Temporal changes in myocardial infarction diagnostic criteria do not account for the lack of long-term beta-blocker effect on cardiovascular prognosis

This commentary refers to ‘Effect of longterm beta-blocker treatment following myocardial infarction among stable, optimally treated patients without heart failure in the reperfusion era: a Danish, nationwide cohort study’, by A. Holt et al., doi:10.1093/eurheartj/ehaa1058. and the discussion piece ‘Long-term beta-blocker treatment in stable patients aftermyocardial infarction: a potential impact due to changes in the diagnosis of myocardial infarction?’, by J.G. van der Stouwe et al., doi:10.1093/eurheartj/ehab257. On behalf of all the original authors, this commentary is a response to ‘Long-term beta-blocker treatment in stable patients after myocardial infarction—a potential impact due to changes in the diagnosis of myocardial infarction?’1 First of all, thank you for your engagement in a relevant discussion. Stouwe, Boeddinghaus, and Mueller highlight some very important considerations when assessing the clinical interpretability of our study.2 We included myocardial infarction (MI) patients during a 16-year period, all undergoing coronary arteriography—but not necessarily percutaneous coronary intervention (PCI). Stouwe et al. underscore that an attending physician could consider the indication for beta-blocker (BB) post-MI differently according to severity and type of MI. Furthermore, this could be extrapolated to a difference in temporal trends of BB prescriptions, i.e. smaller MIs are diagnosed due to improved diagnostic possibilities in the later years of the study period.3 We agree that these assumptions are supported by our data showing a temporal increase in the proportion not prescribed BB from 2003 to 2018. Likewise, patients not undergoing PCI were less likely to be prescribed BB.1 Statistically, we have already adjusted for these probable differences in indication by including interaction terms between BB treatment and calendar year as well as type of procedure during admission.1 Hence, the multivariate model provides treatment effects that closely relate to those which can be obtained via subgroup analyses. Consequently, the standardized average risk difference that we present can be interpreted as a weighted average of the risk difference across subgroups (the larger the group, the larger the weight).4 Nevertheless, we welcome the opportunity to present additional results. Regrettably, the Danish registers do not contain information on cardiac troponin levels; however—and as the authors suggested—this potential bias was sought illustrated by analysing the main outcomes in subgroups undergoing PCI, stratified by calendar periods. Associating BB treatment with cardiovascular prognosis among patients with a suspected type 1 MI requiring revascularization yielded no significant absolute risk differences at the 3-year mark (Table 1). When assessing the effect of BB treatment in all MI patients undergoing PCI, but separately in three consecutive periods according to calendar year, no statistically significant results were found either (Table 1). We lack the power to prove that no association exists, and for a few of the analyses there seem to be a non-significant trend towards a minor benefit of BB treatment. Nonetheless, taking all analyses into account, it seems fair to conclude that any association—if any exists—would be clinically negligible. Cardiovascular risk associated with beta-blocker treatment following patients from 90 days to 3 years after myocardial infarction in subgroups undergoing percutaneous coronary intervention and stratified by calendar periods AR, absolute risk; BB, beta-blocker; CI, confidence interval; CV, cardiovascular; MI, myocardial infarction; PCI, percutaneous coronary intervention. A multiple logistic regression model was fitted to estimate standardized AR and AR differences after 33 months of follow-up (3 years after MI hospitalization). The model included interaction terms between BB treatment and sex and age as well as adjustment for educational level, stroke, diabetes mellitus, hypertension, peripheral arterial disease, acute and chronic kidney disease, liver disease, gastrointestinal bleeding, and cancer (excluding non-melanoma skin cancer). Composite CV outcome included CV death, recurrent MI, heart failure, stroke, angina pectoris, and cardiac procedures. Acute kidney disease could not be adjusted for in this analysis due to too few patients with the condition. Cardiovascular risk associated with beta-blocker treatment following patients from 90 days to 3 years after myocardial infarction in subgroups undergoing percutaneous coronary intervention and stratified by calendar periods AR, absolute risk; BB, beta-blocker; CI, confidence interval; CV, cardiovascular; MI, myocardial infarction; PCI, percutaneous coronary intervention. A multiple logistic regression model was fitted to estimate standardized AR and AR differences after 33 months of follow-up (3 years after MI hospitalization). The model included interaction terms between BB treatment and sex and age as well as adjustment for educational level, stroke, diabetes mellitus, hypertension, peripheral arterial disease, acute and chronic kidney disease, liver disease, gastrointestinal bleeding, and cancer (excluding non-melanoma skin cancer). Composite CV outcome included CV death, recurrent MI, heart failure, stroke, angina pectoris, and cardiac procedures. Acute kidney disease could not be adjusted for in this analysis due to too few patients with the condition. To assess bias introduced by tachyarrhythmias (ICD-10 codes DI47–DI49), we identified all patients diagnosed prior to start of follow-up (n = 462, 1.4% of BB group, 1.6% of no BB group). Neither adding this diagnosis to the multivariate model nor excluding these patients from the analysis changed our main results (data not shown), possibly because patients with the most common tachyarrhythmia (atrial fibrillation)5,6 were already excluded to avoid this bias.1 In conclusion, while we appreciate the critical appraisal by Stouwe et al., and acknowledge that the included cohort of MI patients could be somewhat heterogenous, the lack of association persists—even when restricting the analysis to patients needing revascularization in the earliest days. Conflict of interest: none declared. It is not allowed by Danish law to share the data used for this study.