Diurnal changes in human plasma levels of the fat-soluble vitamins retinol (R), alpha-tocopherol (E), and beta-carotene (CAR) have not been determined. Plasma levels of these three vitamins in 15 healthy volunteers were measured five times over a 24-hour period. Highly sensitive and specific HPLC assays were used. Mean +/- SEM levels for R, E, and CAR were 591 +/- 36 ng/mL, 10.3 +/- 0.6 micrograms/mL, and 271 +/- 28 ng/mL, respectively. Differences between subjects' mean levels were large and highly significant (p less than 0.0001). Relative to the first 8 AM fasting plasma sample, nonfasting plasma levels drawn over the next 24 hours showed no statistically significant or clinically important changes. The percent increases in standard deviation among subjects due to diurnal variation were 1.5%, 3.3%, and 2.3%, respectively, for R, E, and CAR. This implies that sample sizes for cross-sectional studies investigating differences between two treatments in these vitamin plasma levels would need to be increased by 3%, 7%, and 5%, respectively, if measurements are to be taken throughout the day. Plasma levels may be obtained at random during the day, and need not be 8 AM fasting levels. The small addition in variation introduced by diurnal fluctuations would have minimal impact upon the precision of estimates of treatment effect and sample sizes.