Abstract P446: External Validation Of The CAMUNI Score, A Long-term Cvd Prediction Equation For Low-incidence Populations.
Aims: We assess the external validation, on a “new” set of subjects, of the CAMUNI 20-year risk score, developed in a northern Italian population. Although an external validation analysis is recommended before adopting a score in clinical practice, it is rarely performed in long-term prediction equations as it requires similar long-term follow-up procedures in different cohort studies.
Methods: The CAMUNI risk score was developed to estimate the 20-year risk of first coronary or ischemic stroke event, fatal or non-fatal, in 5247 (2574 men) 35-69 subjects free of CVD at baseline enrolled in late 1980s-early 1990s in Brianza, northern Italy (derivation set). The score is based on two gender-specific Cox models, including age, total- and HDL-cholesterol, systolic blood pressure, anti-hypertensive treatment, cigarette smoking and diabetes. As validation set, we considered 5307 (2418 men) subjects enrolled in the Rome area in the same time span (MATISS Study). Both the derivation and the validation cohorts shared the same procedures for baseline risk assessment and follow-up procedures, including MONICA definition of acute events. We evaluated the performance of the CAMUNI score in the validation set; the Framingham CVD risk score was used for comparison. The absolute predicted risk from both scores was re-calibrated to the 20-year risk observed in the validation set. We report the calibration slope as a measure of calibration; a value different from 1 is suggestive of a different strength in predictor effects. The Area Under the ROC-curve (AUC) measured the discrimination ability; the AUC in the validation set was compared to the value in the derivation set, corrected for over-optimism.
Results: The 20-year Kaplan-Meier risk of event was 16.1% (derivation set; 315 events) and 13.2% (validation set; 238 events) in men, and 6.1% (123 events) and 5.6% (119 events) in women. The calibration slope for the CAMUNI score did not significantly differ from 1 in men (1.1; 95% confidence interval 0.9; 1.2) nor in women (1.0; 0.8;1.2). The Framingham score performed equally well in men (1.1; 0.9; 1.2) but worse in women (1.3; 1.1;1.6). In the derivation set, the over-optimism corrected AUC for the CAMUNI model was 0.737 (men) and 0.801 (women); the corresponding figures in the validation set were 0.732 (95% CI: 0.727-0.738) in men, and 0.801 (0.794-0.808) in women. The Framingham score performed less well in men (0.722; 0.717-0.727) and in women (0.705; 0.699-0.711).
Conclusions: Based on these preliminary results, the CAMUNI 20-year risk score seems to be appropriate for long-term risk prediction in Italy and, more generally, in low-incidence populations.
Author Disclosures: G. Veronesi: None. C. Donfrancesco: None. L.E. Chambless: None. F. Gianfagna: None. L. Palmieri: None. G. Mancia: None. G. Cesana: None. S. Giampaoli: None. M.M. Ferrario: None.
- © 2014 by American Heart Association, Inc.