Various studies have reported an inverse association between serum albumin level and incident coronary heart disease (CHD), though biologic mechanisms have not been established. The authors examined the association between serum albumin level and CHD in the Atherosclerosis Risk in Communities cohort, comprising 14,506 White and African-American middle-aged men and women. The mean albumin level in this population was 3.9 g/dl (standard deviation 0.3). During 5.2 years of follow-up, 470 incident CHD events occurred. The hazard ratio for incident CHD associated with a 1-standard deviation decrease in serum albumin level was 1.26 (95% confidence interval (CI): 1.15, 1.38) after adjustment for age, gender, and ethnicity and 1.18 (95% CI: 1.07, 1.30) after additional adjustment for covariates related to CHD. Hazard ratios were similar across gender and ethnic groups. However, there was statistically significant effect modification by smoking status, with hazard ratios of 1.01 (95% CI: 0.84, 1.22) among never smokers, 1.09 (95% CI: 0.92, 1.30) among former smokers, and 1.35 (95% CI: 1.17, 1.54) among current smokers. Further adjustment for factors related to renal disease, nutrition, platelet aggregation, inflammation, use of angiotensin-converting enzyme inhibitors, and hemostasis factors attenuated the albumin-CHD relation only slightly. In this study, serum albumin was inversely associated with incident CHD at the baseline examination in current smokers but not in never or former smokers. Albumin level may be a marker of susceptibility to the inflammatory response that results from smoking.