Objective To investigate the single and combined effectiveness of commonly prescribed secondary preventive medications (post-acute myocardial infarction (AMI)) in reducing overall all-cause mortality and by gender.
Design Population-based longitudinal cohort study.
Setting Western Australia, Australia.
Participants 9580 individuals aged 65 years to 84 years who were admitted to hospital with their first AMI diagnosis between 1 January 1995 and 1 January 2006.
Main outcome measures Time to death from any cause out to 11 years after first AMI, identified from registry data, was the primary outcome measure. Cardiovascular drugs dispensed within 28 days following hospital discharge were identified as main exposure categories.
Results In total, 975 deaths occurred during 1 year follow-up, culminating to 3247 by 11 years. 1-year risk of death was significantly reduced for all drug combinations, but not for drugs dispensed in isolation. Out to 11 years, only combinations of ‘β-blockers and statins’ (with or without ACE inhibitors/angiotensin II receptor blockers (ACEi/ARB)) provided significant reductions in risk of all-cause mortality. In men, the greatest reduction in risk was associated with being dispensed ‘β-blockers and statins’ (HR 0.46, 95% CI 0.36 to 0.58), whereas women benefited most from being dispensed ‘β-blockers, statins and ACEi/ARBs’ (HR 0.77, 95% CI 0.60 to 0.99).
Conclusions The combination of ‘β-blockers and statins’ (with or without ACEi/ARB) dispensed within 28 days postdischarge was associated with the greatest long-term survival following an AMI. Our observations of significantly reduced mortality risk in men (compared with women) who were dispensed ‘β-blockers and statins’, or ‘β-blockers and ACEi/ARBs’, warrants further investigation.
- Acute myocardial infarction
Statistics from Altmetric.com
Advances in prevention strategies and treatment practices for cardiovascular disease (CVD) have led to a corresponding decrease in CVD-related mortality in most high-income countries,1 including Australia.2 Despite better treatment, fatal CVD (of which ischaemic heart disease (IHD) is the most common form) mostly leads national mortality statistics in these same countries.1 In Australia, acute myocardial infarction (AMI) contributed an estimated 35% (55 676) of IHD hospitalisations and approximately 8.5% of all deaths in 2007-2008.2 The majority of evidence relating to the prevention of recurrent AMI supports the use of aspirin, β-blockers (BBs), statins, ACE inhibitors or angiotensin II-receptor blockers (ACEi/ARB), which is highlighted by their inclusion in guidelines for clinical practice.3 ,4 Use of these medications has a temporal correlation with the decrease in AMI incidence and mortality over recent decades.5
However, the very long-term effectiveness of these combined pharmacological treatments remains unclear and evidence is emerging to suggest that the survival benefits may be less pronounced over the longer term.6–8 Additionally, research continues around which drug combinations and doses optimise the reduction in risk of AMI incidence and subsequent death.9 ,10 This longitudinal population study sought to test the single and combined effectiveness of BBs, statins and ACEi/ARBs in reducing all-cause mortality after AMI in Western Australia.
In Western Australia's population of approximately 2.3 million residents, records for all hospital admissions (since 1970) and deaths (since 1969) have been held by the state government. Moreover, records are kept for federally subsidised medications dispensed to individuals. Using these three databases—the Hospital Morbidity Data Collection, the Mortality Register and the Pharmaceutical Benefits Scheme (PBS) register—individuals were linked using the technique of probabilistic matching. The matching procedures are based on full name and address, phonetic compression algorithms and other identifiers, and have been estimated to be 99.89% accurate.11 This ‘linkage’ allowed identification of all incident AMIs (those with no AMI recorded within 5 years prior to inclusion into the study) within the hospitalised population, in addition to dispensed medications and future mortality associated with these individuals.
