Clinical and economic outcomes associated with the use of fluticasone propionate 250 mcg and salmeterol 50 mcg combination versus tiotropium bromide 18 mcg as initial maintenance treatment for chronic obstructive pulmonary disease in managed care

Abstract

Aims: To examine the clinical and economic outcomes associated with the use of long-acting bronchodilators for initial maintenance treatment of chronic obstructive pulmonary disease (COPD) by analyzing health insurance claims data in the US.

Methods: A retrospective, observational, matched cohort study used health insurance claims data (January 2008 to June 2013) to assess COPD-related outcomes for subjects aged ≥40 years. Subjects were assigned to a study cohort according to the first observed prescription fill for a long-acting bronchodilator (fluticasone propionate 250 mcg/salmeterol 50 mcg [FSC] or tiotropium bromide 18 mcg [TIO]). The analysis period for each subject comprised a 1-year pre-index date and 1-year post-index date. Primary outcome measure was total COPD-related costs per-patient per-year (PPPY) during the follow-up period. Secondary outcome measures included COPD-related exacerbations and the components of COPD-related costs.

Results: Overall, 24,040 subjects were identified; the analysis sample consisted of 19,090 subjects (9,545 per cohort) with no significant differences between cohorts. Mean COPD-related total costs PPPY were numerically lower among the FSC cohort; however, the difference was not statistically significant ($2,224 [±4,108] vs $2,352 [±3,721], p = .057). There was no difference between cohorts for COPD-related medical costs (p = .894). COPD-related pharmacy costs were significantly, yet modestly, lower in the FSC cohort compared with the TIO cohort ($1,160 [±1,106] vs 1,275 [±1,110], p < .001). There were no statistically significant differences in the rate or number of exacerbations between the matched cohorts.

Limitations: While propensity scoring achieved balance in baseline characteristics, some residual confounding unobserved in the database may be present.

Conclusions: Few clinical and economic differences between subjects initiating maintenance therapy with FSC or TIO were observed.

Keywords:

• Chronic obstructive pulmonary disease (COPD)
• fluticasone propionate
• salmeterol tiotropium bromide
• maintenance treatment naïve COPD
• healthcare claims
• economic burden
• clinical outcomes

JEL classification codes:

• I11
• I12

Introduction

Chronic obstructive pulmonary disease (COPD) is a life-threatening, chronic lung disease that is preventable and treatable, yet remains a major public health problem due to its high prevalence significant morbidity, and high mortality¹. The condition is characterized by airflow limitation and gradual loss of lung function¹. Common COPD symptoms include breathlessness/dyspnea, abnormal sputum (saliva and mucus in the airway), and a chronic cough, making daily activities difficult (e.g. walking up a short flight of stairs)^1,2. Many patients with COPD experience periodic worsening of their symptoms or exacerbations³. These events reflect an acute deterioration in lung mechanics and airway inflammation, resulting in worsening respiratory symptoms beyond the normal experienced daily variation^1,4–6; contributing to impaired health status, increased hospitalizations, costs, and increased risk of mortality^7–12.

In the US, COPD is the third leading cause of death^13–15. Approximately 15 million US adults have received a physician diagnosis of COPD; however, when considering the number of US adults who also have evidence of impaired lung function, this estimate increases to 27 million^16,17. In the US, COPD-related hospitalizations and emergency department (ED) visits numbered 700,000 and 1.5 million, respectively, with 10.3 million COPD-related outpatient visits occurring in 2010¹⁸. Estimates of COPD’s economic burden in the US rose from $37.2 billion in 2004¹⁹ to $42.6 billion in 2007²⁰ to $49.9 billion in 2010^21,22. Of the latter, $29.5 billion is attributable to direct medical costs ($13.2 billion in hospital care costs, $5.5 billion in physician services, and $5.8 billion in prescription drugs) and $20.4 billion to indirect costs²¹. Another study found a total medical cost estimate for the US in 2010 of $32.1 billion, when adjusting for demographic differences and conditions not thought to be COPD sequelae²².

There is no cure for COPD, and treatment is focused on managing symptoms, reducing the severity and frequency of exacerbations, reducing mortality, preventing disease progression, and improving the patient’s health-related quality-of-life (HRQoL)¹. Despite the availability of pharmacologic interventions with demonstrated efficacy in the clinical trial setting^23–25, studies show that patients with COPD continue to experience COPD-related symptoms^26–28. A number of studies have shown symptomatic patients have higher all-cause healthcare resource utilization and costs compared with those who are asymptomatic^27,29. Thus, COPD management has become a major challenge for healthcare professionals, public and private healthcare payers, and policy-makers³⁰.

