[ Original Research COPD ]

Comparative Effectiveness and Safety of

LABA-LAMA vs LABA-ICS Treatment of
COPD in Real-World Clinical Practice
Samy Suissa, PhD; Sophie Dell’Aniello, MSc; and Pierre Ernst, MD

BACKGROUND: Long-acting b2-agonists (LABAs) and long-acting muscarinic antagonists

(LAMAs) are recommended as initial maintenance treatments for COPD, with their com-
bination (LABA-LAMA) advocated as the disease progresses. Randomized trials comparing
the effectiveness of this combination with the alternative combination of LABA with inhaled
corticosteroid (LABA-ICS) have reported conflicting data, while there are no real-world
comparative effectiveness and safety studies of these regimens in clinical practice settings.
METHODS: We identified a cohort of patients with COPD during 2002-2015, age 55 years or
older, from the United Kingdom’s Clinical Practice Research Datalink. Patients initiating LABA-
LAMA on the same day (no ICS) were matched on time-conditional high-dimensional pro-
pensity scores with patients initiating LABA-ICS on the same day (no LAMA), and monitored
for 1 year for the occurrence of a moderate or severe COPD exacerbation and severe pneumonia.
RESULTS: The cohort included 1,977 initiators of LABA-LAMA matched with 1,977 initiators
of LABA-ICS. The hazard ratio (HR) of moderate or severe COPD exacerbation associated
with LABA-LAMA initiation, relative to LABA-ICS initiation, was 1.04 (95% CI, 0.90-1.20),
while for a severe exacerbation it was 0.94 (95% CI, 0.65-1.36). The incidence of severe
pneumonia requiring hospitalization was lower with LABA-LAMA initiation (HR, 0.66;
95% CI, 0.41-1.05), particularly in the on-treatment analysis (HR, 0.66; 95% CI, 0.50-0.87).
CONCLUSIONS: In a real-world clinical practice setting of COPD treatment, combined LABA-
LAMA inhalers appear to be as effective as combined LABA-ICS inhalers in preventing
COPD exacerbations. However, a LABA-LAMA combination may be preferred because it is
associated with fewer severe pneumonias. CHEST 2019; 155(6):1158-1165

KEY WORDS: COPD treatment; database research; observational study; real world evidence

ABBREVIATIONS: CPRD = Clinical Practice Research Datalink; HR =
hazard ratio; ICS = inhaled corticosteroid; IMPACT = Informing the
Pathway of COPD Treatment; LABA = long-acting b2-agonist;
LAMA = long-acting muscarinic antagonist
AFFILIATIONS: From the Centre for Clinical Epidemiology, Lady
Davis Institute-Jewish General Hospital, and the Department of
Epidemiology, Biostatistics, and Occupational Health, McGill Univer-
sity, Montreal, QC, Canada.
FUNDING/SUPPORT: This research was funded in part by grants from
the Canadian Institutes of Health Research (CIHR) [Grant No. CIHR
MOP-49462], the Canada Foundation for Innovation (CFI) [Grant No.
CFI 94480], and Boehringer-Ingelheim. S. S. is the recipient of the
James McGill Professorship award.
CORRESPONDENCE TO: Samy Suissa, PhD, Centre for Clinical Epide-
miology, Jewish General Hospital, 3755 Cote Ste-Catherine, H-461,
Montreal, QC, Canada H3T 1E2; e-mail: samy.suissa@mcgill.ca
Copyright Ó 2019 American College of Chest Physicians. Published by
Elsevier Inc. All rights reserved.
DOI: https://doi.org/10.1016/j.chest.2019.03.005

