The development of antiviral therapy [highly active antiretroviral therapy (HAART)] for HIV infection represents one of the most remarkable accomplishments in medical history. The goal of HAART is to reduce viral replication to below the limit of detection of standard clinical essays. This limit (50 copies/mL) has firstly been chosen because of technical limits and subsequently validated in many clinical trials and cohort analysis.1–3 The same cut-off value is nowadays recommended by all national and international guidelines.4–6 In patients achieving this goal, clinically significant viral evolution apparently stops, CD4+ cells counts generally increase, and clinical progression of the infection is stopped or highly reduced. However, despite the tremendous success of HAART, major uncertainties still exist. For example, it is not completely clear whether the virus continues a low level replication in patients who respond well to therapy. Enhanced assays for HIV RNA quantitative evaluation with much lower detection limits (1–3 copies/mL) have shown that low-level viremia (LLV) is measurable in a vast proportion of subjects with steady HIV RNA levels <50 copies per milliliter.7,8 The clinical consequences of LLV have still to be clarified. We conducted a large prospective cohort study addressing these topics.
PATIENTS AND METHODS
This was a prospective, single centre, cohort study in asymptomatic HIV-1–infected adults who had achieved suppression of viremia to <50 copies per milliliter (HIV RNA bDNA Versant, Siemens Healthcare Diagnostic, Tarrytown, NY) on a stable HAART regimen for at least 4 months and who had a further HIV RNA test confirming their virologic-suppression status (4 months post a standard HIV RNA test) performed with an ultrasensitive HIV RNA assay with a detection limit of 3 copies per milliliter. The study started in May 2009 after the ultrasensitive HIV RNA test was made available for routine testing in our Centre. The cohort was specifically designed and powered to assess the virologic outcome in patients with a viral load (VL) <3 copies per milliliter compared with those presenting a LLV comprised between 3 and 50 copies per milliliter.
The study protocol was approved by the Ethical Committee of Ospedali Riuniti di Bergamo Hospital, and all patients provided their written informed consent.
Age, gender, nationality, risk factor for HIV infection, HIV RNA blood level before HAART, number and type of previous regimens, total duration of HAART exposure, current HAART regimen and duration, time from the last HIV-RNA level above 50 copies per milliliter, CD4+ cell count per microliter were determined at the time of study entry. Clinical, immunological, and virological assessments were performed thereafter every 4 months (as indicated by Italian guidelines for the management of HIV infection).9 The follow-up period was 12 months.
CD4+ cell counts were determined in whole blood by 3-colour flow cytometry, using standard methods.
Plasma HIV RNA levels were determined with an ultrasensitive HIV RNA test based on a real-time polymerase chain reaction (PCR) 10(HIV RNA 1 1.0 assay kPCR, Siemens Healthcare Diagnostic). The kPCR assay has a range of linearity between 3.7 × 101 and 1.1 × 107 copies per milliliter.11 We performed an internal validation to permit a nominal detection of HIV RNA in the lower part of the standard curve (<50 copies/mL). Plasma from a HIV-infected patient was obtained, the HIV RNA content was determined by both HIV RNA bDNA Versant and HIV-RNA 1 1.0 assay kPCR; and plasma was serially 2-fold diluted in control plasma from healthy HIV-negative donors to obtain RNA concentration between 50 and 0.7 copies per milliliter. Twelve replicate samples (700 μL of plasma) of each dilution were assayed according to manufacturer's indication by the kPCR assay. A quasilinear curve was observed from the original concentration down to 3 copies per milliliter (R2 0.901; adjusted R2 0.899, SE 4.1; P < 0.0001 Anova). The limit of detection of 3 copies per milliliter was set as, at this dilution, the recovery rate was 0.83 with a difference from expected of 0.16 (95% confidence interval (CI) from 0.05 to 0.38) and a z score (compared with the expected result) of 0.074 for a P value of 0.459. At the same dilution, the mean HIV RNA level was 4.3 copies per milliliter (95% CI from 2.7 to 5.8). Viral RNA was inconsistently detected in dilutions containing less than 3 copies per milliliter.