For patients with moderate-to-severe COPD, maintenance therapy with long-acting bronchodilators (e.g. long-acting β₂-adrenergic agonists [LABAs] or long-acting anticholinergics or muscarinic antagonists [LAMAs]) is recommended, with the addition of inhaled corticosteroids (ICS) for those patients with repeated exacerbations and those with more severe disease¹. In some patient populations, the combination of ICS plus LABA for maintenance therapy, including fixed-dose combination fluticasone propionate 250 mcg + salmeterol 50 mcg (FSC), has been associated with a significantly lower risk for COPD-related hospitalizations, ED visits, and healthcare costs compared with short-acting muscarinic antagonists (SAMA)^31–33. Similar findings have been observed for tiotropium bromide 18 mcg (TIO), where the LAMA was shown to reduce exacerbations and COPD-related hospitalizations compared with SAMA^34,35. Studies have also directly compared clinical and economic outcomes of FSC and TIO^36–41. Halpern et al.³⁹ found that patients with COPD initiating maintenance therapy were more likely to be adherent to TIO than to FSC, and that adherence to either TIO or FSC was associated with lower respiratory-related outpatient medical and inpatient costs, and with higher respiratory-related pharmacy costs. Dalal et al.³⁶ observed that initiating maintenance therapy with FSC compared with TIO was associated with significant reductions in the risk of severe exacerbations, healthcare utilization, and COPD-related costs. A limitation of these previous studies is the period of observation, with earlier periods of time selected for analyses. In this regard, the current study was designed to further explore the clinical and economic outcomes associated with the use of FSC and TIO for initial maintenance treatment of COPD using more recent data.

Methods

Study design

A retrospective, matched cohort study was conducted using administrative healthcare claims (medical, pharmacy, and enrollment data) to assess COPD-related outcomes and costs in subjects diagnosed with COPD and initiating maintenance treatment with FSC or TIO. Data from the Truven Health MarketScan Commercial Claims and Encounters and Medicare Supplemental Databases (CCAE-MSD) (January 1, 2008–June 30, 2013) were used for analysis. The data encompassed employees, retirees, their spouses, and dependents covered by employer-sponsored private health insurance, representing more than 100 large employers and 12 health plans throughout the US.

Figure 1 depicts the study design scheme used. Subjects initiating maintenance therapy with FSC or TIO during the enrollment period, January 1, 2009 to June 30, 2012, comprised the study population. The index date was the date of the first prescription fill for FSC or TIO during the enrollment period. The analytical time period for each subject comprised a 1-year pre-index and 1-year post-index period.

Figure 1. Study design scheme. Abbreviations. COPD, chronic obstructive pulmonary disease; FSC, fluticasone propionate 250 mcg + salmeterol 50 mcg; TIO, tiotropium bromide 18 mcg. ^aIndex date = first date of fill for FSC or TIO during enrollment period.

Sample selection

The study population included men and women 40 years of age or older with a COPD diagnosis (International Classification of Diseases, 9^th Edition, Clinical Modification [ICD-9-CM] diagnosis codes: 491.xx [chronic bronchitis], 492.xx [emphysema], 496.xx [chronic airway obstruction]) recorded on ≥1 medical claim for a hospitalization or ED visit in the primary position or in any position on ≥2 outpatient visits with different dates of service. Subjects were naïve to COPD maintenance treatment in the pre-index period. Finally, subjects were required to have continuous health plan and pharmacy benefit eligibility across the pre-index and post-index periods. Subjects were excluded if they switched between or added the alternative study medication or any other maintenance therapies during the first 60 days of the post-index period; and if they had, during the analytical period (pre- or post-index), evidence of other therapies not approved for COPD, other respiratory conditions (e.g. tuberculosis, sarcoidosis, cystic fibrosis), or ≥1 capitated claim.

Subjects were assigned to the FSC or TIO study cohort according to the first observed prescription fill. The study design employed an intent-to-treat (ITT) analysis approach, where subjects who switched to or added other maintenance therapies or index medications after the first 60-days of follow-up or discontinued their index medication and did not have any other maintenance therapies at any time after the index date remained assigned to their index therapy cohort.

Study measures

The primary outcome measure included total COPD-related costs estimated per-patient per-year (PPPY) during the follow-up period and derived from adjudicated medical and pharmacy claims. COPD-related medical costs were defined as payments associated with medical claims having a primary diagnosis of COPD, and COPD-related pharmacy costs were defined as payments associated with prescription fills for COPD-related medications, including maintenance and reliever therapies. Cost data represent actual dollar amounts reimbursed by the payer; therefore, exclude payments made by the patient (i.e. deductible, copayment, or coinsurance). All cost estimates were adjusted to 2013 US dollars using the medical care component of the Consumer Price Index.

The secondary outcomes measures included COPD-related exacerbations and COPD-related medical and pharmacy costs. Exacerbations were defined as one of the following: (1) a physician office visit with a primary diagnosis of COPD and a pharmacy claim for an oral corticosteroid or antibiotic (OCS/AB) within 5 days of the visit, where the date of the physician visit was considered the date of the exacerbation; (2) a hospitalization with a diagnosis of COPD, where the date of admission was considered the date of the exacerbation; or (3) an ED visit with a primary diagnosis of COPD, where the date of the visit was considered the date of the exacerbation. However, a COPD-related ED visit occurring within 14 days prior to a COPD-related hospitalization was deemed a COPD-related hospitalization, with the date of the ED visit set as the date of the exacerbation. Two composite COPD-related exacerbation endpoints were also computed: (1) hospitalization or ED visit; and (2) hospitalization, ED visit, or physician office visit with OCS/AB prescription. The components of COPD-related medical costs were categorized as hospitalizations, ED visits, physician office visits, laboratory, radiology, and “other outpatient”. The components of COPD-related pharmacy costs were categorized as index and non-index medication costs. While pharmacy claims data were available, assessment of drug utilization (e.g. adherence, compliance, medication possession ratio) was not a pre-specified endpoint; and, therefore, not part of the analysis.

