Association of Cardiovascular Risk With Inhaled Long-Acting Bronchodilators in Patients With Chronic Obstructive Pulmonary Disease: A Nested Case-Control Study

Key Points

Question  Does the duration since initial use and new use of inhaled long-acting β₂-agonists (LABAs) or antimuscarinic antagonists (LAMAs) for the treatment of chronic obstructive pulmonary disease (COPD) act as important determinants of the risk of cardiovascular disease?

Findings  In this nested case-control study of more than 280 000 patients with COPD, new use of LABAs or LAMAs is associated with an approximate 1.5-fold increased cardiovascular risk within 30 days of initiation therapy.

Meaning  Health care professionals need to be very vigilant with regard to any cardiovascular symptoms within 30 days of initiating LABA or LAMA treatment for COPD.

Importance  The associations between cardiovascular disease (CVD) and inhaled long-acting β₂-agonists (LABAs) or long-acting antimuscarinic antagonists (LAMAs) in chronic obstructive pulmonary disease (COPD) are greatly debated. Pivotal and relevant randomized clinical trials included prior LABA or LAMA users and excluded patients with baseline CVD; therefore, cardiovascular events arising from first-time LABA or LAMA use, if any, could not be observed. There is an urgent need to examine whether new use of and duration since initiating LABAs and LAMAs could act as important determinants of cardiovascular events.

Objective  To investigate risk of CVD associated with LABAs and LAMAs, focusing on the initiation and duration of LABA and LAMA therapies.

Design, Setting, and Participants  This nested case-control study included 284 220 LABA-LAMA–naïve patients with COPD at least 40 years old (mean age, 71.4 years; 68.9% men), retrieved from the Taiwan National Health Insurance Research Database for health care claims from 2007 to 2011.

Exposure  LABA or LAMA use was measured in the year preceding the event or index date, stratified by duration since initiation of LABA or LAMA treatment, new and prevalent users, concomitant COPD medications, and individual agents.

Main Outcomes and Measures  Cases with inpatient or emergency care visits for coronary artery disease, heart failure, ischemic stroke, or arrhythmia were identified and individually matched to 4 randomly selected controls. Conditional logistic regressions were performed to estimate odds ratios of CVD from LABA and LAMA treatment.

Results  During a mean follow-up of 2.0 years, 37 719 patients with CVD (mean age, 75.6 years; 71.6% men) and 146 139 matched controls (mean age, 75.2 years; 70.1% men) were identified. New LABA and LAMA use in COPD was associated with a 1.50-fold (95% CI, 1.35-1.67; P < .001) and a 1.52-fold (95% CI, 1.28-1.80; P < .001) increased cardiovascular risk within 30 days of initiation, respectively, whereas the risk was absent, or even reduced with prevalent use. Individual LABA agents, LAMA dosage forms, and concomitant COPD regimens did not differ in the CVD risks. The risk persisted in an alternative case-crossover study and remained across subgroups without CVD history or prior exacerbations.

Conclusions and Relevance  New initiation of LABAs or LAMAs in patients with COPD is associated with an approximate 1.5-fold increased severe cardiovascular risk, irrespective of prior CVD status and history of exacerbations.

Chronic obstructive pulmonary disease (COPD) is a chronic and irreversible inflammatory lung disease, presently posing a significant burden to health care systems across the world.^1-6 Long-acting β₂-agonists (LABAs) and long-acting muscarinic antagonists (LAMAs) are the mainstay therapies for COPD^7,8; however, these agents were found to increase the risk of cardiovascular disease (CVD), although the findings varied, concluding that there was no increased risk^9-14 or that there was a 1.1- to 4.5-fold increased cardiovascular risk.^15-18 These studies, even a large randomized clinical trial (RCT),⁹ generally observed few cardiovascular events, excluded patients with severe illness,^9-11 and obtained incomplete medication records^17,18 as well as dropped more than 50% of eligible patients.¹⁵ All of these limitations could potentially weaken the causality or generalizability of the association.

