METHODS

Subjects and source of data

The cohort of patients with COPD was identified using the computerised databases of Saskatchewan Health. These databases were developed to aid in the administration of a universal publicly funded health insurance programme. Over 99% of Saskatchewan residents (approximately 1 million) are covered by this programme and, of these, about 91% are also eligible for outpatient prescription drug benefits. Insured medical, hospital, and outpatient prescription drug services and vital status are recorded for all beneficiaries and may be linked for each individual using a unique health services number. The databases have been used extensively to study the effects of several prescription drugs at the population level.⁸

The cohort was formed from all subjects newly treated for a likely diagnosis of COPD. Subjects were included if they had been dispensed at least three prescriptions for a bronchodilator (β₂ agonist, inhaled anticholinergic, or theophylline) within a 1 year period between 1 January 1980 and 31 December 1997. These three prescriptions had to be dispensed on at least two different dates within the year. The date of the third prescription was taken as the date of cohort entry. Only subjects who were at least 55 years old at cohort entry were included. To confirm the incident nature of the treatment and to exclude asthma patients, subjects were excluded if they had been prescribed any bronchodilators or other asthma drugs including inhaled corticosteroids, nasal corticosteroids, cromones (cromoglycate and nedocromil), and ketotifen during the 5 year period before cohort entry. All subjects were followed until the date of death, the date of emigration from the province, the date of end of coverage of the health insurance plan, or 31 December 1999, whichever occurred first.

Outcome

All cases of acute myocardial infarction that occurred during the cohort follow up were ascertained in two ways: a primary discharge diagnosis of MI (ICD-9 code 410) on the hospital services file or an underlying cause of death of MI (ICD-9 code 410) on the vital statistics file. For subjects with multiple events, the first occurrence of acute myocardial infarction was taken as the case event. Cases occurring during or within a year of the period July 1987–December 1988 (18 months) were excluded because capture of outpatient prescription drug data was incomplete during this 18 month period.

Study design

A nested case-control design was used to permit analysis of the complex time varying nature of the exposure.⁹ Each case was matched to all controls available within the cohort on the basis of three factors: (1) all matched controls for a case had a duration of follow up at least as long as the time to acute MI for the corresponding case (this was achieved by assuring that each control was in the cohort and therefore at risk for the outcome at the date of the acute MI in the case, a date designated as the index date); (2) to control for secular trends in inhaled β agonist use, the matched controls had the same calendar year of cohort entry as the case; (3) the controls were matched to the cases for age within a year of the date of birth.

Exposure to β agonists

All prescriptions of inhaled, oral, or nebulised β agonists that were dispensed to the cases and controls during the year before their index date were identified, as well as at any time between cohort entry and the index date. Each prescription was analysed to identify the number of canisters of inhaled β agonist dispensed concurrently.

Data analysis

The presence of any use of β agonists during the 1 year period before the index date was determined and the number of canisters dispensed during this period was computed. Current exposure to inhaled β agonists was considered as any prescription dispensed in the 2 month period before the index date. To assess the effect of new use of inhaled β agonists we considered currently exposed subjects who had not received β agonists of any form during the 3–12 months before the index date. In addition, first time use of inhaled β agonists was defined as current use preceded by no β agonists since cohort entry.

Conditional logistic regression for matched case-control data was used to estimate adjusted rate ratios of acute MI associated with the use of inhaled β agonists. To assess the quantity-response effect we first stratified use into: no use, ≤1.5 canisters per month, and >1.5 canisters per month. We also examined the quantity response on a continuous basis per canister dispensed in the year before the index date and estimated the risk per 10 canisters of inhaled β agonists dispensed in the previous year. To control for any residual confounding beyond age we adjusted all rate ratio estimates for the number of prescriptions of other asthma drugs, including theophylline and all corticosteroids, dispensed in the year before the index date. We also adjusted for the presence of cardiovascular risk factors as measured by prescriptions of drugs for the treatment of hypertension, congestive heart failure, angina, arrhythmia, hyperlipidaemia, and diabetes dispensed during the year before the index date, as well as sex.

RESULTS

The cohort included 12 090 subjects (60% male) with a mean age at entry to the cohort of 73 years. A total of 1310 subjects with fatal and non-fatal acute MI were identified during follow up. We eliminated 190 of these because they occurred during the period when complete outpatient prescription drug data were not available and seven because no matched controls could be identified. The remaining 1127 cases (406 fatal MI, 721 non-fatal MI) were matched to 10 766 controls. The characteristics of these cases and their matched controls are shown in table 1. As expected, cases were more likely to be male and were receiving more medications for cardiovascular disease than controls.