Statistical analysis

Demographic, economic, and clinical characteristics were evaluated during the pre-index period and included age, sex, geographic region, pulmonologist visit, index year, and season of index date. The Dartmouth-Manitoba adaptation of Charlson Comorbidity Index (CCI)^42–45 score was calculated for each subject based on the presence of ICD-9-CM codes during the 1-year pre-index period. Asthma and cardiovascular disease (CVD) diagnoses were included as separate variables due to their linkage with COPD; therefore, they were excluded from the CCI score calculation. Additional pre-index comorbidity measures included the number of unique prescription drug classes, the number of unique diagnoses (beyond those captured by the CCI), and the presence of other select comorbidities defined as medical claims with a diagnosis or procedure code in any field for asthma, depression, upper or lower respiratory tract infections (URTI or LRTI), and CVD. Finally, COPD disease severity was estimated for each subject via proxy measures including: (1) evidence of pharmacy claims (pre-/post-index) for each of the following: short-acting beta agonists (SABAs), SAMAs, OCS, and AB; (2) home oxygen therapy (pre-index); (3) mechanical ventilation or intubation during hospitalization of ED visit (pre-index); (4) COPD-related exacerbations (pre-index); (5) COPD sub-type based on index date ICD-9-CM diagnosis codes (for subjects with multiple COPD diagnoses codes, the following hierarchy was applied: emphysema, chronic bronchitis, chronic airway obstruction); and (6) COPD-related medical and prescription costs (pre-index).

Descriptive statistics were generated for all analysis variables, including frequency distributions for categorical measures and means and standard deviations (SDs) for continuous measures. Between-group differences were tested using Chi-square tests for categorical measures and Student t-tests or Wilcoxon-Mann Whitney non-parametric tests for continuous measures. Matching was performed, via propensity scores, to balance the distribution of observed confounders between the two cohorts, thereby reducing the possibility of selection bias caused by observed baseline differences between subjects who received FSC and those who received TIO. The propensity score for a study subject was defined as the probability of being treated with FSC vs TIO conditional on the subject’s age, sex, geographic region, pulmonologist visit, index year, season of index, comorbidity measures (CCI and diagnosis of asthma, depression, URTI, LRTI, and CVD), and proxies of COPD severity (SABA, SAMA, OCS, COPD-related exacerbation resulting in hospitalization or ED visit, COPD-related exacerbation resulting in physician office visit and prescription for AB, home oxygen use, mechanical ventilation/intubation during any hospitalization or ED visit and COPD sub-type). Matching was 1:1, using the greedy-matching technique up to three decimal places (0.001). Balance was subsequently assessed by comparing the covariates between the cohorts after matching using standardized differences, with a threshold level of <10% deemed an acceptable level of difference (i.e. supporting the assumption of balance) between the comparison cohorts.

To examine differences in COPD-related costs post-index (primary outcome measure), a generalized estimating equation (GEE) model with a log-link function was used. For the secondary outcome measures, McNemar’s test was used to assess differences in rates of COPD exacerbations and paired t-test to assess the differences in the number of COPD exacerbations in the follow-up period. Finally, a semi-log ordinary least squares (OLS) regression model was used to assess differences in the components of COPD-related medical costs (modeling approach selected based on the underlying characteristics of the data, where semi-log OLS would provide a better fit as compared with a GEE model). All analyses were performed for the overall sample, as well as for the sample stratified by age (40–64 years and ≥65 years). All statistical analyses were conducted using SAS (Statistical Analysis System, version 9.2, SAS Institute, Inc., Cary, NC) with a two-sided a priori significance level of α = 0.05.

Results

The subjects’ disposition and the identification of the study population are shown in Figure 2. A total of 24,040 subjects satisfied the inclusion/exclusion criteria, with 11,141 (46%) subjects in the FSC cohort and 12,899 (54%) subjects in the TIO cohort. The baseline characteristics of the FSC and TIO cohorts are reported in Table 1. Prior to matching, respiratory comorbid burden was significantly higher in the FSC cohort compared with the TIO cohort: asthma (26.8% vs 15.2%, p < .001), URTI (24.8% vs 20.5%, p < .001), and LRTI (39.5% vs 33.5%, p < .001). The FSC cohort had more severe COPD at baseline, with a significantly higher proportion having any COPD-related exacerbation (42.2% vs 35.5%, p < .001), significantly higher COPD-related costs ($2,078 [±$5,470] vs $1,808 [±$5,990], p < .001), and significantly higher rates of reliever medication use (total subjects with a SAMA prescription: 19.1% vs 16.4%; p < .001 and mean number of SAMA prescriptions: 0.58 vs 0.49; p < .001). The FSC cohort also had a significantly higher proportion with an index exacerbation diagnosis of chronic bronchitis (43.2% vs 34.2%, p < .001) than the TIO cohort.