Notably, inclusion of patients with tolerability to the risk of CVD receiving LABA or LAMA therapy is probably a major drawback in previous large RCTs. The 4-year Understanding Potential Long-term Impacts on Function with Tiotropium (UPLIFT) trial¹⁰ and the 3-year Toward a Revolution in COPD Health (TORCH) trial,⁹ which observed a reduced or no increase in CVD risk, included more than 45% and 35% of participants who had received inhaled cholinergics and LABAs, respectively, and the 2 trials excluded patients with a history of recent CVD and life-threatening cardiovascular events. Accordingly, patients who had developed severe CVD with new use of LABAs or LAMAs could have been excluded, and only those LABA or LAMA prevalent users who had probably developed tolerance to the cardiovascular risk were included in the trials. Accordingly, we suspect that new use of and duration since initiating LABAs and LAMAs could act as important determinants of CVD risk, which to date have not been examined in details in previous studies.

The present study investigated use of LABAs and LAMAs associated with the risk of CVD in a nationwide population of patients with COPD, focusing on new use and duration of therapy, individual agents, dosage forms, and concomitant COPD regimens.

We performed a nested case-control study of a nationwide COPD population 40 years or older with the data retrieved from the Taiwan National Health Insurance Research Database (NHIRD) from January 1, 2007, to December 31, 2011, as depicted in eFigure 1 in the Supplement. The NHIRD contains all medical and pharmacy claims records from all medical care settings for more than 99% of the 23 million Taiwanese inhabitants covered under a compulsory and universal national health insurance. The claims are audited quarterly by the National Health Insurance Administration,¹⁹ and multiple disease diagnosis codes in the database have been validated, including several CVD diagnoses.^20-22 The study was exempt from a full review by the institutional review board of Tri-Service General Hospital, National Defense Medical Center.

We identified patients with COPD 40 years or older who had made 2 outpatient visits or an inpatient visit for COPD (International Classification of Diseases, Ninth Revision [ICD-9] codes 491.xx, 492.xx, and 496.xx) in a single year from January 1, 2008, to June 30, 2011, accompanied by a record of filling at least 1 COPD medication at each visit. The cohort entry date was set as the date of the first COPD outpatient visit or the discharge date from a COPD hospitalization. We excluded patients who had received any LABA-LAMA therapy or who lacked continuous health insurance coverage for 1 year preceding cohort entry. We followed-up the remaining patients until the earliest of CVD outcome (defined in the outcome measurement), National Health Insurance program withdrawal, death, or the end of the study period (December 31, 2011). We determined mortality from the NHIRD based on a previously reported approach.²³

We identified CVD cases as patients who had made an inpatient or emergency department (ER) visit with a primary diagnosis of coronary heart disease (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM], codes 410–414), cardiac arrhythmia (code 427), heart failure (code 428), or ischemic stroke (codes 433-434). The adopted coding system for acute myocardial infarction, heart failure, and ischemic stroke has been validated with high accuracy.^20,21,24 We set the first CVD event as the index date.

Each case was matched to 4 randomly selected controls from risk set samples by cohort entry date (±180 days) and disease risk score (DRS) (±0.01) of predicting the occurrence of a cardiovascular outcome during follow-up. Use of the DRS to match cases with controls yields better statistical precision than does exact matching on multiple discrete factors in a nested case-control setting.²⁵ Estimation of the DRS is detailed in eMethods 1 in the Supplement.

We examined all LABA and LAMA prescription records in the year before the index date for both cases and controls. Specifically, drug use was classified into current (≤30 days), recent (31-90 days), past (91-180 days), and remote (>180 days) use based on the most recent prescription date preceding the index date. Current users were further classified as new users if they had no other dispensing records in the 31 to 365 days before the index date and prevalent users otherwise. We also used restricted cubic spline models to analyze the duration of new LABA and new LAMA therapy (as detailed in eMethods 2 in the Supplement). New use of LABAs and LAMAs was also categorized according to various combinations with COPD medication agents, individual LABAs, and different drug-containing devices of LAMAs. Individual agents of LABAs and LAMAs comprised salmeterol, formoterol, and tiotropium. Nonusers of LABAs-LAMAs and new theophylline users were set as the reference group in the main analyses and sensitivity analyses, respectively.

Important determinants of CVD were considered, such as prior CVDs, hypertension, diabetes, hyperlipidemia, use of CVD medications, and agents related to cardiotoxicity. We also considered proxy indicators of COPD severity, including the number of COPD-related outpatient visits accompanied by oral corticosteroids or respiratory antibiotic prescription records (referred to as moderate exacerbation) and the number of severe exacerbation events (defined as any inpatient or ER admission for COPD), as well as the presence of incident or prevalent COPD. Other factors related to LABA-LAMA use or CVD were considered (as detailed in eTable 1 in the Supplement). We first measured these factors in the year preceding cohort entry, and included them in a logistic regression model to estimate the DRS of encountering a CVD outcome during follow-up (eMethods 1 in the Supplement). Second, we treated these confounders as time-varying effects by also assessing concomitant medications and the remaining factors during the 6 months and in the year prior to the index date, respectively.

