Risk of myocardial infarction (MI) and death following MI in people with chronic obstructive pulmonary disease (COPD): a systematic review and meta-analysis

Objectives Cardiovascular disease is an important comorbidity in patients with chronic obstructive pulmonary disease (COPD). We aimed to systematically review the evidence for: (1) risk of myocardial infarction (MI) in people with COPD; (2) risk of MI associated with acute exacerbation of COPD (AECOPD); (3) risk of death after MI in people with COPD. Design Systematic review and meta-analysis. Methods MEDLINE, EMBASE and SCI were searched up to January 2015. Two reviewers screened abstracts and full text records, extracted data and assessed studies for risk of bias. We used the generic inverse variance method to pool effect estimates, where possible. Evidence was synthesised in a narrative review where meta-analysis was not possible. Results Searches yielded 8362 records, and 24 observational studies were included. Meta-analysis showed increased risk of MI associated with COPD (HR 1.72, 95% CI 1.22 to 2.42) for cohort analyses, but not in case–control studies: OR 1.18 (0.80 to 1.76). Both included studies that investigated the risk of MI associated with AECOPD found an increased risk of MI after AECOPD (incidence rate ratios, IRR 2.27, 1.10 to 4.70, and IRR 13.04, 1.71 to 99.7). Meta-analysis showed weak evidence for increased risk of death for patients with COPD in hospital after MI (OR 1.13, 0.97 to 1.31). However, meta-analysis showed an increased risk of death after MI for patients with COPD during follow-up (HR 1.26, 1.13 to 1.40). Conclusions There is good evidence that COPD is associated with increased risk of MI; however, it is unclear to what extent this association is due to smoking status. There is some evidence that the risk of MI is higher during AECOPD than stable periods. There is poor evidence that COPD is associated with increased in hospital mortality after an MI, and good evidence that longer term mortality is higher for patients with COPD after an MI.

INTRODUCTION Cardiovascular disease is a common comorbidity and cause of death in people with chronic obstructive pulmonary disease (COPD), with up to one-third dying of cardiovascular disease. 1 Reducing the cardiovascular disease in this population is an important strategy for reducing the burden of COPD.
Several studies have shown that people with COPD have a higher risk of myocardial infarction (MI) than people without COPD. [2][3][4] One of the reasons for the increased risk of MI in patients with COPD is the shared major risk factor of smoking. In addition, several other cardiovascular risk Strengths and limitations of this study ▪ This systematic review investigated three important areas relating to the relationship between chronic obstructive pulmonary disease (COPD) and cardiovascular disease: (1) the risk of myocardial infarction (MI) associated with COPD; (2) the risk of MI associated with acute exacerbations of COPD; and (3) the risk of death following MI in patients with COPD compared to patient without COPD. ▪ Strengths of this review were the wide search strategy, broad inclusion criteria and rigorous risk of bias assessment of included studies. ▪ We found strong evidence for an increased risk of MI in people with COPD and an increased risk of longer term death after MI for patients with COPD; however, it is unclear how much of this increased risk may be due to smoking status. ▪ We found poorer evidence for an increased risk of MI during periods of acute exacerbation of COPD compared to stable periods, and for an increased risk of death in hospital after MI for patients with COPD. We make recommendations on how future studies can improve our understanding of these relationships. ▪ Due to statistical and clinical heterogeneity, meta-analysis could only be conducted for some of the research questions.
factors, including hypertension, diabetes, inactivity, poor diet, and older age, are also prevalent in patients with COPD. [5][6][7] In addition, several studies have found an association between reduced FEV 1 (forced expiratory volume1 s) and cardiovascular mortality in the general population. 8 However, COPD itself is also thought to be an independent risk factor for MI with increased risk of MI possibly being mediated through increased systemic inflammation or reduced FEV 1 in people with COPD. Acute exacerbations of COPD are events in the natural history of COPD which are characterised by an increase in COPD symptoms such as breathlessness, cough, sputum volume, and sputum purulence. It has recently been suggested that acute exacerbations of COPD (AECOPD) represent a period of increased risk of MI for people with COPD. 9 A subtype of patients with COPD appears to have more frequent exacerbations than others. Frequent exacerbators have been defined as individuals who have two or more treated exacerbations per year. Frequent exacerbators may be at higher risk of MI compared to infrequent exacerbators, even during stable periods.
Several investigators have found that patients with COPD have worse mortality in hospital and following discharge after an MI compared to patient without COPD. [10][11][12] However, the finding that patients with COPD have greater in hospital and short-term mortality has not been found by all investigators. [13][14][15] We aimed to systematically review the literature reporting on: (1) the risk of MI in people with COPD; (2) the risk of MI associated AECOPD, either during AECOPD or that associated with the frequent exacerbator phenotype; and (3) the risk of death after MI in people with COPD. These questions represent the most salient aspects of current research into the relationship between COPD and cardiovascular disease, and no systematic reviews have been published on these topics to date.

