Economic evaluation of single-inhaler triple therapy for chronic obstructive pulmonary disease in Thailand

Abstract

Introduction

Single-inhaler triple therapy (SITT) is known to be a cost-effective intervention for chronic obstructive pulmonary disease (COPD)

in Western countries, but there is no such evidence for resource-limited countries. This study aimed to evaluate the cost-utility of SITT compared with multiple inhaler triple therapy (MITT) for COPD in Thailand.

Methods

A Markov model with a lifetime horizon from a societal perspective was conducted with seven health states, including moderate, severe and very severe, with and without acute exacerbation (AE) and death. Two SITTs, fluticasone furoate/umeclidinium/vilanterol (FF/UMEC/VI) and budesonide/glycopyrrolate/formoterol (BUD/GLY/FOR), were compared with MITT (salmeterol/fluticasone propionate with tiotropium (SAL/FP+TIO). A comprehensive literature review was performed to identify inputs. A quality-adjusted life year (QALY) and incremental cost-effectiveness ratio (ICER) were calculated. A series of sensitivity analyses was performed.

Results

FF/UMEC/VI could improve QALY by 0.014 with an increased lifetime cost of 25 649 Thai baht (THB) ($727) compared with SAL/FP+TIO. In contrast, BUD/GLY/FOR had a lower QALY of −0.007 with an increased total lifetime cost of 34,151 THB ($968). The ICER for FF/UMEC/VI was 1,899,408 THB/QALY ($53,823), while BUD/GLY/FOR was dominated by SAL/FP+TIO. Probabilistic sensitivity analysis indicated that FF/UMEC/VI had a 1.1% chance of being cost-effective at the threshold of 160,000 THB ($4,533). We found that a 30% reduction in FF/UMEC/VI price could lead to a cost-effective option.

Conclusion

Our study indicated that SITTs are not cost-effective at the current price compared with the current MITT for COPD treatment. However, FF/UMEC/VI could be a promising option, with an approximately 30% price reduction.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Single-inhaler triple therapy (SITT) is known to be a cost-effective intervention for chronic obstructive pulmonary disease (COPD) in Western countries.

• No such evidence is available for resource-limited countries.

WHAT THIS STUDY ADDS

• SITTs are not cost-effective for COPD in Thailand at the current price.

• SITT of fluticasone furoate/umeclidinium/vilanterol will be cost-effective with a 30% decreased price.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Policymakers might consider this evidence for the National Formulary selection.

Introduction

Chronic obstructive pulmonary disease (COPD) is a common, preventable and treatable disease characterised by persistent respiratory symptoms and irreversible airflow limitation due to airway and alveolar abnormalities, most often resulting from significant exposure to noxious particles or gases.¹ In 2019, COPD affected approximately 391.9 million individuals worldwide, corresponding to a global prevalence of 10.3% among adults aged 30–79 years and accounting for 3.3 million deaths annually.^{2 3}

According to the 2025 Global Initiative for Chronic Obstructive Lung Disease (GOLD) report, initial pharmacological management of COPD emphasises the use of long-acting bronchodilators, with dual therapy combining a long-acting β2-agonist (LABA) and a long-acting muscarinic antagonist (LAMA) recommended for patients who remain symptomatic or are at increased risk of exacerbations. For those who did not have appropriate response to bronchodilators inhaled corticosteroids (ICS), alongside LABA and LAMA, is recommended, especially in patients with blood eosinophil counts ≥300 cells/µL.⁴ Consolidation of these three agents into a single-inhaler triple therapy (SITT) device has been shown to improve adherence and clinical outcomes compared with multiple-inhaler regimens.⁵,⁷

An economic evaluation in high-income countries has demonstrated the cost-effectiveness of SITT compared with multiple-inhaler triple therapy (MITT). It was reported that FF/UMEC/VI provided greater life-years and QALYs than MITT while reducing total costs by £1,764 per patient, indicating dominance in economic terms.⁸