The covariate balance before and after DRS matching was assessed using the standardized difference, and meaningful imbalances between groups were determined when the standardized difference was greater than 0.1.²⁶ We used conditional logistic regressions to estimate the odds ratio (OR) of CVD with LABA or LAMA use. We expressed the absolute risk of CVD arising from LABA and LAMA treatment as the number needed to harm (NNTH)²⁷ according to a previously reported formula.

Data sets were constructed and analyzed by using SAS (version 9.3; SAS Institute Inc) and STATA (version 13; StataCorp) statistical software, respectively. We defined tests with 2-sided P < .05 as significant.

Multiple additional analyses were performed. First, we adopted a case-crossover study to avoid selection bias and minimize time-invariant confounding (eMethods 3 and eFigure 2 in the Supplement). Second, to mitigate confounding by indication bias, we repeated the analyses by treating patients who began new treatment with theophyllines, another class of bronchodilators used for COPD therapy, in the 30 days before the index date as the reference group. Third, to address protopathic bias, we used a lag-time approach²⁸ that disregarded any prescription records of LABAs and LAMAs in the 7 days before the index date and restricted new LABA and LAMA users to those who received spirometry testing within 30 days before or on the date of initiating these medications to justify respiratory medication use. We also excluded patients who started any cardiovascular medications in the 30 days before the index date and those with any chest pain (ICD-9 code 786.5) or dyspnea/breathing difficulty diagnoses (code 786.0) as possible presymptoms of CVD. Fourth, events of heart failure were excluded from the CVD outcome. Fifth, we also performed stratified analyses by CVD care type, baseline CVD status, CVD severity (fatal vs nonfatal), COPD severity, asthma comorbidity, use of theophyllines, and use of systemic short-acting β₂-agonists. Sixth, we adjusted for all covariates that were measured preceding the index date in Table 1 and evaluated the impact of unmeasured confounding using a rule-out approach,²⁹ as detailed in eFigure 3 in the Supplement.

The study cohort consisted of 284 220 LABA-LAMA-naive patients with COPD with a mean age of 71.4 years, 68.9% of whom were male. During a mean follow-up of 2.0 years, we identified 37 719 patients with severe CVD requiring hospitalization or emergency care, at a rate of 6.6 per 100 person-years, and included 146 139 matched controls (eFigure 4 in the Supplement).

All characteristics were balanced prior to cohort entry between cases and controls (Table 1). During follow-up, most of the covariates remained balanced between the 2 groups, and only a few factors differed significantly higher between cases and controls, for which statistical adjustment was performed.

As indicated in Table 2, overall use of inhaled long-acting bronchodilators was not associated with an increased risk of CVD across different recency of therapy, although a 10% decrease in cardiovascular risk was observed with past LABA use. Among current users, new initiation of LABA and LAMA treatment was associated with a 1.50-fold (95% CI, 1.35-1.67; P < .001) and a 1.52-fold (95% CI, 1.28-1.80; P < .001) increased cardiovascular risk, respectively, while prevalent LABA or LAMA use yielded a 9% to 12% reduction in risk. In addition, new LABA use vs new LAMA use yielded no difference in the risk of CVDs (P = .93) (eTable 2 in the Supplement).

Various clinical usages of LABAs and LAMAs within 30 days of therapy initiation were scrutinized (Table 3). The adjusted ORs ranged minimally from 1.42 to 1.51 across the various LABA medication regimens. Salmeterol and formoterol were found to present similar CVD risks. Different LAMA regimens, including use of tiotropium only, carried similar risks, ranging from 1.39-fold to 1.58-fold increases. Analyses of individual CVD outcomes revealed increased risks of coronary artery disease and heart failure with LABA and LAMA treatment, and an increased risk for cardiac arrhythmias with LAMA therapy (eTable 3 in the Supplement).