Literature search MEDLINE, MEDLINE In-Process & Other Non-Indexed
Citations, EMBASE, BIOSIS & Science Citation Index were searched up to January 2015. A search strategy was devised which would pick up articles relevant to all three research questions. All strategies were based on the MEDLINE search strategy, which is presented in the online supplementary material. In brief, the literature was searched for terms which relate to COPD and terms with relate to MI, and these searches were combined using the AND Boolean logic operator. MeSH terms were combined with natural language searching using truncation where appropriate.

Inclusion and exclusion criteria
Inclusion and exclusion criteria were applied for each of the three research questions. Studies were included if they met the population, exposure, comparator and outcome criteria. These are presented below for each research question. Studies were included from database start date and were not restricted by language.

Risk of MI in people with COPD
The population of interest was the general population. The exposure of interest was diagnosis of COPD. The un-exposed group was people without a diagnosis of COPD. The outcome of interest was acute MI.

Risk of MI associated with AECOPD
The population of interest was people with a diagnosis of COPD. The exposures of interest were either: (1) discrete episodes of AECOPD or periods within 8 weeks of an AECOPD; or (2) frequent exacerbator phenotype. The comparators of interest were either: (1) periods of stable COPD; or (2) infrequent exacerbator phenotype. Studies were included if these reported a relative risk of MI, or if this could be calculated.

Risk of death after MI in people with COPD
The population of interest was those presenting to a hospital with an MI. Studies were included if these compared patients with a diagnosis of COPD to those without a diagnosis of COPD. Outcomes of interest were death in hospital and at any reported time points post discharge. Studies investigating risk of death for patients with COPD after an interventional procedure following an MI (such as percutaneous coronary intervention or coronary artery bypass graft) were specifically excluded under the population criterion.

Selection of included studies
Titles and abstracts, where available, were initially screened for potential inclusion by one reviewer. Full text versions of potentially included studies were then obtained and were screened by two reviewers. Authors were contacted if the information provided in articles was not sufficient to assess whether inclusion criteria were satisfied.

Risk of bias assessment
All included studies, except for those only reported as conference abstracts, were assessed for risk of bias. The risk of bias tool was informed by the Newcastle-Ottawa scale; 16 however, we did not make use of a summary score as this is not advisable. 17 18 Risk of bias was assessed across the key domains related to selection of participants, comparability of groups, and measurement of outcomes. Several items were included under each domain, and were adapted for different study types. Where reports of studies included more than one analysis (eg, a case control as well as a cohort analysis) the risk of bias for these analyses were conducted separately. Risk of bias assessment was completed by one reviewer and checked by another.