Any plasma level >50 copies per milliliter was controlled with a further test performed at least 1 month apart. A genotypic test was performed in the case of confirmed positive viremia. Genotypic test were performed by means of Abbott diagnostic HIV-1 Viroseq version 2 assay (Abbott, Maidenhead, United Kingdom). It is an Food and Drug Administration approved, commercially available assay which has been designed for sequencing of HIV-1 from plasma samples containing at least 1000 copies of HIV-1 RNA per milliliter. We followed indications by Mackie et al12 to analyze sequences in HIV-1 pol in subjects with a low copy number (from 50 to 1000). Briefly, after extraction, RNA is subject to reverse transcriptase–polymerase chain reaction using the primer from the ABI Viroseq HIV-1 genotyping system, version 2.0. Five microliters of product from this first round of PCR are subjected to a second round of PCR using sequencing primers A and H from the Viroseq version 2.0 sequencing module (Applied biosystems, Warrington, United Kingdom). Amplicons generated from the nested PCR reaction are sequenced by PCR amplification with primers A-C, F-H provided in the Viroseq kit. The post-PCR sequencing product is dried and resuspended in template sequencing reagent before separation on an ABI 3130 xl genetic analyzer. The sequence product is analysed by Viroseq version 2.7 HIV-1 analysis software and sequences are subtyped using the Stanford University HIV-1 resistance database.
All data of the present study were collected in an electronic database.
The primary end point was to determine the predictive value of a VL <50 copies per milliliter but still measurable (LLV) on the occurrence of virological failure in the following 4 months. For the purposes of this study, 3 definitions of virologic failure were used as follows:
“broad”, reflecting the definition commonly used in clinical trials, that is 2 consecutive plasma HIV-RNA levels >50 copies per milliliter. The confirmatory test was performed at least 1 month after the first elevated measure;
“restricted”, limited to the current DHHS guidelines 5 definition, that is, 2 consecutive measures of HIV RNA >200 copies per milliliter. In this case, the confirmatory test was performed at least 1 month after the first elevated measure, too;
“comprehensive”, including all patients of the “broad” definition plus all patients lost to follow-up or dead.
The predictive value was calculated for each HIV RNA measure/follow-up period unit that is a measure followed by a 4-month period of observation.
The “broad and restricted” analyses were performed considering the whole follow-up period, too. In this case, being HIV RNA a measure subject to variation over time, patients had to be grouped into classes. A first distinction was made considering the baseline VL and patients were grouped into 2 classes: those presenting a VL <3 copies per milliliter and those with a VL between 3 and 50 copies per milliliter. The second grouping definition was based on the variation of HIV RNA level over time in the single patients. Three classes were derived as follows: patients with a steadily controlled viremia with HIV RNA values <3 copies per milliliter at all sampling times; subjects presenting variable values having some of the samples <3 copies per milliliter and others with detectable LLV (<50 copies/mL, but >3 copies/mL); and finally individuals with steadily detectable LLV between 3 and 50 copies per milliliter.
Secondary endpoints of the study were to verify if any recorded individual parameter including demographics and pretreatment virological and immunological variables could influence VL dynamics during the study follow-up or if these variation and the presence of HIV RNA levels <3 copies per milliliter could be associated with any antiretroviral drug combination.
The study power was calculated to assess superiority in the group of patients with a VL <3 copies per milliliter compared with those presenting a plasma HIV RNA between 3 and 50 copies per milliliter under the assumption that, in the former group (HIV RNA <3 copies/mL), the primary end-point (broad definition) would be observed in a proportion of subjects <5% and that the odds ratio (OR) for patients with higher VL would be greater than 3. According to these assumptions, a sample size of 1200 patients would assure a power (1-beta) of 95% at a significance level of 0.01 (alpha = 1%).
