Case fatality in coronary heart disease: the art of counting

This editorial refers to ‘One in four major ischaemic disease events are fatal and 60% are pre-hospital deaths: a national data-linkage study (ANZACS-QI 8)’, by C. Grey et al., on page 172. Coronary heart disease (CHD) mortality is declining in many European countries.1 This positive development is to a great extent due to decreasing levels of risk factors in the population.2 Despite the general increase in body weight and girth, slightly paradoxically, levels of blood pressure and serum cholesterol in the population are decreasing. This, in turn, is probably due to massive research aiming at detecting risk factors for CHD. With this in mind it is slightly surprising that CHD can be defined, and measured, in so many different ways. Common and important concepts in cardiovascular epidemiology such as incidence, prevalence, and mortality are difficult enough, but when it comes to anything that has to rely on a ratio between any two different numbers, comparing studies is quite a challenge. In measuring incidence, then, one has first to decide who is a case. Some studies, for instance the Multinational MONItoring of trends and determinants in CArdiovascular disease (MONICA), use very stringent criteria for acute myocardial infarction—by now, with the advent of highly sensitive markers of myocardial damage, very restrictive and limited to fairly severe cases.3 Unstable angina is usually not included but, while the rates of hospitalization for angina in Sweden essentially were reduced by two-thirds between 1998 and 2014 (from 374 to 126 per 100 000), the rates of acute myocardial infarction (AMI) decreased much less (from 274 to 209 per 100 000), probably reflecting much more changes in diagnostic practice and criteria for defining AMI rather than in biology (http://www.socialstyrelsen.se/statistik/statistikdatabas/diagnoserislutenvard). When considering case fatality, things become even more complicated.4,5 These days, surviving a heart attack is thankfully usually not accompanied by any severe and incapacitating symptomatology, and, apart from the need to take several medications on a daily basis, may pass relatively unnoticed. Case fatality in hospitalized cases has decreased substantially after the introduction of modern treatment and is now at an all-time low. However, a substantial part of deaths from CHD occur among patients who are not hospitalized, which is one of the more frightening aspects of CHD, because this often occurs in previously healthy people, with sudden death as the first presenting symptom. Cohort studies collecting follow-up data of fatal and non-fatal coronary events over a period of time allow for careful characterization, but are usually limited as to the actual number of events. Hospital-based registries, usually with details on clinical presentation, treatment, and complications, keep track of at least short-term and usually also long-term case fatality, but not of events occurring out of hospital, which is a serious limitation given that a large proportion of coronary deaths are sudden and not in a hospital setting. Population-based registries, many of them initiated within the framework of MONICA, provide data on out-of-hospital fatal events. However, in using criteria that may not be applicable to AMI as we define it today, the denominator in calculating case fatality may be too restricted, inflating case fatality. Even so, keeping to strict criteria is highly valuable in ensuring comparability over time, and in a study combining six population-based European registries on a total of 78 128 AMI events in persons aged 35–64 years, all populations except one (Tallinn in Estonia) showed significantly decreasing trends in total 28-day case fatality between 1985 and 2010.6 Of note, they adjusted for changing AMI definitions with the advent of highly sensitive cardiac biomarkers. Still, in-hospital case fatality, even though presented only in a graph, was considerably higher than in recent hospital-based populations in Sweden and the UK. For example, in men and women aged 65–74 years, inhospital 28-day case-fatality in AMI was about 20% in 2009, whereas corresponding figures for Sweden and the UK in 2004 to 2010 were less than half of that, based on data from >100 000 patients in this age bracket alone.7 This indicates that more severe infarctions are captured by the MONICA criteria than by the criteria used in the hospital-based registries, and such factors will obviously affect case fatality estimates in a major way. Administrative registers are increasingly being used in data-linkage studies in order to analyse trends in incidence and in survival. These registers, however, have pitfalls of their own. Apart from the possibility of miscoding, i.e. mistakenly providing an erroneous code, there is a very real possibility of misdiagnosis, because even though criteria for AMI exist, there is no guarantee that these criteria will be applied consistently by busy clinicians in a hospital setting. With respect to persons who die outside hospital, these events also offer choices on how to define them. For instance, selecting a wider group of deaths due to any coronary diagnosis may be preferred to selecting deaths attributed to AMI alone because a developing AMI may not be recognized at autopsy. The lack of consistency in defining out-of-hospital case fatality from CHD is conspicuous. One aspect that up till now has been overlooked is that not all deaths from CHD that are not linked to an admission for AMI occur without a previous hospitalization. Some CHD deaths follow a non-CHD-related hospitalization and that has usually not been considered. In the article in the present issue of the journal,8 the authors noted, slightly to their surprise, that there is currently no consistent approach to defining fatal and non-fatal CHD events in data-linkage studies, which has led to large variation in reported estimates of case fatality. Given the variation in defining cases detailed above, however, both numerators and denominators will be apt to vary considerably. Additionally, they argue that if CHD deaths, and not only AMI deaths, are included in the denominator, using different definitions for hospitalized and fatal events may create populations of hospitalized and non-hospitalized patients that are too dissimilar and that including all non-fatal CHD cases in the nominator would make for a more balanced approach. The study cohort that they investigated was constructed using linkage of routinely collected national health data in New Zealand, with the cohort defined as all CHD events (CHD-coded hospitalizations and deaths) occurring during 2 years from 1 December 2008. All hospitalizations with CHD as the primary discharge diagnosis were included, and encompassed with a few exceptions most diagnoses defined in the International Classification of Diseases as I20–I25. With a balanced approach using all CHD events and CHD deaths as the denominator for the fatalities, they estimated that about a quarter of all events were fatal within 28 days, of which ∼15% occurred within 28 days of a CHD hospitalization, 25% within 28 days of a non-CHD hospitalization, and ∼60% in persons who had not been hospitalized, a lower estimate for out-of-hospital death than in other reports.4 An unbalanced approach, with hospitalized AMI only and CHD deaths forming the denominator, in contrast, provided an estimate of 39% fatal events, with 61% alive after 28 days. Of those who died within a period of 28 days, roughly two-thirds had not been hospitalized. This is important and new information; however, many of the CHD deaths that were defined as hospitalized for a non-CHD diagnosis will also have occurred unexpectedly, out of hospital, and after discharge. Careful attention to risk factor management in these patients could help reduce these deaths, especially in those with a previous history of ischaemic heart disease. Two out of five of these hospitalizations were due to other circulatory conditions, emphasizing the importance of secondary prevention, but the rest were due to miscellaneous other conditions, and a potential window of opportunity for prevention. Future research will have to clarify to what extent and how these deaths might be prevented. Accordingly, identifying and characterizing these patients and the circumstances of their deaths represents an important next step. Mortality from coronary heart disease in hospital (within 28 days) and out of hospital per 100 000 population aged 35–84 in Sweden 1991–2006. Reprinted with kind permission from Dudas et al.9 Because diagnostic practices and systems for reimbursement differ between countries, and sometimes even within countries, harmonizing definitions for CHD presents considerable difficulties, particularly in view of the continuous changes in clinical presentation of this disease, sometimes as a result of new diagnostic methods, but also in biology and treatment. Even so, as emphasized in this work,8 in interpreting the literature in this area it is important to keep in mind both the numerator and the denominator. The research of A.R. is supported by grants from the Swedish state, under the agreement between the Swedish government and the county councils concerning economic support of research and education of doctors (ALF-agreement), the Swedish Heart and Lung Foundation, the Swedish Research Council (SIMSAM), and the Swedish Council for Working Life and Social Research. Conflict of interest: none declared.