Figure 1 presents the results of duration-response analysis, which indicated that the cardiovascular risks peaked at around the 30th day after new initiation of LABA or LAMA therapy; waned from 31 to 60 days of therapies, and reduced to a level even lower than the baseline risk from 71 to 240 days.

Our main findings remained robust in most of the sensitivity analyses (Figure 2, eTables 4 and 5, eFigures 2 and 3 in the Supplement). The increased risk of CVD with new LABA and new LAMA use persisted in a case-crossover analysis (eTables 4-5, eFigure 2 in the Supplement) and in comparison with new use of theophyllines. The results of the analyses performed to address protopathic bias corresponded closely to the main findings. The LABA- or LAMA-associated risk of CVDs remained significant regardless of patients’ CVD history and COPD exacerbations. In addition, an unmeasured confounder could fully account for our findings only if the confounder increased the CVD risk by 4-fold and was more prevalent in new LABA and LAMA users by 3.5 and 3.6 times, respectively, than in nonusers (eFigure 3 in the Supplement).

One severe CVD event requiring hospitalization or ER care occurred for every 406 (95% CI, 303-580) new LABA users and 391 (95% CI, 254-725) new LAMA users during the first 30 days of therapy (eTable 6 in the Supplement).

This large observational study of more than 280 000 patients with COPD reported an approximate 1.5-fold increased risk of a severe CVD event within 30 days of initiation of LABA or LAMA therapy, but the risk was absent, or even reduced, with prevalent use of these medications. The findings were replicated in a case-crossover design analysis and unaffected by patients’ baseline CVD status. Collectively, we have provided the first evidence to indicate that new use and duration since initiation of inhaled long-acting bronchodilators are associated with the therapy-related risk of CVD in patients with COPD.

Previous studies^15-18,30,31 that observed an increased risk of CVD from LABAs or LAMAs evaluated elderly patients with COPD,¹⁵ identified few CVD events,^18,30 or included patients with prior LABA or LAMA usage.^15-18,30,31 None of these studies^15-18,30,31 examined drug use within a time period as short as 30 days after the initiation of therapy. The ability to observe an acute cardiovascular effect with inhaled long-acting bronchodilators in our study primarily resulted from the adoption of a new-user design and examination of a nationwide population of patients with COPD.

Our observed duration-response associations may explain why previous relevant RCTs did not find an increased risk of CVD with LABA or LAMA use. We report herein that the greatest risk of CVD emerged at around the 30th day following initiation of LABA and LAMA therapy. Nevertheless, a substantial proportion of participants with prior LABA or LAMA usage was included in previous RCTs.^9-12,32 For instance, more than one-third of the enrolled patients received LABAs at baseline in the TORCH trial⁹; therefore, cardiovascular events arising following first use of LABAs or LAMAs, if any, could not be observed. In addition, patients with severe CVD or life-threatening cardiovascular events at baseline were excluded in prior trials.^9-12 Both factors could have led to the inclusion of patients with tolerability to cardiovascular events in the RCTs.

The phenomenon of depletion of susceptibles effect may contribute to the observed duration-response associations. Depletion of susceptibles effect has been described as an increased rate of an event from the initial exposure to a medication, followed by a decreased incidence rate with a longer drug exposure,^33,34 such as the temporal sequence of hormone replacement therapy and its relationship with venous thromboembolism.³⁴ We suspect that there may exist a subgroup of patients with COPD who are particularly at risk of CVD with initial exposure to LABAs or LAMAs owing to predisposing factors that could amplify sympathetic overactivation or systemic inflammation with inhaled long-acting bronchodilators. Future studies should identify and characterize the most vulnerable subgroups of patients with COPD. We also suspect that after depletion of susceptible patients who experienced CVD events early on, the remaining patients with COPD who continued to use LABAs or LAMAs may exhibit improved systemic inflammation,³⁵ or may have a more stable lung function status,³⁶ which could lessen the CVD burden in these patients.

Our findings contribute significantly to investigation of the cardiovascular safety of inhaled long-acting bronchodilators, in which the risk window of adverse cardiovascular events from inhalation therapies is specified. Based on our findings, we suggest that the use of inhaled long-acting bronchodilators in COPD need to be carefully assessed, and a thorough cardiovascular physical examination, especially heart rate measurement and electrocardiograms, need to be performed when prescribing LABAs and LAMAs to patients. Health care professionals should be vigilant for any cardiovascular symptoms during the first 30 days of inhalation therapy. Given that CVD is highly prevalent among patients with COPD, clinicians should also pay attention to the management of CVD risk factors throughout the duration of LABA or LAMA therapy. We also suggest that physicians assess patients’ cardiovascular risk before initiation of LABAs or LAMAs, and, if needed, a preventive therapy for CVD should be considered during the initial treatment of inhaled long-acting bronchodilators.