1158 Original Research [ 155#6 CHEST JUNE 2019 ]

COPD is a leading cause of morbidity and mortality
throughout the world.1 Long-acting bronchodilator
medications, which include long-acting b2-agonists
(LABAs) and the long-acting muscarinic antagonists
(LAMAs) such as the anticholinergic tiotropium, have
become central maintenance therapy to the management
of COPD, with inhaled corticosteroids (ICSs) added
with increasing severity.2 As a result, the treatment of
COPD increasingly involves using several of these drugs,
with some of their combinations now formulated into
single inhalers. While the reduction of exacerbations
with these drugs is a main focus, the risk of
pneumonia associated with the ICS component remains
an issue.3
The head-to-head trials that compared the effectiveness
of LABA-LAMA with LABA-ICS combinations on the
prevention of COPD exacerbations reported
inconsistent findings.4 While some trials found that
patients receiving the LABA-LAMA combination had
fewer exacerbations than those receiving the LABA-ICS
Methods
Data Source
This study was conducted using the Clinical Practice Research Datalink
(CPRD), a primary care database from the United Kingdom that
contains primary care medical records for over 10 million people
enrolled from over 600 practices. Trained participating general
practitioners record medical information, including demographic
data, lifestyle factors, and medical diagnoses, using the Read
classification. Prescriptions are automatically transcribed, using the
UK Prescription Pricing Authority Dictionary. For over half of the
practices, the CPRD can be linked to the Hospital Episode Statistics
(HES) database that provides hospitalization data. The recorded
information on medications and diagnoses has been validated and
shown to be of high quality.14–16
Study Design
The base cohort was formed by identifying all patients from the CPRD,
linkable to HES databases, with a diagnosis of COPD and who received
at least one prescription for a long-acting bronchodilator, either a
LABA or LAMA, or for an ICS from January 1, 2002 until
December 31, 2015. We restricted the study LAMA to tiotropium
because it accounted for 97% of all LAMA prescriptions during this
period. To increase the likelihood of a diagnosis of COPD, we
included only patients 55 years of age or older on the date of this
initial prescription and excluded all patients with a prior diagnosis of
asthma.
We used an incident new-user cohort design, a special case of the
prevalent new-user design, with time-conditional propensity scores
to match the two comparison groups.17 The study cohort was
formed by first identifying all members of the base cohort with the
first occurrence of prescriptions for LABA and LAMA (tiotropium)
on the same date, but no ICS. For each subject initiating
LABA-LAMA treatment, a matched comparator subject initiating
LABA-ICS was identified using a time-conditional propensity score-
matched approach. The LABA-ICS comparator subjects were
chestjournal.org
combination,5–7 others did not find a difference between
these combinations,8–11 and the recent IMPACT
(Informing the Pathway of COPD Treatment) trial
found the reverse association, namely that LABA-ICS
was associated with fewer exacerbations.12 These
conflicting findings may be due in part to the compared
combinations involving different LABAs. Besides, the
compulsory discontinuation of maintenance treatments
before randomization does not permit to emulate the
real-world clinical setting. The sole observational study
on this question, using data from a United States-based
claims database, reported no difference in exacerbation
rates between these combinations but did not address
the risk of pneumonia.13
We therefore assessed the effectiveness of a combination
of LABA-LAMA compared with LABA-ICS on the
incidence of COPD exacerbation and the risk of severe
pneumonia requiring hospitalization, using a large
population-based cohort formed from real-world clinical
practice data.
selected from the base cohort by forming time-based exposure sets
(2 months wide) defined by the calendar year of entry into the base
cohort for the LABA-LAMA and LABA-ICS simultaneous
prescription users. We computed high-dimensional time-conditional
propensity scores by identifying all available data (eg, diagnoses,
signs, symptoms, procedures, medications, up to 200 variables) in
the 1-year period prior to the date of cohort entry for each time-
based exposure set, using Cox proportional hazards regression.18
Then, for each LABA-LAMA-exposed subject, and starting
chronologically with the first one, we identified a matched
comparator subject from the members of the subject’s time-based
exposure set. The matched comparator subject was then selected as
the one with the closest time-conditional propensity score in the
matched set, after matching on sex and trimming within each
exposure set for the positivity assumption.
Subjects needed 1 year of medical history information prior to the date
of treatment initiation (cohort entry date) to allow the identification of
new use and the measurement of baseline covariates. Study subjects
were monitored for up to 1 year from the date of combined
treatment initiation, or until the earliest of treatment
discontinuation, the addition of ICS for LABA-LAMA or addition of
LAMA for LABA-ICS, the date of death, March 31, 2016, or the end
of coverage in the practice, whichever occurred first.
Outcome Events
The primary outcome was the first moderate or severe COPD
exacerbation to occur after cohort entry. A severe exacerbation was
defined as a hospitalization for COPD and a moderate exacerbation
as a new prescription for prednisolone, defined as a prescription
with no prior prescription in the previous 30 days. Secondary
outcomes included the number of exacerbations during follow-up
and the occurrence of the first hospitalization for community-
acquired pneumonia (serious pneumonia). The diagnostic and
medication codes to identify these outcomes have been shown to be
valid for these outcomes and used in several studies using the
