Cardiovascular Safety of Inhaled Long-Acting Bronchodilators in Individuals With Chronic Obstructive Pulmonary Disease

Importance  Chronic obstructive pulmonary disease (COPD) is a common and deadly disease. Long-acting inhaled β-agonists and anticholinergics, first-line medications for COPD, have been associated with increased risk of cardiovascular outcomes. When choosing between the medications, patients and physicians would benefit from knowing which has the least risk.

Objective  To assess the association of these classes of medications with the risk of hospitalizations and emergency department visits for cardiovascular events.

Design  We conducted a nested case-control analysis of a retrospective cohort study. We compared the risk of events between patients newly prescribed inhaled long-acting β-agonists and anticholinergics, after matching and adjusting for prognostic factors.

Setting  Health care databases from Ontario, the largest province of Canada, with a multicultural population of approximately 13 million.

Participants  All individuals 66 years or older meeting a validated case definition of COPD, based on health administrative data, and treated for COPD from September 1, 2003, through March 31, 2009.

Exposure  New use of an inhaled long-acting β-agonist or long-acting anticholinergic.

Main Outcome and Measures  An emergency department visit or a hospitalization for a cardiovascular event.

Results  Of 191 005 eligible patients, 53 532 (28.0%) had a hospitalization or an emergency department visit for a cardiovascular event. Newly prescribed long-acting inhaled β-agonists and anticholinergics were associated with a higher risk of an event compared with nonuse of those medications (respective adjusted odds ratios, 1.31 [95% CI, 1.12-1.52; P < .001] and 1.14 [1.01-1.28; P = .03]). We found no significant difference in events between the 2 medications (adjusted odds ratio of long-acting inhaled β-agonists compared with anticholinergics, 1.15 [95% CI, 0.95-1.38; P = .16]).

Conclusions and Relevance  Among older individuals with COPD, new use of long-acting β-agonists and anticholinergics is associated with similar increased risks of cardiovascular events. Close monitoring of COPD patients requiring long-acting bronchodilators is needed regardless of drug class.

Chronic obstructive pulmonary disease (COPD), a largely preventable and manageable respiratory condition, affects more than 1 in 4 adults older than 35 years, is the third leading cause of death in the United States, and is a leading cause of hospitalization for chronic disease.^1-3 Medications are a mainstay of COPD management. Long-acting β-agonists (LABAs, consisting of salmeterol xinafoate and formoterol fumarate) and long-acting anticholinergics (LAAs, of which tiotropium bromide monohydrate is the only one currently available) have shown benefit in large randomized clinical trials (RCTs),^4,5 and are currently recommended interchangeably for moderate to severe disease.⁶

Although little controversy remains as to the effectiveness of LABAs and LAAs, their cardiovascular safety is a matter of debate. Long-acting anticholinergics are believed to suppress parasympathetic control, whereas LABAs are believed to stimulate sympathetic control, both causing an increased risk of tachyarrhythmias, myocardial ischemia, stroke, and death.^7,8 Meta-analyses and observational studies suggest that both medicines increase the risk of cardiovascular events,^8-11 whereas large RCTs suggest they do not.^4,5 However, because both types of studies have limitations¹² (eg, limited generalizability and inclusion of patients already using the study medications in RCTs vs lack of randomization in observational studies), the issue remains unresolved. Two industry-sponsored observational studies found trends toward but no statistically significant increases in cardiovascular events in LAAs compared with LABAs; however, these studies may have been underpowered and included subjects without COPD.^13,14 In addition, record keeping of potential confounders was incomplete in 1 study.¹³ Our group also examined the comparative cardiovascular effects of these drugs; cardiovascular disease was part of a composite outcome, and when it was added to other parts of that outcome, it reduced the association observed, thus also suggesting no difference in cardiovascular safety.¹⁵ Nonetheless, patients with COPD and their physicians are faced with choosing between LABAs and LAAs.⁶ Given their similar effectiveness, knowing which medicine has the least risk is important to making informed decisions.^15,16 We conducted the present study to clarify the relative cardiovascular risk of LABA compared with LAA in older patients with COPD.