Descriptive results are presented as means ± standard deviation and percentages with 95% CIs. Inferential statistics using either parametric or nonparametric tests were used, as appropriate for the data type. The χ2 was used to analyze all categorical variables. Analysis of Variance test and Student t test were used to analyze continuous variables. Binary and multinomial logistic regressions were used to analyze the relationship among individual parameters and the magnitude of viral suppression.
All tests were 2-sided, and a P value inferior to 0.05 was regarded as significant. All analyses were performed with the SPSS statistical software package for Windows, version 17.0.
Baseline Characteristics and Patients' Disposition
A total of 1214 patients of whom 287 females (23.6%) were enrolled. Eight patients died during the follow-up period due to causes not related to HIV and steadily presenting an HIV RNA <3 copies per milliliter. Causes of death were end-stage liver disease and hepatocarcinoma (2 cases each), carcinoma of the kidney, uterus, and pancreas (1 case each), and myelodysplasia (1 case).
Twenty-five patients were lost to follow-up. Deaths and patients lost to follow-up were included in the “comprehensive” analysis of the risk of virological failure (and counted as failures), but excluded from all other analysis performed, therefore, on 1181 patients.
At baseline, the mean age was 45.5 years (SD ± 8). As expected, most patients were Italian (1106; 91.1%), whereas subjects from sub-Saharan Africa counted for 6.3% of cases and patients from other European Countries or South America were roughly 1% each. The most common risk factor for HIV acquisition was heterosexual sex (45.6%), followed by intravenous drug use (36.7%) and homosexual contacts (15.1%). Other risk factors counted for 2.6%. The mean pre-HAART HIV RNA level was 5.18 log copies per milliliter (SD ± 5.24). Overall, 273 patients (22.5%) were on their first HAART regimen, whereas the remaining patients had previous treatment experiences although only 3.8% of them were heavily experienced (≥10 HAART regimens). On average, they were exposed to 3.7 different regimens (SD ± 3).
The average exposure to antiretroviral drugs was 95.7 months (SD ± 46), whereas the mean time spent on their actual regimen was 38.6 months (SD ± 29). The mean time elapsing between the last HIV RNA measure >50 copies per milliliter and the enrolment into the study was 54.1 months (SD ± 37). Most patients were taking a nonnucleoside reverse transcriptase inhibitor (NNRTI)–based HAART (666; 54.8%). Protease inhibitor (PI)–based HAART regimens counted for a further 37.1%, whereas therapeutic combinations containing drugs of other classes counted for the remaining 8.1%. Most patients on a PI-based HAART (87.4%) received a ritonavir boosting too. As far as the HAART backbone is concerned, beside lamivudine/emtricitabine that were taken by most patients, tenofovir (43.9%), abacavir (18.3%), and zidovudine (17.5%) were the most commonly used drugs. Most therapies were kept constant throughout the whole study period. Eighteen patients changed therapy because of virologic failure and 13 because of adverse events or according to a simplification strategy. At enrollment, the mean CD4+ count was 611 cells per microliter (SD ± 273).
Plasma HIV RNA Variations
The mean follow-up was 378 days (SD ± 69). Over time, the proportion of patients with a plasma HIV RNA <3 copies per milliliter constantly resulted above 69% and was rather similar at each time point. More in detail, at baseline, 71.5% of subjects were below the limit of detection; after 4 months, the proportion was 69.0%; and after 8 and 12 months, 71.7 and 72.9%, respectively. However, only 480 patients (40.6%) showed a steadily controlled viremia with HIV RNA values <3 copies per milliliter at all sampling times; 628 patients (53.2%) presented variable values having some of the samples <3 copies per milliliter; and others with detectable LLV (<50 copies ml, but >3 copies/mL). The remaining 73 subjects (6.2%) presented steadily detectable LLV between 3 and 50 copies per milliliter. That means that a hypothetical patient with a measure of HIV RNA <3 copies per milliliter had an 80% chance to remain below the detection limit of the test at the subsequent time point. His risk of showing a low-level detectable viremia after 4 months was 18.4% and the chance of a VL >50 copies per milliliter was as low as 1.6%. On the other hand, a subject with a detectable low-level HIV RNA test had a chance of 50.6%, after 4 months, to result below the detection limit, but a risk of still presenting LLV or a HIV RNA level >50 copies per milliliter of 41.7 and 7.6%, respectively.