CPRD.19–23
1159
Covariates
Covariates included age, sex, BMI, smoking status, and excessive
alcohol consumption. Measures of COPD severity included the
occurrence and number of COPD exacerbations and pneumonia
during the 1-year baseline period. An exacerbation in the 30 days
prior to cohort entry was also identified as a marker of COPD
exacerbation risk. Lung function was determined on the basis of the
last FEV1 measurement taken among all patients with available data,
calculated as the percent predicted value using age, sex, and height,
but assuming all patients were white as the CPRD does not include
data on race. Other respiratory drugs during the baseline period,
used as a measure of COPD severity, included the number
of prescriptions for short-acting b-agonists, short-acting
anticholinergics, methylxanthines, and antibiotics for a respiratory
condition. Moreover, for each of the cohorts, baseline comorbidity
was measured using diagnoses and prescriptions for various
conditions often present with COPD, namely, cardiovascular disease,
diabetes, and cancer, all in the year prior to cohort entry.
Data Analysis
The primary comparative analysis used a Cox proportional hazard
regression model to compare current use of LABA-LAMA with
LABA-ICS on the incidence of exacerbation and pneumonia during
Results
The base cohort included 56,460 patients with a
diagnosis of COPD and a prescription for LABA,
LAMA, or ICS from January 1, 2002 to December 31,
2015, after excluding patients with a diagnosis of asthma
and less than 55 years of age (Fig 1). There were 2,066
initiators of LABA-LAMA on the same date at some
point during follow-up. After computing time-
conditional propensity score, 1,977 initiators of LABA-
LAMA were matched to 1,977 initiators of LABA-ICS.
The baseline characteristics of these matched subjects
show good balance between the two groups, except for
prior use of the components of the study combination
treatments (Table 1). The mean blood eosinophil count
quantified as a percentage of WBC counts, among the
82% of patients for whom these data are available, was
equal at 3.2% in the two groups. The mean percent
predicted FEV1 was also comparable at 53.9% for
LABA-LAMA and 52.7% for LABA-ICS.
The mean treated duration of the LABA-LAMA group
was 3.3 months, truncated for adding ICS (23%) and
discontinuing one of the two components (44%)
during the 1-year follow-up. For the 1,977 LABA-ICS
initiators, the mean treated duration was 2.8 months,
primarily due to adding LAMA (38%) and
discontinuing one of the two components (34%). The
LABAs used during the follow-up were mainly
salmeterol (81%) and formoterol (15%) in the LABA-
LAMA group and salmeterol (74%) and formoterol
(25%) in the LABA-ICS group. An antibiotic was
1160 Original Research
the first year after combination treatment initiation. Current use was
defined as exposure within the 60-day period after the prescription
date. The model was further adjusted for patient characteristics
found to be imbalanced after matching on propensity scores, as well
as the deciles of propensity score. The rate of exacerbation was
analyzed using the negative binomial regression model, with the
same approach to covariate adjustment as the primary analysis.
Several sensitivity analyses were performed. First, we also used an
intent-to-treat analysis, which compares the two long-acting
bronchodilators at cohort entry during the 1-year follow-up,
irrespective of treatment switching or discontinuations. Second,
the 60-day period used to define current use in the as-treated
analyses was changed to 30- and 90-day periods. Third, analyses
were stratified by prior exacerbations and by whether double
treatment was initiated or stepped up. In addition, the primary
analysis was stratified by the baseline blood eosinophils
(percentage of WBC counts) as < 2%, 2-4%, 4-6%, and > 6%.
Finally, the primary analysis was stratified by sex. All analyses
were conducted with SAS version 9.4 (SAS Institute). The study
protocol was approved by the Independent Scientific Advisory
Committee of the CPRD (Protocol 17_159R) and the Ethics
Committee of the Jewish General Hospital (JGH Protocol #17-
137), Montreal, Quebec, Canada.
present in over 60% of the prednisolone prescriptions
used to define a moderate exacerbation.
The incidence rate of a first moderate or severe
exacerbation was 76.3 per 100 per year for the LABA-
LAMA initiators, compared with 74.1 per 100 per year
for the reference LABA-ICS initiators (Table 2). The
adjusted hazard ratio of a first moderate or severe
exacerbation associated with LABA-LAMA relative to
LABA-ICS from the as-treated analysis was 1.04
(95% CI, 0.90-1.20), while for severe exacerbation the
hazard ratio was 0.94 (95% CI, 0.65-1.36). The
cumulative incidence of a first moderate or severe
exacerbation over 1 year was similar between the two
treatment groups (Fig 2).
Table 2 also shows that the incidence rate of the
frequency of moderate or severe exacerbations was 90.7
per 100 per year with LABA-LAMA compared with 83.2
per 100 per year with LABA-ICS. The adjusted rate ratio
of the frequency of moderate or severe exacerbations
associated with LABA-LAMA relative to LABA-ICS
from the as-treated analysis was 1.07 (95% CI, 0.92-
1.25), while for severe exacerbations the rate ratio was
0.85 (95% CI, 0.55-1.33).
The incidence rate of a first pneumonia event requiring
hospitalization was 5.1 per 100 per year with LABA-
LAMA compared with 7.7 with LABA-ICS (Table 3).
The resulting adjusted hazard ratio of a first severe
pneumonia event associated with LABA-LAMA relative
to LABA-ICS from the as-treated analysis was 0.66
[ 155#6 CHEST JUNE 2019 ]
 Figure 1 – Flowchart of cohort selection. The
Users of LABA, LAMA or ICS between January 1, 2002 and LAMAs included exclusively tiotropium.
December 31, 2015 from UTS practices linkable to HES HES ¼ Hospital Episode Statistics; ICS ¼
N = 669,626 inhaled corticosteroid; LABA ¼ long-acting b2-
agonist; LAMA ¼ long-acting muscarinic
antagonist; UTS ¼ up to standard.

