ if antibiotics are included no change if antibiotics are included SEK 1342 if antibiotics are included	This combination represents a clinical and economic benefit to the healthcare system This combination represents a clinical and economic benefit to the healthcare system This combination represents a clinical and economic benefit to the healthcare system This combination represents a clinical and economic benefit to the healthcare system
Oba87 2007	1 year		tiotropium placebo salmeterol placebo tiotropium	$835 (incremental) $1066 (incremental) $330 (incremental)	QALYs QALYs QALYs	0.032 (incremental) 0.026 (incremental) 0.036 (incremental)	$26,094 $41,000 −$391	Range: $11,780–$77,214 1-way ICERs: $5160–$53,571 Range: $23,650–$98,750 1-way ICERs: $10,825–$56,000 1-way: could save $244–$967 and gain 0.011–0.060 QALYs	Cost-effective Cost-effective Cost-effective (cost saving)
Oba37 2009	3 years	base case	placebo S FP FP/S	$2780 $5832 $6927 $9598	QALYs QALYs QALYs QALYs	1.444 1.498 1.510 1.575	dominated dominated $52,046	Insensitive to changes in drug costs, hospitalization rates, mortality rates and utility values	Seemed to be ‘most cost-effective’ in the base case
Onukwuga et al.90 2008	1 year	full sample mild moderate severe very severe hospitalized not hospitalized 1 emergency room visit 2 emergency room visits >2 emergency room visits	tiotropium vs ipratropium tiotropium vs ipratropium tiotropium vs ipratropium tiotropium vs ipratropium tiotropium vs ipratropium tiotropium vs ipratropium tiotropium vs ipratropium tiotropium vs ipratropium tiotropium vs ipratropium	not reported not reported not reported not reported not reported not reported not reported not reported not reported	exacerbation avoided (hospitalization or emergency room visit)	not reported not reported not reported not reported not reported not reported not reported not reported not reported	$2360 $4691 $5,559 $2073 −$1818 −$4472 $1271 $684 $454	sensitive to effectiveness rates, costs	unlikely to be cost-effective in whole sample could save money if restricted to very severe COPD and patients with COPD hospitalizations
Oostenbrink et al.90 2004	1 year		tiotropium ipratropium	€1721 €1541	exacerbations/patient patients with ≥4 units improvement on SGRQ exacerbations/patient patients with ≥4 units improvement on SGRQ	0.74 51.2% 1.01 34.60%	€667 €1084	Dominant in 24% of patients; 80% cost-effective at a WTP of €2000 Dominant in 25% of patients; 72% cost-effective at a WTP of €2000	Tiopropium improves health outcomes and increases costs
Oostenbrink et al.92 2005	1 year	Netherlands Canada	tiotropium salmeterol ipratropium tiopropium salmeterol ipratropium	€1760 €1802 €1930 €1309 €1306 €1307	QALMs QALMs QALMs QALMs QALMs QALMs	8.42 8.17 8.11 8.42 8.17 8.11	dominant dominated dominated dominant dominated dominated	results unstable Probability sensitivity analyses: equivocal Probability sensitivity analyses: equivocal results unstable Probability sensitivity analyses: equivocal Probability sensitivity analyses: equivocal	No substantial difference among drugs No substantial difference among drugs
Price et al.93 2011	3 years		indacaterol tiotropium indacaterol tiotropium	€2067 €2415 €2043 €2179	QALYs QALYs QALYs QALYs	2.13 2.12 2.13 2.12	dominant dominant	Dominant in 89.7% of simulations Variables most sensitive = time horizon, mortality rates and benefit from indacaterol Dominant in 77.8% of simulations	Indacaterol is cost-effective compared to tiopropium and salmeterol
Price et al.94 2013	3 years	150 mg 150 mg 300 mg	indacaterol tiotropium indacaterol salmeterol indacaterol tiotropium	£4534 £4781 £4583 £4692 £4501 £4760	QALYs QALYs QALYs QALYs QALYs QALYs	2.158 2.150 2.158 2.149 2.162 2.151	dominant dominant dominant	84% of simulations dominant and 89% cost-effective Baseline model population was most influential variable 72% of simulations dominant and 82% cost-effective	indacaterol is cost-effective