In total, 70 patients (5.8%) presented, during the follow-up period, an unconfirmed viral blip >50 copies per milliliter. A single patient had 2 episodes.
According to the “broad” definition of virologic failure, we observed 43 events (3.6% of patients); when a more restrictive limit to define virologic failure (200 copies/mL) was considered, the risk lowered to 2.6% (31 events). Considering patients lost to follow-up and deaths as failures raised the risk to 6.2% (76 patients).
Predictive Value of LLV
We obtained 3562 HIV RNA/follow-up units. According to the “broad” definition of virologic failure (confirmed VL >50 copies/mL), the risk of failing the current regimen in the following 4 months for patients with an HIV RNA <3 copies per milliliter was 0.4% compared with a 3.2% risk for those with any value of LLV (P < 0.0001; OR: 7.52, 95% CI from 3.8 to 15.0). There was a significant (P < 0.0001) linear relationship between the HIV RNA level and the concurrent risk of virologic failure (Fig. 1). According to the receiver operating curve (ROC) analysis (Fig. 2), the presence of LLV significantly (P < 0.0001) predicted the risk of failure as showed by an area under the curve (AUC) of 0.764 (95% CI from 0.68 to 0.84). Similarly, when the more conservative limit of 200 copies per milliliter was used to define virologic failure, patients with a HIV RNA <3 copies per milliliter showed a risk of 0.4% compared with a 2.0% risk for those with any value of LLV (P < 0.0001; OR: 4.64, 95% CI from 2.2 to 9.7). In this case, the ROC AUC was 0.71 (95% CI from 0.61 to 0.82) (P < 0.0001). Finally, including patients lost to follow-up and deaths as failures, we obtained a risk of failure of 0.9% in patients with an HIV RNA <3 copies per milliliter compared with a risk of 3.4% in those with any value of LLV (P < 0.0001; OR: 3.96, 95% CI from 2.1 to 7.6). In this last case, the ROC AUC was slightly lower (0.70; 95% CI from 0.61 to 0.77), but still statistically relevant (P < 0.0001).
Beside the time interval between 2 consecutive HIV RNA determinations, the risk of virological failure was calculated for the whole follow-up period too.
In this case, patients with a steadily undetectable VL had a risk of failure (>50 copies/mL) of 1.2%; those with variable HIV RNA levels showed a risk of 1.9% that increased to 34.2% for patients with steady HIV RNA levels between 3 and 50 copies per milliliter (P < 0.0001). When a “restricted” definition of failure was applied, the same values resulted 1.2%, 1.9%, and 17.8% (P < 0.0001), respectively. Similarly, a striking and significant difference for the risk of failure was observed when the baseline VL was used as grouping variable. According to the “broad” definition of failure, the risk was 1.4% for subjects with HIV RNA <3 copies per milliliter and 9.3% for those with HIV RNA between 3 and 50 copies per milliliter (P < 0.0001). These figures lowered to 1.4% and 5.7% (P < 0.0001) considering the “restricted” definition of failure.
Further, the risk of showing an unconfirmed viral blip was higher in patients with LLV (3.9%) than in those with a HIV RNA below the detection limit (1.1%) (P < 0.0001; OR: 3.56, 95% CI from 2.2 to 5.9). Thirty-seven percent of blips were preceded by a VL <3 copies per milliliter, whereas the remaining 63% were observed in subjects that before the blip episode had a VL between 3 and 50 copies per milliliter. In this case, the mean HIV RNA preceding the blip episode was 19.3 copies per milliliter (SD ± 12).
