A nationwide causal mediation analysis of survival following ST-elevation myocardial infarction.

OBJECTIVE
International studies report a decline in mortality following ST-elevation myocardial infarction (STEMI). The extent to which the observed improvements in STEMI survival are explained by temporal changes in patient characteristics and utilisation of treatments is unknown.


METHODS
Cohort study using national registry data from the Myocardial Ischaemia National Audit Project between first January 2004 and 30th June 2013. 232 353 survivors of hospitalisation with STEMI as recorded in 247 hospitals in England and Wales. Flexible parametric survival modelling and causal mediation analysis were used to estimate the relative contribution of temporal changes in treatments and patient characteristics on improved STEMI survival.


RESULTS
Over the study period, unadjusted survival at 6 months and 1 year improved by 0.9% and 1.0% on average per year (HR: 0.991, 95% CI: 0.988 to 0.994 and HR: 0.990, 95% CI: 0.987 to 0.993, respectively). The uptake of primary percutaneous coronary intervention (PCI) (HR: 1.025, 95% CI: 1.021 to 1.028) and increased prescription of P2Y12 inhibitors (HR: 1.035, 95% CI: 1.031 to 1.039) were significantly associated with improvements in 1-year survival. Primary PCI explained 16.8% (95% CI: 10.8% to 31.6%) and 13.2% (9.2% to 21.9%) of the temporal survival improvements at 6 months and 1 year, respectively, whereas P2Y12 inhibitor prescription explained 5.3% (3.6% to 8.8%) of the temporal improvements at 6 months but not at 1 year.


CONCLUSIONS
For STEMI in England and Wales, improvements in survival between 2004 and 2013 were significantly explained by the uptake of primary PCI and increased use of P2Y12 inhibitors at 6 months and primary PCI only at 1 year.


TRIAL REGISTRATION NUMBER
NCT03749694.


AbsTrACT
Objective international studies report a decline in mortality following sT-elevation myocardial infarction (sTeMi). The extent to which the observed improvements in sTeMi survival are explained by temporal changes in patient characteristics and utilisation of treatments is unknown. Methods cohort study using national registry data from the Myocardial ischaemia national audit Project between first January 2004 and 30th June 2013. 232 353 survivors of hospitalisation with sTeMi as recorded in 247 hospitals in england and Wales. Flexible parametric survival modelling and causal mediation analysis were used to estimate the relative contribution of temporal changes in treatments and patient characteristics on improved sTeMi survival. results Over the study period, unadjusted survival at 6 months and 1 year improved by 0.9% and 1.0% on average per year (hr: 0.991, 95% ci: 0.988 to 0.994 and hr: 0.990, 95% ci: 0.987 to 0.993, respectively). The uptake of primary percutaneous coronary intervention (Pci) (hr: 1.025, 95% ci: 1.021 to 1.028) and increased prescription of P2Y 12 inhibitors (hr: 1.035, 95% ci: 1.031 to 1.039) were significantly associated with improvements in 1-year survival. Primary Pci explained 16.8% (95% ci: 10.8% to 31.6%) and 13.2% (9.2% to 21.9%) of the temporal survival improvements at 6 months and 1 year, respectively, whereas P2Y 12 inhibitor prescription explained 5.3% (3.6% to 8.8%) of the temporal improvements at 6 months but not at 1 year. Conclusions For sTeMi in england and Wales, improvements in survival between 2004 and 2013 were significantly explained by the uptake of primary Pci and increased use of P2Y 12 inhibitors at 6 months and primary Pci only at 1 year.

