Objectives:To assess the specific contributions of assay variation and biological variation to the total variation of plasma HIV-1 RNA measured by the Roche Monitor assay and the extent to which batch assays reduced both assay variability and total variability compared with real-time determinations.Design:A retrospective analysis of data obtained from three trials conducted by the Adult and Pediatric AIDS Clinical Trials Groups (ATCG), the Women and Infants Transmission Study (WITS) and the NIAID-sponsored Virology Quality Assurance Program.Methods:Within-subject variation was assessed from stored, serially collected plasma samples from 663 subjects enrolled in the ACTG and WITS studies. Interassay and intra-assay variation were estimated from two of the clinical trials and 22 laboratories that participated in a quality assurance program and were used to estimate the effect of real-time testing on total variation.Results:The total variation (standard deviation) from a random effects model was 0.26 log10 RNA copies/ml. The estimated interassay variation was 0.08 log10 and intra-assay variation was 0.12 log10 RNA copies/ml. Biological variation accounted for 56-80% of total variation. The effect of real-time testing compared with batch testing was minimal.Conclusion:Our estimates of total within-subject HIV-1 RNA variation support the current recommendation to obtain at least two specimens, preferably obtained less than 2 weeks apart, for viral RNA measurement before starting therapy. The major contribution of biological variation to the total variation supports the use of real-time HIV-1 RNA assays, provided that consistent specimen collection procedures are followed and acceptable assay proficiency is maintained.