A population sample of 1,645 men aged 40-59 at entry examination has been followed up for 25 years. When hypertensives (160 and/or 95 mm Hg or more; n = 432) were divided into fat and lean on the basis of body mass index (cut off at 24.9), the latter group showed a nonsignificant higher risk of dying during the subsequent 25 years (relative risk = 1.15). Univariate and multivariate analyses suggest that among lean people there is an excess of smokers partly explaining the excess risk of lean hypertensives. Both univariate and multivariate analyses also proved that the relationship of fatness-leanness indicators to total mortality is U-shaped and applies to hypertensives as well as to nonhypertensives, and to the whole population. Evidence has been given that changing the definition of fatness-leanness (cutoff of body mass index of 27.9) the relative risk of lean versus fat hypertensives may reverse (0.93). This suggests that a large part of the problem is artificial and should not be tackled without considering the parabolic relationships of indicators of obesity to mortality.