Taken together, the results from these studies clearly indicate that alcohol and cigarette smoking may both decrease serum or plasma beta-carotene levels after controlling for dietary intake and lipid profile. The variation among studies in the magnitude of the association and statistical significance may reflect, in part, the joint distribution of smoking and alcohol intake in the populations. For example, in the supplementation study reported by Costantino, 89% consumed alcohol, considerably higher than population averages. Further, one-third of the men in that study consumed more than 65 ml (52 gm) of pure alcohol per day. This contrasts with the data from the Health Professionals Follow-up Study, where the mean intake of alcohol was 12.3 gm/day with a standard deviation of 15.8. Even lower alcohol intakes are reported in women, further attenuating the relation and the statistical power to detect a relation between alcohol and beta-carotene levels. Measurement error in assessing dietary intake of beta-carotene could explain some of the association between smoking, alcohol and serum beta-carotene levels because of incomplete control (residual confounding) of dietary intake. Methods for statistically adjusting for error in measurement have not yet been implemented in analyses of this nature.