Factors Influencing LLV and Outcome
Most demographic and epidemiological variables (gender, age, ethnicity, risk factor for HIV, pre-HAART VL, number of previous HAART regimens) did not influence the outcome. However, the total exposure to antiretroviral agents (P = 0.016) and some characteristics of ongoing HAART were significantly associated the risk of virologic failure. Patients treated with a NNRTI-based HAART had lower risk to fail (1.4%), throughout the study period, compared with those receiving a PI-based HAART (6.7%) or a regimen based on other drug classes (3.6%) (P = 0.001). The direct comparison of NNRTIs with boosted PIs or non-boosted PIs showed an increment of the overall risk of failure of 1.4% to 5.3% to 15.8%, respectively (P < 0.0001). Nevertheless, the strongest predictors of the risk virologic failure remained the time elapsing between the last HIV RNA measure >50 copies per milliliter and the start of the study period (time below detection limit of conventional tests)(P = 0.008) and the presence of a current VL <3 copies per milliliter (P = 0.003). Shorter time below detection and higher VLs were indicative of greater risk.
These observations were further confirmed when the extent of viral suppression was considered. Patients treated with a NNRTI-based HAART had a higher chance to maintain a viremia <3 copies per milliliter throughout the study period (45.2%) compared with those receiving a PI-based HAART (32.7%) or a regimen based on other drug classes (35.5%) (P < 0.0001). When NNRTIs were directly compared with boosted PIs or non-boosted PIs, the chance of having a constant viremia <3 copies per milliliter was 45.2%, 33.1%, and 29.8% (P < 0.0001), respectively. Interestingly, the chance to maintain a viremia <3 copies per milliliter throughout the study period was significantly influenced by the pre-HAART VL (P = 0.004) and by the time below detection limit of conventional tests (P < 0.0001) (Fig. 3). Restricting the analysis to those individuals presenting measurable LLV, a shorter period on HAART (P = 0.012), a shorter time below detection limit of conventional tests (P = 0.024), being on a PI-based HAART (P = 0.0005), and presenting a steadily detectable LLV between 3 and 50 copies per milliliter (P < 0.0001) were negative prognostic factors for virological failure.
Because of these results, 2 post hoc subanalysis were performed on selected patients.
The first analysis included all patients who never failed a previous or their current HAART. This group consisted in 528 individuals who either were on the first HAART or did switch to a second-line regimen because of simplification or toxicity. The analysis was performed to verify if previous virological failures could influence the predictive value of the ultrasensitive HIV RNA test. As one might expect, in this group of patient, the overall risk of virological failure was lower than in the comprehensive cohort, however, the predictive value of the HIV RNA test was maintained. Patients with a steady HIV RNA level <3 copies per milliliter did not present any confirmed rebound above 50 copies per milliliter (0%), patients with variable values having some of the samples <3 copies per milliliter and others with detectable LLV (<50 copies ml, but >3 copies/mL) showed a failure risk of 1.4% and finally individuals with steadily detectable LLV between 3 and 50 copies per milliliter a risk of 18.5% (P < 0.0001).
The second analysis was performed to explore the effect of the time below detection limit of conventional tests as variable able to influence the results. Particularly, we wanted to analyze if a survivor bias could exist. To do this, we limited the analysis to those patients who had been suppressed for a longer time before entering the study. Only patients who presented a suppression time greater than the median value of our cohort (44 months) were entered into this analysis. The group consisted of 600 patients. Once again, patients with a steady HIV RNA level <3 copies per milliliter did not present any confirmed rebound above 50 copies per milliliter (0%), patients with variable values having some of the samples <3 copies per milliliter, and others with detectable LLV (<50 copies ml, but >3 copies/mL) showed a failure risk of 1.1% and finally individuals with steadily detectable LLV between 3 and 50 copies per milliliter a risk of 31.3% (P < 0.0001).
In these 2 post hoc analysis, the use of a “restricted” virological end point did not alter the significance of the results (data not shown).
