The aim of this longitudinal study was to investigate 3-year change in serum albumin concentration as a determinant of incident cardiovascular disease (CVD) and all-cause mortality. Data were from 713 respondents of the Longitudinal Aging Study Amsterdam initially aged 55-85 years. Serum albumin was measured at baseline (1992/1993) and after 3 years. At the 6-year follow-up, incident CVD (among 456 respondents with no prevalent CVD at the 3-year follow-up) and all-cause mortality were ascertained. Overall, 18.9% developed CVD and 10.9% died. After adjustment for potential confounders, a higher level of serum albumin at the 3-year follow-up was associated with a lower risk for incident CVD (relative risk = 0.88, 95% confidence interval (CI): 0.79, 0.98). The risk of incident CVD was 0.88 (95% CI: 0.78, 0.99) per unit (g/liter) increase in change in albumin between 3-year follow-up and baseline. Chronic low serum albumin (<or=43 g/liter at baseline and 3-year follow-up) was not associated with incident CVD (p = 0.22). A clinically relevant decrease in serum albumin (>or=1 standard deviation (2.5 g/liter) between baseline and 3-year follow-up) tended to be associated with a twofold risk (relative risk = 2.00, 95% CI: 0.91, 4.39). For all-cause mortality, no associations were observed. These findings suggest that older persons with a decrease in serum albumin concentration, even within the normal range, might be at increased risk of incident CVD. Change in serum albumin may be used as an early marker for CVD risk.