Figure 2. Identification of study population. Abbreviations. COPD, chronic obstructive pulmonary disease; Dx, diagnosis; FSC, fluticasone propionate 250 mcg + salmeterol 50 mcg; TIO, tiotropium bromide 18 mc.

Table 1. Baseline characteristics before and after propensity score matching.

	Before propensity score matching				After propensity score matching
Characteristic	Total (n = 24,040)	FSC (n = 11,141)	TIO (n = 12,899)	p-value^d	Total (n = 19,090)	FSC (n = 9,545)	TIO (n = 9,545)	p-value^e
Demographic characteristics
Age, mean (SD)	67.3 (11.8)	66.9 (12.1)	67.7 (11.5)	<.001*	67.4 (11.8)	67.4 (11.9)	67.5 (11.6)	.404
Females, n (%)	12,804 (53.3%)	6,076 (54.5%)	6,728 (52.2%)	<.001*	10,126 (53.0%)	5,085 (53.3%)	5,041 (52.8%)	.529
Geographic region, n (%)
Northeast	4,853 (20.2%)	2,131 (19.1%)	2,722 (21.1%)	<.001*	3,714 (19.5%)	1,874 (19.6%)	1,840 (19.3%)	N/A^f
North Central	8,254 (34.3%)	3,701 (33.2%)	4,553 (35.3%)		6,533 (34.2%)	3,244 (34.0%)	3,289 (34.5%)
South	7,736 (32.2%)	3,763 (33.8%)	3,973 (30.8%)		6,282 (32.9%)	3,136 (32.9%)	3,146 (33.0%)
West	2,723 (11.3%)	1,322 (11.9%)	1,401 (10.9%)		2,180 (11.4%)	1,098 (11.5%)	1,082 (11.3%)
Unknown	474 (2.0%)	224 (2.0%)	250 (1.9%)		381 (2.0%)	193 (2.0%)	188 (2.0%)
Pulmonologist visit, n (%)	6,304 (26.2%)	2,533 (22.7%)	3,771 (29.2%)	<.001*	4,635 (24.3%)	2,325 (24.4%)	2,310 (24.2%)	.795
Index year
2009	8,287 (34.5%)	3,999 (35.9%)	4,288 (33.2%)	<.001*	6,750 (35.4%)	3,364 (35.2%)	3,386 (35.5%)	N/A^f
2010	6,872 (28.6%)	3,299 (29.6%)	3,573 (27.7%)		5,489 (28.8%)	2,736 (28.7%)	2,753 (28.8%)
2011	6,264 (26.1%)	2,720 (24.4%)	3,544 (27.5%)		4,845 (25.4%)	2,428 (25.4%)	2,417 (25.3%)
2012	2,617 (10.9%)	1,123 (10.1%)	1,494 (11.6%)		2,006 (10.5%)	1,017 (10.7%)	989 (10.4%)
Season of index date, n (%)
Spring (March–May)	7,196 (29.9%)	3,352 (30.1%)	3,844 (29.8%)	<.001*	5,754 (30.1%)	2,889 (30.3%)	2,865 (30.0%)	N/A^f
Summer (June–August)	5,183 (21.6%)	2,266 (20.3%)	2,917 (22.6%)		3,994 (20.9%)	1,990 (20.9%)	2,004 (21.0%)
Fall (September–November)	4,692 (19.5%)	2,154 (19.3%)	2,538 (19.7%)		3,755 (19.7%)	1,875 (19.6%)	1,880 (19.7%)
Winter (December–February)	6,969 (29.0%)	3,369 (30.2%)	3,600 (27.9%)		5,587 (29.3%)	2,791 (29.2%)	2,796 (29.3%)
Comorbidity
CCI^a, mean (SD)	1.15 (1.50)	1.18 (1.50)	1.13 (1.50)	0.003*	1.2 (1.5)	1.2 (1.5)	1.2 (1.5)	.639
Unique Rx classes, mean (SD)	9.9 (6.1)	10.2 (6.2)	9.6 (6.0)	<.001*	10.0 (6.1)	10.0 (6.1)	9.9 (6.1)	.050
Unique Dx codes^b, mean (SD)	13.0 (8.7)	13.2 (8.9)	12.8 (8.6)	<.001*	12.9 (8.7)	13.0 (8.8)	12.9 (8.6)	.604
Specific comorbidities, n (%)
Asthma	4,944 (20.6%)	2,985 (26.8%)	1,959 (15.2%)	<.001*	3,758 (19.7%)	1,888 (19.8%)	1,870 (19.6%)	.657
Depression	3,952 (16.4%)	1,818 (16.3%)	2,134 (16.5%)	0.638	3,103 (16.3%)	1,534 (16.1%)	1,569 (16.4%)	.501
URTI	5,409 (22.5%)	2,764 (24.8%)	2,645 (20.5%)	<.001*	4,341 (22.7%)	2,170 (22.7%)	2,171 (22.7%)	1.000
LRTI	8,723 (36.3%)	4,402 (39.5%)	4,321 (33.5%)	<.001*	7,061 (37.0%)	3,532 (37.0%)	3,529 (37.0%)	.975
CVD	15,587 (64.8%)	7,085 (63.6%)	8,502 (65.9%)	<.001*	12,311 (64.5%)	6,163 (64.6%)	6,148 (64.4%)	.829