This study comprised all patients who were admitted to a public or private hospital in Western Australia during the period 1 January 1995 to 1 January 2006. Of the 215 285 patients aged 65–84 years (inclusive) who were hospitalised for any reason during this period, a first AMI was identified in the primary diagnosis category for 11 247 patients using International Classification of Disease-version 9 (ICD-9) codes 410–411 (before July 1999) or ICD-10 codes I21–I23 (after July 1999), recorded in the Hospital Morbidity Data Collection. These patients had no prior AMI recorded in any of the 21 diagnosis fields within 5 years prior to the date of their first identified AMI during the study period. A lower age restriction of 65 years enabled identification of concession card holders (>90% of Australians) who pay less for their medications on the PBS, including aspirin. Patients with AMI were excluded if they died while in hospital (n=1286) or within 28 days of the date of discharge (n=324) as there would have been inadequate time available to establish the postdischarge pattern of exposure to pharmacotherapy. The exposure period for medication use was defined as up to 28 days following the date of discharge from hospital. Upon discharge from hospital after an AMI, individuals would have received a 5-day supply of cardioprotective medications, a PBS prescription to be dispensed at a community pharmacy, direction to the family physician for the prescription of medications, or in rare circumstances left without a medication prescription. Initiation or continuation of guideline-recommended medications required periodic review by the individual's treating physician and a PBS prescription. A further 57 patients were excluded because their 28-day postdischarge exposure period ended after 1 January 2006. The remaining 9580 individuals comprised the final study cohort.
PBS item codes were used to identify treatments that patients with AMI received during the 28-day exposure period postdischarge for their primary AMI hospitalisation (see online supplementary table S1). The drug classes identified during the exposure window were: BBs (atenolol, metoprolol, oxprenolol, pindolol, carvedilol, bisoprolol), statins (simvastatin, pravastatin, fluvastatin and atorvastatin), ACEis (captopril, fosinopril, enalapril, ramipril, quinapril, lisinopril, trandolapril and perindopril), ARBs (irbesartan, candesartan, telmisartan and eprosartan).
Patients were then grouped into exposure categories according to which combination of BB, statins and ACEi/ARB they received. The eight mutually exclusive exposure categories constructed were ‘no drugs’, ‘BB only’, ‘statins only’, ‘ACEi/ARB only’, ‘BB and statins’, ‘BB and ACE/ARB’, ‘statins and ACE/ARB’, ‘BB and statins and ACE/ARB’. The primary outcome was all-cause mortality.
Covariates in our study were selected a priori and at the time of presentation with incident AMI included: age, sex, history of IHD (ICD-9 410–414 or ICD-10 I20–I25 in any of the remaining 21 diagnostic categories in the dataset within the last 5 years), use of BB, statins or ACEi/ARBs within the past 60 days, coronary artery revascularisation procedures (CARP) (during and within 28 days of discharge or within the last 5 years), heart failure (ICD-9 428–429 or ICD-10 I50) within the last 5 years, Charlson comorbidity score (based on a modified Dartmouth-Manitoba algorithm12) within the last 5 years, and use of aspirin or clopidogrel within 28 days postdischarge.
Cox proportional hazards regression models were used to estimate HRs with 95% CIs for all-cause mortality, with the ‘no drug’ group as the reference. Crude and adjusted models were performed for two separate end points—(i) 1 year following discharge date plus 28 days and, (ii) until 1 January 2006 (mean survival time of 4 years). The proportional hazards assumption that the ratio of mortality rates according to the exposure variable remained constant for the adjusted models was tested by inclusion of an interaction term with ln(1+survival time) in the model. This identified a violation of the proportional hazards where follow-up was observed until 2006. We further tested this violation by plotting HR estimates over time (including the interaction term) and found no clinically significant differences in estimates. We therefore presented the HR estimates for the entire period rather than stratified by year of hospital discharge.
Propensity score matching was also used to investigate for confounding. A list of candidate variables for propensity scoring was compiled based on their likely association with the outcome. The list included: age, previous other IHD (within 5 years prior admission), previous heart failure (within 5 years prior admission), previous CARP (during hospitalisation or within 28 days of discharge or within 5 years prior admission), previous BB dispensed (within 2 months prior admission), previous statins dispensed (within 2 months prior admission), previous ACE/ARB dispensed (within 2 months prior admission), previous aspirin or clopidogrel dispensed (within 2 months prior admission) and Charlson index. For those exposure groups where a significant difference was observed, propensity scores were generated from logistic regression models for the likelihood of being placed in one of the eight exposure groups with the ‘no drug’ group as the comparator. A forward selection model (p<0.05) was used to identify potential confounders from the candidate variables and interactions terms. A 1:1 matching without replacement algorithm was used to match cases with controls within 0.25 SDs of the logits of the cases’ propensity scores. Matched analyses of these pairs were then performed using Cox proportional hazards regression to estimate HRs with 95% CIs for the likelihood of death for patients on either ‘ACEi/ARB only’, ‘BB and statins’ or ‘all three medications’, compared with patients on ‘neither BB, statins, nor ACEi/ARB’.