In Thailand, no SITT has been listed in the Thai National List of Essential Medicines (NLEM). It serves as the essential list of medications. The Thai government will pay for the medication listed in NLEM for all three major public health insurance, including civil servant medical benefit schemes, social security insurance and universal coverage schemes. SITT, which is not listed in NLEM, has limited access for patients with COPD. Thus, in real-world clinical practice in Thailand, physicians usually use the MITT of salmeterol/fluticasone propionate (SAL/FP) with tiotropium (TIO) for patients with COPD, which might lead to lower adherence^{6 9} and a higher rate of acute exacerbation.^{6 10} There is a need to consider whether SITT should be listed in NLEM with regard to various factors, including safety, efficacy and economics. Even though some studies reported efficacy and safety of SITT over MITT, economic evidence of SITT compared with MITT is limited and is not transferable from one country to another. No economic evidence of SITT compared with MITT in Thailand. Evidence on the economic evaluation of SITT compared with MITT in the Thai context could help national policymakers to decide whether SITT should be considered in NLEM. In addition, such evidence could help policymakers at the institutional level to consider supporting the use of SITT for some patients with COPD who need the medication but might not be able to afford it. Therefore, this study aimed to evaluate cost-effectiveness of SITT (FF/UMEC/VI or budesonide/glycopyrrolate/formoterol; BUD/GLY/FOR) compared with MITT (SAL/FP+TIO), which is currently listed in NLEM, among patients with COPD group E, who did not respond to dual therapy from a societal perspective, according to 2025 GOLD guideline.⁴

Methods

Overall description

A cost-utility analysis from a societal perspective with a lifetime horizon was conducted to mimic COPD progression. A state transition Markov model with a 1 year cycle length was developed using Microsoft Excel 365. Patients with COPD group E, with the staging of moderate, severe and very severe, according to the 2025 GOLD guideline,⁴ were simulated. The starting age of 68 years was used according to the average age of COPD patients in Thailand.^{11 12} The proportion of starting cohorts among moderate, severe and very severe was applied based on a previous clinical trial.¹³

Interventions and comparator

The SITTs of FF/UMEC/VI (100/62.5/25 µg) once-daily and BUD/GLY/FOR (160/9/4.8 µg) twice-daily were considered because they are available in the Thai pharmaceutical market. The comparator was MITT of SAL/FP+TIO (50/250 twice daily+18 µg once daily). Patients were assumed to use theophylline control-release 200 mg once daily as another controller; however, the proportion of patients receiving theophylline was based on the severity of COPD. In addition, all patients were assumed to use ipratropium/fenoterol as a reliever. Patients were assumed to stay on their controllers regardless of COPD progression.

Model structure and assumptions

A total of seven health states was constructed based on the 2025 GOLD guideline including (1) moderate COPD with no recent acute exacerbation (AE), (2) moderate COPD with recent AE, (3) severe COPD with no recent AE, (4) severe COPD with recent AE, (5) very severe COPD with no recent AE, (6) very severe COPD with recent AE and (7) death (figure 1).

Figure 1. Markov model for patients with COPD group E who did not respond to dual therapy. AE, acute exacerbation; COPD, chronic obstructive pulmonary disease.

Patients entered the model distributionally among the COPD health states. The patients would only progress to worse COPD health states or death, with no consideration of COPD regression to better health states; however, patients could stay in the same severity but with a different history of AE. For example, patients could move from severe COPD with no recent AE to severe COPD with recent AE if they had an AE in the previous model cycle and could move from severe COPD with recent AE to severe COPD with no recent AE if they had no AE in the previous model cycle. However, patients in a severe COPD health state could not move to a moderate COPD health state. The number of annual AE and standardised mortality ratios was applied to the model based on the severity of COPD. The risk of pneumonia as an important adverse event of the SITTs and MITT was also applied based on the risk of each intervention.

Due to the lack of data on COPD progression between SITTs and MITT, the baseline COPD progression from better health states to worse health states was assumed to be equal among the interventions. Still, the risk of AE was different among the interventions.

Model inputs

Treatment effects of each intervention on acute exacerbation

A narrative review was conducted to identify the effects of FF/UMEC/VI and BUD/GLY/FOR compared with SAL/FP+TIO. A systematic review and network meta-analysis that compared the efficacy of all interventions on AE among patients with COPD was first considered. Thus, a systematic review and network meta-analysis conducted by Ismaila AS¹⁴ was selected because it provided the most updated evidence of the efficacy of all interventions of interest on AE. A pulmonologist in our research team verified the appropriateness of the inputs. Based on the network meta-analysis, FF/UMEC/VI had a lower risk of AE compared with SAL/FP+TIO and BUD/GLY/FOR. The risk of AE in patients with FF/UMEC/VI compared with SAL/FP+TIO was 0.71 (0.38–1.31), while that compared with BUD/GLY/FOR was 0.62 (0.45–0.86). The details of the inputs are presented in table 1.

