The classification of arterial hypertension (HT) to define metabolic syndrome (MS) is unclear in that different cutoffs of blood pressure (BP) have been proposed. We evaluated the categorization of HT most qualified to define MS in relationship with coronary heart disease (CHD) mortality at a population level. A total of 3257 subjects aged > or =65 years were followed up for 12 years. MS was defined according to the criteria of the National Education Cholesterol Program using three different categories of HT: MS-1 (systolic blood pressure (SBP) > or =130 and diastolic blood pressure (DBP) > or =85 mm Hg), MS-2 (SBP > or =130 or DBP > or =85 mm Hg) and MS-3 (pulse pressure (PP) > or =75 mm Hg in men and > or =80 mm Hg in women). Gender-specific adjusted hazard ratio (HR) with 95% confidence intervals (CI) for CHD mortality was derived from Cox analysis in the three MS groups, both including and excluding antihypertensive treatment. In women with MS untreated for HT, the risk of CHD mortality was always significantly higher than in those without MS, independent of categorization; the HR of MS was 1.73 (CI 1.12-2.67) using MS-1, 1.75 (CI 1.10-2.83) using MS-2 and 2.39 (CI 3.71-1.31) using MS-3. In women with MS treated for HT, the HR of CHD mortality was significantly increased only in the MS-3 group (1.92, CI 1.1-2.88). MS did not predict CHD in men. In conclusion, MS can predict CHD mortality in elderly women with untreated HT but not in those with treated HT; in the latter, PP is the most predictive BP value.