For all analyses, a two-sided p<0.05 was considered statistically significant. The statistical software SAS V.9.3 was used to perform all analyses.
Some 5892 (61.5%) men and 3688 women (38.5%) aged 65–84 years were admitted with a first AMI in Western Australia between the years 1995 and 2006. Of these patients and in the 5 years prior to their incident AMI, 2341 (24.4%) had another IHD-related admission, 2197 (22.9%) had a CARP and 905 (9.4%) had a recorded admission for heart failure. Within 60 days prior to AMI admission, 1777 (18.6%) patients were dispensed a BB, 2161 (22.6%) were dispensed a statin and 3100 (32.4%) patients had ACEi/ARB dispensed. In contrast, within 28 days following discharge for their AMI, the number of patients who were dispensed these medications were 4739 (49.5%), 4720 (49.3%) and 5381 (56.2%) for BB, statins and ACEi/ARB, respectively.
Table 1A,B present the baseline characteristics of the study population according to preventive drugs dispensed within 28 days following the incident AMI. Patients taking ‘ACEi/ARB only’ tended to be older and female compared with patients using ‘BB only’ or ‘statins only’. They were more likely to have had a previous other IHD and heart failure, and have used ACEi/ARB prior to AMI-admission. Aside from those in the ‘no drugs’ group, the ‘ACEi/ARB only’ group had the highest mortality rate within 1 year and throughout the study period (table 1A). For patients using a combination of drugs, the 1-year mortality and full-study mortality were highest among those using all three drug groups despite previous other IHD and heart failure being lowest in this group (table 1B).
All-cause mortality following an AMI admission
The risk of all-cause death within 1 year and until 2006 for patients dispensed different cardiovascular drugs within 28 days following discharge for an AMI admission are shown in table 2. Results suggested a decreased risk for all-cause mortality within 1 year was associated with drugs used in combination, compared with being on ‘no drugs’. However, drugs used singly showed no overall decrease in risk of 1 year mortality. Following adjustment for potential confounding factors, significant protective effects remained for each of the four drug combinations (table 2). The two drug combinations exerting the greatest protective effect were ‘BB+statins’ (adjusted HR 0.46, 95% CI 0.32 to 0.64) and ‘BB+statins+ACEi/ARB’ (adjusted HR 0.50, 95% CI 0.40 to 0.63). Gender-specific estimates suggested differences in risk for men and women associated with some drug categories measured but a lack of adequate power in this 1 year follow-up gender-stratified group limited interpretation.
Adjusted HR estimates for the full study period (to 2006) reflected the significant protective effects of ‘BB+statins’ and ‘BB+statins+ACEi/ARB’ seen in the 1 year follow-up (table 3). BB and statins dispensed in combination within 28 days postdischarge again afforded the largest decrease in risk for all-cause-mortality (adjusted HR 0.58, 95% CI 0.48 to 0.69). This protective effect was more pronounced in men (adjusted HR 0.46, 95% CI 0.36 to 0.58) compared with women (adjusted HR 0.77, 95% CI 0.60 to 0.99) with a significant (p=0.0046) interaction effect between gender and ‘BB+statins’ dispensed on risk for all-cause mortality. A significant drug-gender interaction (p=0.0289) was also observed for dispensed ‘BB+ACEi/ARB’, with men in this group possessing a decreased risk for all-cause mortality (adjusted HR 0.82, 95% CI 0.70 to 0.96) whereas women did not (adjusted HR 1.11, 95% CI 0.91 to 1.37).
Overall, dispensed ‘BB+statins+ACEi/ARB’ drugs were associated with a 34% decreased risk (adjusted HR 0.66, 95% CI 0.58 to 0.75) for all-cause death in men and women combined. In contrast, a significantly increased risk (adjusted HR 1.30, 95% CI 1.17 to 1.46) for all-cause death (in combined men and women) was associated with being dispensed ‘ACEi/ARB only’ within 28 days postdischarge.