Exclusion:
• No prescriptions after 55 years of
age (n = 411,293)
• COPD not documented after 1995
(n = 119,194)
• No prescriptions after COPD diagnosis
(n = 10,642)

Eligible patients
(1st LABA, LAMA, or ICS Rx preceded by COPD diagnosis)
n = 128,497

Exclusion:
1) With diagnosis for asthma (n = 69,850)
2) Less than 55 years of age (n = 2,187)

Base cohort
n = 56,460

Incident users of LABA-LAMA (no ICS)

n = 2,066

(95% CI, 0.41-1.05). An on-treatment analysis using any
current LABA-LAMA and LABA-ICS exposures during
the entire 1-year follow-up resulted in a similar
reduction in pneumonia incidence but tighter
confidence interval for the hazard ratio of a first severe
pneumonia event associated with LABA-LAMA relative
to LABA-ICS (hazard ratio [HR], 0.66; 95% CI, 0.48-
0.92).
Sensitivity analyses indicate that the intent-to-treat
analyses and the various stratified analyses did not alter
the results (e-Tables 1,2). The stratification by baseline
blood eosinophil count, as a percentage of WBC counts,
suggests a trend for a higher incidence of exacerbation
with LABA-LAMA in patients with > 6% count (HR,
1.44; 95% CI, 0.88-2.36), although the number of
patients is too low to provide sufficient precision
(e-Table 3).
Discussion
In this observational study in the real-world setting of
COPD treatment, we found that patients treated with a
LABA and a LAMA, both as initial treatment or adding
one to the other, have a similar incidence of
exacerbations as those treated with a LABA and an ICS,
over the first year of use. There was a trend for a higher
incidence of exacerbation with LABA-LAMA in patients
with a > 6% blood eosinophil count. However, the
incidence of pneumonia was lower with the LABA-
LAMA treatment. This is the first observational study to
compare these combinations in a real-world clinical
setting on both effectiveness and safety.
Our study results differ from those of some of the head-
to-head randomized trials conducted to date.5–7 For
example, the FLAME (Effect of Indacaterol-
Glycopyrronium vs Fluticasone-Salmeterol on COPD
Exacerbations) trial reported a lower incidence of
moderate-severe exacerbation with the LABA-LAMA
combination compared with LABA-ICS over a 1-year
follow-up (HR, 0.78; 95% CI, 0.70-0.86).7 On the other
hand, the IMPACT trial reported the reverse finding,
namely a higher rate of these exacerbations with the
LABA-LAMA combination (1.21 per year) compared
with LABA-ICS (1.07 per year) for a rate ratio of 1.13