Table 1. Input parameters.

Inputs	Value (SE or 95% CI)	Distribution	References
Proportion of the starting cohort
Recent AE
Moderate	0.277 (-)	Dirichlet	13
Severe	0.196 (-)	Dirichlet	13
Very severe	0.052 (-)	Dirichlet	13
No recent AE
Moderate	0.173 (-)	Dirichlet	13
Severe	0.199 (-)	Dirichlet	13
Very severe	0.086 (-)	Dirichlet	13
Proportion of patients with AE for MITT	0.439 (0.008)	Beta	6
Transitional probabilities
Moderate COPD without AE to Severe COPD	0.041 (0.008)	Beta	13
Moderate COPD with AE to Severe COPD	0.088 (0.008)	Beta	13
Severe COPD without AE to Very severe COPD	0.078 (0.008)	Beta	13
Severe COPD with AE to Very severe COPD	0.143 (0.008)	Beta	13
Acute exacerbation rate for MITT	0.71 (0.017)	Gamma	10
Treatment effect of moderate-to-severe AE
FF/UMEC/VI vs SAL/FP+TIO	0.71 (0.38 to 1.13)	Log-normal	14
FF/UMEC/VI vs BUD/GLY/FOR	0.62 (0.45 to 0.86)	Log-normal	14
Risk of pneumonia
FF/UMEC/VI	0.09 (0.07 to 0.11)	Beta	16
SAL/FP+TIO	0.09 (0.07 to 0.11)	Beta	16
Risk ratio of pneumonia (FF/UMEC/VI vs BUD/GLY/FOR)	1.00 (0.91 to 1.10)	Beta	17
Utility
Moderate COPD	0.76 (0.04)	Beta	20
Severe COPD	0.75 (0.04)	Beta	20
Very severe COPD	0.55 (0.07)	Beta	20
Disutility for acute exacerbation	0.15 (-)	Fixed	23 32
Standardised mortality ratio
Moderate COPD	1.43 (1.35 to 1.51)	Log-normal	19
Severe COPD	1.78 (.166 to 1.90)	Log-normal	19
Very severe COPD	2.13 (1.94 to 2.34)	Log-normal	19
Age-specific mortality	Varied by age	Fixed	18
Price of interventions (THB/unit)
Price of FF/UMEC/VI	1605 (±20%)	Gamma	24
Price of SAL/FP	462 (±20%)	Gamma	24
Price of TIO (Handihaler)	500 (±20%)	Gamma	24
Price of TIO (Respimat)	1348 (±20%)	Gamma	24
Price of BUD/GLY/FOR	1605 (±20%)	Gamma	24
Cost of COPD treatment (2024 value; THB/year)
Moderate-to-severe AE	6838 (±20%)	Gamma	25
Moderate COPD	6620 (±20%)	Gamma	25
Severe COPD	8899 (±20%)	Gamma	25
Very severe COPD	10 087 (±20%)	Gamma	25
% cost of medication	0.544 (-)	Fixed	26
Cost of co-interventions (THB/unit)
Cost of Ipratropium	163 (±20%)	Gamma	24
Cost of theophylline	1.34 (±20%)	Gamma	24
% patients receiving theophylline
Moderate COPD	0.315	Fixed	25
Severe COPD	0.175	Fixed	25
Very severe COPD	0.095	Fixed	25
Direct non-medical cost (2024 value; THB/visit)
Additional food cost	307	Fixed	28
Transportation cost	40	Fixed	28
Caregiver cost	638	Fixed	28
Number of visits per year
Moderate COPD	2.52 (2.37 to 2.67)	Gamma	36
Severe COPD	3.56 (3.28 to 3.84)	Gamma	36
Very severe COPD	3.95 (2.98 to 4.90)	Gamma	36
Cost of pneumonia treatment (THB/admission)	24 795 (±20%)	Gamma	27

AE, acute exacerbation; BUD, budesonide; COPD, chronic obstructive pulmonary disease; DNMC, direct non-medical cost; FF, fluticasone furoate; FOR, formoterol; FP, fluticasone propionate; GLY, glycopyrrolate; HB, Thai baht; MITT, multiple-inhaler triple therapy; SAL, salmeterol; TIO, tiotropium; UMEC, umeclidinium; VI, vilanterol.