Sensitivity testing and propensity matching analysis
A large disparity was observed between adjusted HR estimates (table 2) and crude HR estimates for the ‘ACEi/ARB only’ group, for the 1 year (crude HR 1.37, 95% CI 1.14 to 1.65) and 2006 (crude HR 1.79, 95% CI 1.62 to 1.98) follow-up periods. In contrast, differences between crude and adjusted HR estimates for the other drug exposure categories varied by 0.1 or less despite many of these exposure categories containing fewer observations respectively, suggesting the ‘ACEi/ARB only’ exposure group was particularly vulnerable to effects of confounding.
However, we found that neither restriction of age groups to those aged less than 75 years, nor removal of those with previous heart failure, other IHD or past use of ACEi/ARB medication notably altered the significantly increased risk associated with being on ‘ACEi/ARB only’ for the 2006 follow-up period. Similarly, there were no effective differences observed for the apparent protective exposure groups (‘BB+statins’ and ‘BB+statins+ACEi/ARB’). Since a substantial number of individuals did not possess a suitable match during the propensity score matching process, we opted for standard adjusted Cox regression analysis as a primary analytic strategy. Nevertheless, propensity-score matched Cox regression analyses provided estimates similar to those we report for the 2006 follow-up period (‘ACEi/ARB only’, HR 1.42, 95% CI 1.18 to 1.70; ‘BB+statins’, HR 0.70, 95% CI 0.51 to 0.97; ‘BB+statins+ACEi/ARB’, HR 0.70, 95% CI 0.53 to 0.93).
In this population-based cohort study of 9580 AMI-hospitalised individuals, we found significant reductions in risk for all-cause mortality at 1 year and up to 11 years following initial AMI if combinations of either ‘BBs and statins’ or ‘BB, statins and ACEi/ARB’ were dispensed within 28 days of discharge, compared with none of these drugs being dispensed. This protective effect was not observed for each of the three drug types when used singly. Differences in the magnitude of reduced mortality risk appeared to exist between genders for particular drug combinations. A 54% reduction in risk was observed in men who were dispensed ‘BBs and statins’, whereas the comparable female group only achieved a 23% reduced risk for all-cause mortality. Men also appeared to benefit from an 18% decreased risk associated with the use of ‘BBs and ACEi/ARBs’, compared with an unchanged risk in women on ‘BBs and ACEi/ARBs’. Tests for interaction confirmed that the risk associated with dispensed ‘BBs and statins’ or ‘BB and ACEi/ARBs’ on mortality differed between genders.
Our study includes a number of strengths. The ability to link the AMI cohort with the PBS register provides a real-world example of the effectiveness of these drugs in reducing mortality over 11 years follow-up. This is important because many studies have limited follow-up periods and interventional trials of these drugs typically under-represent women, older people and people with existing comorbidities.13–15 The population dataset also provides a large enough sample size to investigate gender as an effect modifier of drug effectiveness, and importantly also allowed us to compare seven possible drug combinations against a reference of ‘no drugs’. Importantly, since data were available for all hospitals within Western Australia, we are likely to have included the vast majority of non-fatal AMIs between 1995 and 2006 as very few are treated in the community. Likewise, mortality records are near-complete with estimates of over 99%.16 We expect coverage and financial access to BBs, statins and ACEi/ARBs to be high in our cohort. In 2004-2005, 90% of those aged 65–75 years and 95% of those aged 75+ years held a government concession card.17 We expect an AMI cohort to have even higher coverage than this. Given the heavy subsidies available to concession card holders, we do not consider significant financial barriers exist to patient access to these secondary preventive medications.
Administrative data do have limitations, including the inability to accurately determine aspirin use. Aspirin is included on the PBS register and may be prescribed by clinicians, but the dispensed aspirin estimates are likely under-represented since it is also possible to purchase aspirin over the counter. Medication dose and frequency data are not recorded in the PBS, limiting interpretation of results in relation to dose. Given the size of the cohort studied, it is unlikely that individual differences in dosage would appreciably affect the results.
