Safety analysis of long-acting dual bronchodilator after market approval: a real-world study from FDA Adverse Event Reporting System (FAERS) database

Abstract

Background

Long-acting muscarinic antagonist (LAMA) and long-acting β2-agonist (LABA) combinations are widely prescribed for chronic obstructive pulmonary disease (COPD). This study evaluated post-marketing safety signals for four dual bronchodilators (formoterol/glycopyrronium, indacaterol/glycopyrronium, vilanterol/umeclidinium, and olodaterol/tiotropium).

Methods

A retrospective pharmacovigilance study of four bronchodilators was conducted using the data from Food and Drug Administration Adverse Event Reporting System (FAERS) database between January 1, 2014 and March 31, 2025. The association between drugs and adverse events (AEs) was evaluated using reporting odds ratio (ROR), proportional reporting ratio (PRR), Bayesian confidence propagation neural network (BCPN), and multi-item gamma Poisson shrinker (MGPS). Venn analysis was performed to identify drug-specific disproportionality signals for the four bronchodilators.

Results

A total of 18,677 cases were included. Venn analysis identified 13 common safety signals and various drug-specific disproportionality safety signals, including 56 important medical events (IMEs) that met signaling thresholds for only one drug. Urinary retention emerged as a common IME among four bronchodilators. Indacaterol/glycopyrronium was associated with distinct cardiovascular risks, including arrhythmia, cardiac arrest, myocardial infarction, angina pectoris, and coronary artery disease. Olodaterol/tiotropium showed signals for ileus, retinal vein occlusion, cataract, and hip fracture, and exhibited the longest latency period among the other three bronchodilators.

Conclusions

This pharmacovigilance study identified significant safety signals associated with four bronchodilators, highlighting both common and drug-specific disproportionality signals and thereby providing essential evidence to support drug monitoring in clinical practice.

Highlight box.

Key findings

• This study evaluated the post-marketing safety signals of four dual bronchodilators (formoterol/glycopyrronium, indacaterol/glycopyrronium, vilanterol/umeclidinium, and olodaterol/tiotropium) for chronic obstructive pulmonary disease (COPD). Analysis of data from the Food and Drug Administration Adverse Event Reporting System between January 1, 2014, and March 31, 2025, included 18,677 cases.

What is known and what is new?

• Long-acting muscarinic antagonist/long-acting β2-agonist dual bronchodilators are cornerstone maintenance therapies for COPD, effectively improving lung function and symptoms. Urinary retention was identified as an important medical event shared by all four bronchodilators as a common safety signal.

• Using real-world data, this study systematically compares the safety signal profiles of these four commonly used dual bronchodilators and quantifies their differences for the first time. The most significant new insight is the identification of distinct safety signals among different bronchodilators, particularly the cardiovascular signals associated with indacaterol/glycopyrronium and the non-respiratory signals and longer latency observed with olodaterol/tiotropium. This suggests that safety profiles are not homogeneous within this drug class.

What is the implication, and what should change now?

• To enhance awareness of drug-specific disproportionality signals, clinicians should actively monitor for cardiovascular events when prescribing indacaterol/glycopyrronium, and vigilantly observe for gastrointestinal, ocular, and musculoskeletal issues with olodaterol/tiotropium, integrating these monitoring points into patient education.

Introduction

Chronic obstructive pulmonary disease (COPD) stands as a predominant contributor to the global burden of mortality and morbidity (1). The etiology of this pervasive respiratory affliction is multifaceted, with inhalation of noxious particulates, such as tobacco smoke and environmental pollutants, cited as principal risk factors (2). Inhalation therapy serves as a cornerstone in the management of COPD, providing targeted and effective treatment to alleviate symptoms, improve lung function, and enhance the quality of life for patients (3). In the clinical treatment, the strategic pairing of long-acting muscarinic antagonists (LAMAs) with long-acting β2-agonists (LABAs) in fixed-dose combinations (FDCs) has become a widely embraced standard for symptomatic relief in COPD patients. This LAMA/LABA formulation has been recognized for its enhanced clinical efficacy, surpassing the benefits achievable with LAMA or LABA monotherapies (4-6).

The U.S. Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS) is a pivotal database compiling reports on adverse events (AEs) and medication errors, strategically supporting the FDA post-marketing safety surveillance program to ensure the ongoing safety of drugs and therapeutic biologics (7,8). Mining the FAERS database offers the advantage of cost and time savings, particularly for the detection of rare but clinically significant AEs (9). Moreover, data mining may uncover previously unknown yet clinically important associations, providing us with valuable insights to guide clinical decisions (10).

Given that long-acting dual bronchodilators are widely used for COPD, we conducted disproportionality analysis to identify the potential safety signals of four dual bronchodilators (formoterol/glycopyrronium, indacaterol/glycopyrronium, vilanterol/umeclidinium, olodaterol/tiotropium) in practice using real-world pharmacovigilance data from FAERS. We present this article in accordance with the READUS-PV reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0224/rc).

Methods

Data source

The FAERS database was used to conduct a retrospective pharmacovigilance study (9-11). In this study, FAERS data collected between January 1, 2014 and March 31, 2025 were analyzed. The AE data of formoterol/glycopyrronium, indacaterol/glycopyrronium, vilanterol/umeclidinium and olodaterol/tiotropium were cleaned and analyzed. Formoterol/aclidinium was excluded from this study due to the insufficient number of reports in the FAERS database during the study period, which prevented stable disproportionality analysis. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Study design

The study design flowchart is presented in Figure 1. As data in the database were collected through voluntary reporting, some duplicate, withdrawn, or deleted reports were present. Therefore, the official FDA guidance document provided rules for data deduplication and a list of reports that needed to be deleted (12). This study strictly followed the data cleaning procedure outlined in the guidance document on the official FDA website.