Evidence synthesis
Characteristics and findings of studies were tabulated and compared. Data on severity of COPD was extracted as GOLD stage or FEV 1 % predicted, where available. Information was also extracted on smoking status and previous cardiovascular disease. Estimates of effect were extracted or calculated, and are presented as ORs, risk ratios (RR), incidence rate ratios (IRR) or HRs.
Where included studies were reasonably statistically and clinically similar, we pooled results using random effects meta-analysis. We used the generic inverse variance method to pool maximally adjusted effect estimates. Analysis was conducted in Review Manager 5.3. Where studies were too statistically (I 2 over 75%) or clinically heterogeneous, meta-analysis was not conducted, but study summary results were graphed on forest plots without pooling the results. Studies which were not adjusted at all were not included in forest plots. For the question on risk of MI associated with COPD, studies were stratified by adjustment for smoking status (yes or no) and study design (cohort or case-control). For the question on risk of death following MI in COPD compared to patient without COPD, studies were stratified by outcome time point (in-hospital mortality or follow-up mortality). For follow-up mortality, studies were further stratified by analysis.

Identified studies
Literature searches yielded 8362 records. After title and abstract screening, 49 records were selected for full-text assessment, which resulted in the inclusion of 24 studies. The inclusion and exclusion process is summarised in figure 1. Of the 24 included studies, 9 investigated the risk of MI in patients with COPD compared to patient without COPD; 2 investigated the risk of MI associated with AECOPD; no studies were found which investigated the risk of MI associated with the frequent exacerbator phenotype; and 12 investigated outcomes after MI for patients with COPD compared to patient without COPD. Summary characteristics of included studies are presented in tables 1-3.
All of the included studies which investigated risk of MI in people with COPD used data from either routine clinical or administrative databases. COPD was defined using diagnostic codes; these varied between COPD diagnosis in primary care, outpatient departments, hospital admission or discharge codes, and cause of death codes. Three studies also required that patients with COPD had been prescribed COPD medicines. One of the studies, Rodriguez et al, 19 included only patients with a recent diagnosis of COPD and followed up to 5 years after this diagnosis to identify MI. Only one study 3 reported a summary of COPD severity, and only two reported prevalence of current smokers. Four studies reported a cohort analysis only. Two studies 4 19 reported a cohort analysis as well as a case-control analysis. One study reported the results of a cohort analysis and an analysis of period prevalence. One study 20 compared rates of MI in patients with COPD to standardised populations rates of MI.
Two studies 9 21 were identified which investigated the risk of MI associated with AECOPD. Both studies defined risk periods after the onset of AECOPD and used within person designs to compare the risk to a baseline period.
Nine studies reported mortality for patients with COPD after an MI compared to patient without COPD. Five studies 11 12 14 15 22 reported a comparison of in-hospital mortality after an MI between patients with COPD and patient without COPD. Eight studies 10 12 13 23-27 used a time-to-event analysis to investigate death after discharge from a hospital admission for MI.

Risk of bias assessment
The proportion of studies (or analyses, where appropriate) which were assessed as having either lower, unclear or higher risk of bias for each of the research questions is presented in figure 2. Detailed results from the risk of bias assessment for individual studies are presented in the appendix.