InTrOduCTIOn
There has been a global decline in mortality and non-fatal complications following acute myocardial infarction. 1 For ST-elevation myocardial infarction (STEMI), the adoption of new health technologies such as primary percutaneous coronary intervention (PPCI) as well as the availability of novel pharmacotherapies has been identified as driving improvements in clinical outcomes. 2 However, the extent to which population-based temporal improvements in outcomes from STEMI are due to the uptake of, say, PPCI compared with other guideline-indicated treatments or changes in patient characteristics is not known. Resolving the knowledge gap around the effectiveness of STEMI treatments on temporal outcomes could help future healthcare planning for developing countries with, or predicted to have, a high burden of cardiovascular disease.
Notably, there is a paucity of large-scale cohorts that are of sufficient duration to enable a detailed evaluation of the association of baseline risk and guideline-indicated therapies with temporal trends in STEMI mortality. [3][4][5] Where there have been studies of treatments and outcomes for STEMI, analyses have quantified associations and not necessarily reported explanatory (causal) factors. The Myocardial Ischaemia National Audit Project (MINAP) is a whole country registry of hospitalised cases of acute coronary syndrome (ACS), representing all hospitals in a single health system (the National Health Service of England and Wales) with prospective collection of detailed information about quality of care and clinical outcomes of patients for more than 15 years. 6 7 Our objective was to investigate whether temporal improvements in survival were associated with changes in patients' baseline clinical risk or use of guideline-indicated treatments for the management of STEMI, and to determine the extent to which associations explained the temporal improvements in survival.

MeThOds data and patients
The analyses were based on data from MINAP, a comprehensive registry of ACS hospitalisations started in 2000 and mandated by the Department of Health in England and Wales. 6 7 Data were collected prospectively at each hospital, electronically encrypted and transferred online to a central database. Data entry is subject to routine error checking and a mandatory annual data validation exercise. Patient-level data concerning demographics, cardiovascular risk factors, medical history and clinical characteristics at the time of hospitalisation were extracted from MINAP and (if applicable) date of death from linkage to the Office for National Statistics. Further details of MINAP have been published elsewhere. 6 7 The diagnosis of STEMI was based on guidelines from the European Society of Cardiology (ESC), American College of Cardiology and American Heart Association, Coronary artery disease and determined at local level by the attending Consultant on discharge from hospital. 8 The analytical cohort (n=232 353) was drawn from 272 263 patients with STEMI admitted to one of 247 hospitals between first January 2004 and 30 June 2013 (figure 1). For multiple admissions, we used the earliest record. As discharge medication was a key exposure, we excluded 23 504 (8.6%) who died in hospital; 16 406 (6.0%) patients with missing mortality data were also excluded. The primary outcome was all-cause mortality at 1 year following discharge from hospital. For care interventions, patients were classified as ineligible if a treatment was contraindicated, not indicated, not applicable, if the patient declined treatment as recorded in MINAP or if the admission preceded the inclusion of the treatment in guidelines (online supplementary eTable 1).

statistical analyses
Baseline characteristics were described using numbers and percentages for categorical data and means and SD or medians and interquartile ranges for normally and non-normally distributed continuous variables. Temporal trends of patient and treatment characteristics were summarised by comparing data from the start of the study (2004)(2005) to the end of the study (2012-2013) using χ 2 tests, t-tests and Wilcoxon rank sum tests.
Royston-Parmar flexible parametric survival models 9 were fitted to explore the association between temporal changes in patient demographics (age, sex, socioeconomic deprivation (Index of Multiple Deprivation score), comorbidities and risk factors (diabetes, hypercholesterolaemia, hypertension, smoking status, asthma/chronic obstructive pulmonary disease, chronic renal failure, chronic cardiac failure, cerebrovascular disease, peripheral vascular disease, angina, previous myocardial infarction, PCI and coronary artery bypass graft surgery), pharmacological treatments (secondary prevention pharmacotherapies prescribed at hospital discharge: aspirin, statins, P2Y 12 inhibitors, ACE inhibitors (ACEi)/ARBs and β-blockers), cardiac rehabilitation and reperfusion strategy (defined as receipt of PPCI) with temporal changes in 1-year survival from hospital discharge. Flexible parametric survival models were selected in favour of Cox regression models to overcome violation of the proportional hazards assumption. Improvements in model fit at each stage were determined by minimising the Akaike information criterion (AIC) and Bayesian information criterion (BIC) (online supplementary eTable 2). The scale (proportional hazards, proportional odds or normal) and complexity (number of degrees of freedom) for flexible parametric survival models were checked on the full multivariable model for each imputation. The baseline hazard on the normal scale with five degrees of freedom produced the optimal model through minimisation of the AIC and BIC (online supplementary eTable 2).
Initially, to determine the overall temporal trend in 1-year survival, an unadjusted model comprising only the year of admission was fitted. Subsequently, to determine the impact of patient demographics, PPCI, comorbidities and risk factors, aspirin at discharge, statin at discharge, P2Y 12 inhibitors at discharge, ACEi/ARBs at discharge, β-blockers at discharge and cardiac rehabilitation on the temporal trend in survival, each of these factors were added individually to the univariate year model. These models were then adjusted for age, sex and deprivation to form the primary analysis.
Secondary analyses included; a full model including all of the explanatory and adjustment variables stated above. This strategy allowed the influence of each factor on the temporal survival trend to be assessed in both an unadjusted and adjusted manner. To confirm that the identified factors had an association with survival trends, two flexible parametric survival models were fit, one including only the subset of variables found to have an influence on the survival trends (model A) and the other including only the subset variables found not to have an influence on survival trends (model B).
A causal mediation analysis was conducted to determine the relative contribution of patient and treatment variables on the survival trend. This analysis allowed the assessment of potential causal pathways that linked year of admission to temporal improvements in STEMI survival, such as through changes in clinical factors or therapeutic strategies, beyond simple point estimates. The magnitude of the contribution of the mediating variables was adjusted for confounding variables (online supplementary section 2) and presented as proportions.
To mitigate potential bias caused by missing data, we used multiple imputation by chained equations to create 10 datasets from 20 iterations, of which the resultant model estimates for each were combined using Rubin's rules (Section 3, eTable three in the Supplement for the imputation strategy). As a secondary outcome, we repeated the above analytical methods for temporal changes in 6-month survival. All tests were two-sided, and statistical significance was considered p<0.05. Statistical analyses were performed in Stata V.14 (http://www. stata. com/) and R V.1.2 (https:// cran. r-project. org/ bin/ windows/ base/).