Figure 1.

Flowchart of study design. DEMO, demographic and administrative data; DRUG, drug information; REAC, adverse drug reactions; Q1, Januray to March.

The process involved deduplicating based on FDA methods, selecting key fields, and retaining reports with the highest values in FDA_DT and PRIMARYID for identical CASEIDs (11). To address the heterogeneity of drug names in the FAERS database, a systematic standardization process based on RxNorm was implemented. First, all raw drug entries were extracted from the DRUGNAME field and preprocessed to remove special characters, extraneous spaces, and dosage information. The cleaned drug names were then mapped to standardized RxNorm Concept Unique Identifiers (RxCUIs) using RxNorm’s approximate matching algorithms, which are designed to automatically account for spelling variations, word order differences, and spelling discrepancies. For combination products, the ingredient relationship table within RxNorm was utilized to deconstruct them into their individual components, ensuring accurate identification of reports containing multiple target ingredients. A manual review of a random sample of mapped entries was conducted to validate the accuracy of the standardization process. This approach ensured comprehensive identification of the study drugs and enhanced the comparability of the data for subsequent analysis (13). Only reports where the target drugs were identified as the “primary suspect” (PS) were included in this study.

To ensure statistical robustness, excluded preferred terms (PT) included those labeled as “chronic obstructive pulmonary disease”, “emphysema”, “no adverse event” and excluded system organ classes (SOCs) included those labeled as “product issues” and “social circumstances” along with all their subordinate PT.

In this study, AEs were categorized using the PT and then classified into SOC, following the standardized approach outlined in the Medical Dictionary for Regulatory Activities (MedDRA), version 28.1 (14). To mitigate potential channeling bias, sensitivity analyses were conducted by stratifying the disproportionality signals by age (<65 and ≥65 years) and sex for the four bronchodilators. Additionally, the MedDRA IME list (version 28.0) was used to identify IMEs related to four long-acting dual bronchodilators.

Statistical analysis

In this study, the relationship between four long-acting dual bronchodilators and AEs was meticulously evaluated through the application of both Bayesian and Frequentist statistical approaches. Data mining was conducted using a suite of methods, as detailed in Table 1. Corresponding algorithms and formulas are presented in Table 2. A positive signal must meet the following conditions simultaneously: reporting odds ratio [ROR, significant signal: the lower limit of the 95% CI of the ROR (ROR₀₂₅) >1, N≥3], proportional reporting ratio (PRR, significant signal: PRR ≥2, χ²≥4, N≥3), Bayesian confidence propagation neural network [BCPN, significant signal: the lower limit of the 95% confidence interval (CI) of the information component (IC₀₂₅) >0], and multi-item gamma Poisson shrinker [MGPS, significant signal: empirical Bayesian geometric mean fifth percentile (EBGM₀₅) >2] (15,16).

Table 1. Four grid table of proportional imbalance method.

Drug group	Target adverse events	Other adverse events	Total
Target drugs	a	b	a + b
Other drugs	c	d	c + d
Total	a + c	b + d	a + b + c + d

Table 2. Algorithms, formulas, and positive signal criteria of disproportionality analysis.

Method	Computational formula	Threshold value
ROR	$ROR=\frac{\left(a/c\right)}{\left(b/d\right)}=\frac{ad}{bc}$	ROR025 >1, N≥3
	$95\%CI=e^{Ln\left(ROR\right)\pm 1.96\sqrt{\left(\frac{1}{a}+\frac{1}{b}+\frac{1}{c}\right)}}$
PRR	$PRR=\frac{a/\left(a+b\right)}{a/\left(c+d\right)}$	PRR ≥2, χ2≥4, N≥3
	${\chi }^2=\frac{{\left(ad-bc\right)}^2\left(a+b+c+d\right)}{\left(a+b\right)\left(a+c\right)\left(c+d\right)\left(b+d\right)}$
BCPNN	$IC={\log }_2\frac{p\left(x,y\right)}{p\left(x\right)p\left(y\right)}={\log }_2\frac{a\left(a+b+c+d\right)}{\left(a+b\right)\left(a+c\right)}$	IC025 >0
	$I{C}_{025}=IC-2\sqrt{V\left(IC\right)}$
MGPS	$EMGB=\frac{a\left(a+b+c+d\right)}{\left(a+c\right)\left(a+b\right)}$	EBGM05 >2
	$95\%CI=e^{Ln\left(EMGB\right)\pm 1.96\sqrt{\left(\frac{1}{a}+\frac{1}{b}+\frac{1}{c}\right)}}$

BCPNN, Bayesian confidence propagation neural network; CI, confidence interval; EBGM, empirical Bayes geometric mean; EBGM₀₅, EBGM fifth percentile; IC, information component; IC₀₂₅, the lower limit of the 95% CI of the IC; MGPS, multi-item gamma Poisson shrinker; PRR, proportional reporting ratio; ROR, reporting odds ratio; ROR₀₂₅, the lower limit of the 95% CI of the ROR.

Weibull’s shape parameter (WSP) analysis was employed to delineate the onset times of events (17). To avoid misleading interpretations regarding long-term safety, the analysis was stratified by SOC, allowing for the differentiation of temporal patterns across various event types. The Weibull distribution, a continuous probability distribution, is determined by two parameters: the scale parameter (α) and the shape parameter (β), offering a complete description of the data (18). The analysis of the latency period, utilizing the shape parameter, forecasted the risk of AEs over time, classifying them as early, random, or wear-out failures. When the shape parameter and its 95% CIs are both less than 1, it denotes a waning risk, classified as early failure. A parameter close to 1 with a CI that includes 1 suggests a steady risk, or random failure. Conversely, a parameter above 1 with a CI above 1 indicates a growing risk, referred to as wear-out failure. The Kaplan-Meier method was utilized to evaluate the cumulative proportion of drug-induced AEs, while the log-rank test examined the temporal onset profiles of these AEs.

