Aims.
Recent US guidelines advocate the introduction of lifetime or long-term absolute risk prediction for primary prevention of cardiovascular events, especially for young people and women. Therefore, long-term prediction models might be specially beneficial in population considered at low incidence. We aim to develop a 20-year absolute risk prediction equation in a Northern Italy population.
Methods.
Four independent population-based cohorts were enrolled between 1986 and 1994 from the Brianza population (Northern Italy), adopting standardized MONICA procedures. The study sample comprises n=2574 men and 2673 women, aged 35 to 69 years and free of CVD at baseline. Participants were followed-up for incidence of first coronary and ischemic stroke events (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. We compared several gender-specific Cox Proportional Hazards models: the basic one includes age, total cholesterol, HDL-cholesterol, systolic blood pressure, anti-hypertensive treatment, cigarette smoking and diabetes. Candidates to model addition were diastolic blood pressure, triglycerides, BMI, family history of CHD, and education. Model calibration was tested using the Grønnesby-Bogan goodness-of-fit statistic. The Area Under the ROC-Curve (AUC) was a measure of discrimination, corrected for over-optimism via bootstrapping. Changes in discrimination (Δ-AUC) and reclassification (Net Reclassification Improvement, NRI) defined the improvement from the basic model due to an additional risk factor. Intermediate risk was defined as 20-year risk between 10% and 40%.
Results.
We observed n=286 events in men (incidence rate 7.7 per 1000 person-years) and n=108 in women (2.6 per 1000 person-years). All risk factors included in the basic model were predictive of first cardiovascular event in both genders; discrimination was 0.725 and 0.802 in men and women, respectively. Average specificity in the top risk quintile (cut-off value: 23% in men and 8.5% in women) was similar in men and women (85% vs. 83%), while sensitivity was higher in women (63% vs. 46%). All the models were well-calibrated (p-values >0.05). The addition of a positive family history of CHD in men (Hazard Ratio: 1.6; 95%CI 1.2-2.1) and of diastolic blood pressure in women (HR: 1.4 for 11 mmHg increase; 1.1-1.8) significantly improved discrimination (Δ-AUC=0.01; 95%CI 0.002-0.02 [men] and Δ-AUC=0.005; 95%CI 0.0001-0.01 [women]) and reclassification of subjects at intermediate risk (NRI=8.4%;1.7%-19.1% [men]; and NRI=11.7%; -3.2%-33.5% [women]).
Conclusions.
Traditional risk factors are predictive of cardiovascular events after 20 years, with good discrimination. The addition of family history of CHD may contribute to model improvement, at least among men; the role of diastolic blood pressure in women should be carefully evaluated.