COPD severity
SAMA (total), n (%)	4,243 (17.7%)	2,122 (19.1%)	2,121 (16.4%)	<.001*	3,468 (18.2%)	1,726 (18.1%)	1,742 (18.3%)	.777
SAMA Rx (total), mean (SD)	0.53 (1.8)	0.58 (1.9)	0.49 (1.7)	<.001*	0.6 (1.9)	0.6 (1.9)	0.6 (1.8)	.866
SABA, n (%)	8,958 (37.3%)	4,380 (39.3%)	4,578 (35.5%)	<.001*	7,169 (37.6%)	3,570 (37.4%)	3,599 (37.7%)	.666
SABA canisters, mean (SD)	1.5 (4.7)	1.6 (4.8)	1.4 (4.6)	<.001*	1.5 (5.0)	1.5 (4.9)	1.5 (5.0)	.917
OCS, n (%)	9,253 (38.5%)	4,727 (42.4%)	4,526 (35.1%)	<.001*	7,504 (39.3%)	3,721 (39.0%)	3,783 (39.6%)	.339
OCS Rx, mean (SD)	0.77 (1.5)	0.86 (1.5)	0.70 (1.4)	<.001*	0.8 (1.5)	0.8 (1.5)	0.8 (1.5)	.369
Antibiotics, n (%)	15,826 (65.8%)	7,657 (68.7%)	8,169 (63.3%)	<.001*	12,673 (66.4%)	6,376 (66.8%)	6,297 (66.0%)	.218
Antibiotic Rx, mean (SD)	1.7 (2.1)	1.8 (2.1)	1.6 (2.0)	<.001*	1.7 (2.1)	1.7 (2.0)	1.7 (2.1)	.610
Home oxygen therapy, n (%)	3,479 (14.5%)	1,566 (14.1%)	1,913 (14.8%)	0.089	2,768 (14.5%)	1,383 (14.5%)	1,385 (14.5%)	.984
Mechanical ventilation/intubation in hospitalization/ED visit, n (%)	219 (0.9%)	96 (0.9%)	123 (1.0%)	0.455	176 (0.9%)	88 (0.9%)	88 (0.9%)	1.000
Exacerbations (any), n (%)	9,273 (38.6%)	4,700 (42.2%)	4,573 (35.5%)	<.001*	7,536 (39.5%)	3,776 (39.6%)	3,760 (39.4%)	.815
Exacerbations (all subjects), mean (SD)	0.48 (0.70)	0.52 (0.70)	0.44 (0.70)	<.001*	0.5 (0.7)	0.5 (0.7)	0.5 (0.7)	.538
Exacerbations (subjects with ≥1 exacerbation), mean (SD)	1.24 (0.62)	1.23 (0.62)	1.25 (0.62)	0.307	1.2 (0.6)	1.2 (0.6)	1.3 (0.6)	N/A^g
Exacerbations (Hosp + ED), n (%)	5,206 (21.7%)	2,762 (24.8%)	2,444 (19.0%)	<.001*	4,221 (22.1%)	2,109 (22.1%)	2,112 (22.1%)	.971
Exacerbations (Hosp + ED), mean (SD)	0.24 (0.50)	0.27 (0.50)	0.21 (0.50)	<.001*	0.25 (0.50)	0.24 (0.50)	0.25 (0.50)	.360
COPD sub-type^c, n (%)
Emphysema	3,346 (13.9%)	1,271 (11.4%)	2,075 (16.1%)	<.001*	2,407 (12.6%)	1,195 (12.5%)	1,212 (12.7%)	N/A^f
Chronic bronchitis	9,219 (38.4%)	4,813 (43.2%)	4,406 (34.2%)		7,564 (39.6%)	3,772 (39.5%)	3,792 (39.7%)
Chronic airway obstruction	11,475 (47.7%)	5,057 (45.4%)	6,418 (49.8%)		9,119 (47.8%)	4,578 (48.0%)	4,541 (47.6%)
COPD-related costs, mean (SD)
Total	$1,933 (5,757)	$2,078 (5,470)	$1,808 (5,990)	<.001*	$1,987 (6,016)	$1,962 (5,332)	$2,013 (6,630)	.556
Medical	$1,757 (5,722)	$1,888 (5,428)	$1,644 (5,961)	<.001*	$1,808 (5,982)	$1,782 (5,292)	$1,833 (6,601)	.554
Prescription	$176 (430)	$190 (459)	$164 (403)	<.001*	$179 (428)	$180 (440)	$179 (417)	.992

Abbreviations. CCI, Charlson Comorbidity Index; COPD, chronic obstructive pulmonary disease; CVD, cardiovascular disease; Dx, diagnosis; ED, emergency department; FSC, fluticasone propionate 250 mcg/salmeterol 50 mcg; Hosp, hospitalizations; LRTI, lower respiratory tract infection; NA, not applicable; NEC, not elsewhere classified; OCS, oral corticosteroids; Rx, prescription; SABA, short-acting beta agonists; SAMA, short-acting muscarinic agent; SD, standard deviation; TIO, tiotropium bromide 18 mcg; URTI, upper respiratory tract infection.