Utilizing online tools, Venn analysis was conducted to discern both the common and drug-specific disproportionality signals (defined as signals unique to one drug group within the confines of this study) associated with each drug (19). Data were meticulously cleaned, deduplicated, processed, and subjected to rigorous statistical analyses using the software R (version 4.4.2), while the software GraphPad Prism v.9.5.0 and R were employed to expertly prepare the graphical representations.

Results

Descriptive analysis

During the study period, 1,215 cases of formoterol/glycopyrronium, 2,607 cases of indacaterol/glycopyrronium, 11,471 cases of vilanterol/umeclidinium, and 3,384 cases of olodaterol/tiotropium were included in the analysis (Table 3). The proportion of men receiving formoterol/glycopyrronium, indacaterol/glycopyrronium, vilanterol/umeclidinium, and olodaterol/tiotropium is 36.1%, 46.0%, 37.0% and 39.8%, respectively. The distribution of reports by health professional for formoterol/glycopyrronium, indacaterol/glycopyrronium, vilanterol/umeclidinium, and olodaterol/tiotropium stands at 33.1%, 29.0%, 7.4% and 21.6%, respectively. Indacaterol/glycopyrronium exhibited the highest mortality rate (24.8%) among the four bronchodilators, followed by formoterol/glycopyrronium (15.3%).

Table 3. Information of reports with long-acting dual bronchodilator inhalation as the primary suspected drug from the FAERS database.

Variables	Formoterol/glycopyrronium (n=1,215)	Indacaterol/glycopyrronium (n=2,607)	Vilanterol/umeclidinium (n=11,471)	Olodaterol/tiotropium (n=3,384)
Sex
Female	606 (49.9)	843 (32.3)	5,916 (51.6)	1,783 (52.7)
Male	439 (36.1)	1,200 (46.0)	4,247 (37.0)	1,348 (39.8)
NA	170 (14.0)	564 (21.6)	1,308 (11.4)	253 (7.5)
Age, years
<18	0	2 (0.1)	21 (0.2)	19 (0.6)
18–64.9	140 (11.5)	311 (11.9)	818 (7.1)	428 (12.6)
65–85	375 (30.9)	691 (26.5)	2,584 (22.5)	1,370 (40.5)
>85	42 (3.5)	84 (3.2)	329 (2.9)	139 (4.1)
NA	658 (54.2)	1,519 (58.3)	7719 (67.3)	1,428 (42.2)
Reporters’ role
Health-professional	402(33.1)	758 (29.0)	849 (7.4)	730 (21.6)
Non-health-professional	655 (53.9)	1,699 (65.2)	10,526 (91.8)	2,487 (73.5)
NA	158 (13.0)	150 (5.8)	96 (0.8)	167 (4.9)
Outcome
Death	186 (15.3)	647 (24.8)	382 (3.3)	299 (8.8)
Disability	7 (0.6)	49 (1.9)	51 (0.4)	28 (0.8)
Hospitalization	117 (9.6)	651 (25.0)	1,257 (11.0)	579 (17.1)
Life-threatening	12 (1.0)	95 (3.6)	57 (0.5)	59 (1.7)
Other	893 (73.5)	1,165 (44.7)	9,724 (84.8)	2,419 (71.5)
Reporters’ country
United State	1,158 (95.3)	107 (2.0)	10,585 (92.3)	2,772 (81.9)
Japan	16 (1.3)	227 (8.7)	153 (1.3)	126 (3.7)
Germany	14 (1.2)	160(6.1)	58 (0.5)	52 (1.5)
United Kingdom	5 (0.4)	51 (2.0)	157 (1.4)	27 (0.8)
Brazil	0	523 (20.1)	44 (0.4)	48 (1.4)
France	0	123 (4.7)	26 (0.2)	18 (0.5)
Other	22 (1.81)	1416(54.3)	448(3.9)	341(10.1)
Reported year
2014	0	0	199 (1.7)	0
2015	0	84 (3.2)	852 (7.4)	170 (5.0)
2016	12 (1.0)	477 (18.3)	1,321 (11.5)	452 (13.4)
2017	292 (24.0)	776 (29.8)	1,275 (11.1)	540 (16.0)
2018	181 (14.9)	478 (18.3)	1,711 (14.9)	376 (11.1)
2019	166 (13.7)	284 (10.9)	1,648 (14.4)	288 (8.5)
2020	137 (11.3)	151 (5.8)	1,548 (13.5)	279 (8.2)
2021	120 (9.9)	143 (5.5)	946 (8.2)	356 (10.5)
2022	75 (6.2)	82 (3.1)	703 (6.1)	348 (10.3)
2023	103 (8.5)	92 (3.5)	583 (5.1)	309 (9.1)
2024	112 (9.2)	40 (1.5)	558 (4.9)	217 (6.4)
2025Q1	17 (1.4)	0	127 (1.1)	49 (1.4)

Data are presented as n (%). FAERS, Food and Drug Administration Adverse Event Reporting System; NA, not applicable; Q1, January to March.

Among the included regions, AEs reported for formoterol/glycopyrronium, vilanterol/umeclidinium, and olodaterol/tiotropium are primarily from the United States, accounting for 95.3%, 92.3%, and 81.9% of the total submissions, respectively. In addition, Brazil has the highest number of AEs reported for indacaterol/glycopyrronium (20.1%).