Viral Genotype of Failing Patients
An attempt to genotype all true failing patients was made. Results are reported in Table 1. In 1 case, genotype was not performed because the sample could not be amplified (patient 4570). In most other cases, we observed a wild-type virus or just minor mutations were detected (30/43 patients). Patient 11,039 relapsed with a virus already genotyped in a previous occasion and with mutations not related to his ongoing therapy. Nevertheless, 13 patients (30.2% of cases) developed mutations able to alter the efficacy of the current HAART and that could, to different extents, reduce their future therapeutic options. Of note, all patients with viral rebound above 10,000 copies per milliliter presented a wild-type virus.
Current guidelines state that the goal of HAART is to suppress and maintain HIV VL below 50 copies per milliliter,4–6 although recent studies have explored the effect of drugs using much lower cutoffs.13
It has been shown that consistently measurable HIV RNA levels, even if low (<1000 copies/mL) may negatively influence the clinical progression of HIV infection. For example, data from the SMART study14 showed that patients with HIV RNA levels ≥400 copies per milliliter were more than twice as likely to develop a clinical event than those with a VL <400 copies per milliliter. Similarly, data from the Multicenter AIDS Cohort study15 showed that the risk of disease progression or death was increased in patients with HIV RNA ranging from 501 to 3000 copies per milliliter compared with those patients having a VL <500 copies per milliliter. These studies used different tests to quantitatively determine VL, included patients that could present differences from ours and had a longer follow-up, but nevertheless invariably strongly indicated the need to set a stringent cutoff to define virological failure. Further, HIV genetic evolution has been observed with RNA levels above 6.5 copies per milliliter.16
Based on this evidence, the endpoint of this study was defined as the rise of HIV RNA above the 50 copies per milliliter threshold. To rule out possible isolated viral blips,17 the measure had to be confirmed in 2 separate tests so to indicate persistence of the measurable viremia.
Generally speaking, the goal of the 50 copies per milliliter is reached through a tri-phasic curve. In 1995, studies by Ho et al18 showed that plasma HIV RNA levels decrease rapidly when patients started assuming a potent antiretroviral drug. This phase is mainly due to the decay of productively infected cells and particularly activated CD4+ T cells that have a short half-life of approximately 1–2 days.19 After this exponential decay, a second, slower phase of the decay curve takes place and the infected cells responsible for it are mainly infected macrophages that produce much less virus than activated CD4 cells but have a much slower decay rate too.
The first clue of the existence of a third phase of the decay curve came from the observation that several patients on stable HAART have transient HIV RNA elevations above the 50 copies per milliliter threshold (viral blips).20 Newer, more sensitive assays for the quantitation of HIV RNA have revealed that up to 50% of patients with a VL <50 copies per milliliter, still have detectable viremia >3 copies per milliliter.7,8
Cells responsible for this third phase of viral decay are the latently infected resting CD4+ T cells. These cells are rare, as only 1 per million resting CD4 carries a stably integrated transcriptionally silent form of viral genome that can result in the production of virus once the cell becomes activated.21 This third phase cannot be described with standard clinical VL assays as the level of viremia resulting from the release of virus from this stable reservoir is below their limit of detection.22 Several studies have shown an intrinsic stability of the latent reservoir.21–23 According to these data, plasma LLV could be the result of virus release from the reservoir without the occurrence of a complete cycle of replication as concomitant HAART would prevent it.21–23 With this respect, it is important to understand that all drugs used to treat HIV infection prevent new cells from being infected, but do not block virus production and release by cells that already have an integrated provirus. Some of our data seem to confirm this theory. According to our data, patients more likely to have a steady level of HIV RNA <3 copies per milliliter were those who showed a lower baseline pre-HAART VL and who presented the longer time without any (according to the available limit of detection) active viral replication. Both these events could explain a smaller HIV reservoir and therefore a reduced chance of occasional virus release. However, there is a vigorous debate about the meaning of LLV, and the stability of the latent reservoir could be explained by a replenishment of it by low-level viral replication. This is a disturbing idea as ongoing replication in the presence of drugs is indication of, at least partial, inefficacy leading to the selection of resistance. Although contrasting results have been published,24 most evidence on the effect of HAART intensification in patients with LLV does not support the hypothesis of an ongoing replication.25,26 It seems plausible that the 2 models to explain low-level residual viremia may coexist.
