Two important changes occurred in the time between the Third National Health and Nutrition Examination Survey (NHANES III) (1991-1994) and the later survey (NHANES 1999+) regarding total homocysteine (tHcy), i.e., a change in matrix from serum to plasma and a change in analytical methods. The goals of this study were to determine the magnitude of potential differences between plasma and serum with regard to tHcy concentrations, and between the two analytical methods used in these surveys. Optimally prepared plasma, serum allowed to clot for 30 and 60 min at room temperature and serum allowed to clot for 30 and 60 min and subjected to four freeze-thaw cycles, prepared from blood samples collected from 30 healthy people, were analyzed by both methods. Serum samples had significantly higher tHcy concentrations than plasma samples, and the difference increased with longer clotting time. Freeze-thaw cycles had little or no effect on the variability or bias in the serum sample results. The tHcy results produced by the two analytical methods were significantly different, but consistent across sample types. On average, the results of the method used in NHANES III were lower by 0.64 micromol/L; however, the relative bias varied with tHcy concentration. The tHcy results determined in surplus serum from NHANES III overestimated tHcy concentrations by approximately 10% compared with optimally prepared plasma. The average method bias was 6% between the two analytical methods. On the basis of changes in matrix and methodology, direct comparison of tHcy results between the two surveys is inappropriate.