Signals of disproportionality in the SOC

Disproportionality analysis was conducted to assess the signal strength of 24 SOC for four long-acting dual bronchodilators (Figure 2). “Respiratory, thoracic and mediastinal disorders” was observed for formoterol/glycopyrronium (n=676, ROR₀₂₅ =3.89), indacaterol/glycopyrronium (n=2408, ROR₀₂₅ =5.88), vilanterol/umeclidinium (n=4762, ROR₀₂₅ =4.34) and olodaterol/tiotropium (n=1,965, ROR₀₂₅=6.99). “Eye disorders” was identified for formoterol/glycopyrronium (n=107, ROR₀₂₅ =1.36), vilanterol/umeclidinium (n=746, ROR₀₂₅ =4.21) and olodaterol/tiotropium (n=249, ROR₀₂₅ =1.44). “Infections and infestations”, “cardiac disorders”, “ear and labyrinth disorders” were observed for indacaterol/glycopyrronium (n=719, ROR₀₂₅ =1.17; n=605, ROR₀₂₅ =2.68; n=60, ROR₀₂₅ =1.02, respectively) and olodaterol/tiotropium (n=475, ROR₀₂₅ =1.03; n=262, ROR₀₂₅ =1.47; n=55, ROR₀₂₅ =1.25, respectively). “Injury, poisoning and procedural complications” was detected for formoterol/glycopyrronium (n=1,030, ROR₀₂₅ =2.33) and vilanterol/umeclidinium (n=7,669, ROR₀₂₅ =2.88). “Surgical and medical procedures” was significantly associated with vilanterol/umeclidinium (n=666, ROR₀₂₅ =1.59).

Figure 2.

Safety signals of SOCs in four long-acting dual bronchodilator. Safety signal of patients receiving (A) formoterol/glycopyrronium, (B) indacaterol/glycopyrronium, (C) vilanterol/umeclidinium, and (D) olodaterol/tiotropium. CI, confidence interval; ROR, reporting odds ratio; SOC, system organ class.

Signals of disproportionality in formoterol/glycopyrronium

A total of 638 different safety signals were identified, and after statistical analysis, 45 of these were found to be significant. A forest plot was used to represent the top 20 reported AEs and their ROR values for the formoterol/glycopyrronium (Figure 3A, Table S1). Among the top 20 AEs, the top five ranked by report numbers were dyspnoea (n=226), death (n=167), cough (n=103), dysphonia (n=27), and chest discomfort (n=27). The top five AEs in terms of ROR were body height decreased (n=26, ROR₀₂₅ =28.59, PRR =41.81, IC₀₂₅ =3.5, EBGM₀₅ =28.28), feeling jittery (n=14, ROR₀₂₅ =8.53, PRR =14.37, IC₀₂₅ =2.18, EBGM₀₅ =8.49), respiration abnormal (n=7, ROR₀₂₅ =6.21, PRR =13.03, IC₀₂₅ =1.36, EBGM₀₅ =6.2), urinary retention (n=17, ROR₀₂₅ =5.28, PRR =8.47, IC₀₂₅ =1.90, EBGM₀₅ =5.26), and candida infection (n=10, ROR₀₂₅ =4.22, PRR =7.83, IC₀₂₅ =1.4, EBGM₀₅ =4.2). To assess the robustness of these findings, sensitivity analyses stratified by sex and age were conducted, with results shown in Tables S2-S5.

Figure 3.

Forest plot displaying the safety signals of top 20 adverse events in (A) formoterol/glycopyrronium, (B) indacaterol/glycopyrronium, (C) vilanterol/umeclidinium, and (D) olodaterol/tiotropium. CI, confidence interval; PT, preferred term; ROR, reporting odds ratio.

Signals of disproportionality in indacaterol/glycopyrronium

A total of 1,452 different safety signals were identified, and after statistical analysis, 146 of these were found to be significant (Table S6). Figure 3B showed the top 20 AEs by report numbers were dyspnoea (n=483), cough (n=311), malaise (n=217), pneumonia (n=187) and lung disorder (n=97). The top five AEs in term of ROR were choking (n=66, ROR₀₂₅ =15.88, PRR =20.12, IC₀₂₅ =3.61, EBGM₀₅ =15.7), lung disorder (n=97, ROR₀₂₅ =9.57, PRR =11.6, IC₀₂₅ =3.09, EBGM₀₅ =9.47), lung neoplasm malignant (n=84, ROR₀₂₅ =8.34, PRR =10.27, IC₀₂₅ =2.89, EBGM₀₅ =11.56), productive cough (n=69, ROR₀₂₅ =6.19, PRR =7.80, IC₀₂₅ =2.48, EBGM₀₅ =6.15), arrhythmia (n=51, ROR₀₂₅ =5.57, PRR =7.3, IC₀₂₅ =2.30, EBGM₀₅ =5.53). Sensitivity analyses stratified by sex and age for indacaterol/glycopyrronium are shown in Tables S7-S10.

Signals of disproportionality in vilanterol/umeclidinium

A total of 1,753 different safety signals were identified, and after statistical analysis, 117 of these were found to be significant (Table S11). The top 20 AEs by the ROR and report numbers were shown in Figure 3C. The top five positions among the top 20 AEs, based on the report numbers, were dyspnoea, cough, dysphonia, oropharyngeal pain, and productive cough (n=1,516, 736, 188, 185, 168, respectively). The top five AEs in term of ROR were urinary retention (n=142, ROR₀₂₅ =8.7, PRR =10.22, IC₀₂₅ =3.01, EBGM₀₅ =8.61), productive cough (n=168, ROR₀₂₅ =6.51, PRR =7.54, IC₀₂₅ =2.63, EBGM₀₅ =6.45), dysphonia (n=188, ROR₀₂₅ =6.25, PRR =7.17, IC₀₂₅ =2.58, EBGM₀₅ =6.18), dyspnoea (n=1,516, ROR₀₂₅ =6.14, PRR =6.16, IC₀₂₅ =2.54, EBGM₀₅ =5.83), throat irritation (n=117, ROR₀₂₅ =5.05, PRR =6.04, IC₀₂₅ =2.26, EBGM₀₅ =5.02). The results of sensitivity analyses were presented in Tables S12-S15.