Price et al.95 2014	1 year 3 years 5 years 10 years lifetime		indacaterol/glycopyrronium indacaterol + glycopyrronium FP/S indacaterol/glycopyrronium indacaterol + glycopyrronium FP/S indacaterol/glycopyrronium indacaterol + glycopyrronium FP/S indacaterol/glycopyrronium indacaterol + glycopyrronium FP/S indacaterol/glycopyrronium indacaterol + glycopyrronium FP/S	SEK 27,635 SEK 28,403 SEK 30,379 SEK 91,788 SEK 93,906 SEK 100,642 SEK 149,464 SEK 152,772 SEK 163,402 SEK 273,053 SEK 278,685 SEK 300,548 SEK 500,248 SEK 508,951 SEK 543,281	QALYs QALYs	assumed equal (0.001 more than FP/S) assumed equal (0.006 more than FP/S) assumed equal (0.015 more than FP/S) assumed equal (0.047 more than FP/S) assumed equal (0.200 more than FP/S)	dominant dominated dominated dominant dominated dominated dominant dominated dominated dominant dominated dominated dominant dominated dominated	Dominant in 100% of all iterations	indacaterol/glycopyrronium is cost-effective
Rutten-van Mölken et al.97 2007	5 years	societal NHS Base case	tiotropium salmeterol ipratropium tiotropium salmeterol ipratropium	€6574 €6125 €5545 €6424 €5869 €5181	QALYs QALYs QALYs QALYs QALYs QALYs	3.15 3.02 3.00 3.15 3.02 3.00	€3483 €35,158 €4118 €38,931	tiopropium had highest net benefit above €7076/QALY below €11,000/QALY, ipratropium had the highest probability of being cost-effective; above that value, tiopropium had highest probability of being cost-effective
Rutten-van Mölken et al.98 2007	1 year	societal National Health Service	roflumilast placebo roflumilast placebo	€1637 €1401 €1418 €1242	moderate or severe exacerbations relevant improvement in SGRQ moderate or severe exacerbations relevant improvement in SGRQ moderate or severe exacerbations relevant improvement in SGRQ moderate or severe exacerbations relevant improvement in SGRQ	0.96 0.19 1.06 0.14 0.96 0.19 1.06 0.14	€2356 €4712 €1755 €3510	Probability of cost-effectiveness = 10% at a WTP = 0, 70% at €5000, 90% at €50,000 Probability of cost-effectiveness = 10% at a WTP = 0, >70% at €20,000 1-way ICERS: €375–€2817 1-way ICERS: €2363–€5633	Increases costs but lowers healthcare use potential to be cost saving
Samyshkin et al.99 2013	lifetime		(1) roflumilast + LAMA LAMA (2) roflumilast + LABA/ICS LABA/ICS (3) roflumilast + LABA/ICS + LAMA LABA/ICS + LAMA	CHF 86,754 CHF 83,364 CHF 91,470 CHF 88,161 CHF 99,364 CHF 95,564	QALYs QALYs QALYs QALYs QALYs QALYs	6.466 6.191 6.479 6.191 6.468 6.191	CHF 12,313 CHF 11,456 CHF 13,671	Probabilistic: 79% probability of cost-effective at a WTP of CHF 70,000 Roflumilast + LAMA dominated in 1 case; otherwise ICER = dominant to CHF 17,196 Probabilistic: 96% probability of cost-effective at a WTP of CHF 70,000 Roflumilast + LABA/ICS dominant to CHF 84,682 Probabilistic: 96% probability of cost-effective at a WTP of CHF 70,000 Roflumilast + LABA/ICS + LAMA dominant to CHF 85,689	Roflumilast cost-effective and reduces burden
Samyshkin et al.50 2014	30 years	Base case	roflumilast + LABA LABA	£19,358 £16,161	QALYs QALYs	5.615 5.451	£19,505	CI: £364–£38,646 82% probability of cost-effectiveness at a WTP of £30,000 1-way ICERs: £12,829–£53,435	can be cost-effective ICERs most influenced by exacerbation and mortality rates
Sin et al.100 2004	3 years 3 years lifetime lifetime	assuming a mortality benefit assuming no mortality benefit assuming a mortality benefit assuming no mortality benefit	no ICs ICs Stage 3 only ICs Stages 2 and 3 ICs all stages no ICs ICs Stage 3 only ICs Stages 2 and 3 ICs all stages no ICs ICs Stage 3 only ICs Stages 2 and 3 ICs all stages no ICs ICs Stage 3 only ICs Stages 2 and 3 ICs all stages	NR NR NR NR NR NR NR NR NR NR NR NR NR NR NR NR	QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs QALYs	2.71 2.72 2.72 2.75 2.71 2.72 2.72 2.75 NR NR NR NR NR NR NR NR	$11,100 $17,000 $46,200 $30,200 $34,100 $123,700 $2000 $2900 $4600 $15,000 $21,200 $26,200	84% probability of being cost-effective 75% probability of being cost-effective 57% probability of being cost-effective Treating Stage 2–3 non-responders for only 6 months: ICER = $18,100 Treating Stage 2–3 non-responders for only 6 months: ICER = $36,400	Cost-effective Cost-effective
Spencer et al.101 2005	25 years	Base case	FP/S	$25,780 $6416	QALYs usual care	4.21 4.08	$74,887	CI: $21,985–$128,671	Cost-effective at a WTP of $100,000/QALY
van der Palen et al.102 2006	6 months	Base case	FP placebo	€511 €456	exacerbations hospital admissions exacerbations hospital admissions	0.87 0.073 1.37 0.116	€110 € 1,286	Cost/exacerbation CI: −€52 to €331 Cost/hospitalization CI: −€1411 to €3069 1-way sensitive to clinical rates, costs	Overall cost savings but increased exacerbations and hospitalizations Note: ICERs both exceed the average cost to treat the outcome