^a COPD, asthma, and CVD are excluded from the CCI.

^b A unique count of disease states beyond those used to calculate the CCI and specific comorbidities listed in the table.

^c Some subjects may have ≥1 Dx type. So, in that case we will use only 1 Dx, and will order in this ranking with emphysema first, chronic bronchitis, and then chronic airway obstruction NEC.

^d T-test for continuous variables and Chi-square analysis for categorical variables.

^e Paired t-test for continuous variables and McNemar’s test for dichotomous variables.

^f McNemar’s test is computed only for 2 × 2 tables.

^g Paired t-test could not be performed, as pairs with either one or both subjects without an exacerbation were excluded.

* p-values <.05.

After propensity score matching, the analysis sample consisted of 9,545 subjects in each cohort (Table 1). Post-propensity score matching, the mean COPD-related total costs PPPY (mean [SD]) among the FSC cohort were numerically lower; however, the difference ($128) failed to reach statistical significance ($2,224 [±4,108] vs $2,352 ± 3,721], p = .057) (Table 2). Mean COPD-related medical costs PPPY were similar between the two cohorts ($1,083 [±$3,828] vs $1,076 [±$3,457], p = .894). The FSC cohort, however, incurred significantly lower mean PPPY COPD-related prescription costs than the TIO cohort ($1,160 [±1,106] vs 1,275 [±1,110], p < .001). When stratifying by age (i.e. 40–64 years and ≥65 years), results were similar to the analysis of all patients; i.e. no statistically significant differences in the mean PPPY COPD-related total and medical costs; and a statistically significant lower mean PPPY COPD-related prescription cost for the FSC cohort compared to the TIO cohort (see supplemental material).

Table 2. Per-patient per-year costs and exacerbations results for subjects initiating maintenance therapy with FSC vs TIO.

Outcomes	Total (n = 19,090)	FSC (n = 9,545)	TIO (n = 9,545)	Difference (FSC – TIO)	p-value^a
Primary outcomes
COPD-related costs^b, mean (SD)
Total	$2,298 (3,920)	$2,244 (4,108)	$2,352 (3,721)	–$108	.057
Medical	$1,080 (3,647)	$1,083 (3,828)	$1,076 (3,457)	$7	.894
Prescription	$1,218 (1,109)	$1,160 (1,106)	$1,275 (1,110)	–$115	<.001*
Secondary outcomes
Exacerbations^c, n (%)
Hospitalization	779 (4.1%)	399 (4.2%)	380 (4.0%)	0.2%	.510
ED visit	1,199 (6.3%)	619 (6.5%)	580 (6.1%)	0.4%	.261
Phy + Rx	3,560 (18.7%)	1,787 (18.7%)	1,773 (18.6%)	0.1%	.809
Hospitalization/ED visit	1,832 (9.6%)	947 (9.9%)	885 (9.3%)	0.7%	.135
Hospitalization/ED visit/Phy + Rx	4,897 (25.7%)	2,498 (26.2%)	2,399 (25.1%)	1.0%	.104
Exacerbation rate^c, mean (SD)
Hospitalization	0.05 (0.3)	0.05 (0.3)	0.05 (0.2)	0.00	.442
ED visit	0.08 (0.3)	0.08 (0.3)	0.07 (0.3)	0.01	.808
Phy + Rx	0.25 (0.6)	0.24 (0.6)	0.25 (0.6)	–0.01	.205
Hospitalization/ED visit	0.12 (0.4)	0.13 (0.4)	0.12 (0.4)	0.01	.533
Hospitalization/ED visit/Phy + Rx	0.37 (0.8)	0.37 (0.8)	0.38 (0.8)	–0.01	.532
Components of COPD-related medical cost^d, mean (SD)
Hospitalization	$415 (2,874)	$435 (2,985)	$396 (2,758)	$39	.470
ED	$59 (478)	$62 (510)	$56 (444)	$6	.159
Physician visits	$153 (350)	$143 (405)	$163 (284)	–$20	<.001*
Laboratory	$4 (50)	$3 (49)	$4 (50)	$0	.178
Radiology	$11 (125)	$12 (143)	$11 (104)	$0	.231
Other outpatient	$437 (1,594)	$428 (1,662)	$446 (1,523)	–$18	.225
Components of COPD-related Rx cost^d, mean (SD)
Index maintenance therapy	$929 (838)	$848 (794)	$1,010 (872)	–$162	<.001*
Non-index therapies (includes all COPD-related rescue  and maintenance agents)	$289 (603)	$312 (621)	$266 (583)	$47	<.001*

Abbreviations. COPD, chronic obstructive pulmonary disease; ED, emergency department; FSC, fluticasone propionate 250 mcg/salmeterol 50 mcg; GEE, generalized estimating; OCS, oral corticosteroids; OLS, ordinary least squares; Phy + Rx, COPD-related Physician/Outpatient visit with OCS or antibiotic Rx within 5 days of Physician/Outpatient visit; Rx, prescription; SD, standard deviation; TIO, tiotropium bromide 18 mcg.