We conducted a population-based, nested case-control study using health administrative data from Ontario, a province of Canada with a diverse, multicultural population of approximately 13 million, including about 1.8 million people 66 years or older, from September 1, 2003 (when LAAs were first commonly used) through March 31, 2009. Older patients constituted the study population because they carry the highest burden of COPD and because population-wide medication records were only available for those 66 years or older. We obtained ethics committee approval from Sunnybrook Health Sciences Centre.

Residents of Ontario have universal public health insurance under the Ontario Health Insurance Plan, the single payer for all medically necessary services across providers and hospitals. Service details are captured in health administrative databases that can be linked on an individual level to provide a complete health services profile for each resident. The Registered Persons Database contains basic demographic information and date of death for all insured individuals. The Ontario Drug Benefit Program database contains prescription medication dispensation records for all residents 65 years or older. The Canadian Institute of Health Information Discharge Abstract Database contains information on all discharges from acute care hospitals. The Canadian Institute of Health Information National Ambulatory Care Reporting System contains information on all emergency department and same-day surgery visits. The Ontario Health Insurance Plan physician claims database contains information about all services provided by fee-for-service physicians and “shadow-billings” for physicians compensated under alternate payment plans.

We also used the Ontario COPD database, a validated registry of all Ontario residents with COPD generated using the Ontario Health Insurance Plan and Canadian Institute of Health Information databases described above. Individuals with COPD are identified using a previously validated COPD case definition, described in detail elsewhere and used in previous studies,^1,15,17 which yielded a positive predictive value of 81.3% in adults 35 years or older and 86.3% in adults 65 years or older¹⁷ (A.G., Jun Guan, MSc, T.T., unpublished data, April 2008) when compared with physician diagnosis.

All patients in the Ontario COPD database 66 years or older from September 1, 2003, through March 31, 2009, were entered into the study once they received at least 1 prescription for a short- or long-acting β-agonist or anticholinergic, an inhaled corticosteroid, or a methylxanthine. Patients not using these medications, who would have been more likely to have milder disease and fall into the control group, were excluded. Subjects who were not eligible for public health insurance for at least 5 years before their index date were also excluded to ensure complete data during the retrospective period.

The primary outcome was a hospital or an emergency department visit for a cardiovascular event, defined as an acute coronary syndrome (including acute myocardial infarction), heart failure, ischemic stroke, or cardiac arrhythmia.¹⁸ The index date was the date of the first admission or the first visit after study entry. Secondary outcomes were first hospitalization or emergency department visit for each of these 4 cardiovascular events separately.

All subjects who had a cardiovascular event were cases. Their index date was the first date after study entry that they had a hospitalization or an emergency department visit for a cardiovascular event. Potential control subjects for each case were all patients with COPD who were still at risk for a cardiovascular event on the case’s index date in that they were alive, living in the province, and free of cardiovascular events (cases were eligible to be controls before their event date). For each case, 1 control was randomly selected, matched for age (±1 year), sex, duration of COPD, heart failure, and history of hospitalization for acute coronary syndrome, heart failure, ischemic stroke, cardiac arrhythmia, and acute respiratory disease (ie, acute exacerbation of COPD, pneumonia, influenza, or acute bronchitis). Controls were assigned the index dates of their respective cases.

New use of an LABA or an LAA (delivered as dry tiotropium inhalation powder [Spiriva Handihaler; Boehringer Ingelheim, Inc]) constituted the primary exposure. This exposure was defined as receiving a prescription within 90 days of the index date and not having received any prescriptions for the same medication in the previous year. Current medication use was not examined to avoid studying patients less likely to encounter adverse events, as evidenced by the fact that they had continued to fill prescriptions and presumably tolerated the medication.