Probability, mortality and health utility

A comprehensive literature review was conducted to identify inputs to inform the model. Studies conducted in the Thai population with suitable data were first selected, followed by studies in low- and middle-income countries in Asia, and international studies were further considered hierarchically.

According to our approach, no study in Thailand and Asia has provided suitable data on the transitional probabilities of COPD progression by a history of recent exacerbation. Thus, we decided to use the transitional probability from a previous cost-effectiveness study,¹³ while the proportion of patients with AE was derived from a study in China.⁶ As the risk of AE was not separately reported by severity, the proportion of AE by severity from the ECLIPSE study¹⁵ was used to adjust the risk of AE.

The risk of pneumonia admission of FF/UMEC/VI was derived from the IMPACT study.¹⁶ The risk of pneumonia admission of MITT was assumed from the same study,¹⁶ with ICS/LABA intervention to ensure similar baseline characteristics of the population between FF/UMEC/VI and SITT. Due to the lack of a head-to-head randomised controlled trial between FF/UMEC/VI and BUD/GLY/FOR, evidence of the comparative effect of FF/UMEC/VI and BUD/GLY/FOR on pneumonia admission was derived from an observational direct comparison study, which reported the comparative effect of 1.0 (0.91–1.11).¹⁷

Age-specific mortality of the Thai general population from the WHO¹⁸ was used to estimate the risk of death among patients with COPD, along with the standardised mortality ratio of 1.43, 1.78 and 2.13 for moderate, severe and very severe COPD, respectively.¹⁹ The health utility of each health state was derived from a previous study,²⁰ while the utility decrement of AE was from other studies.^{21 22} All clinical inputs are reported in table 1.

Cost

Direct medical costs and direct non-medical costs were identified to inform the model. The cost of productivity loss was not included according to the Thai health technology assessment guideline.²³

Prices of interventions and comparators were from the Drug and Medical Supply Information Centre, Ministry of Public Health.²⁴ The price of co-interventions, including theophylline and ipratropium/fenoterol, was also derived from the Drug and Medical Supply Information Centre.

Costs of COPD treatment and AE were derived from a previous Thai study.²⁵ However, the study reported the overall cost of COPD treatment by severity, but no report on the proportion of the cost of COPD controller medication by severity. Applying the reported cost of COPD treatment to the model directly might lead to double-counting the cost. Thus, we used the proportion of the cost of COPD medication as 54.5% of the overall direct medical cost to subtract the overall cost of COPD treatment to avoid double-counting.²⁶ The cost of pneumonia admission was also from a public online report on the website of the Health Administration Division, Ministry of Public Health.²⁷ Direct non-medical costs, including additional food costs, transportation costs and caregiver costs, were derived from the Thai standard cost list.²⁸ All cost inputs are shown in table 1. All costs were adjusted to the 2024 value using the consumer price index and converted to US dollars ($) using the average currency exchange rate of 2024 as 35.29 Thai baht (THB) per $.²⁹

Model validation

One pulmonologist and one clinical pharmacist in our research team validated the model concept and clinical inputs during two meetings. In addition, a coding and programming validation was performed by a health economist in our research team who did not construct the model.