Risk of MI in people with COPD
Of nine included studies, eight found a higher risk of MI in patients with COPD compared to patient without COPD. Six studies estimated the ratio of incidence rates of MI in patients with COPD compared to patient without COPD. Five studies 4 19 20 28 29 estimated this for all MIs, this ranged from IRR 1.18 (95% CI 0.81 to 1.71) to 5.31 (4.54 to 6.21). One study 2 30 estimated the IRR for hospitalisation due to MI (IRR 1.49, 95% CI 0.71 to 3.13) and fatal MIs (1.51, 1.14 to 2.01). Two studies 31 32 estimated the ratio of hazard of MI in patients with COPD compared to patient without COPD one study   4). Meta-analysis was not conducted for the studies which did not adjust for smoking as heterogeneity was too high (I 2 =93%). These results are graphically summarised in figure 5. Some studies investigated whether the effect of COPD on the risk of MI was different in terms of age and severity of airflow obstruction. Feary et al 3  Risk of MI associated with AECOPD Donaldson 2010 9 conducted a self-controlled case series using data from The Health Improvement Network (THIN). They used prescription of antibiotics and steroids in patient with COPD to identify AECOPD, and report an increased risk of MI in the 1-5 days following the onset of AECOPD (IRR 2.27, 95% CI 1.10 to 4.70). No difference in the risk of MI was found for the period 6-49 days, or at any time point when the alternative definitions of AECOPD of prescription of steroids alone or antibiotics alone were used. Halpin et al 21 reported a secondary analysis of the UPLIFT trial, which was an RCT comparing inhaled tiotropium and placebo in patients with COPD with a primary outcome of reduction in FEV 1 decline. Time to first AECOPD was a secondary outcome. AECOPD were identified using a symptombased definition and were reported to trial staff at AECOPD, acute exacerbation of COPD; COPD, chronic obstructive pulmonary disease; CVD, cardiovascular disease; GOLD, Global Initiative for Chronic Obstructive Lung Disease; FEV1, forced expiratory volume in 1 s; FVC, forced vital capacity; IRR, incidence rate ratios; MI, myocardial infarction; RCT, randomised control trial.  Age, serum creatinine, systolic blood pressure, troponin elevation, heart failure or cardiogenic shock at presentation, ST-segment changes, heart rate and prior peripheral arterial disease.   . These results are graphically summarised in figure 6. Owing to different exposure time periods, the results for within person studies investigating the risk of MI associated with AECOPD were not pooled in meta-analysis.

Risk of death after MI in people with COPD
Of the studies investigating differences in in-hospital mortality after an MI, two 12    did not find an increased in-hospital mortality for patients with COPD after a STEMI (OR 1.05, 95% CI 0.95 to 1.17), but did find increased in-hospital death for patients with COPD after a non-STEMI (OR 1.21, 95% CI 1.11 to 1.33). Meta-analysis of adjusted results showed weak evidence for an increased risk of in-hospital death for patients with COPD (OR 1.13, 95% CI 0.97 to 1.31) (figure 7). One study 14 reported mortality at 30 days for patients with COPD compared to patient without COPD. This study found increased mortality for patients with COPD (OR 1.31, 1.10 to 1.58). Another study 33 reported mortality at 8 months, and in an unadjusted analysis, found increased mortality for patients with COPD compared to patients without COPD (OR 2.69, 95% CI was not reported and could not be calculated). One study 22 also found, on unadjusted analysis, that mortality was greater for patients with COPD at 1 year (RR 1.34, 95% CI 1.16 to 1.55) and 5 years (RR 1.28, 95% CI 1.18 to 1.40) after MI.
Eight studies 10 12 13 23-27 reported results of survival analysis of mortality during follow-up after an MI. All of the studies reported higher mortality for patients with COPD compared to patients without COPD during follow-up after discharge following an MI. HRs ranged from 1.15 (95% CI 1.04 to 1.28) to 2.15 (95% CI 1.30 to 3.55). However, one of these studies 13 found no evidence of a difference in mortality when restricting the time period to the first 30 days following discharge (HR 0.89, 95% CI 0.68 to 1.11). Meta-analysis of studies which reported adjusted results showed an increased risk of death after discharge following MI for patients with COPD compared to patients without COPD (HR 1.26, 1.13 to 1.40) (figure 8). Four of the studies included under this question were excluded from meta-analysis for methodological 12 33 or clinical heterogeneity. 25 27

Main findings
Most studies which investigated the risk of MI in people with COPD found that those with COPD have higher risk of MI than people who do not have COPD; however,   it is unclear how much of this increased risk is due to smoking status. The included cohort study which adjusted for smoking status showed an increased risk of MI in people with COPD, but this was not apparent in pooled analysis of the case-control studies which adjusted for smoking status. Both of the included studies which investigated the risk of MI associated with AECOPD found an increased risk of MI in the weeks following AECOPD. Most studies which investigated mortality after an MI for patients with COPD as compared to patients without COPD found that mortality after discharge was greater for those with COPD, and an increased risk of death was found on pooled analysis. However, findings on in-hospital mortality after an MI were mixed, and there was only weak evidence for increased risk of death in hospital for patients with COPD on pooled analysis.

