Abstract
Discrepancies between HIV-1 RNA results assayed by different FDA-approved platforms have been reported. Plasma samples collected from 332 randomly selected clinical trial participants during the second year of antiretroviral treatment were assayed with three FDA-approved platforms: UltraSensitive Roche Amplicor Monitor, v1.5 (Monitor), the Abbott RealTime HIV-1 test on the m2000 system (Abbott), and the Roche TaqMan HIV-1 test, v2.0 (TaqMan). Samples from 61 additional participants with confirmed HIV-1 RNA levels of >50 copies/ml during trial follow-up were also included. Endpoints were HIV-1 RNA quantification of ≤50 copies/ml versus >50 copies/ml at an individual-sample level (primary) and determination of confirmed virologic failure (VF) from longitudinal samples. A total of 389 participants had results obtained from all assays on at least one sample (median = 6). Proportions of results of >50 copies/ml were 19% (Monitor), 22% (TaqMan), and 25% (Abbott). Despite indication of strong agreement (Cohen's kappa, 0.76 to 0.82), Abbott was more likely to detect HIV-1 RNA levels of >50 copies/ml than Monitor (matched-pair odds ratio [mOR] = 4.2; modified Obuchowski P < 0.001) and TaqMan (mOR = 2.1; P < 0.001); TaqMan was more likely than Monitor (mOR = 2.6; P < 0.001). Despite strong agreement in classifying VF across assay comparisons (kappa, 0.75 to 0.92), at a 50-copies/ml threshold, differences in the probability of VF classification (in the same direction as primary) were apparent (all McNemar's P < 0.007). At a 200-copies/ml VF threshold, no differences between assays were apparent (all P > 0.13). Despite strong agreement among assays, significant differences were observed with respect to detecting HIV-1 RNA levels of >50 copies/ml and identifying VF at the 50-copies/ml threshold. This has important implications for the definition of VF in clinical trials and clinical practice.