Signals of disproportionality in olodaterol/tiotropium

A total of 1,237 different safety signals were identified, and after statistical analysis, 92 of these were found to be significant (Table S16). Forest plot used to represent the top 20 reported AEs and their ROR values for the olodaterol/tiotropium (Figure 3D). The report numbers ranking showed the top five were dyspnoea (n=646), cough (n=244), pneumonia (n=104), wheezing (n=69), and asthma (n=62). The ROR ranking showed the top five were oxygen saturation (n=6, ROR₀₂₅ =73.16, PRR =164.83, IC₀₂₅ =1.64, EBGM₀₅ =70.87), infective exacerbation of chronic obstructive airways disease (n=6, ROR₀₂₅ =33.39, PRR =74.71, IC₀₂₅ =1.59, EBGM₀₅ =32.89), sputum retention (n=5, ROR₀₂₅ =17.43, PRR =42.01, IC₀₂₅ =1.24, EBGM₀₅ =17.28), cor pulmonale (n=3, ROR₀₂₅ =12.2, PRR =37.99, IC₀₂₅ =0.44, EBGM₀₅ =12.11), halo vision (n=4, ROR₀₂₅ =11.87, PRR =31.71, IC₀₂₅ =0.85, EBGM₀₅ =11.79). The results of the sensitivity analyses, stratified by sex and age for olodaterol/tiotropium, were presented in Tables S17-S20.

Comparison of significant safety signals of four bronchodilators

The Venn analysis method has successfully identified 13 common safety signals among the safety signals reported for the four long-acting dual bronchodilators (Figure 4A,4B). Among 13 common safety signals, urinary retention was an IME following treatment with four bronchodilators.

Figure 4.

Comparative of significant safety signals in four long-acting dual bronchodilator. (A) Venn analysis and (B) the common safety signal in four long-acting dual bronchodilator. (C) The drug-specific disproportionality signals with IME. Drug-specific disproportionality signals, defined as signals unique to one drug group within the confines of this study. CI, confidence interval; F, formoterol; G, glycopyrronium; I, indacaterol; IME, important medical event; O, olodaterol; PT, preferred term; ROR, reporting odds ratio; T, tiotropium; U, umeclidinium; V, vilanterol.

Based on the Venn analysis, several safety signals reached the statistical significance threshold exclusively for specific bronchodilators. However, these drug-specific signals may reflect differences in statistical power due to varying report volumes rather than true clinical exclusivity. The Venn diagram showed that formoterol/glycopyrronium, indacaterol/glycopyrronium, vilanterol/umeclidinium, and olodaterol/tiotropium had 7, 90, 60, and 33 drug-specific disproportionality signals (Table S21), respectively. Among these drug-specific disproportionality signals, 56 IMEs were identified (Figure 4C).

For formoterol/glycopyrronium, both death (n=167, ROR₀₂₅ =2.53, PRR =2.87, IC₀₂₅ =1.28, EBGM₀₅ =2.46) and pulmonary congestion (n=5, ROR₀₂₅ =2.65, PRR =6.37, IC₀₂₅ =0.57, EBGM₀₅ =2.65) were drug-specific disproportionality signals that were also categorized as IMEs based on the Venn diagram. For indacaterol/glycopyrronium, it exhibited drug-specific disproportionality signals and IME for cardiac disorders, such as arrhythmia (n=51, ROR₀₂₅ =5.57, PRR =7.3, IC₀₂₅ =2.3, EBGM₀₅ =5.53), cardiac arrest (n=38, ROR₀₂₅ =2.48, PRR =3.4, IC₀₂₅ =1.21, EBGM₀₅ =2.47), cardiac fibrillation (n=5, ROR₀₂₅ =9.94, PRR =23.95, IC₀₂₅ =1.13, EBGM₀₅ =9.88), infarction (n=24, ROR₀₂₅ =15.74, PRR =23.47, IC₀₂₅=3.04, EBGM₀₅ =15.61), angina pectoris (n=21, ROR₀₂₅ =3.57, PRR =5.47, IC₀₂₅ =1.57, EBGM₀₅ =3.56), myocardial ischaemia (n=59, ROR₀₂₅ =2.7, PRR =3.48, IC₀₂₅ =1.36, EBGM₀₅ =2.69), coronary artery stenosis (n=5, ROR₀₂₅ =4.09, PRR =9.83, IC₀₂₅ =0.81, EBGM₀₅ =4.08), arteriosclerosis coronary artery (n=5, ROR₀₂₅ =2.76 PRR =6.63, IC₀₂₅ =0.59, EBGM₀₅ =2.75), aortic arteriosclerosis (n=4, ROR₀₂₅ =4.1, PRR =10.95, IC₀₂₅ =0.58, EBGM₀₅ =4.09), coronary artery insufficiency (n=3, ROR₀₂₅ =32.41, PRR =102.04, IC₀₂₅ =0.49, EBGM₀₅ =31.53). In addition, safety signal of dementia Alzheimer’s type (n=8, ROR₀₂₅ =2.65, PRR =5.31, IC₀₂₅ =0.88, EBGM₀₅ =2.65), senile dementia (n=6, ROR₀₂₅ =26.39, PRR =59.09, IC₀₂₅ =1.56, EBGM₀₅ =25.96) and deafness (n=15, ROR₀₂₅ =2.03, PRR =3.36, IC₀₂₅ =0.83, EBGM₀₅ =2.02) were found in the indacaterol/glycopyrronium. For vilanterol/umeclidinium, necrotising ulcerative gingivostomatitis (n=3, ROR₀₂₅ =23.5, PRR =74.78, IC₀₂₅ =0.46, EBGM₀₅ =22.48) was observed. For olodaterol/tiotropium, ileus (n=6, ROR₀₂₅ =2.09, PRR =4.65, IC₀₂₅ =0.52, EBGM₀₅ =2.09), cataract (n=24, ROR₀₂₅ =2.13, PRR =3.17, IC₀₂₅ =0.97, EBGM₀₅ =2.12), retinal vein occlusion (n=3, ROR₀₂₅ =3.28, PRR =10.17, IC₀₂₅ =0.18, EBGM₀₅ =3.27), hip fracture (n=17, ROR₀₂₅ =2.77, PRR =4.46, IC₀₂₅ =1.22, EBGM₀₅ =2.77), as drug-specific disproportionality signals with IME, was found in this study.