Limitations of included studies and future work
One common limitation among the included studies, particularly those which investigated the risk of MI associated with COPD, was missing information on smoking status. As smoking is very strongly associated with COPD and risk of MI, it is likely to be a major confounder in all studies investigating this association. All of the studies in this review which investigated this association used either clinical or administrative routine data sources. Routine data are a potentially rich source of information about huge numbers of patients. However, data on smoking are not routinely recorded in all administrative databases. Indeed, all of those studies which did not have data for smoking in this question used administrative databases. Future studies on the association between COPD and cardiovascular disease should use data sources which contain reliable information on smoking status.
Further studies should be carried out to confirm findings that AECOPD are periods of increased risk of MI for people with COPD. These studies should ensure they use validated exposure measures and are adequately powered. Possible reasons for an increased risk of MI during AECOPD include increased inflammation and the potential cardiovascular effects of the drugs used to treat AECOPD. If indeed the finding of increased risk during AECOPD is confirmed, future studies should attempt to disentangle the reasons for increased risk of MI. In addition, studies should investigate factors which might modify this relationship, such as drugs used for treatment of COPD and cardiovascular prevention. Another potential bias in studies which investigate the relationship between AECOPD and MI, which could explain some of the increased risk of MI after AECOPD, is differential misclassification of episodes of angina as AECOPD.
No studies were found which investigated the risk of MI associated with the frequent exacerbator phenotype. The frequent exacerbator phenotype may prove to be a useful characteristic for stratifying cardiovascular risk among patients with COPD. Future cohort studies of cardiovascular disease in people with COPD should, where possible, phenotype participants and investigate the relationship between exacerbator phenotype and risk of MI. Few included studies assessed the influence of severity of COPD on risk of MI; further research should investigate Figure 8 Forest plot showing risk of death after discharge following MI for patients with COPD compared to patients without COPD. Cis may vary slightly from those quoted in tables due to transformation during meta-analysis. COPD, chronic obstructive pulmonary disease; MI, myocardial infarction. Figure 7 Forest plot showing risk of in-hospital death following MI for patients with COPD compared to patients without COPD. CIs may vary slightly from those quoted in tables due to transformation during meta-analysis. COPD, chronic obstructive pulmonary disease; MI, myocardial infarction. this relationship as well as the influence of severity of COPD on risk of death following MI.
A further limitation of several of the included studies on death following MI was availability of information on cause of death. Collection of information on cause of death in future studies would allow investigators to draw more confident conclusions about the reasons for increased risk of death following MI for people with COPD.
Strengths and limitations of this review This review benefitted from using a comprehensive search strategy which covered several bibliographic databases. As the relationship between AECOPD and MI has not been extensively studied, the inclusion criteria for this research question were kept purposively broad. This allowed all information pertaining to this relationship to be included in the evidence synthesis. One potential limitation of systematic reviews is publication bias. The potential for publication bias was highest for the review of outcomes after MI. In order to reduce the risk of this bias, we only included studies which specifically investigated the risk of COPD on MI rather than several different potential prognostic factors, as studies which investigated several factors which did not find an association between COPD and MI may not have reported this in the abstract or even in the text. Owing to clinical and statistical heterogeneity, meta-analysis could only be conducted for some of the research questions. Where meta-analysis was conducted, statistical heterogeneity was generally high, and this may limit the generalisability of pooled estimates.

CONCLUSIONS
There is good evidence of an increased risk of MI in people with COPD; however, it is unclear to what extent this association is due to smoking status.
There is some evidence that among people with COPD, AECOPD represent periods of increased risk of MI. However, further larger studies using validated exposure methods are needed to support this finding.
There is weak evidence that in-hospital mortality is higher for people with COPD after an MI. There is good evidence that postdischarge mortality after an MI is higher for people with COPD.