chestjournal.org 1161
TABLE 1 ] Baseline Characteristics of the Study Cohort of 1,977 Initiators of LABA-LAMA and Their Propensity
Score-Matched 1,977 Initiators of LABA-ICS
Characteristic LABA-LAMA LABA-ICS
No. of patients 1,977 1,977
Age at cohort entry, y, mean (SD) 71.9 (8.5) 71.8 (8.6)
Female sex, No. (%) 778 (39.4) 778 (39.4)
Year of cohort entry, No. (%)
Before 2008 849 (42.9) 856 (43.3)
2008-2011 541 (27.4) 529 (26.8)
2012-2015 587 (29.7) 592 (29.9)
Smoker, No. (%) 1,867 (94.4) 1,837 (92.9)
Obesity status, No. (%)
Obese 469 (23.7) 451 (22.8)
Nonobese 1,239 (62.7) 1,253 (63.4)
Missing data 269 (13.6) 273 (13.8)
Alcohol abuse, No. (%) 24 (1.2) 25 (1.3)
a
Blood eosinophil count %, mean (SD) 3.2 (2.8) 3.2 (2.2)
FEV1 (% predicted),b mean (SD) 53.9 (18.9) 52.7 (19.4)
Respiratory events and medications in year prior to cohort entry, No. (%)
Hospitalization for COPD 166 (8.4) 154 (7.8)
Moderate or severe COPD exacerbation 805 (40.7) 818 (41.4)
Pneumonia hospitalization 76 (3.8) 87 (4.4)
Short-acting b-agonists 1,674 (84.7) 1,627 (82.3)
Short-acting antimuscarinic 572 (28.9) 499 (25.2)
Inhaled glucocorticoids 731 (37.0) 1,043 (52.8)
Long-acting b-agonist 1,507 (76.2) 831 (42.0)
Long-acting antimuscarinic 1,524 (77.1) 970 (49.1)
Prednisolone 724 (36.6) 747 (37.8)
Methylxanthines 103 (5.2) 95 (4.8)
Respiratory antibiotics 1,376 (69.6) 1,411 (71.4)
Comorbidity diagnoses and medications
in year prior to cohort entry, No. (%)
Cancer 104 (5.3) 113 (5.7)
Heart failure 59 (3.0) 65 (3.3)
Myocardial infarction 18 (0.9) 21 (1.1)
Stroke 25 (1.3) 14 (0.7)
ACE inhibitors 510 (25.8) 532 (26.9)
ARBs 210 (10.6) 187 (9.5)
b-Blockers 288 (14.6) 312 (15.8)
CCBs 486 (24.6) 528 (26.7)
Thiazide diuretics 313 (15.8) 331 (16.7)
Antiarrhythmic agents 59 (3.0) 57 (2.9)
Antidiabetic medications 164 (8.3) 190 (9.6)
Statins 756 (38.2) 755 (38.2)
PPIs 698 (35.3) 665 (33.6)
NSAIDs 416 (21.0) 441 (22.3)
Opioids 766 (38.7) 793 (40.1)

ACE ¼ angiotensin-converting enzyme; ARBs ¼ angiotensin receptor blockers; CCBs ¼ calcium channel blockers; ICS ¼ inhaled corticosteroid; LABA ¼
long-acting b2-agonist; LAMA ¼ long-acting muscarinic antagonist; NSAIDs ¼ nonsteroidal antiinflammatory drugs; PPIs ¼ proton pump inhibitors.
a
Blood eosinophil count as percentage of WBC count, based on available data from 1,625 users of LABA-LAMA and 1,621 users of LABA-ICS.
b
Based on available data from 1,581 users of LABA-LAMA and 1,478 users of LABA-ICS.

1162 Original Research [ 155#6 CHEST JUNE 2019 ]

TABLE 2 ] Crude and Adjusted Hazard Ratios of COPD Exacerbation, Comparing LABA-LAMA Initiation With
LABA-ICS Initiation in Patients With COPD in the First Year After Treatment Initiation, From the
As-Treated Analysis
No. of No. of Rate per Crudea
Treatment Patients Events Person-Years 100 per Year HR Adjustedb HR (95% CI)
First COPD exacerbation
Moderate or severe exacerbation
LABA-LAMA 1,977 412 540 76.3 1.04 1.04 (0.90-1.20)
LABA-ICS 1,977 344 464 74.1 1.00 1.00 (Reference)
Severe exacerbation
LABA-LAMA 1,977 60 625 9.6 0.94 0.94 (0.65-1.36)
LABA-ICS 1,977 55 531 10.4 1.00 1.00 (Reference)
Frequency of COPD exacerbations
Moderate or severe
exacerbations
LABA-LAMA 1,977 575 634 90.7 1.07 1.07 (0.92-1.25)
LABA-ICS 1,977 449 540 83.2 1.00 1.00 (Reference)
Severe exacerbations
LABA-LAMA 1,977 71 634 11.2 0.87 0.85 (0.55-1.33)
LABA-ICS 1,977 71 540 13.2 1.00 1.00 (Reference)

