The risk functions incorporated in Riscard 2002: A software for the prediction of cardiovascular risk in the general population based on Italian data

Background. The purpose of this analysis was to produce risk functions for the prediction of cardiovascular diseases based on Italian epidemiological data and suitable for the use in a PC program dedicated to the estimate of risk. Methods. Three studies were used for the purpose: the Italian Rural Areas of the Seven Countries Study, the Gubbio Population Study and the ECCIS study, for a total of 9771 men and women aged 35 to 74 years and followed for a period lasting 5 to 6 years. The risk factors used for the prediction of cardiovascular events were sex, age, body mass index (derived from height and weight), mean blood pressure (derived from systolic and diastolic blood pressures), non-HDL cholesterol (derived from total and HDL cholesterol), HDL cholesterol, diabetes (yes-no), heart rate, and daily cigarette consumption. The endpoints were the first major coronary event, the first major cerebrovascular event, and the first major cardiovascular event (either one between the previous two plus major peripheral artery diseases). The model employed for the analysis was the accelerated failure time model. Results. Having excluded those already presenting with a cardiovascular disease and those with missing values, a total of 9089 subjects were included in the models. In a period lasting 5 or 6 years, a total of 211 coronary, 64 cerebrovascular and 269 cardiovascular events occurred and were considered for analysis. Coefficients from the coronary model suggested a significant association of all risk factors except body mass index and diabetes (marginal significance). Coefficients from the cerebrovascular model suggested a significant association limited to age and mean blood pressure. Coefficients from the cardiovascular model suggested a significant association of all risk factors except body mass index. The discrimination between cases and non-cases was satisfactory with proportions of 37.0, 52.3 and 37.8% of observed cases in decile 10 of the distribution of the estimated risk for the three endpoints respectively. Conclusions. The three models were used as a mathematical core for the construction of a PC software for the prediction of major cardiovascular events in Italy