Analysis

A discount rate of 3% was applied to both costs and outcomes. The total lifetime cost and quality-adjusted life years (QALYs) were estimated. The ICER per QALY was calculated. The willingness-to-pay (WTP) 160,000 THB per QALY ($4,533/QALY) was used as the cost-effectiveness threshold according to the Thai HTA guideline. A series of scenario analyses was performed. First, we changed the dosage form of TIO from Handihaler to Respimat to reflect the uncertainty of the different uses of MITT (Scenario 1). Second, we changed the perspective from a societal perspective to a healthcare payer perspective (Scenario 2). Last, we incorporated adherence to estimate the real-world effectiveness of the medications (Scenario 3). A previous real-world study³⁰ showed that the percentage of patients with full adherence (proportion of days covered; PDC ≥80%) of SITT was 34.4%, while that of MITT was 14.9% in 12 months. A previous meta-analysis indicated that patients with full adherence had 0.715 times lower risk of acute exacerbation than patients with poor adherence (PDC<80%).³¹ We used the information to adjust treatment effects of the FF/UMEC/VI and BUD/GLY/FOR using the following steps. First, we calculated the weighted adherence factor for each medication using equation (1).

$W_{drug}=\left(P\times Ef{f}_{adherence}\right)+\left(1-P\right)\times Ef{f}_{non-adherence}$(1)

where P is the proportion of patients with good adherence, and Eff is the effectiveness in the adherence group

Second, we adjusted the treatment effect of each medication using equation (2).

$T{E}_{adj}=T{E}_{original}\times \frac{W_A}{W_B}$(2)

where TE_adj is the adjusted treatment effect by adherence, W_A is the weighted adherence factor of the intervention, and W_B is the weighted adherence factor for the comparator.

Third, we applied the adjusted treatment effect as the treatment effect of the medications. According to the calculation, the treatment effect of FF/UMEC/VI changed from 0.71 to 0.67, while that of BUD/GLY/FOR changed from 1.14 to 1.08.

One-way sensitivity and probabilistic sensitivity analyses were performed to explore the effects of the inputs’ uncertainty. The one-way sensitivity analysis’s finding is presented as a tornado diagram, while the findings of probabilistic sensitivity analysis are presented as a cost-effectiveness analysis plane and a cost-effectiveness acceptability curve. In addition, a threshold analysis of the price of interventions was performed when the intervention was not cost-effective to determine which price should be considered in the negotiation process.

Results

Base-case and scenario analysis results

The total lifetime cost of SAL/FP+TIO was 118,607 THB ($3,361) with a total QALY of 2.648. The cost of SAL/FP+TIO was approximately 39.5% of the total cost, while other direct medical costs were 41.7% of the total cost (figure 2). The total lifetime costs of FF/UMEC/VI and BUD/GLY/FOR were 1 44 255 THB ($4088) and 1 52 758 THB ($429), respectively. The cost of FF/UMEC/VI was approximately 54.4% of the total lifetime cost, while the cost of BUD/GLY/FOR was 51.2% of the total lifetime cost (figure 2). Total QALYs of FF/UMEC/VI and BUD/GLY/FOR were 2.662 and 2.642, respectively (table 2).

Figure 2. Cost components for each intervention. BUD, budesonide; DNMC, direct non-medical cost; FF, fluticasone furoate; FOR, formoterol; FP, fluticasone propionate; GLY, glycopyrrolate; SAL, salmeterol; TIO, tiotropium; UMEC, umeclidinium; VI, vilanterol.

Table 2. Cost-effectiveness analysis results.

Intervention	Total cost (THB)	Total life-years	Total QALY	Incremental cost (THB)	Incremental QALY	Icer (THB/QALY)	Icer ($/QALY)
Base-case analysis (MITT as SAL/FP+TIO handihaler)
SAL/FP+TIO	118 607	4.063	2.648	Reference	Reference	Reference	Reference
FF/UMEC/VI	144 255	4.072	2.662	25 649	0.0135	1 899 408	53 823
BUD/GLY/FOR	152 758	4.059	2.642	34 151	−0.0067	Dominated	Dominated
Scenario analysis 1 (MITT as SAL/FP+TIO Respimat)
SAL/FP+TIO	159 981	4.063	2.648	Reference	Reference	Reference	Reference
FF/UMEC/VI	144 255	4.072	2.662	−15,725	0.0135	Dominant	Dominant
BUD/GLY/FOR	152 758	4.059	2.642	−7,223	−0.0067	Lower cost but lower QALY	Lower cost but lower QALY
Scenario analysis 2 (Healthcare payer perspective)
SAL/FP+TIO	105 586	4.063	2.648	Reference	Reference	Reference	Reference
FF/UMEC/VI	131 246	4.072	2.662	25 660	0.0135	1 900 271	53 847
BUD/GLY/FOR	139 731	4.059	2.642	34 145	−0.0067	Dominated	Dominated
Scenario analysis 3 (Adding adherence factor)
SAL/FP+TIO	118 607	4.063	2.648	Reference	Reference	Reference	Reference
FF/UMEC/VI	143 883	4.072	2.663	25 276	0.0141	1 787 784	50 659
BUD/GLY/FOR	151 485	4.061	2.645	32 879	−0.0037	Dominated	Dominated