Our data indicate that low-level viral replication may be the driving force of LLV, at least in some patients. According to our results, it seems that the average steady-state HIV RNA is below 3 copies per milliliter, although a consistent proportion of patients (approximately 30%) have, at any time, a measurable LLV. However, this risk is unevenly distributed and, as previously reported with higher cut-off levels,27 patients with a previous viremia <3 copies per milliliter have a much higher chance to maintain this result.
We demonstrated that a HIV RNA level >3 copies per milliliter is highly predictive of virological failure and that a linear relationship exists between the entity of residual viremia and the risk of virological failure. These results confirm recently published data28 on 1247 patients. In this cohort, patients with completely suppressed viremia had a risk of virologic failure over a year of 4% compared with a risk of 11.3% for patients <40 copies per milliliter and of 34.2% for those with VL between 40 and 49 copies per milliliter. On the contrary, Gianotti et al29 excluded the influence of low-level residual viremia on the risk of 1 year virological rebound. However, their results were derived from a cohort of 739 patients that did confer to the study, according to actual results, a power of 25%. This limited statistical power should induce to a very careful evaluation of any negative result.
The results of the present study and of that by Doyle et al28 may be questioned for biases. It is known that patients who have had virologic suppression for longer periods of time are less likely to present viral rebound compared with those suppressed for a shorter period.30 It may take longer to get <3 copies per milliliter than <50 copies per milliliter and in our study the time below the detection limit of a conventional HIV RNA assay before entering the study was an independent predictor of the chance of showing a complete viral suppression. Further, in both our and Doyle study,28 the time below the detection limit was a predictor of virological failure. In both studies, conclusions could be influenced by an excess of patients, in the group with low but detectable viremia, still in the third phase of viral decay which is estimated to last 9–15 months after initiation of therapy7 and, therefore, not yet at the steady state. To rule out this survivor bias,31 we performed a post hoc subanalysis limited to those patients with longer time below the detection limit (above the median of our cohort, 44 months). The long HIV RNA suppression in this group of patients should ensure to be far beyond the third phase of viral decay. The analysis confirmed the predictive value in terms of risk of virogical failure of the low-level residual viremia as patients with steady HIV RNA level <3 copies per milliliter did not present any confirmed rebound above 50 copies per milliliter, whereas failure was recorded in 1.1% of patients having some of the samples <3 copies per milliliter and others with detectable LLV and in 31.3% of individuals with steadily detectable LLV between 3 and 50 copies per milliliter (P < 0.0001).
A second question could rise as a consequence of the study design. Both studies (ours and Doyle) where cohort analyses enrolling patients with different treatment histories. Previous failures may have interfered with the ability of current HAART to completely suppress viral replication and thus bias the final results. To address this possible limit, we performed a second post hoc analysis limited to patients who had never failed a previous HAART regimen. Once again, in this selected subset of patients, the risk of virological failure, although relatively lower, in absolute terms than that observed in the general cohort was strictly linked to the level of residual viremia. Patients with steady HIV RNA level <3 copies per milliliter did not present any confirmed rebound above 50 copies per milliliter, whereas failure was recorded in 1.4% of patients having some of the samples <3 copies per milliliter and others with detectable LLV and in 18.5% of individuals with steadily detectable LLV between 3 and 50 copies per milliliter (P < 0.0001).
Investigate on causes of LLV was beyond the aims of these study. We found that some patients' characteristics are associated with or predispose to the persistence of LLV. However, other causes, not addressed in this study, may play a relevant role. Adherence may be one of these if not the most important of them. LLV could be an early indicator of less-than-perfect adherence, although not yet demonstrated,28 and we cannot rule out that reduced adherence was the cause of LLV in some of our patients. However, it is difficult to think that the consequences of LLV, as described in this work, may vary according to the causes leading to the presence of LLV itself.
