Introduction Infection with HIV-1 results in progressive immunodeficiency and death from opportunistic infection or cancer. Current antiretroviral therapy is effective in suppressing plasma viremia to levels below the detection limit of FDA-approved assays (50–75 copies HIV-1 RNA/ml), restoring immune function, and reducing morbidity and mortality. Antiretroviral therapy does not cure HIV-1 infection, however, and, at a minimum, eradication of HIV-1 infection will require complete suppression of its replication as well as elimination of latent viral reservoirs. Although some reports have shown the persistence of low-level viremia in patients on suppressive therapy [1–6], the determinants of this viremia, and whether it results from ongoing replication cycles or release from latent reservoirs, are not well defined [2,7–9] due in part to the limited sensitivity of prior HIV-1 RNA assays. To investigate these issues, we developed a real-time PCR-based method (single-copy assay, SCA) capable of detecting and reliably quantifying HIV-1 RNA with a limit of one copy per ml plasma [10]. Using this assay, we have found that more than 80% of patients on currently recommended antiretroviral therapies have quantifiable viremia for at least 2 y after initiation of therapy. These results have important implications for understanding the mechanism of HIV-1 persistence despite long-term antiretroviral therapy. Results/Discussion We first measured plasma HIV-1 RNA with both an FDA-approved assay (bDNA; detection limit 75 copies/ml) and SCA in three patients initiating antiretroviral therapy. As expected [11–13], therapy produced a rapid decline in plasma HIV-1 RNA (Figure 1), reaching undetectable levels within 50–260 d. HIV-1 RNA values from the two assays were similar at levels that were detectable by both assays, but the SCA continued to detect HIV-1 RNA throughout the sampling period, well below the limit of detection of the bDNA assay. These initial observations suggested that low-level viremia can persist for years in patients receiving suppressive antiretroviral therapy. To investigate this phenomenon in a larger patient population and to compare the effects of different treatment regimens, we analyzed specimens from study M98–863, a Phase III randomized clinical trial comparing lopinavir/ritonavir (LPV/r) and nelfinavir (NFV), each in combination with stavudine and lamivudine, in previously antiretroviral-naïve HIV-1-infected individuals [14]. We studied a subset of 145 patients (see Figure S1) whose plasma HIV-1 RNA declined to less than 50 copies/ml within 24 wk of initiating therapy and remained at this level at all time points through 60 wk. SCA results were available from 130 patients (63 on LPV/r and 67 on NFV). As shown in Figure 2, HIV-1 RNA values at week 60 in the two arms combined ranged from 250 wk (mean 111.2 wk) prior to SCA sampling. All patients were well and without history of recent intercurrent illness at the time of phlebotomy, had hemoglobin levels ≥12 g/dl, and provided informed consent for phlebotomy and for research sample storage. HIV-1 RNA determination. The SCA for HIV-1 RNA detection was performed as described previously [10], starting with 7 ml plasma, except for the M98–863 samples, from which only 3 ml plasma was available. As a result, the lower limit of quantitation for these samples was 0.6 copies HIV-1 RNA/ml, compared to 0.3 copies/ml when 7 ml plasma was used. In all analyses, values below the assay quantitation limit for study M98–863 (0.6 copies/mL) were considered to be equal to the assay limit. Plasma was obtained from whole blood samples within 2–4 h of phlebotomy and immediately frozen at −70 °C. [10,31]. For each sample, three separate aliquots of the cDNA product were assayed for HIV-1 RNA and two aliquots for the recombinant avian retrovirus internal standard RNA using real-time PCR of conserved sequences within gag as described [10]. About 10% of M98–863 patients were excluded because SCA and Amplicor assays on pretherapy samples were significantly discordant due to inefficient amplification by SCA (Figure S2), probably a result of polymorphism in the probe or primer sequences (A. Wiegand and S. Palmer, unpublished data). In all other samples, there was a close correlation between commercial Amplicor RT-PCR and SCA values for pretherapy samples (r 2 = 0.61, Figure S2). Further details of extraction, optimum amplification conditions, and performance characteristics, as well as quality control procedures to prevent artifactual amplification, have been described [10]. Study M98–863 used the PCR-based Amplicor HIV-1 MONITOR assay version 1.0 for plasma HIV-1 RNA quantitation, performed according to manufacturer's specifications (Roche Diagnostics, http://www.roche.com). For patients taking NNRTI-containing regimens, plasma HIV-1 RNA was quantified using the bDNA-based VERSANT HIV-1 RNA assay version 3.0 according to manufacturer's specifications (Bayer Diagnostics, http://diagnostics.siemens.com) in a manufacturer-certified site using the semiautomated Bayer System 340 unit. Versant HIV-1 RNA version 3.0 is approved by the FDA for clinical use with a limit of quantification of 75 copies HIV-1 RNA/ml plasma, although our reported experience with this assay indicates a quantification limit of 50 copies/ml [32,33]. Statistical analysis. For log-transformed baseline viral RNA determinations, comparisons between Amplicor HIV-1 RNA assays and SCA were conducted using linear regression and Pearson's correlation, as were comparisons between log-transformed viral RNA determinations at baseline and week 60. Mean SCA values were compared between groups using a one-way analysis of variance; comparisons of distributions of SCA values were conducted using the Kolmogorov-Smirnov test. A linear mixed effects regression model with a spatial linear correlation structure to account for correlation between repeated measurements within participants was used to assess the relationship between time and log-transformed viral determinations using SCA. In all analyses, values below the assay quantitation limit for study M98–863 (0.6 copies/mL) were considered to be equal to the assay limit. Sensitivity analyses using other imputation methods did not alter results meaningfully. Supporting Information Figure S1 Schematic Describing Patient Inclusion for Study of HIV-1 Viremia in M98–863 Patients (21 KB PDF) Click here for additional data file. Figure S2 Correlation between Pretherapy Plasma HIV-1 RNA Levels as Determined by Amplicor HIV-1 Monitor Assay and SCA Pretherapy samples from 157 participants were analyzed by both assays. 12 samples (triangles) had SCA values significantly lower than Amplicor and were excluded from further analysis. Remaining samples (n = 145, diamonds) were highly correlated (r2 = 0.61, broken lines represent ± 0.5 log10 copies/mL from Y = X line); corresponding participants were included in subsequent analyses. (32 KB PDF) Click here for additional data file.