BUD, budesonide; FF, fluticasone furoate; FOR, formoterol; FP, fluticasone propionate; GLY, glycopyrrolate; ICER, incremental cost-effectiveness ratio; MITT, multiple inhaler triple therapy; QALY, quality-adjusted life year; SAL, salmeterol; THB, Thai baht; TIO, tiotropium; UMEC, umeclidinium; VI, vilanterol.

Compared with SAL/FP+TIO Handihaler, FF/UMEC/VI could improve 0.009 life-years and 0.014 QALYs. However, it required an additional total lifetime cost of 25 649 THB ($727), resulting in the ICER of 1 899 408 THB/QALY ($53 823/QALY). On the other hand, BUD/GLY/FOR did not improve patients’ life-years and QALYs compared with SAL/FP+TIO Handihaler. BUD/GLY/FOR had a lower QALY of −0.007 with an increased total lifetime cost of 34,151 THB ($968).

A scenario analysis by changing the dosage form of TIO from Handihaler to Respimat showed different findings from the base-case analysis findings. FF/UMEC/VI could improve 0.014 QALY with a decrease in total lifetime cost of −15,725 THB ($−446). In contrast, BUD/GLY/FOR had a lower QALY of −0.007, with a decrease in total lifetime cost of −7,223 THB ($−205) (table 2).

Sensitivity analysis

One-way sensitivity analysis indicated that the price of FF/UMEC/VI was the most influential input for FF/UMEC/VI for the incremental cost and treatment effect of moderate-to-severe AE of FF/UMEC/VI, compared with SAL/FP+TIO, which was the most influential input for the incremental QALY. For BUD/GLY/FOR, the most influential input for both incremental cost and incremental QALY was the treatment effect of moderate-to-severe AE of FF/UMEC/VI compared with SAL/FP+TIO. Tornado diagrams for all one-way sensitivity analyses are presented in online supplemental figure S1.

Probabilistic sensitivity analysis of FF/UMEC/VI compared with SAL/FP+TIO indicated that 85.9% fell into the right-upper quadrant of the cost-effectiveness analysis plane, while 13.6% fell into the left-upper quadrant (online supplemental figure S2). The cost-effectiveness acceptability curve analysis showed that 1.1% of FF/UMEC/VI was likely to be cost-effective at the current WTP threshold. FF/UMEC/VI could be more cost-effective than SAL/FP+TIO at the WTP of 2,500,000 THB ($70,842) (figure 3).

Figure 3. Cost-effectiveness acceptability curve. BUD, budesonide; FF, fluticasone furoate; FOR,formoterol; FP, fluticasone propionate; GLY, glycopyrrolate; SAL, salmeterol; THB, Thai baht; TIO, tiotropium; UMEC, umeclidinium; VI, vilanterol.

For BUD/GLY/FOR, probabilistic sensitivity analysis showed that the most iterations fell into the left-upper quadrant (65.1%), followed by the right-upper quadrant (34.8%) (online supplemental figure S2). The cost-effectiveness acceptability curve analysis revealed 0.2% of BUD/GLY/FOR being cost-effective at the current WTP threshold. (figure 3)

Threshold analysis

According to the base-case analysis result, FF/UMEC/VI was not cost-effective, while BUD/GLY/FOR was dominated by SAL/FP+TIO (MITT) with a higher cost and a lower QALY. Therefore, a threshold analysis of FF/UMEC/VI should be performed, while that of BUD/GLY/FOR might not be warranted. Our threshold analysis indicated that the price of FF/UMEC/VI should be reduced from 1,605 THB per canister ($45.5) to 1124 THB per canister ($31.9) or approximately 30.0% to reach an ICER of 160,000 THB/QALY ($4,533/QALY).