Table 2 shows the crude and adjusted rate ratios of acute MI associated with the use of inhaled β agonists during the 1 year period before the index date, as well as for current, new, and first time use. In all instances the rate ratios are close to unity and not significant. However, although not significant, the quantity-response analysis suggests a small increase in the risk of acute MI of 11% (95% CI –1 to 23) for every 10 canisters of inhaled β agonists dispensed during the year before the index date.

Table 3 presents the analyses of current, new, and first time use stratified by the presence of cardiovascular disease or risk factors before the index date. Here again, the rate ratios are close to unity and not significant. In particular, the rate ratio for first time use of inhaled β agonists among patients with concomitant cardiovascular disease or risk factors was 0.99 (95% CI 0.56 to 1.73). Moreover, the rate ratios were practically unchanged among the patients who did not use β blockers.

Table 4 shows that the rate ratios are close to unity and not significant for other forms of β agonists (nebulised and oral) irrespective of the quantity or timing of their use.

DISCUSSION

Beta agonists, including inhaled, nebulised, and oral forms, are not associated with a risk of fatal or non-fatal acute MI in patients with COPD. In addition, first time use of inhaled β agonists by patients with and without pre-existing concomitant cardiovascular disease was found to be unrelated to acute MI.

Our findings contradict the one potentially troubling result of the case-control study by Au et al⁶—namely, that first time use of inhaled β agonists in people with existing cardiovascular disease increases the risk of MI. That study, however, had several shortcomings in design that could have biased their finding and explain their divergent results. Firstly, it included a majority of patients who did not have asthma or COPD, the usual indication for inhaled β agonists. Indeed, the reference group of never users of β agonists probably contained few patients with COPD and thus only a small minority of the 1444 cases and 4094 controls in that study were relevant to the risk estimate. More importantly, however, since COPD is a known risk factor for acute MI, the analysis should have adjusted for this confounder or, as our study was, been restricted to patients with COPD who are users of these drugs. Secondly, the study by Au and colleagues failed to assess the concurrent possible protective effect of β blockers. The antagonist effect of β blockers may neutralise the effects of β agonists, so the risk of MI associated with β agonists may be different in subjects already using β blockers than among non-users of β blockers. Since statistical “adjustments” do not resolve such effect modification, we instead re-estimated the risk strictly among non-users of β blockers and found no increased risk. We also felt it important to account for the use of oral and nebulised forms of β agonists. We consider these to be important potential confounders since they have been shown to be associated with acute cardiac death⁵ and are likely to be prescribed more frequently among subjects prescribed the usual metered dose formulations. Moreover, studying these higher dose formulations is instrumental in assessing dose-response effects of β agonists in general. Our analysis of these formulations found no increased risk.

We believe that the more plausible explanation for the observed increased risk found by Au et al soon after one time use of β agonists is a misdiagnosis, as asthma or COPD, of angina or other respiratory symptoms associated with an imminent MI. That isolated prescription for an inhaled β agonist interpreted as “first time use” was probably prescribed for such symptoms to a patient who did not have asthma or COPD. Indeed, chart reviews revealed the presence of symptoms such as chest pressure, atypical chest pain and pressure, or pleuritic chest pain. We feel that, in view of our unequivocal findings, the observation by Au et al for first time use is probably the result of diagnostic bias and confounding.

We observed a trend towards a small increase in risk of MI of 11% for each 10 canisters dispensed within the previous year for frequent or high dose use of β agonists. The most likely explanation is that subjects with greater utilisation of these medications have more severe COPD with more profound airflow limitation which is itself the explanation for the increase in risk.⁷ Since we did not have access to measures of lung function, it is likely that adjustments made for severity of COPD were insufficient and that residual confounding is the most likely explanation for this small increase in risk of MI with increasing number of metered dose inhalers of β agonists dispensed in the previous year.

Our study has both strengths and limitations. Since the cohort was formed from administrative databases, all that is known is that inhaled β agonists were dispensed on a given date, but not whether they were actually taken. We can expect, however, that most were taken as COPD is a disease that requires significant assistance with bronchodilation. This is probably more accurate for subjects who were dispensed a large number of canisters during the year before the index date. Even these subjects had no increased risk of acute MI. The outcome of acute MI has been validated in the Saskatchewan databases and found to be accurate.¹⁰

In conclusion, we found that short acting β agonists in any form did not increase the risk of acute MI in a large cohort of patients with COPD. First time use, in particular, also did not increase the risk.