Findings from trials and registries
Some large randomised studies have been conducted over extended follow-up periods, but their results differ somewhat. In 2006, Dagenais et al18 performed a meta-analysis of three large randomised trials of ACE inhibitors—the Heart Outcomes Prevention Evaluation, the European Trial on Reduction of Cardiac Events with Perindopril in Stable Coronary Artery Disease, and the Prevention of Events with Angiotensin-Converting-Enzyme inhibition trial—with follow-up periods of 4.5 years (median), 4.2 years (mean) and 4.8 years (median), respectively. While the meta-analysis found a significant reduction in all-cause mortality associated with the use of ACEi, initial analysis of the Prevention of Events with Angiotensin-Converting-Enzyme inhibition trial in isolation found no significant decrease in all-cause death or any of their cardiovascular-related outcomes.19 Similarly, a recent study of BB use in patients post-AMI found no significant reductions in risk for all-cause death associated with BB use.6
A number of registry-based studies have provided evidence of the importance of using combination therapies for secondary prevention of CVD (including AMI).10 ,20–22 This evidence, coupled with clinical trials data on effectiveness of the various drugs in isolation, has led to current trials being undertaken into the use of ‘poly-pills’ for primary and secondary prevention of cardiac events.23 Nevertheless, the optimal combinations and dosage remain controversial and the effectiveness of various drug combinations over extended follow-up periods in unrestricted populations is not fully elucidated,6 ,7 with many studies having follow-up times limited to less than 2 years post-AMI.6 ,10 ,15 ,21 ,22 ,24
One feature of most previous studies is the lack of reported outcomes associated with specific drug combinations. Most researchers control for baseline or concurrent secondary prevention medications and some studies compared the number of concurrent drugs (single, dual, triple or quadruple combinations). However, few studies report risks stratified by all specific drug combinations, rather reporting aggregated ‘numbers’ of drug types in combination.10 ,18 ,21 ,22 For example, Zeymer et al10 reported increased risk of 1 year mortality for patients on 0–1 or 1–2 drugs (in addition to BBs) compared with individuals on a three-drug regime. We observed marked differences between specific combinations of drugs in addition to the ‘number’ of drugs in combination. Our results suggest that ‘BBs and statins’ confer a similar risk to being on ‘BBs, statins and ACEi/ARB’ concurrently.
There is a paucity of evidence related to cardiovascular drug effectiveness by subpopulation, including gender.14 ,15 ,25–27 A meta-analysis conducted in 2003 suggested men given ACEi following heart failure had a reduced risk of mortality compared with women, although not statistically significant (p=0.07).13 In 2009, Dey et al28 reported that women were at a 24% (95% CI 1.14 to 1.34) higher risk than men of having adverse outcomes (death, AMI, stroke and rehospitalisation) at 6 months following coronary angiography. This observation may partly be explained by women being less likely to receive BBs and/or statins compared with men.29 A recent Danish registry study30 found no differences between sexes in all-cause mortality following an initial AMI, despite the similarities between genders in secondary preventive medical treatment (aside from lipid-lowering drugs which were significantly more prevalent in women). In our study the gender ratios in different exposure categories were reasonably well balanced, suggesting treatment similarities between genders. Possible explanations for the contradiction between our findings and those of the Danish study might include our relatively long follow-up (11 years compared with 3 years), the lipid-lowering drug imbalance between genders in the Danish study and differences in gender-related risk factors between populations.
In summary, our population-based investigation into the use of secondary prevention cardiovascular medications suggests that the use of ‘BBs and statins’, with or without ACEi/ARB drugs, within 28 days postdischarge provides the greatest reduction in risk for all-cause mortality over extended periods (up to 11 years post-AMI). This supports previous evidence that combination therapy is more effective than single therapies and that some combinations may influence survival over the very long term. Furthermore, our results suggest that risks associated with different drug combinations vary by gender. The largest risk reduction in all-cause death out to 11 years was associated with the use of ‘BBs and statins’ in men whereas a combination of all three drugs (BBs, statins and ACEi/ARBs) conferred the largest reduction in risk for women. This result has potential implications for treatment regimes and should be verified in other populations.
We wish to thank the Australian Department of Health and Ageing, the WA Department of Health and the Registrar Generals Office of WA for providing the data used for this investigation. We are furthermore grateful to the Data Linkage Branch of the WA Department of Health for extracting and linking the data.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online table
Contributors ASG, KE, FS, CDJH and TB conceived and designed the study. KE analysed the data. ASG wrote the first draft of the manuscript. All authors contributed to interpretation of results and writing of the manuscript. ICMJE criteria for authorship read and met by all authors.
Funding The research was supported by a project grant from Australia's National Health and Medical Research Council, (403928) in addition to salary funding from the Edith Cowan University Survey Research Centre. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests None.
Ethics approval The study was approved by the Human Research Ethics Committees of The University of Western Australia and the Western Australian Department of Health (ethics approval RA/4/1/1211).
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.