HR ¼ hazard rate. See Table 1 legend for expansion of other abbreviations.

a
Crude, after matching on high-dimensional propensity scores and sex.
b
After matching on high-dimensional propensity scores and sex, adjusted further for decile of propensity score.

and a similar incidence of a first exacerbation.12 Our
findings are concordant with other trials that did not
find such differences between these combinations.8–11
There could be several reasons for these discrepancies.
First, most of the trials used combinations that involved
different LABAs, such as indacaterol, formoterol, or
vilanterol, in the LABA-LAMA combination group and
another LABA, salmeterol, in the comparator LABA-ICS
group. Such a design strategy does not permit to
distinguish whether the difference in outcomes between
1.00
Incidence of a moderate-severe
the groups is due to a more effective LABA in the
LABA-LAMA group, a less effective one in the LABA-
ICS group, or some combination of the two effects. In
our study, the LABAs used in the two combination
groups were the same, mainly salmeterol and
formoterol, which is more representative of the real-
world clinical context. The only trial to date that used
the same LABA in both groups, namely IMPACT,
reported a slightly higher rate of exacerbation with
LABA-LAMA and a similar incidence of a first
exacerbation.12 Another limitation of the randomized
trials is that the majority of the patients were already

0.90 LABA-ICS LABA-LAMA using these combinations before randomization, which

0.80 they had to discontinue and enter a run-in period. This
0.70 context does not emulate closely the clinical setting, and
exacerbation

0.60 run-in periods can be subject to bias.24

0.50
0.40 With respect to safety, all trials showed a lower
0.30 incidence of pneumonia with the LABA-LAMA
0.20 combination (pooled HR, 0.57; 95% CI, 0.42-0.79).4 The
0.10 IMPACT trial also reported a similarly lower incidence
0.00 with the LABA-LAMA combination (our calculation:
0
30
60
90

0
0
0
0
0
0
0
0
0

rate ratio, 0.63; 95% CI, 0.51-0.79).12 These findings are

12
15
18
21
24
27
30
33
36

Time since treatment initiation (d)

Figure 2 – One-year cumulative incidence of the first moderate or severe
COPD exacerbation, comparing current treatment with LABA-LAMA
and LABA-ICS, estimated using the Kaplan-Meier method. See Figure 1
legend for expansion of abbreviations.
consistent with our results of the on-treatment analysis
of a lower incidence of hospitalization for pneumonia
with the LABA-LAMA combination (HR, 0.66; 95% CI,
0.48-0.92).