Latency period analysis

The onset patterns of AEs associated with the four bronchodilators are presented in Figure 5 and Table 4. Due to data sparsity, reliable Weibull distribution fitting could not be performed for all SOC. Therefore, the WSP calculation was not feasible for some SOCs (Tables S22-S25). Kaplan-Meier analysis with the log-rank test revealed a significant difference in the cumulative probability of AEs among the four bronchodilators (log-rank test, P<0.005). The median time to AE onset for formoterol/glycopyrronium, indacaterol/glycopyrronium, vilanterol/umeclidinium, and olodaterol/tiotropium was 29 [interquartile range (IQR): 7–102], 49 (IQR: 12–239), 32.5 (IQR: 5–181), and 127 (IQR: 4–141) days, respectively. WSP analysis demonstrated that all four bronchodilators exhibited an “early failure” pattern, indicating that the majority of AEs occurred early during treatment and subsequently declined over time (formoterol/glycopyrronium, β=0.68, 95% CI: 0.56–0.80; indacaterol/glycopyrroniu, β=0.60, 95% CI: 0.56–0.65; vilanterol/umeclidinium, β=0.47, 95% CI: 0.45–0.50; olodaterol/tiotropium, β=0.53, 95% CI: 0.50–0.57). In the analysis stratified by SOC, most evaluable SOCs for the four bronchodilators were classified as exhibiting an early failure pattern (Tables S22-S25). Notably, for olodaterol/tiotropium, the “Eye disorders” (β=0.49) and “Musculoskeletal and connective tissue disorders” (β=0.60) SOCs also demonstrated early failure patterns, indicating that AEs in these categories tend to occur early after treatment initiation. It is worth noting that despite olodaterol/tiotropium having the longest median onset time (127 days), its overall AE profile remained consistent with an “early failure” pattern, suggesting that while some events may emerge later, the hazard of AE occurrence is highest in the initial treatment period.

Figure 5.

Latency period from administration of four bronchodilators to adverse event onset. (A) Latency period of adverse events of the four bronchodilators; (B) cumulative reporting proportion of adverse events since the initiation of four bronchodilators. Latency period: time from long-acting dual bronchodilator inhalation start to PT diagnosis. F, formoterol; G, glycopyrronium; I, indacaterol; O, olodaterol; PT, preferred term; T, tiotropium; U, umeclidinium; V, vilanterol.

Table 4. The Weibull parameter of long-acting dual bronchodilator-related AEs in FAERS.

Drug	Case, n	TTO (days), median [IQR]	Weibull distribution		Failure type
			Scale parameter, α (95% CI)	Shape parameter, β (95% CI)
Formoterol/glycopyrronium	76	29 [7–102]	61.07 (39.71–82.42)	0.68 (0.56–0.80)	Early failure
Indacaterol/glycopyrronium	394	49 [12–239]	116 (96.51–136.65)	0.60 (0.56–0.65)	Early failure
Vilanterol/umeclidinium	863	32.5 [5–181]	89 (76.10–102.89)	0.47 (0.45–0.50)	Early failure
Olodaterol/tiotropium	432	127 [4–141]	65.27 (53.11–77.44)	0.53 (0.50–0.57)	Early failure

AE, adverse event; CI, confidence interval; FAERS, Food and Drug Administration Adverse Event Reporting System; IQR, interquartile range; TTO, time to onset.

Discussion

To our knowledge, this study is the most comprehensive to investigate the safety signal of four long-acting dual bronchodilators (formoterol/glycopyrronium, indacaterol/glycopyrronium, vilanterol/umeclidinium, olodaterol/tiotropium) using the FAERS database. We have identified promising and distinctive signals, targeting the reduction of potential harm from the inhalation therapy of four long-acting dual bronchodilators. Dyspnoea and cough had the highest reporting proportions among safety signals associated with the four bronchodilators. Moreover, 13 common safety signals and 56 IMEs with drug-specific disproportionality signals were identified among the four bronchodilators by Venn analysis. Urinary retention was the only IME identified among the common safety signals. Indacaterol/glycopyrronium exhibited a distinct signal for cardiovascular IMEs within its safety signal profile, including arrhythmia, cardiac arrest, cardiac fibrillation, infarction, angina pectoris and coronary artery disease. Dementia Alzheimer’s type, senile dementia, deafness were also identified in indacaterol/glycopyrronium. A safety signal for ileus, retinal vein occlusion, cataract, and hip fracture was found in olodaterol/tiotropium. For the latency periods, olodaterol/tiotropium had the longest latency period for safety signals. In contrast, the latency periods for formoterol/glycopyrronium, indacaterol/glycopyrronium, and vilanterol/umeclidinium were all within 50 days. Moreover, the reporting proportion of safety signals peaked early in the prescribing period and gradually decreased thereafter.