Two potential mechanisms have been proposed to underpin the cardiovascular risk arising from LABA and LAMA use. Acting as autonomous nervous system agents, LABAs and LAMAs are believed to cause sympathetic overactivation by activating sympathetic β₂-adrenergic receptors (β2R) and suppressing parasympathetic muscarinic-3 receptors (M3R), respectively.^6,37,38 However, LABA and LAMA use in COPD has been observed to increase the inflammatory cytokine levels, such as an increased level of interleukin-8,^39,40 which could lead to an increased risk of CVD.

Several important strengths of our study merit discussion. We investigated the whole spectrum of timing and duration effects of LABA and LAMA usage on the development of severe CVD events in a population with COPD, which is a clinically important aspect that, to our knowledge, no study has addressed previously. In addition, this study was the first to examine the effect of LABA-LAMA regimens with various concomitant COPD medications, which provided pivotal information given that COPD care usually involves multiple therapies. Furthermore, a new-user design served to minimize tolerance effects in prevalent drug users. Moreover, our findings can be generalized to the entire COPD population with multiple coexisting comorbidities.

Several study limitations of the current report need to be addressed. First, worsening of COPD could have prompted the use of LABAs or LAMAs and caused the observed CVD. To control for this confounding effect, we adjusted for multiple proxies of COPD severity and compared new LABA users and new LAMA users, respectively, with new theophylline users, which yielded consistent results. Second, selection bias may have been present, although this bias was expected to be minimal, because we balanced all covariates at cohort entry between cases and controls, treated them as time-varying effects, and obtained consistent findings with an adopted case-only design. Third, protopathic bias could have existed. Nevertheless, we reached the same conclusions when adopting multiple approaches to address this bias, such as the use of a lag-time approach. Fourth, patients’ CVD status at baseline may have confounded our findings. However, we believe this confounding effect not to be substantial because all baseline CVD statuses and CVD medications were balanced between the 2 groups, and the LABA- or LAMA-associated CVD risk remained significant among patients without a CVD history. Fifth, several important determinants of CVD, for example, smoking and alcohol consumption, were unavailable, which could have confounded our findings. However, their effects were unlikely to fully explain our findings, as assessed by rule-out analysis, especially for highly prevalent confounders such as smoking. Sixth, newer LABAs or LAMAs were unable to be assessed, such as aclidinium.

New use of LABAs or LAMAs was associated with a 1.5-fold increased cardiovascular risk in patients with COPD within 30 days of therapy initiation. We caution physicians to closely monitor new users of LABAs or LAMAs for cardiovascular symptoms.

Corresponding Author: Meng-Ting Wang, PhD, School of Pharmacy, National Defense Medical Center, 9 F, No. 161, Section 6, Min–Chuan East Road, Taipei 114, Taiwan (wmt@mail.ndmctsgh.edu.tw).

Accepted for Publication: November 3, 2017.

Published Online: January 2, 2018. doi:10.1001/jamainternmed.2017.7720

Author Contributions: Dr Wang had full access to all of the study data and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: M.-T. Wang, Liou, Lin, Tsai, Y.-H. Wang.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: M.-T. Wang, Liou, Lin.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Y.-H. Wang, Lai.

Obtained funding: M.-T. Wang.

Administrative, technical, or material support: M.-T. Wang, Liou, Lin, Tsai, Hsu.

Study supervision: M.-T. Wang.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by grants from the Ministry of Science and Technology, ROC (MOST 103-2320-B016-010).

Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Meeting Presentation: Part of the findings in the current study was presented in a poster presentation at the Seventh Asian Association of Schools Pharmacy Conference; October 30, 2015; Taipei, Taiwan.

Additional Contributions: The analyzed databases were provided by the National Health Research Institutes (NHRI), but the interpretation and conclusions contained herein did not represent those of NHRI. We sincerely thank for the technical supports from Bi-Juan Wu, ADN, School of Pharmacy, National Defense Medical Center, who received no compensations.