chestjournal.org 1163
TABLE 3 ] Crude and Adjusted Hazard Ratios of Severe Pneumonia Comparing LABA-LAMA Initiation With
LABA-ICS Initiation in Patients With COPD in the First Year After Treatment Initiation, From the
As-Treated Analysis
Treatment Exposure No. of Patients No. of Events
As-treated exposure
LABA-LAMA 1,977 32
LABA-ICS 1,977 41
On-treatment exposurec
LABA-LAMA 1,977 49
LABA-ICS 1,977 143
See Table 1 and 2 legends for expansion of abbreviations.
a
Crude, after matching on high-dimensional propensity scores and sex.
b
After matching on high-dimensional propensity scores and sex, adjusted further for the decile of propensity score.
c
On-treatment exposure based on analysis of current use during the entire 1-year follow-up, allowing patients to switch or add treatments.
A strength of the present study design is that it identified
only initiators of the combinations, namely patients
initiating the two inhalers on the same day or patients
adding one of the inhalers to a prior single inhaler
treatment. This new-user approach allows the study to
reflect outcomes of treatment as of the initial usage,
which is more relevant for clinical practice. Second, the
outcome events used definitions that were validated,
including hospitalization records, and are thus less
subject to misclassification.21–23,25 However, milder
exacerbations treated with antibiotics only were not
included, which may or may not have been prescribed
for COPD, but restricted our analysis to moderate and
severe exacerbations treated with oral corticosteroids or
requiring hospitalization. Exacerbations identified only
by antibiotics have lower positive predictive value than
those identified by oral corticosteroids with or without
antibiotics.21 In our study, 62% of exacerbations defined
by prednisolone had an antibiotic prescribed on the
same day.
This study has limitations, including the exposure
measures to the two combinations that were based on
written prescriptions, not necessarily dispensed, and
could thus be subject to some misclassification, which
would tend to dilute the results. Second, the continuous
combination treatment durations were cut short because
patients either discontinued one of the components or
1164 Original Research
Rate per
Person-Years 100 per Year Crudea HR Adjustedb HR (95% CI)
629 5.1 0.66 0.66 (0.41-1.05)
535 7.7 1.00 1.00 (Reference)
907 5.4 0.65 0.66 (0.48-0.92)
1,778 8.0 1.00 1.00 (Reference)
added another agent, with only around one-third of the
subjects continuing for the 1-year follow-up, which
likely reflects the real-world clinical situation. Third,
while the base cohort included over 56,000 patients,
there were only 2,066 who initiated LABA-LAMA at
some point in time. This clinical reality limited the
power of this study to estimate the effectiveness in
patient subgroups, such as those with elevated blood
eosinophils. Finally, our cohort aimed to include only
patients with COPD and, despite using a cutoff of 55
years of age, our definition based on physician diagnosis
may have included some patients with asthma more so
for users of LABA-ICS than for LABA-LAMA. This
would tend to underestimate the effectiveness of LABA-
LAMA, even with the use of high-dimensional
propensity scores, so that residual confounding cannot
be ruled out, despite the comparable FEV1 distribution
in the two groups. Moreover, this age cutoff excludes the
patients diagnosed before 55 years of age, which
somewhat restricts the real-world generalizability of this
study.
In all, in this real-world setting study of the treatment of
COPD, maintenance treatment with a LABA-LAMA
combination appears to be equally effective as a LABA-
ICS combination in preventing COPD exacerbations.
However, the LABA-LAMA combination tends to be
associated with fewer hospitalizations for pneumonia.
[ 155#6 CHEST JUNE 2019 ]
Acknowledgments
Author contributions: P. E. participated in
study design, data interpretation, and writing
of the manuscript. S. D’A. participated in
data analysis and writing of the manuscript.
S. S. participated in data acquisition, study
design, data interpretation, and writing of the
manuscript, and acts as guarantor of this
entire manuscript.
Financial/nonfinancial disclosures: The
authors have reported to CHEST the
following: S. S. has received research grants
from Boehringer Ingelheim and Novartis and
has participated in advisory board meetings
or as speaker for AstraZeneca, Boehringer-
Ingelheim, and Novartis. None declared (P.
E. and S. D’A.).
Role of sponsors: The sponsors were not
directly involved in the design and conduct of
the study. Boehringer-Ingelheim was invited
to comment on the study protocol and the
manuscript, with any modifications resulting
from comments made independently by the
authors on the basis of scientific and editorial
merit.
Additional information: The e-Tables can
be found in the Supplemental Materials
section of the online article.
References