Most reported safety signal

In this study, cough was the most reported safety signal associated with four bronchodilators. The causes of cough include medication-related factors and disease-related factors. For the medication-related factors, four bronchodilators were reported to cause cough in the drug label. For disease-related factors, cough serves as a crucial airway defense mechanism, yet it is also a predominant symptom of various respiratory diseases, including asthma and COPD. Previous studies have shown that most patients experience an improvement in coughing after using inhalers (20). Therefore, this report may be subject to bias, given that over 50% of the respondents were non-health professionals.

IME with common safety signals

In our study, urinary retention was identified as an IME among 13 common safety signals through Venn analysis. It is well-recognized that the pathogenesis of urinary retention is closely linked to the administration of inhaled anticholinergic agents (21,22). In the previous study, this association was not dose-dependent, but rather varied with the method of administration, with the highest risk identified in nebulizer delivery. Additionally, the correlation was most significant among male patients presenting with both COPD and benign prostatic hyperplasia (23). However, elderly male patients with benign prostatic hyperplasia are at a higher risk of developing acute urinary retention (24).

Drug-specific disproportionality signals with IME

Cardiac disorders represent a critical potential risk that requires close monitoring during treatment with LABA/LAMA. Based on Venn analysis, indacaterol/glycopyrronium was associated with the highest number of cardiovascular IMEs. Arrhythmia, cardiac arrest, cardiac fibrillation, myocardial infarction, angina pectoris, and coronary artery disease have been documented as safety signals associated with indacaterol/glycopyrronium treatment. Our stratified analyses by sex and age further revealed that these cardiovascular signals were more pronounced in the male subgroup and were even more significant in patients aged over 65 years (Tables S2-S5,S7-S10,S12-S15,S17-S20). Notably, indacaterol/glycopyrronium aligns with its male-predominant (46.0%) usage in this study. Additionally, cor pulmonale has been reported in patients receiving olodaterol/tiotropium therapy. Pulmonary congestion has also been observed in association with glycopyrronium/formoterol use. The cardiovascular safety of LAMA/LABA therapy remains controversial. Compared with inhaled corticosteroids (ICS)/LABA, LAMA/LABA therapy increases the risk of major adverse cardiovascular events in patients with COPD (25). In previously FAERS study (26), LAMAs are not free from cardiac AE risk, but indacaterol/glycopyrronium, vilanterol/umeclidinium and formoterol/glycopyrronium showed no significant cardiovascular disease (CVD) safety signal compared with glycopyrronium or umeclidinium. However, LABA/LAMA did not show a higher cardiovascular risk compared to the use of ICS/LABA (27). The controversy is driven by the interaction between disease progression and pharmacological actions. In the real-world study, COPD is frequently complicated by CVD, and disease exacerbations elevate the risk of acute cardiovascular events. Notably, a high prevalence of CVD exists even among clinically stable COPD patients (28,29), though it is often overlooked. This cardiovascular risk increases substantially during acute exacerbations (30,31). Overall, indacaterol/glycopyrronium suggested elevated signals for atrial fibrillation, arrhythmia, and coronary artery disease, while formoterol/glycopyrronium and olodaterol/tiotropium showed signals that may be associated with heart failure.

Interesting, dementia Alzheimer’s type (n=8, ROR₀₂₅ =2.65, PRR =5.31, IC₀₂₅ =0.88, EBGM₀₅ =2.65) and senile dementia (n=6, ROR₀₂₅ =26.39, PRR =59.09, IC₀₂₅ =1.56, EBGM₀₅ =25.96) was observed in indacaterol/glycopyrronium. In the previous studies, anticholinergics have been demonstrated to increase the risk of dementia (32,33). In sensitivity analyses stratified by sex and age, these dementia signals did not show a clear distinction between males and females. However, they became more pronounced in patients aged over 65 years, suggesting that age may be a significant modifying factor in this association. Moreover, cumulative exposure to anticholinergic drugs is a critical risk factor for dementia (34). Approximately 10% of indacaterol/glycopyrronium is absorbed through the gastrointestinal tract into systemic circulation, leading to elevated systemic drug levels that may increase the risk of dementia (21). Additionally, deafness (n=15, ROR₀₂₅ =2.03, PRR =3.36, IC₀₂₅ =0.83, EBGM₀₅ =2.02) was observed in the indacaterol/glycopyrronium. The mechanism by which indacaterol/glycopyrronium induces hearing loss or deafness remains unclear, but it may involve effects on M₁ receptors, leading to auditory impairment or deafness (35). Despite the relatively low number of reported cases and generally low systemic exposure via inhalation, the robust safety signals for both dementia and deafness associated with indacaterol/glycopyrronium underscore the need for extended safety monitoring and further investigation into the underlying mechanisms.