Information on potential confounding variables was derived from the health administrative data. Socioeconomic status was obtained from neighborhood income, based on subjects’ postal codes.¹⁹ Comorbidity was based on health records in the 2 years before the index date and characterized using The Johns Hopkins University Adjusted Clinical Group Case-Mix System.²⁰ Severity of COPD and cardiovascular history and risk factors were determined using information on health services and prescription drug use.

Baseline characteristics of cases and controls were compared using standardized differences. Differences of less than 10% were accepted as adequate balance.^21-23 We also used unpaired 2-tailed t tests for normally distributed variables, Wilcoxon rank sum tests for ordinal variables, and χ² tests for categorical variables.

The association of cardiovascular hospital or emergency department visits and new use of each LABA and LAA was evaluated using conditional logistic regression, taking matching into account and adjusting for all baseline characteristics.²⁴ The reference group consisted of nonusers of either medication in the previous year. We contrasted the cardiovascular risk associated with new LABA and LAA use. All statistical tests were 2-sided, with statistical significance defined as P < .05. Analyses were performed using commercially available software (SAS, version 9.2; SAS Institute, Inc).

A number of additional analyses were conducted to further explore and evaluate the robustness of the results. To see whether results differed by sex, preexisting cardiovascular disease or COPD severity (as indicated by a previous hospitalization for acute respiratory disease), we tested the significance of interaction terms between each of these factors and new LABA and LAA use in the regression models. To explore the relative contribution of each of the 4 cardiovascular events in the primary outcome, we conducted analyses of each separately.

Because we used health administrative data rather than direct clinical evaluation to identify individuals with COPD, we conducted 2 sensitivity analyses to determine how potential misclassification might have affected the results. First, we examined whether having spirometry (the clinical reference standard test for COPD) influenced the result. Second, we conducted a sensitivity analysis to assess the potential effects of misclassification. We assumed that all misclassified subjects (13.7% as per the positive predictive value of the validation study¹⁷) were in the same medication group and had asthma, because this condition is most likely to be mistaken for COPD and is associated with a lower cardiovascular risk than COPD. We then recalculated the odds ratio (OR) of cardiovascular events associated with new LABA compared with new LAA use under a range of assumptions about the relative cardiovascular risk of asthma compared with COPD. To determine whether confounding could explain the findings, we used an array approach to determine the OR and the balance that a potential unmeasured confounder needed to have to significantly alter the OR of cardiovascular events associated with new LABA compared with new LAA use.²⁵

Almost two hundred thousand (N = 191 005) subjects 66 years or older and undergoing treatment for COPD were eligible for analysis, of whom 53 532 (28.0%) had a cardiovascular event. Of those, 26 628 (49.7%) were matched to a control. As expected, cases were more likely than controls to have a history of diabetes mellitus and hypertension, to be taking some medications, and to have had a recent hospitalization and/or emergency department visit. Baseline characteristics of cases and controls are described in Table 1. Also as expected, matched cases had lower and matched controls had higher rates of previous cardiovascular disease and cardiac risk factors (Table 2) than unmatched cases and controls, respectively (data not shown). Characteristics of COPD by case or control group and medication use are described in Table 3 and Table 4, respectively.

Compared with people not taking either medication, new users of LABAs and LAAs were more likely to have a hospitalization or an emergency department visit for a cardiovascular event (adjusted OR for LABAs, 1.31 [95% CI, 1.12-1.52; P < .001]; for LAAs, 1.14 [1.01-1.28; P = .03]). We found no evidence that new users of LABAs were more likely to have cardiovascular events than new users of LAAs (adjusted OR, 1.15 [95% CI, 0.95-1.38; P = .16]) (Table 5). Cardiovascular events in cases newly taking LABAs or LAAs appeared to be highest in the first 2 to 3 weeks after the medication was prescribed (Figure). Cardiovascular risk associated with new use of LAA relative to LABA did not appear to differ by sex, preexisting cardiovascular disease, or COPD severity.