1. Rabe KF, Hurd S, Anzueto A, et al. Global
Initiative for Chronic Obstructive Lung
Disease. Global strategy for the diagnosis,
management, and prevention of chronic
obstructive pulmonary disease: GOLD
executive summary. Am J Respir Crit Care
Med. 2007;176(6):532-555.
2. Vogelmeier CF, Criner GJ, Martinez FJ,
et al. Global strategy for the diagnosis,
management, and prevention of chronic
obstructive lung disease 2017 report:
GOLD executive summary. Eur Respir J.
2017;49(3):1700214.
3. Ernst P, Saad N, Suissa S. Inhaled
corticosteroids in COPD: the clinical
evidence. Eur Respir J. 2015;45(2):525-
537.
4. Horita N, Goto A, Shibata Y, et al. Long-
acting muscarinic antagonist (LAMA)
plus long-acting b-agonist (LABA) versus
LABA plus inhaled corticosteroid (ICS)
for stable chronic obstructive pulmonary
disease (COPD). Cochrane Database Syst
Rev. 2017;2:CD012066.
5. Vogelmeier CF, Bateman ED, Pallante J,
et al. Efficacy and safety of once-daily
chestjournal.org
QVA149 compared with twice-daily
salmeterol-fluticasone in patients with
chronic obstructive pulmonary disease
(ILLUMINATE): a randomised, double-
blind, parallel group study. Lancet Respir
Med. 2013;1(1):51-60.
6. Zhong N, Wang C, Zhou X, et al.
LANTERN Investigators. LANTERN: a
randomized study of QVA149 versus
salmeterol/fluticasone combination in
patients with COPD. Int J Chron
Obstruct Pulmon Dis. 2015;10:1015-
1026.
7. Wedzicha JA, Banerji D, Chapman KR,
et al. FLAME Investigators. Indacaterol-
glycopyrronium versus salmeterol-
fluticasone for COPD. N Engl J Med.
2016;374(23):2222-2234.
8. Rabe KF, Timmer W, Sagkriotis A, Viel K.
Comparison of a combination of
tiotropium plus formoterol to salmeterol
plus fluticasone in moderate COPD.
Chest. 2008;134(2):255-262.
9. Donohue JF, Worsley S, Zhu C-Q,
Hardaker L, Church A. Improvements in
lung function with umeclidinium/
vilanterol versus fluticasone propionate/
salmeterol in patients with moderate-to-
severe COPD and infrequent
exacerbations. Respir Med. 2015;109(7):
870-881.
10. Singh D, Worsley S, Zhu C-Q,
Hardaker L, Church A. Umeclidinium/
vilanterol versus fluticasone propionate/
salmeterol in COPD: a randomised trial.
BMC Pulmon Med. 2015;15:91.
11. Vogelmeier C, Paggiaro PL, Dorca J, et al.
Efficacy and safety of aclidinium/
formoterol versus salmeterol/fluticasone: a
phase 3 COPD study. Eur Respir J.
2016;48(4):1030-1039.
12. Lipson DA, Barnhart F, Brealey N, et al.
IMPACT Investigators. Once-daily single-
inhaler triple versus dual therapy in
patients with COPD. N Engl J Med.
2018;378(18):1671-1680.
13. Samp JC, Joo MJ, Schumock GT,
Calip GS, Pickard AS, Lee TA.
Comparative effectiveness of long-acting
b2-agonist combined with a long-acting
muscarinic antagonist or inhaled
corticosteroid in chronic obstructive
pulmonary disease. Pharmacotherapy.
2017;37(4):447-455.
14. Jick SS, Kaye JA, Vasilakis-Scaramozza C,
et al. Validity of the General Practice
Research Database. Pharmacotherapy.
2003;23(5):686-689.
15. Herrett E, Thomas SL, Schoonen WM,
Smeeth L, Hall AJ. Validation and
validity of diagnoses in the General
Practice Research Database: a
systematic review. Br J Clin Pharmacol.
2010;69(1):4-14.
16. Herrett E, Gallagher AM, Bhaskaran K,
et al. Data resource profile: Clinical
Practice Research Datalink (CPRD). Int J
Epidemiol. 2015;44(3):827-836.
17. Suissa S, Moodie EE, Dell’Aniello S.
Prevalent new-user cohort designs for
comparative drug effect studies by time-
conditional propensity scores.
Pharmacoepidemiol Drug Saf. 2017;26(4):
459-468.
18. Schneeweiss S, Rassen JA, Glynn RJ,
Avorn J, Mogun H, Brookhart MA. High-
dimensional propensity score adjustment
in studies of treatment effects using health
care claims data. Epidemiology.
2009;20(4):512-522.
19. Suissa S, Dell’Aniello S, Ernst P. Long-
acting bronchodilator initiation in
COPD and the risk of adverse
cardiopulmonary events. Chest.
2017;151(1):60-67.
20. Suissa S, Dell’Aniello S, Ernst P.
Comparative effectiveness of
LABA-ICS versus LAMA as initial
treatment in COPD targeted by blood
eosinophils: a population-based cohort
study. Lancet Respir Med. 2018;6(11):
855-862.
21. Rothnie KJ, Müllerová H, Hurst JR, et al.
Validation of the recording of acute
exacerbations of COPD in UK primary
care electronic healthcare records. PLoS
One. 2016;11(3):e0151357.
22. Rothnie KJ, Müllerová H, Thomas SL,
et al. Recording of hospitalizations for
acute exacerbations of COPD in UK
electronic health care records. Clin
Epidemiol. 2016;8:771-782.
23. Meropol SB, Metlay JP. Accuracy of
pneumonia hospital admissions in a
primary care electronic medical record
database. Pharmacoepidemiol Drug Saf.
2012;21(6):659-665.
24. Suissa S. Run-in bias in randomised trials:
the case of COPD medications. Eur Respir
J. 2017;49(6):1700361.
25. Quint JK, Mullerova H,
DiSantostefano RL, et al. Validation of
chronic obstructive pulmonary disease
recording in the Clinical Practice Research
Datalink (CPRD-GOLD). BMJ Open.
2014;4:e005540.
1165