Olodaterol/tiotropium was associated with ileus (n=6, ROR₀₂₅ =2.09, PRR =4.65, IC₀₂₅ =0.52, EBGM₀₅ =2.09), which was identified as both an IME and a drug-specific disproportionality signal. Although ileus has been documented as an AE in the prescribing information for olodaterol/tiotropium, this AE remains drug-specific disproportionality signal when compared to three long-acting dual bronchodilators. Tiotropium may contribute to the development of ileus by modulating both M₂ and M₃ receptors in gastrointestinal smooth muscle (36). The occurrence of this AE may be associated with the slower dissociation rate of tiotropium from the M₃ receptor compared to glycopyrronium and umeclidinium (37). Hip fracture (n=17, ROR₀₂₅ =2.77, PRR =4.46, IC₀₂₅ =1.22, EBGM₀₅ =2.77) was also found in olodaterol/tiotropium. In the sensitivity analysis stratified by sex and age (Tables S17-S20), this signal was observed exclusively in the female subgroup (n=13, ROR₀₂₅ =3.07, PRR =5.39, IC₀₂₅ =1.24, EBGM₀₅ =3.06), but was not detected in the male subgroup or prominently in the age-stratified analysis. Several factors may contribute to this finding. First, the olodaterol/tiotropium group had a higher proportion of elderly females (Table 3), a population with an inherently elevated baseline risk of osteoporosis. Second, patients treated with long-acting bronchodilators often have a history of systemic corticosteroid use, a well-established risk factor for reduced bone mineral density and fractures. Third, anticholinergic burden may play a role: previous studies have reported a positive correlation between Anticholinergic Risk Scale (ARS) scores and fracture risk (38), and a higher anticholinergic burden has been independently associated with an increased risk of hip fracture (39). In elderly hip fracture patients, anticholinergic exposure has also been linked to higher 30-day and one-year mortality (40). The hip fracture signal observed with olodaterol/tiotropium may reflect a combination of direct drug effects and the high-risk characteristics of the treated population. Clinicians should remain vigilant regarding this potential safety signal when prescribing olodaterol/tiotropium, particularly in elderly female patients. Notably, new safety signals not listed in the prescribing information were identified for olodaterol/tiotropium, specifically cataract (n=24) and retinal vein occlusion (n=3), both of which were characterized as IMEs. In the sensitivity analysis stratified by sex and age (Tables S17-S20), cataract was identified in the male subgroup (n=11, ROR₀₂₅ =2.35, PRR =4.23, IC₀₂₅ =0.90, EBGM₀₅ =2.34) but was not observed in the female subgroup or prominently in the age-stratified analysis. The underlying mechanism for olodaterol/tiotropium-related eye disorders remains uncertain, and therefore, clinicians should be aware of these potential vision-related AEs.

A disproportionality signal for necrotising ulcerative gingivostomatitis was observed with vilanterol/umeclidinium (n=3, ROR₀₂₅ =23.5, PRR =74.78, IC₀₂₅ =0.46, EBGM₀₅ =22.48). Biologically, this condition is closely associated with oral microbiome dysbiosis (41), which can often be secondary to drug-induced dry mouth (42). Methodologically, however, this finding warrants caution. The extremely low case count (n=3) and wide CI suggest it may reflect reporting variability or statistical instability rather than a true safety signal. This case highlights that drug-specific disproportionality signals should be interpreted cautiously, as they can be influenced by low statistical power due to sparse data in comparator groups. Notably, dry mouth, a well-recognized anticholinergic adverse effect, also demonstrated a disproportionality signal for vilanterol/umeclidinium (n=112, ROR₀₂₅ =2.73, PRR =3.28, IC₀₂₅ =1.41, EBGM₀₅ =2.72). While it has been suggested that umeclidinium’s slow dissociation from the M₃ receptor may contribute to anticholinergic effects, any association with oral ulcerative conditions remains speculative and is not supported by the present data.

Latency period

Latency period analysis serves as a valuable tool for healthcare professionals to predict and detect AEs. In this study, due to the limitations of the FAERS database, only 9.45% of the collected reports were ultimately included in the time-to-event analysis. Among the four bronchodilators, olodaterol/tiotropium exhibited the longest median time to AE onset [median 127 (IQR: 4–141) days], whereas for formoterol/glycopyrronium, indacaterol/glycopyrronium, and vilanterol/umeclidinium, most AEs occurred within 50 days of treatment initiation. WSP analysis revealed distinct temporal patterns across different SOCs. For all four bronchodilators, “Cardiac disorders” exhibited an early failure pattern. Interestingly, for olodaterol/tiotropium, the “Eye disorders” (β=0.49) and “Musculoskeletal and connective tissue disorders” (β=0.60) SOCs also demonstrated early failure patterns, indicating that AEs in these categories tend to occur early during treatment and subsequently decline. Although these findings suggest that most AEs associated with olodaterol/tiotropium occur early, the relatively prolonged median onset time (127 days) underscores the importance of continued vigilance. Clinicians should be aware that some AEs may emerge after several months of treatment, and long-term periodic monitoring remains warranted, particularly for patients on olodaterol/tiotropium therapy.

Limitation

Firstly, the FAERS database, which operates on a voluntary reporting basis, may not accurately reflect the true situation (43). It is challenging to establish a clear relationship between the drug and AEs. The inability to calculate true clinical incidence due to the lack of a denominator is a major limitation of using the FAERS database. Secondly, bias is acknowledged in this study as an inherent element that cannot be entirely eliminated. The safety signals can be influenced by missing data, such as age, gender, concomitant medication use, reporter, disease symptoms, disease progression, and treatment duration. The observed safety signal for “respiratory, thoracic, and mediastinal disorders” may be confounded by the underlying pathophysiology of COPD and its associated complications, which could potentially obscure drug-related safety signals. In addition, the drug-specific disproportionality signals identified through Venn analysis represent signals that reached statistical significance specifically within one drug cohort. This may be influenced by differences in sample size and statistical power across the groups rather than representing true clinical exclusivity. For rare events with low case counts and wide CIs, such as necrotising ulcerative gingivostomatitis, the lack of a signal in other cohorts does not definitively rule out their occurrence. Future large-scale prospective studies are needed to confirm these drug-specific disproportionality signal associations. Fourth, the data sparsity for specific AEs limited our ability to perform a refined, stratified temporal analysis. The inherent reporting bias and missing time-to-onset data in spontaneous reporting databases often preclude the fitting of Weibull distribution, particularly for rare events, which remains a common challenge in pharmacovigilance studies.

Conclusions

This pharmacovigilance study identified significant safety signals associated with four dual bronchodilators, highlighting both common and drug-specific disproportionality signals and providing essential evidence to inform their clinical monitoring.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.