Concordance between molecular assays may be suboptimal at low HIV-1 viremia levels (<1,000 copies/ml); therefore, it may be difficult to define and compare virologic endpoints for successful and failed therapy. We compared two commercial assays (the Abbott RealTime HIV-1 and the Roche Cobas AmpliPrep/TaqMan HIV-1 version 2.0) for their ability to detect and quantify low viral loads. A comparison was performed using 167 residual clinical samples (with values ranging from "not detected" to 1,000 copies/ml, as measured by the Abbott assay) and the National Institute and Biological Standards and Control (NIBSC) HIV-1 RNA working reagent 1 for nucleic acid amplification techniques (NAT) assays (serially diluted to a range from 1 to 1,000 copies/ ml). Quantitative results were compared using Lin's concordance correlation coefficient and a Bland-Altman plot. Concordance with the qualitative results was measured by Cohen's kappa statistic. With clinical samples, the degree of interassay concordance of the qualitative results at a 40-copies/ml HIV-1 RNA threshold was substantial ( ؍ 0.762); the correlation among the quantified samples was suboptimal (concordance correlation coefficient, 0.728; P < 0.0001); the mean difference of the values between the Roche and Abbott assays was 0.193 log 10 copies/ml. Using the HIV-1 RNA working reagent 1 for NAT assays, the results provided by the Roche assay were, on average, 3 times higher than expected, while the Abbott assay showed high accuracy. The Roche assay was highly sensitive, being able to detect a level as low as 3.5 copies/ml HIV-1 RNA with 95% probability. The performance characteristics of each molecular assay should be taken into account when HIV-1 RNA threshold values for "virologic suppression," "virologic failure," "persistent low viral loads," etc., are defined and indicated in the support of clinical decisions.