Components of biological variation can be used to assess the usefulness of reference values, to evaluate the significance of changes in serial results from an individual and to define objective analytical goals. The aim of the study was to assess, in 15 healthy subjects studied at regular monthly intervals over a period of 6 consecutive months, the biological variation of interleukin-1beta (IL-1beta), interleukin-8 (IL-8), and tumor necrosis factor-alpha (TNF-alpha). Biological variation data (within-subject and between-subject coefficient of variation (CV)) were determined using a simple nested analysis of variance. Derived parameters (index of individuality, reliability coefficient and critical diferences) were calculated from within-subject and between-subject CV. The mean and standard deviation (SD), within-subject CV, between-subject CV, index of individuality and reliability coefficient were as follows: for IL-1beta, 0.67 (0.32) pg/ml, 30%, 36%, 0.85, and 0.76; for IL-8, 3.68 (1.45) pg/ml, 24%, 31%, 0.85 and 0.75; and for TNF-alpha, 3.14 (1.87) pg/ml, 43%, 29%, 1.56 and 0.50, respectively. We conclude that between-subject variation and within-subject variation are quite similar for IL-1beta and IL-8 and are relatively high for the three cytokines studied. Index of individuality is less than 1.4 for IL-1beta and IL-8, and thus reference intervals based on population studies are of limited value. On the contrary, the index of individuality for TNF-alpha is greater than 1.4 and reference values can be used for diagnosis. Quality goals for imprecision are easily achieved for the three cytokines with current methodology.