In a prospective study of 7735 middle-aged British men (British Regional Heart Study) 660 died during an average follow-up period of 9.2 years. There was a marked increase in mortality rate with decreasing serum albumin concentrations that persisted even after adjustment for age, social class, town of residence, cigarette smoking, serum total cholesterol, systolic blood pressure, serum total calcium, and forced expiratory volume in 1 s. When serum albumin concentration was less than 40 g/l, the mortality rate was 23/1000 per year compared with 4/1000 per year for a concentration equal to or above 48 g/l. A similar pattern was observed for cardiovascular, cancer, and other deaths. The associations persisted for cardiovascular disease and cancer even when deaths within the first five years of follow-up were excluded. These results must be interpreted with caution since there was no prior hypothesis concerning serum albumin. However, the strength of the association between serum albumin and mortality seems to be comparable with that for cigarette smoking.