CI, 95% confidence interval; FP, fluticasone propionate; IC, inhaled corticosteroids; NR, not reported; QALM, quality-adjusted life-month; QALY, quality-adjusted life-year; S, salmeterol; SFD, symptom-free day; SGRQ, St. George’s Respiratory Questionnaire; WTP, willingness-to-pay.

*Expressed as cost per day.

†Negative ICER mean cost savings.

Discussion

In this review, we were able to use data from 14 studies that examined patients with asthma and 39 that researched patients with COPD. Our search did not find studies that examined patients who suffered from both asthma and COPD simultaneously. Furthermore, with the exception of a few studies, our search did not locate studies that took into account the severity of the diseases. One reason could be the difficulty in defining the different levels of severity. Furthermore, few drugs are expressly indicated for severe cases of either COPD or asthma and, as such, economic evaluations with this perspective have not been undertaken.

In both asthma and COPD-related economic evaluations, predictive decision-analytic type models were predominant, with Markov models being most commonly used. Given the chronic nature of either asthma or COPD, use of Markov and discrete event simulation models may be most appropriate.

However, study parameters, for the most part, were not validated. Validation of models may include face validation, where experts would confirm the model structure used, data sources cited, and assumptions made. Other types of validation such as internal, external, or predictive validity were also largely absent from the studies included in this review. The joint task force of the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) and the Society for Medical Decision Making (SMDM) recommend use of validation in their Good Practices Research report34.

