Comparing short-term and long-term absolute cardiovascular risk prediction models in a low-incidence Country. Insights from the CAMUNI absolute risk equation in Northern Italy

Aims. The 10-year absolute risk of cardiovascular events, a current standard for treatment indication in CVD primary prevention, has been recently shown to inadequately estimate the risk in young subjects and women, even in presence of a cluster of risk factors. We aim to compare a 10- and a 20-year absolute risk of CVD prediction models in a low incidence population. Methods. Four independent population-based cohorts were enrolled between 1986 and 1994 in Northern Italy, adopting standardized MONICA procedures. N=2574 men and n=2673 women, 35-69 years old and free of CVD at baseline, were followed-up to ascertain the first coronary or ischemic stroke event (fatal and non-fatal; all MONICA validated) for a median time of 15 years (IQ range: 12-20) and up to the end of 2008. The 10- and the 20-year absolute risk (AR) of event were estimated from gender-specific Cox models including age, total cholesterol, HDL-cholesterol, systolic blood pressure, anti-hypertensive treatment, cigarette smoking and diabetes. Model discrimination at year 10 and 20, defined as the Area Under the ROC Curve, as well as sensitivity and specificity in the top quintile, were estimated taking censoring into account; the AUC was adjusted for over-optimism via bootstrapping. We assessed the Net Reclassification Improvement (NRI) for the 20-year risk model compared to the 10-year risk model re-calibrated to the Kaplan-Meier estimate of 20-year survival. We report NRI in young men and women as well as in subjects with AR(10)<5% but 2 or more major risk factors (total cholesterol>240 mg/dl; HDL-cholesterol <40 [men] or <50 [women] mg/dl; systolic blood pressure >160 mmHg; smoking; diabetes). Results. The Kaplan-Meier estimates of AR(10) and AR(20) were 7.1% and 14.0% in men; and 2.2% and 5.3% in women. After 20 years from measurement, major risk factors were still predictive of CV events; although in comparison to the 10-year model the hazard ratio for total cholesterol in men fell from 1.40 to 1.34; and for HDL-cholesterol in women reduced from 0.64 to 0.71, both calculated for 1 SD increase. In the top risk quintile, the 20-year model had higher specificity (85% vs. 82% in men; 83% vs. 81% in women) and lower sensitivity (46% vs. 50% in men; 63% vs. 69% in women) than the 10-year model; the overall discrimination was similar (0.725 vs. 0.732 [men], and 0.802 vs. 0.813 [women]). NRI in young subjects was 10.5% in men (15.4% in cases) and 5% in women. Six per cent of men and 15% of women had 2 or more major risk factors but an AR(10)<5%. Among them, NRI was 7% in men and 13% in women (18% in cases). Conclusions. In our low-incidence population, long-term prediction of first major ischemic cardiovascular event is more suitable to identify high CVD risk in young men and in women with a cluster of risk factors, overcoming the limits of the 10-year model. Preventive strategies based on long-term absolute risk should be evaluated.