Patient involvement
This study did not involve patients/service users/carers/lay people in its design or for the development of outcome measures.

Temporal trends in clinical characteristics
Over the study period, the proportion of STEMI who had a previous myocardial infarction (12.  (table 1).

Association between changing risk profile and improved outcomes
One   declined over the study period, we conducted a sensitivity analyses stratified according to prior statin use in which our findings were substantiated (online supplementary eTables 4 and 5).

Causal mediation analysis
The uptake of PPCI explained 13.2% (95% CI: 9.2% to 21.9%; average across the 10 imputed datasets) of the 1-year and 16.8% (95% CI: 10.8% to 31.6%) of the 6-month survival improvements (online supplementary eTables 6 and 7). The increased prescription of P2Y 12 inhibitors at hospital discharge explained 5.3% (95% CI: 3.6% to 8.8%) of the improvement at 6 months, but not at 1 year (online supplementary eTables 6 and 7). A sensitivity analysis including in-hospital deaths was carried out and the results were consistent with the main findings (online supplementary eTables 8 and 9).

dIsCussIOn
In this national cohort study of the management and outcome of nearly a quarter of a million patients hospitalised with STEMI, we found that temporal improvements in 1-year survival between 2004 and 2013 were associated with the uptake of PPCI and increased prescription of P2Y 12 inhibitors at hospital discharge. Similar findings were observed for temporal improvements in 6-month survival. At the end of the study, the use of guideline-indicated therapies for STEMI was high, such that over three-quarters of patients with STEMI received PPCI, and the prescription of evidence-based pharmacotherapies reached more than 95%. While the majority of treatment and patient-related factors studied were significantly associated with 6-month and 1-year survival, it was only temporal trends in PPCI and P2Y 12 inhibitors at hospital discharge that significantly explained the temporal trend in survival. The association between guideline recommended care and improved survival after myocardial infarction is recognised and supported by robust evidence from randomised clinical trials. 10 For STEMI, PPCI has been shown to be associated with a decline in mortality, [11][12][13] as has the use of antithrombotic and secondary prevention medications. 3 4 14 15 While others have investigated whether these or other factors are associated with the decline in mortality over time, [3][4][5] to our knowledge this is the largest study to date and first investigation using causal medication analysis Coronary artery disease to determine the relative contribution of factors on survival improvements in STEMI.
Our findings support and extend the existing literature. [3][4][5] A recent study of 105 674 patients with STEMI recorded in the Swedish Web-System for Enhancement and Development of Evidence-Based Care in Heart Disease Evaluated According to Recommended Therapies registry found that changes in reperfusion and PPCI were each associated with improved in-hospital outcomes, whereas an increase in the prescription of discharge medications was associated with improved outcomes at 1 year. Our study builds on this, finding that the nationwide introduction of PPCI explained over one-tenth of the improvements in 1-year survival for STEMI.
We found no significant mediation effect for improvements in 1-year survival by increased prescription of P2Y 12 inhibitors at hospital discharge. By contrast, a significant effect was identified for survival improvements at 6 months. Debate remains as to the most appropriate duration of dual antiplatelet therapy following myocardial infarction [16][17][18] and in the UK variation exists as to whether this is from 3 to 12 months or longer. 19 Thus, it is possible that persistence with therapy influenced the duration of the causal effect. Alternatively, it may be that a second antiplatelet agent following STEMI has limited benefit for longerterm all-cause mortality. 