Compared with nonuse of either medication, LABA and LAA use were associated with an increased risk of acute coronary syndrome and heart failure but not cardiac arrhythmias or stroke. A protective effect for ischemic strokes was observed with new LAA but not new LABA use, resulting in a 73% increased risk of stroke in new users of LABAs compared with new users of LAAs (Table 6).

Having or not having prior spirometry did not significantly alter the absolute or relative cardiovascular risk associated with LABA or LAA use (P = .42 for the interaction between medication use and prior spirometry). If all misclassified subjects were found in the LAA group and had asthma and a risk of cardiovascular events that was zero, the OR of cardiovascular events associated with new LABA compared with new LAA use would still not be statistically significant (ie, so that its 95% confidence interval did not include 1.00). Alternatively, if all misclassified subjects were found in the LABA group and had asthma, their risk of cardiovascular events would have to be less than 0.83 the risk of those with true COPD to obtain statistically significant results (Table 7). A theoretical unmeasured confounder, uncorrelated with any of the variables included in the analyses and affecting only 1 of the 2 medication groups, would have to have a prevalence of at least 20% or a relative risk ratio of at least 0.75 (if found only among LABA users) or 1.22 (if found only among LAA users) to render the observed OR in new LABA users relative to new LAA users significant (Table 8).

We conducted an observational, population-based study to compare the cardiovascular risk associated with LAA and LABA use in older individuals with COPD and found increased risks of hospitalization and emergency department visits for cardiovascular disease in new users of both medications but no evidence of differing risk between medications. This lack of a difference was independent of sex, preexisting cardiovascular disease, and COPD severity. To the best of our knowledge, this study is the first to compare the cardiovascular safety of LABAs and LAAs directly in a large, inclusive, real-world population of older patients with COPD.

Our results agree with those of previous observational studies that found no significant differences in cardiovascular risk associated with LABA and LAA use in patients seen in primary care practices in the United Kingdom.^13,14 The results are also consistent with an RCT that demonstrated comparable cardiovascular event rates in individuals with COPD taking LABAs and LAAs in its secondary safety analysis.¹⁶ Finally, the results are consistent with our own previous study¹⁵ in which the comparative association of LABAs and LAAs on the composite outcome of death, hospitalization, or an emergency department visit for COPD or a related respiratory disease was decreased when a hospitalization or an emergency department visit for cardiovascular disease was added to the outcome.

Our study also found significantly increased cardiovascular risk associated with the new use of LABAs and the new use of LAAs relative to nonuse of either drug. Higher cardiovascular risk associated with LAA has been demonstrated in 2 previous meta-analyses.^9,10 In addition, the increased risk associated with short-acting anticholinergics may also apply to the long-acting forms.²⁶ Contrary to our study, Lee et al²⁷ found no increase in cardiovascular risk associated with LABA use in their study of patients with COPD. However, differences in the setting, the population studied, and the reference group may account for their different results. Most notably, their study excluded patients with a codiagnosis of asthma, and their reference group contained a higher proportion of patients taking short-acting β-agonists. We are not aware of other large-scale studies that have specifically evaluated the cardiovascular safety of LABAs in individuals with COPD, although increased LABA-associated mortality has been found in individuals with asthma.²⁷ Evaluation of short-acting β-agonist–associated cardiovascular safety in individuals with COPD is also lacking, although studies have suggested an association in patients with obstructive airways disease in general.^8,11,28 Our results, however, contradict previous RCTs that did not find increased risk with use of these medications.^4,5,29 This difference may represent several reasons. First, previous RCTs excluded patients with severe disease and/or comorbidity who commonly use these medications. Second, previous RCTs used volunteers, who were unlikely to participate in a study requiring them to take a medication to which they had previously had an adverse reaction. Third, previous RCTs did not reflect real-world clinical settings; participants likely had better health care access than real-world patients, allowing early adverse events to be addressed before they became acute events. Fourth, previous RCTs were not powered to detect adverse outcomes. Last, in previous RCTs, information on adverse events was not available after the intervention ended—often when patients withdrew from the study—and therefore some events may have been missed.

