A Bayesian Meta-Analysis approach to address the effectiveness of statins in preventing death after an initial myocardial Infarction

It is interesting to note that the effect of cholesterol lowering drugs such as statins and their ability to actually lower the cholesterol or some component such as the low density lipids (LDL) is fairly well established. The benefits appear to be present regardless of patient age, gender, or baseline cholesterol. However, the use of statins post myocardial infarction (MI) and their efficacy for reducing the likelihood of death can still be a topic of discussion. Despite the fact that they are an effective cardio event preventive therapy, the number of patients receiving statin therapy as a secondary prevention may remain suboptimal to some investigators. There are several retrospective cohort studies of patients discharged from the hospital following an acute myocardial infarction and many investigators have sought to look at the population impact of statin prescribing patterns. There have been conflicting studies concerning the efficacy of this intervention and meta analyses have been performed to determine the benefits and risks of intervention for individuals with an initial MI. Basically the question considered concerns the reduction in mortality in MI subjects due to this type of intervention. The general results have produced some conflicting data. This, of course, is controversial. The controversy usually arises from the fact that the patient cohort in most studies include unselected consecutive survivors of a first recorded MI from a large number of different hospital or institution types. The only exclusion criterion used, for example, may be age, since the recorded information focuses on coronary artery disease and, therefore, data on other important comorbidities influencing survival in various populations may not be available. There are usually no exclusions due to presence or absence of specific risk factors, comorbidities, anticipated adverse effects, participation in clinical trials, or contraindications to certain medications. Many representative cohorts are also strengthened by the inclusion of all patients with MI from the general population at institutions or clinics with different levels of care within an entire country or various countries of the world.. Compared with the National Registry of Myocardial Infarction in the United States, for example, the Swedish registry does not focus on thrombolytic therapy but includes all types of MI patients and a wider selection of background characteristics and treatments, which allows for adjustment of a large number of confounding factors. Some investigators will even contend that the definitions and detection of non fatal MI and adjudication of cause of death are not always straight forward particularly in large multi center trials. One should try to be aware of as many of these contentions as possible. However, this may not be possible. With these limitations in mind, the goal of this study is to examine the six known intervention trials which addressed this issue with conflicting results, combine the statistics in a rigorous Bayesian meta-analytic format with little or no bias and reach a conclusion concerning the efficacy of intervention in reducing post MI mortality. The approach here is to apply a Markov Chain Monte Carlo strategy to coherently combine prior diffuse information on the binary response (death versus alive) with the distribution of the logit of the response probability in the 6 studies and derive a posterior log odds of death for treatment (cholesterol lowering agents) versus no treatment (no intervention). We will examine the credible regions for the parameters and the convergence properties as well. The posterior values of the parameters of the six studies as well as the combined posterior parameters of the combined meta-analytic approach will be evaluated and examined. The interesting point is that results may conflict in part depending on the type of statin or population being treated. Our goal here is to investigate the models closely and determine the discrepancies and reasons for them. We then follow this process with another study to determine what consistency, if any, is there in studying the merits of statins after MI.