Abstract
A high risk of cardiovascular disease (CVD) is usually the consequence of a combination of moderately elevated risk factors rather than very high levels of single risk factors. Therefore, a consensus was reached during the 1990s that the identification of high-risk individuals has to be based on a multivariable assessment of the total risk of getting CVD or dying from it during a defined time period, usually 5 or 10 years. Accordingly, most international and national guidelines on the prevention of CVD have recommended some formal multivariable system for the assessment of absolute risk of either coronary heart disease (CHD) or CVD [1–6]. To begin with, a risk prediction system based on equations derived from the Framingham Heart Study [7] came into wide use, because it was most readily available. Framingham-based coronary risk charts were also provided in the guidelines on cardiovascular disease prevention created by the Joint European Societies in 1994 and 1998 [1, 2] cautioning, however, that the absolute risk may vary considerably between populations. The Framingham research group had already initially expressed that caveat [7] and later several studies validating the Framingham risk prediction systems in European populations showed an overprediction, not only in low-risk southern populations but also in populations with higher risk [8, 9].
To overcome the problems in the risk prediction based on non-European data, a collaborative project, called SCORE, was launched with the aim to provide European risk prediction systems. The SCORE project was based on 12 European cohort studies from 11 countries and included data on more than 200 000 persons representing 2.7 million person-years of follow-up [10]. Risk equations were calculated for the 10-year risk of fatal atherosclerotic CVD. CVD death was chosen as the outcome variable, because information on non-fatal CVD endpoints had not been collected in a uniform way in different cohorts. Age was used in the Weibull proportional hazards model as a measure of exposure time rather than as risk factor. The simple set of risk factors included smoking status, systolic blood pressure and total cholesterol or alternatively, total cholesterol/high-density lipoprotein cholesterol ratio. Separate risk prediction models and risk charts were provided for regions with high and low CVD mortality rates because of marked differences in CVD mortality within Europe. The survival functions for the cohorts from Denmark, Finland and Norway, combined with the risk-factor coefficients from the whole data set, were used to develop the high-risk model, while the survival functions for the cohorts from Belgium, Italy and Spain were used similarly to develop the low-risk region model.
The Third Joint European Societies' guidelines on cardiovascular disease prevention in clinical practice published in 2003 [5] recommend the use of SCORE risk charts. A 10-year risk of 5% or greater is proposed to denote a high risk, but in the cells of the charts numeric estimates of the risk are given to emphasize that the risk estimate is a continuous variable. For younger people, whose 10-year risk is low irrespective of their risk factor levels, extrapolation to the age of 60 years is recommended to emphasize the life-time risk. In charts for high-risk regions the risk estimates for middle-aged and older people with the same age and sex and risk characteristic combinations are almost twice as high as in charts for low-risk regions. The SCORE high-risk and low-risk models are also available as interactive web-based versions [11].
Already at the time of the preparation of the Third Joint European Societies' guidelines it was recognized that the two risk prediction models and charts would not take full account of the extent of CVD mortality differences between European countries. This led to the initiation of HeartScore project [11], which, in addition to the development of web-based versions, offers assistance in the recalibration of the risk models and charts for individual countries using current national data available on risk factor distributions and CVD mortality. Such country-specific risk prediction models and charts have already been developed for Sweden [12], Belgium [13], and Germany [14] and recalibration is in progress for several other countries.
In this issue of the European Journal of Cardiovascular Prevention & Rehabilitation, Lindman and her co-authors [15] have applied the SCORE high-risk model for the calculation of the 10-year risk of CVD death in the data from recent population surveys in Norway and have compared the outcome with the observed 10-year mortality for the years 1999–2003 calculated on the basis of national mortality statistics. The SCORE high-risk model overpredicted the CVD mortality risk markedly in men in all age groups and in older age groups of women. The main explanation for the discrepancy between the predicted and observed CVD mortality is that in Norway the CVD mortality has declined markedly from the time, years 1974–1994, when the data from the Norwegian cohort study participating in the SCORE were collected. Therefore, the original SCORE high-risk models will classify almost 80% of Norwegian men aged 60 years, and projecting age to 60 years, also the majority of younger men, in the high risk category. Interestingly, only 14% of Norwegian women aged 60 years were classified in the high-risk category, and projecting age to 60 years only a small proportion of younger women had high risk. Two other Norwegian research groups also recently pointed out that application of the SCORE high-risk model to other recent population-based studies in Norway would classify a large part of asymptomatic middle-aged and older population as high-risk, particularly applying extrapolation to age 60 years, and thus would lead to a marked increase in the use of antihypertensive and lipid-lowering drugs [16, 17].
Norwegian epidemiologists have to be congratulated for their work, but the next step in their country should be adaptation and updating cardiovascular risk prediction systems to their current conditions or development of own models based on good population-based data available. Because a substantial decline in cardiovascular mortality has occurred in many other European countries, a more general lesson from their findings is that further concerted efforts are needed to adapt and update the risk prediction systems to current national circumstances.