Develoment, validation and clinical utility of a long-term cardiovascular disease risk prediction model in the italian population

We illustrate the development of a 20-year prediction model of first major coronary or ischemic stroke event in a Northern Italian population of men and women aged 35 to 69 years at baseline. The model included age, blood lipids, systolic blood pressure, anti-hypertensive treatment, smoking habits and diabetes. The discrimination ability of the model was high as 0.736 in men and 0.801 in women. The model has been internally and externally validated using a different cohort study of subjects enrolled in Latina. Based on the external validation analysis, the risk score seems to be appropriate for long-term risk prediction in Italy and, more generally, in low-incidence populations. The clinical utility of the risk score in stratifying subjects in risk categories has been evaluated considering two strategies for the identification of “high-risk” subjects with contrasting public health goals, either to decrease the fraction of missed events or to decrease un-necessary treatment. These can be implemented by choosing threshold values for the predicted risk driven by either sensitivity or by specificity, respectively. The risk stratification based on 20-year absolute predicted risk had a higher clinical utility than any stratification based on the number of risk factors. Finally, we discussed from the statistical perspective the concept of “improvement” in risk prediction through the paradigmatic analysis of two indicators of disease heritability and social status, i.e. family history of coronary heart disease and educational level, added to the initial model. A new SAS package, Risk Estimation in Survival Analysis using SAS 9.2 [reSAS], detailed in the appendix, has been specifically developed from the author.