18 The finding that the uptake of P2Y 12 inhibitors and PPCI significantly explained the temporal trends in survival is not surprising. P2Y 12 inhibitors are indicated with a Class I recommendation and Level A evidence 10 for all STEMI before, or at the time of PPCI. As such, one would expect a moderating relationship between PPCI and P2Y 12 inhibitors. Yet, the uptake of PPCI and increased prescription of P2Y 12 inhibitors did not explain all of the improvements in survival. Earlier research found that organisational factors, as well as national and local infrastructure are key elements in the delivery of a STEMI service and, therefore, clinical outcomes. 2 This study has wider implications for international cardiovascular health. In line with the WHO Global Action Plan for noncommunicable disease, we have identified factors including and beyond PPCI that are associated with improved survival over time for patients with STEMI-therefore helping identify where in a healthcare system the provision of essential treatments is required to reduce premature death. Furthermore, our results may be extrapolated to other developed and developing countries which lag behind Northern Europe and North America in their provision of care and where greater gains in cardiovascular health maybe realised. 20 21 MINAP is the largest whole-country, single-health-system, prospective observational cohort of the quality of care and clinical outcomes for ACS. It is designed to be representative of the management of ACS and has standardised criteria for defining case mix and treatments. To our knowledge, this study is the first to quantify the relative contribution mediators of temporal improvements in survival for STEMI. However, our study has limitations. (1) We were reliant on the accurate recording of data in MINAP. (2) MINAP collects the majority, but not all cases of STEMI in England and Wales. (3) Missing data could have biassed our estimates; to mitigate this we studied the nature of the missing data and used imputation algorithms. (4) Other factors beyond the hospital stay (such as drug adherence and compliance, and primary care visits) may have influenced temporal changes in survival. (5) We studied all-cause mortality, when non-cardiovascular deaths may not be attributable to temporal improvements in STEMI care. 22 (6) Given that the determined mediators did not fully explain the survival improvements implies unmeasured mediators exist out with MINAP data fields. (7) MINAP lacks information on temporal change in stent platforms, that is, shift from bare metal stents to first-generation and second-generation drug eluting stents. (8) Door-to-balloon and total ischaemic times were not investigated as potential mediators due to poor recording of this information within our MINAP extract.

COnClusIOn
Among 232 353 patients hospitalised with STEMI in England and Wales, improvements in all-cause mortality between 2004 and 2013 were significantly associated with, and explained by, the national uptake of PPCI and increased prescription of P2Y 12 inhibitors, and not entirely related to changes in comorbidities or increased use of other pharmacological therapies.

Key messages
What was already known on this subject? ► Temporal improvements in survival following ST-elevation Myocardial Infarction (STEMI) have been observed and the improvements have been attributed to adoption of new health technologies such as primary percutaneous coronary intervention (PPCI) as well as the availability of novel pharmacotherapies. However, the relative contributions of temporal improvements in treatments and patient characteristics on improved STEMI survival are unknown.
What might this study add? ► While the majority of treatment and patient-related factors studied were significantly associated with STEMI survival, it was only temporal improvements in P2Y 12 inhibitors prescription and national uptake of PPCI that significantly explained the temporal trend in STEMI survival observed between 2004 and 2013.
how might this impact on clinical practice? ► Identifying factors associated with improved survival over time for patients with STEMI helps identify where in healthcare systems the provision of essential treatments is required to reduce premature death from cardiovascular disease.