Our secondary analyses suggest that LABAs and LAAs were associated with similar risks of acute coronary syndrome, congestive heart failure, and arrhythmia but not stroke, for which LAAs were found to be protective. After a 2008 US Food and Drug Administration warning of a possible association between stroke and tiotropium, physicians may have withheld LAA therapy from individuals at higher risk of stroke.³⁰

Our study has limitations that need to be recognized. First, we cannot be sure that confounding did not occur. Without clinical detail, we were unable to rule out protopathic bias, or symptoms of presumed worsening COPD (prompting a step-up in medication therapy) being mistaken for cardiovascular instability. However, we believe that this limitation is unlikely to have biased our main outcome because LABAs and LAAs are recommended interchangeably, because cardiovascular history is unlikely to influence whether an LABA or an LAA is prescribed and because our results did not differ in subgroup analysis of individuals without cardiovascular disease. Using very conservative assumptions that were unlikely to occur, failing to account for a potential unmeasured confounder that affects only people taking LABAs or LAAs (but not both) and is uncorrelated with any of the other variables could plausibly mask an increased cardiovascular risk of LABA compared with LAA, although the overall association would be weak at best.

A second important limitation is the possible misclassification of COPD. The results, however, were not influenced by previous receipt of spirometry, suggesting that misclassification did not play an important role. Without data on the cardiovascular risks experienced by individuals with asthma, we cannot be certain of the true impact of misclassification, but the extreme scenario in which all misclassified individuals had asthma and were in the LABA group could plausibly mask a true higher risk related to LABA compared with LAA use.

A third important limitation is that of death as a competing risk. Our study did not consider the difference in nonhospital deaths among people newly starting therapy with LAAs and LABAs. However, when we used our data to estimate the number of individuals in each medication group who met our inclusion criteria and died during this short time, we found the difference to be small and to have little influence on the association between LAAs and LABAs (data not shown).

In summary, we found that new users of LABAs and LAAs in a population-based study of older individuals with COPD had similar but increased risk of cardiovascular events. These results support the need for close monitoring of all patients with COPD who require long-acting bronchodilators regardless of drug class.

Corresponding Author: Andrea Gershon, MD, MS, Institute for Clinical Evaluative Sciences, G1 06, 2075 Bayview Ave, Toronto, ON M4N 3M5, Canada (andrea.gershon@ices.on.ca).

Accepted for Publication: January 31, 2013.

Published Online: May 20, 2013. doi: 10.1001/jamainternmed.2013.1016

Author Contributions: Dr Gershon and Ms Croxford had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Gershon, Croxford, Stanbrook, Upshur, and Stukel.

Acquisition of data: Gershon and Calzavara.

Analysis and interpretation of data: All authors.

Drafting of the manuscript: Gershon, Croxford, and Upshur.

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

Statistical analysis: Croxford, Calzavara, Stanbrook, and Stukel.

Obtained funding: Gershon.

Administrative, technical, and material support: Gershon and Upshur.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by the government of Ontario; by the Institute for Clinical Evaluative Sciences (ICES), which is funded by an annual grant from the Ontario Ministry of Health and Long-term Care; by a Career Scientist Award from the Ontario Ministry of Health and Long-term Care (Dr Gershon); by a New Investigator Award funded by team grant OTG-88591 from the Canadian Institutes of Health Research Institute of Nutrition, Metabolism, and Diabetes (Dr Gershon); and by The Dales Award in Medical Research from the University of Toronto (Dr To). Brogan Inc, Ottawa, permitted use of their Drug Product and Therapeutic Class Database.

Role of the Sponsors: The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.

Disclaimer: The opinions, results, and conclusions reported in this study are those of the authors and are independent from the funding sources. No endorsement by the ICES or the Ontario Ministry of Health and Long-term Care is intended or should be inferred.

Previous Presentation: This study was presented at the 2011 American Thoracic Society International Conference; May 16, 2011; Denver, Colorado.