Registries have become extremely fashionable in cardiology: a PubMed search with the terms “acute myocardial infarction” and “registry” retrieved only 18 articles in 1995, compared with 112 in 2010. One may therefore wonder why such a sudden interest in what is only, after all, observational medicine.
To answer this question, one needs to determine what registries can (and cannot) bring to clinicians. To this purpose, we will describe the experience from the French registries in acute myocardial infarction (AMI), and particularly the French registry on Acute ST-elevation and non-ST-elevation Myocardial Infarction (FAST-MI) .
The first level of information brought by registries consists of an epidemiological description of disease characteristics, management and outcomes. The FAST-MI registry included patients admitted within 48 hours of symptom onset over a 1-month period at the end of 2005, in centres participating on a voluntary basis throughout France . FAST-MI had been preceded by two nationwide registries based on a similar methodology–USIK 1995 and USIC 2000 . These three registries had the participation of 60–80% of French institutions taking care of patients with AMI, and provided a unique opportunity to monitor the characteristics and management of patients admitted to hospital for AMI in France. In terms of the population description, probably the two most striking changes over the 10-year period, from 1995 to 2005, were the increased proportion of younger women (≤ 50 years) among patients with ST-elevation myocardial infarction (STEMI), from 3.7% in 1995 to 11.2% in 2005 ( Fig. 1 ), and the striking progression of body mass index (BMI), with a 50% increase in the prevalence of obesity in patients admitted with AMI, from 14 to 21%. Although the second observation was expected because of the constant increase in BMI in the general population, the first came as much more of a surprise, and therefore should generate questions and probably also specific measures in terms of public health information, targeted to a public of younger women.
Registries also interact with the guidelines from learned societies in a two-way process. First, registries can determine whether the guidelines are actually applied in the ‘real world’ and can help to promote their application in clinical practice. For instance, in France, the use of reperfusion therapy in STEMI patients has increased from 49% in 1995 to 63% in 2005, and this increase coincides with a reduction in early mortality (decline in 5-day mortality from 8.6% in 1995 to 6.4% in 2000 to 4.1% in 2005). Likewise, the prescription of recommended medications at hospital discharge has increased to levels that can now be considered rather adequate, although some efforts to improve these rates remain necessary (96% for antiplatelet agents, 84% for statins, 69% for angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, and 78% for beta-blockers). But registries can also verify that recommendations issued from guidelines are actually applicable. The 2003 guidelines on the management of STEMI of the European Society of Cardiology recommended that primary percutaneous coronary intervention (PCI) be used if the time-to-balloon was expected to be < 90 min . In FAST-MI, however, as in numerous other registries, the median time from first call to reperfusion was much longer : 170 min, and it was still 120 min (interquartile range 90, 170 min) in the patients who had called the SAMU– Services d’Aide Médicale Urgente –directly, meaning that three of four patients treated with primary PCI for STEMI were outside the recommended timeframe. The 2003 guidelines therefore appeared unrealistic ; the 2008 guideline revision took into account this fact and recommended a time-to-balloon of < 120 min (the 90-min goal remaining for those patients seen very early) .
Registries can also contribute significantly to scientific knowledge, and they can be included in the generation of evidence-based medicine. Although many still think of randomized controlled trials (RCTs) as synonymous with evidence-based medicine, there is no question that evidence should go far beyond that brought by such trials. Populations included in RCTs are far from representative of the populations with a given disease seen in everyday practice. Indeed, many exclusion criteria apply in RCTs ; in particular, the presence of concomitant diseases or conditions such as chronic renal failure result in the exclusion of important, and often more at risk, populations. In this regard, elderly patients are often excluded, per protocol or de facto, from RCTs, so that it may be difficult to determine whether the conclusions drawn from these trials actually apply to elderly populations. Evidence must therefore be sought from other sources, and well-designed registries probably provide the best complementary source of data.
Many such unresolved clinical questions were thus tackled using the FAST-MI data. In AMI, France has gathered a long-standing experience in the use of prehospital therapy, and prehospital fibrinolysis has become the standard of care, when this reperfusion method is chosen ; most of the patients are sent for coronary angiography and PCI soon afterwards. This pharmaco-invasive approach is quite different from stand-alone fibrinolysis, which has been shown in RCTs to be clinically inferior to primary PCI. The results of the pharmaco-invasive strategy used in France were therefore compared with the results of primary PCI .
Statistical methods can help to determine whether the outcomes observed in a registry cohort as a whole are valid, and are not fraught with biases, in particular prescription bias. Multivariable analysis, for example, adjusts for known potential confounders. Another statistical technique is the propensity score, which involves the building of a probability score for getting one type of treatment over another, according to the patient’s profile ; the score is then used to match one patient having actually received one type of treatment with another patient who received the alternative treatment. In the case of the pharmaco-invasive approach to reperfusion, two cohorts were built, each comprising 448 patients, with the same likelihood of receiving either type of reperfusion therapy, and consequently having similar baseline characteristics ( Table 1 ); the clinical outcomes in these two groups were found to be very similar ( Fig. 2 ). Similar techniques can be used to address a number of questions that are unresolved by RCTs. As the totality of medications used, whether cardiovascular or non-cardiovascular, was recorded in FAST-MI, many issues could be examined. Thus, a strong inverse correlation was found between the early prescription of statins and the occurrence of atrial fibrillation at the acute stage, suggesting a protective effect of these medications vis-à-vis the development of acute episodes of atrial fibrillation . We also investigated the outcomes of diabetic patients on sulfonylureas before the AMI, comparing those treated with glibenclamide, a medication known to block the mechanisms of ischaemic preconditioning, with those receiving the pancreatic-specific sulfonylureases gliclazide or glimepiride; in-hospital mortality and complications were more frequent in the patients on glibenclamide, a finding which suggested a true protective role of ischaemic preconditioning in patients developing AMI . Recently, we investigated the influence of the co-prescription of proton pump inhibitors on clinical outcomes in patients treated with clopidogrel, and found no deleterious effect on the occurrence of ischaemic events . Finally, the collection of serum and DNA in a large proportion of the patients gave us the opportunity to test new biological markers and to assess the interactions of gene variants with the efficacy of medications, opening a door into the field of pharmaco-genetics; for instance, in the FAST-MI population, patients with two variant alleles of the cytochrome CYP2C19, an enzymatic system needed for the transformation of clopidogrel into its active metabolite, were at higher risk of developing ischaemic events, both at the acute stage and during follow-up, suggesting a reduced efficacy of clopidogrel in this subgroup of patients .