Specifically, only Norman et al.12 (who focused on asthma) discussed validation of their model in detail. They did extensive testing to assure the model integrity and that it generated accurate results. Some authors used existing published models, assuming that validation had been done; however, in many cases, the original publications did not discuss validation. For example, van Nooten et al.32 indicated that they used ‘the validated and published Markov 5-state omalizumab model’, citing Brown et al.21 and Dewilde et al.25 (the original publication) as references. However, neither the word ‘validate’ nor ‘validation’ appears in the paper by Dewilde et al. Since there is overlap in the authorship of those three papers, they would presumably have had access to that information, but failed to include it in their articles. Therefore, there is a need for authors to describe their validation process explicitly and completely.

On the other hand, Hoogedorn et al.163 identified seven studies and requested that these authors validate their own models using pre-identified criteria. They found a general consistency in structure across models. Outcomes varied somewhat, but the variation could be explained by differences in input values for disease progression and mortality. There was a high probability (90–100% for five models, 70% and 50% for the others) of lifetime cost-effectiveness at a willingness-to-pay of €50,000/QALY. Note that this threshold is quite high compared to those in many jurisdictions. Those authors concluded that mortality was the important factor in determining differences between ICERs.

We identified at least three studies35–37 that did not analyze and present their findings according to well-accepted standard procedures37^,38. For example, Dal Negro et al.35 compared each drug to placebo, which is not appropriate when some alternatives are dominated. According to Drummond et al.38, all dominated drugs must first be eliminated from consideration in the analysis and non-dominated alternatives analyzed incrementally, starting with the choice having the lowest cost. Reporting results that include dominated alternatives may affect what is considered to be within accepted threshold parameters. In their analyses, Dal Negro et al. failed to eliminate the placebo and fluticasone arms, which were strongly dominated, as well as formoterol/budesonide, which was extendedly dominated. Nevertheless, in the final analysis, their drug was still cost-effective. Earnshaw et al.36 also compared each drug to placebo, reporting an ICER for fluticasone/salmeterol vs placebo of $33,865 per additional QALY gained for which they claimed cost-effectiveness, citing a threshold of $50,000. However, when calculating incrementally from salmeterol (next in line), the ICER would be ∼$69,889, which far exceeds their limit for cost-effectiveness. A similar situation occurs with the study by Oba37. In his Table 1, he recognized that two drugs were ‘dominated through extended dominance’. Therefore, they should both have been eliminated, but, despite that, Oba calculated an incremental ICER for fluticasone/salmeterol, comparing it with fluticasone alone (a dominated choice). The reported (incremental) ICER was $41,092; however, had it been properly compared with the only non-dominated choice (i.e., placebo), the ICER would have been ∼$52,046. If the same criterion used by Earnshaw et al. had been used to judge this result (i.e., $50,000/QALY gained), it would also fail to achieve cost-effectiveness. Thus, there is an apparent need for appropriate reporting of results from economic models.

Conclusions

We have summarized the economic analyses that examined drug use in severe asthma and COPD. Very few biologic drugs have been studied in these patients, and more options are needed. In asthma-related studies, there was not a large number that were found to be cost-effective, and often at a quite high level of willingness-to-pay. Controversy concerning some results was uncovered in this study, which needs to be avoided in the future. More research is needed to fill gaps in knowledge.

As well, there is a need for more complete and accurate reporting of results. The authors should be required to address the issue of model validation and provide adequate documentation of how it was done. Insufficient attention has been placed on the calculation and presentation of results when there are multiple options being compared. Authors, reviewers, and editors need to assure that manuscripts adhere to guidelines for reporting results.

Transparency

Declaration of funding

This study was funded by Janssen Pharmaceutica.

Declaration of financial/other relationships

TRE has received sponsorship from Janssen Pharmaceutica and has declared consultancy/advisory interests from Janssen-Cilag. JVL is an employee of Janssen Pharmaceutica NV and MEHH was an employee at the time of writing. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgments

This review was funded by Janssen-Cilag BV Drs. Einarson and Bereza received financial consultation for this research and manuscript At the time of writing, Nielsen, Van Laer, and Hemels were all employees of Janssen.