^a p-value obtained from statistical models specified for each outcome.

^b p-value obtained from GEE model with gamma distribution and log-link accounting for the matched cohorts.

^c p-value obtained from paired t-test for continuous variables and McNemar’s test for categorical variables.

^d p-value obtained from semi-log OLS models accounting for the matched cohorts.

* p-values <.05.

There were no statistically significant differences in the rate or number of exacerbations across the matched FSC and TIO cohorts. However, once stratified by age group, differential patterns in exacerbation rates and numbers were apparent (see supplemental material). Among the 40–64-year-old subjects, the number of exacerbations (any type) was significantly lower among the FSC cohort (0.34 [±0.7] vs 0.38 [±0.8], p = .026). This result was driven by the lower number of physician office visit with OCS/AB prescriptions within the FSC cohort (p = .024). Among those ≥65 years old, the rate of exacerbation was significantly higher among the FSC cohort than the TIO cohort (27.0% vs 24.6%, p = .005).

Among the components of COPD-related medical costs, physician visit costs PPPY were significantly lower among the FSC cohort than the TIO cohort ($143 [±405] vs $163 [±284], p < .001). The other medical cost components were similar across the two cohorts. Among the components of COPD-related prescription costs, the PPPY costs for the index medication were significantly lower for the FSC cohort than the TIO cohort ($848 [±794] vs $1,010 [±872], p < .001); however, the costs for non-index medications (including all COPD-related rescue and maintenance agents) were significantly higher among the FSC cohort ($312 [±621] vs $266 [±583], p < .001).

Discussion

The objective of this study was to compare the real-world clinical and economic outcomes associated with the use of two long-acting bronchodilators (FSC: fluticasone propionate 250 mcg + salmeterol 50 mcg and TIO: tiotropium bromide 18 mcg) for initial maintenance treatment of COPD. Prior to propensity score matching, there were significant baseline differences between the FSC and TIO cohorts, with the respiratory burden being significantly higher in the FSC cohort. After matching, balance was achieved between the cohorts with respect to the baseline demographic and clinical characteristics observed. Our results indicate that the mean PPPY COPD-related total (FSC vs TIO: –$128; p = .057) and medical costs (FSC vs TIO: $7; p = .894) were not statistically significantly different between the two cohorts. Significant differences in the pharmacy costs were observed, where the FSC cohort incurred significantly lower mean COPD-related prescription costs than the TIO cohort (FSC vs TIO: –$115; p < .001). The analysis further indicated that initiating COPD maintenance therapy with FSC was not associated with significant differences in COPD-related healthcare costs, rate of COPD exacerbations, or number of exacerbations when compared with initiating therapy with TIO in the first year following treatment start before stratifying by age.

Study results are consistent with clinical trials, but are somewhat counter to previously published observational studies. From the clinical trial perspective, INSPIRE (Investigating New Standards for Prophylaxis in Reducing Exacerbations) was a randomized, double-blind, parallel-group, 2-year trial comparing the effects of the fixed-dose combination of fluticasone propionate 500 mcg/salmeterol 50 mcg (FSC 500/50) and TIO (tiotropium bromide 18 mcg) on the rate of moderate and/or severe exacerbations in patients with COPD³⁷. Results from the INSPIRE clinical trials showed that no differences were observed between FSC 500/50 and TIO in annual exacerbations rates (1.28 vs 1.32, rate ratio = 0.967; 95% confidence internal [CI] = 0.836–1.119; p = .656). Additional outcomes from the study showed that the probability of withdrawing from the study was 29% greater with TIO than FSC 500/50 (p = .005), improved HRQoL associated with FSC 500/50 compared with TIO (St George’s Respiratory Questionnaire difference = 2.1 units; 95% CI = 0.1–4.0; p = .038), and a significantly lower rate of mortality in the FSC 500/50 group compared with the TIO group (3% vs 6%; p = .032).

