Abstract Objective To derive and validate a set of functions to predict coronary heart disease (CHD) and stroke, and validate the Framingham-REGICOR function. Method Pooled analysis of 11 population-based Spanish cohorts (1992–2005) with 50,408 eligible participants. Baseline smoking, diabetes, systolic blood pressure (SBP), lipid profile, and body mass index were recorded. A ten-year follow-up included re-examinations/telephone contact and cross-linkage with mortality registries. For each sex, two models were fitted for CHD, stroke, and both end-points combined: model A was adjusted for age, smoking, and body mass index and model B for age, smoking, diabetes, SBP, total and HDL cholesterol, and for hypertension treatment by SBP, and age by smoking and by SBP interactions. Results The 9.3-year median follow-up accumulated 2973 cardiovascular events. The C-statistic improved from model A to model B for CHD (0.66 to 0.71 for men; 0.70 to 0.74 for women) and the combined CHD-stroke end-points (0.68 to 0.71; 0.72 to 0.75, respectively), but not for stroke alone. Framingham-REGICOR had similar C-statistics but overestimated CHD risk. Conclusions The new functions accurately estimate 10-year stroke and CHD risk in the adult population of a typical southern European country. The Framingham-REGICOR function provided similar CHD prediction but overestimated risk.