Probabilistic sensitivity analysis of FF/UMEC/VI at the threshold price (1,124 THB; $31.9) showed that 49.6% fell into the right-upper quadrant, 35.9% into the right-lower quadrant and 13.7% into the left-upper quadrant (online supplemental figure S3). Cost-effectiveness acceptability curve analysis showed 45.9% of FF/UMEC/VI to be cost-effective at the current WTP threshold (online supplemental figure S3).

Discussion

This study assessed the cost-effectiveness of SITTs compared with the current practice of MITT (SAL/FP+TIO) for COPD patients who did not adequately respond to dual therapy in Thailand. We found that both SITTs (FF/UMEC/VI and BUD/GLY/FOR) are not cost-effective at the current market price. Breaking down to specific treatments, FF/UMEC/VI could provide better QALY with an increased cost; in contrast, BUD/GLY/FOR does not provide better QALY but requires additional lifetime cost compared with SAL/FP+TIO.

Our findings contradict a previous UK cost-effectiveness study.⁸ The study indicated that SITT of FF/UMEC/VI could improve QALY and decrease total lifetime cost compared with non-ELLIPTA MITT. It might be due to several inputs and model aspects. First, the UK study used the efficacy of FF/UMEC/VI from the INTREPID study,⁵ which assessed the effect of SITT over non-ELLIPTA MITT on COPD assessment test (CAT) score, while our study used the efficacy evidence of SITTs (both FF/UMEC/VI and BUD/GLY/FOR) from the most updated network meta-analysis of randomised controlled trials, which could provide trustworthy evidence on AE.¹⁴ Second, the UK study used the GALAXY model, which is the common economic model for COPD to evaluate the cost-effectiveness[30], while our model was based on the model concept from Martin et al,¹³ which is similar to the model used in cost-effectiveness studies for COPD in Thailand.^{32 33} Finally, the overall health systems between the UK and Thailand are different, which could affect healthcare costs, such as AE cost, cost of COPD and cost of pneumonia treatment.

Our sensitivity analysis indicated that the prices of both SITTs are important inputs that could affect the findings. The prices of the interventions could be changed over time, especially when they can be reimbursed. Thus, we addressed this uncertainty by performing a threshold analysis to determine the prices of interventions that could be the cost-effective options. According to the base-case findings that FF/UMEC/VI provided better health outcomes but BUD/GLY/FOR did not, we performed the threshold analysis for only FF/UMEC/VI. We found that the prices of FF/UMEC/VI should be decreased by at least 30% of the current price to reach the WTP threshold. In addition, we conducted a probabilistic sensitivity analysis of FF/UMEC/VI at the threshold price and found that 45.6% of FF/UMEC/VI is cost-effective. This information is crucial for the Thai government in the price negotiation process with the pharmaceutical company if the Thai government considers listing the medication in the NLEM.

The price of SAL/FP+TIO was another important input of the model. We performed a scenario analysis to determine whether the TIO is changed from Handihaler to Respimat. We found that FF/UMEC/VI turned out to be a dominant option, while BUD/GLY/FOR had a lower cost but lower QALY. These findings showed the great uncertainty of the use of TIO as an available LAMA for current practice. To our knowledge, TIO Handihaler is still the treatment of choice for LAMA among patients with COPD, but the use of TIO Respimat is emerging. Thus, health policy decision makers should consider it an important issue.

One important influential input of the model was the treatment effect of SITTs compared with MITT. Even though there is no direct head-to-head randomised controlled trial to compare SITTs to MITT, a network meta-analysis could provide important information for our model.¹⁴ However, there are some limitations to the network meta-analysis, such as the lack of severe AE outcomes, which limit our analysis on this aspect. Therefore, further randomised controlled trials addressing the efficacy and safety of SITTs compared with MITT are still warranted.

Our study was conducted from a societal perspective; however, the cost of productivity (indirect cost) was not included in our analysis. This exclusion could potentially affect the findings and conclusions. However, the Thai policy decision-makers used cost-effectiveness data from a cost-utility analysis study under a societal perspective as the primary data according to the Thai Health Technology Assessment guideline,²³ which recommends excluding the cost of productivity for the analysis to avoid double-counting of indirect cost and utility. Thus, we conducted this study according to the guidelines.