From the observational study perspective, a number of studies have compared FSC and TIO^{28,36,38–41,46}, as well as the ICS/LABA combination budesonide/formoterol (BFC) and TIO^47,48. In one of the first studies to compare FSC and TIO, Dalal et al.³⁸ assessed FSC and SAMA/LAMA (ipratropium bromide/albuterol [IPA], ipratropium bromide [IPR] and TIO) among Medicare beneficiaries. Treatment with FSC was associated with total COPD-related costs savings ranging from $110 (TIO; p < .05) to $1,235 (IPR; p < .05), albeit at a higher pharmacy cost ranging from $169 (TIO; p < .05) to $398 (IPA; p < .05). A similar study comparing initial maintenance therapy with FSC and SAMA/LAMA (TIO and IPR) was conducted in a commercially-insured COPD population and found an increased risk of COPD-related hospitalization and/or ED visit with TIO vs FSC (hazard ratio = 1.29, 95% CI = 1.17–1.41) and lower mean adjusted 12-month COPD-related total healthcare costs ($2,408 vs $2,068; p < .05)⁴¹. This design was repeated in a study comparing initial maintenance therapy, FSC and TIO, with similar findings of lower costs and COPD-related resource utilization associated with FSC when compared with TIO³⁶. Halpern et al.³⁹ examined adherence and respiratory-related costs among patients with COPD initiating therapy on TIO or FSC and found that the TIO cohort as compared with the FSC cohort were 31.6% less likely to discontinue therapy (95% CI = 0.64–0.73) and had 2.25-times higher odds of being adherent to therapy (95% CI = 1.85–2.73). However, similar to the current study, there were no significant differences in costs when comparing the TIO and FSC cohorts. Finally, Kern et al.⁴⁷ evaluated BFC and TIO, comparing costs and outcomes in an administrative claims dataset. The authors found that BFC-treated patients had lower total all-cause and COPD-related healthcare costs, which were driven by lower inpatient expenditures. These results were consistent with a later study evaluating BFC and TIO among patients with COPD new to therapy, where BFC was associated with improvements in health and economic outcomes compared with TIO⁴⁸.

One hypothesis for the differences observed in the current study and those previously published may be attributed to the study population. Specifically, comparison of the baseline subject characteristics in the current study with those identified by Trudo et al.⁴⁸ suggests the population for the current study may be less severe. The current study population had fewer SABA or SAMA, OCS, and AB prescriptions; and an exacerbation rate that was approximately a third of that observed in the Trudo et al.⁴⁸ study. A similar observation is found when comparing the current study results and those of Kern et al.⁴⁹. In general, the current study population appears to be less severe with fewer SABA, SAMA, OCS, and AB prescriptions; and a mean annual exacerbation rate that was approximately half of that observed in the Kern et al.⁴⁹ study. The Trudo et al.⁴⁸ and Kern et al.⁴⁹ studies are the most relevant comparator studies, given the similar study periods (2009–2012). Comparisons with earlier studies are more difficult to interpret, due to vastly different time periods; and, by extension, different treatment guidelines. Ultimately, the severity of the study population, in particular whether or not the population represents an exacerbating or non-exacerbating population, will likely influence study results and the subsequent interpretation.

Limitations

The study is not without limitations, and most are inherent limitations of research using claims data. Healthcare claims afford a large sample size and evaluation of real-world use, but lack clinical data to assess COPD severity, and some clinical and subject characteristics that might influence physician treatment decisions. While propensity scoring achieved balance in baseline characteristics, some residual confounding unobserved in the database may be present, leading to the possibility that differences in cost and exacerbations seen are due to underlying differences in the health status of the cohorts, rather than to the therapeutic agent. In addition, the study uses a modified ITT design, allowing for therapy changes beyond the first 60 days of follow-up, likely resulting in crossover bias. COPD-related medical costs were defined as costs on claims with a primary diagnosis of COPD, and do not include costs of claims with other respiratory conditions as the primary diagnosis. This approach ensures that the costs are more likely to be attributed to the disease (more specificity), but it may also under-estimate costs incurred due to COPD (less sensitivity). However, the cost differences between study cohorts are not likely to be affected, as the same approach was used for both study cohorts. It is likely that some patients may use their index maintenance therapy on an “as needed” basis, rather than on a maintenance basis as is recommended, resulting in sub-optimal adherence. However, this pattern is not expected to differ between the two index medications. Finally, the results are generalizable to a COPD population that are naïve to maintenance therapy and who are insured through employer-sponsored plans. Thus, the results may not be generalized to the entire US population or to individuals covered by other federal insurance programs (e.g. Medicare, Medicaid, and Veterans Affairs), and should be applied to individual care settings cautiously. In summary, the study results should be interpreted with caution and within the context of the aforementioned limitations.

Conclusion

This retrospective study compares costs and outcomes for subjects initiating maintenance therapy with FSC or TIO in a propensity score matched population from a large administrative claims database in the US. Results demonstrated that initiating maintenance therapy with either FSC or TIO was not associated with significant differences in COPD-related healthcare costs, rate of COPD exacerbations, or number of exacerbations in the first year following treatment start.

Transparency

Declaration of funding

This manuscript was funded by GlaxoSmithKline (GSK) (HO-14-14965/201512).

Declaration of financial/other relationships

CFB is an employee of GSK receiving salary, stock, and benefits. ADC, EF, TL, and PL-B are employed by Xcenda LLC, a consulting company that has received a research grant from GSK to conduct this study. A JME peer reviewer on this manuscript declares that they have received research support from GSK for a COPD project not related to any particular medication. A JME peer reviewer on this manuscript declares that they have participated in consulting, advisory boards, speaker panels, and received travel reimbursement from Amphastar, Astra Zeneca, Mylan, Novartis, Oriel, Pearl, Sunovion, Teva, and Theravance. This reviewer has also conducted multi-center clinical research trials for ∼40 pharmaceutical companies. The remaining JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgments

Editorial assistance was provided by Katy Tucker at Fishawack Indicia Ltd, UK, and this service was supported by GSK.