Our scenario analyses by changing to a healthcare payer perspective and adding adherence to adjust treatment effect showed similar results compared with the base-case analysis results. The results from the healthcare payer perspective had lower incremental lifetime costs than those from the societal perspective; however, the conclusion was the same, as the FF/UMEC/VI and BUD/GLY/FOR were not cost-effective at the current prices. Similarly, adding adherence to adjust treatment effect could result in lower incremental lifetime cost and better incremental QALY for both medications, but the conclusion remained the same that both medications were not cost-effective. Our analyses showed the robustness of the results.

The pneumonia rate was an important factor in the model. The rate of adverse events could increase the overall lifetime cost of each intervention. In our study, we used the pneumonia rate of FF/UMEC/VI from the IMPACT study.¹⁶ We decided to use the pneumonia rate of FF/VI from the IMPACT study as a proxy for SAL/FP+TIO to reduce the heterogeneity between the comparisons. In addition, instead of the use of the pneumonia rate of BUD/GLY/FOR from the landmark ETHOS study,³⁴ we decided to use the relative effect of FF/UMEC/VI compared with BUD/GLY/FOR on pneumonia from an observational study.¹⁷ This approach aimed to reduce the heterogeneity between the IMPACT and the ETHOS studies. Even though we applied the overall pneumonia rate observed from the abovementioned studies, we acknowledged that pneumonia could occur without any treatment, which was observed in the ECLIPSE study.¹⁵ In our analysis, we did not take into account the pneumonia rate due to the nature of the disease.

Although the incremental life-years and QALYs of FF/UMEC/VI compared with SAL/FP+TIO were relatively modest, the findings should be interpreted alongside other clinical outcomes, which are also relevant to patients with COPD. FF/UMEC/VI provides better symptom reduction and control, health-related quality of life and disease stability, which affect patients’ daily activities.³⁵ FF/UMEC/VI – SITT consists of ultra-long bronchodilators, which could support more stable symptom control than SAL/FP+TIO – MITT due to the fluctuations in bronchodilation and symptom control. Therefore, healthcare policymakers should consider our cost-effectiveness findings along with other clinical outcomes and whether SITT should be listed in NLEM or the institutional drug formulary.

Several limitations should be discussed. First, because of the lack of evidence on transitional probabilities among COPD stages in Thai patients, we need to use data from a randomised controlled trial, which might not directly reflect the real-world epidemiologic data in Thailand. Thus, further epidemiological studies on the progression of COPD in Thailand are needed. Second, we used the cost of COPD treatment from a Thai study,²⁵ which was conducted in 2015. The treatment patterns at present might be different from what was observed in 2015. It might also affect the overall cost of COPD treatment. However, to our knowledge, no updated Thai study estimates the cost of COPD treatment by GOLD stages in Thailand. Thus, we believe the data is acceptable for our study. Third, the Thai COPD cost study²⁵ did not comprehensively provide the cost structure of COPD treatment; we need to adjust the cost of COPD controller use using the proportion of medication cost to the overall total COPD treatment cost from the APBORD observational study.²⁶ This might lead to the large uncertainty around the cost of COPD treatment. However, we performed a one-way sensitivity analysis and found that the cost of COPD treatment did not have a significant impact on the findings. Thus, this approach could be acceptable. Finally, the unit cost used in this study was based on a standard cost list that estimated the healthcare costs in the year 2009. Even though we adjusted the cost to 2024 value using the consumer price index, the cost might not accurately reflect the cost in the year 2024 because of the changes in healthcare technologies and healthcare systems over the period.

Our study provides economic evidence for the value for money of SITT compared with MITT for Thai health technology assessment and the national policymakers to consider SITT in NLEM. In addition, our study provides empirical evidence for other countries, especially low-and-middle income countries, where health technology assessment capacity is limited. However, the transferability of our findings to other countries should be carefully considered.

In conclusion, our study indicated that neither SITT (FF/UMEC/VI or BUD/GLY/FOR) is cost-effective at the current price compared with the current MITT (SAL/FP+TIO) for COPD treatment. However, FF/UMEC/VI could be a promising option, especially with an approximately 30% price reduction.

Data availability statement

Data are available upon reasonable request.