Interestingly, as indicated by previous studies,8,28,29,32,33 the type of ongoing antiretroviral regimen was an independent predictor of both the chance to obtain and maintain a VL <3 copies per milliliter and virological failure. Patients treated with a NNRTI-based HAART had a higher chance to maintain a viremia <3 copies per milliliter throughout the study period compared with those receiving a boosted-PI or non-boosted PI. Similarly, the overall risk of failure was NNRTIs <boosted PIs <non-boosted PIs. Several hypothetical explanations of these observations may exist. NNRTIs may have specific pharmacokinetic properties, which may influence penetration into “sanctuaries” and enhance the complete suppression of viral replication or their simplified dosing schedule and high tolerability profile might positively influence adherence rates.34 The most probable explanation, in our opinion, depends on the “forgiveness” of NNRTI-based regimens that may masque minor reductions of adherence.35
The high predictive value of a HIV RNA levels <3 copies per milliliter on a clinically relevant outcome such as a confirmed virologic failure indicates the opportunity to reconsider the goal of antiretroviral therapy to a lower cut-off than 50 copies per milliliter. This indication is further strengthened by the observation that, in a fairly relevant proportion of failing patients, the development of viral mutations leading to resistance is possible. Several studies36–38 have demonstrated that the selection of viral mutants may occur with very low levels of viral replication and that they are most frequent at VLs between 300 and 10,000 copies per milliliter. In our experience, 30% of failing patients developed drug resistance to all or part of their ongoing HAART, often reducing their future drug options. All the documented resistance-inducing mutations were selected in patients with HIV RNA values <10,000 copies per milliliter. A possible explanation of this observation is that patients with greater VL rebound could be those with a worst adherence to HAART and, therefore, a reduced selective pressure. As the risk of selecting for resistance-inducing mutations is greater at lower VLs, it is of paramount importance to use sensitive tools able to indicate an increased probability of virological failure so to reduce the accumulation of mutations and the development of cross-resistance.36
Our findings open new and interesting clinical problems.
Despite the clear indication that a measurable viremia was associated with a significant risk of virological failure, these events were quite rare. That might indicate that an alternative explanation for LLV, such as virus release from reactivated latently infected T cells, may be true in several patients. An effective approach, such as HAART intensification, would therefore implicate the possibility to discriminate those patients whose LLV is supported by active replication from those in whom it recognize a different pathogenesis. A limit of our study is that we could not explore several potential determinants of LLV either because not available as routine tests to run on a large number of patients (eg, integrated HIV DNA)33,39 or because unpractical and expensive in a large cohort such as adherence monitoring with electronic devices. For the same reasons, we were unable to study other surrogate markers, such as immune-activated T cells or soluble markers of immune activation, that could help in discriminating those patients most likely to benefit of HAART intensification.39
To date, in our knowledge, such a differentiation is not feasible. If an individualized therapeutic approach, based on the causes of LLV, may be beyond the current clinical possibilities, at least a differentiated clinical management of patients with LLV may be advisable. We demonstrated that a shorter period on HAART, a shorter time below detection limit of conventional tests, being on a PI-based HAART and presenting a steadily detectable LLV between 3 and 50 copies per milliliter were all independent negative prognostic factors for virological failure.
Patients presenting some or all these characteristics should deserve a careful investigation of all possible causes leading to LLV including, but not limited to: virologic potency of the current regimen, its efficacy in viral sanctuaries, its pharmacokinetic properties, or patient's adherence. Further, a more efficient follow-up based on more frequent controls may be advisable to better define the individual risk and to limit the consequences (mostly in terms of viral resistance) of a possible therapeutic failure. On the other hand, a less aggressive management could be reserved to patients with steadily undetectable HIV RNA whose probability to maintain this situation over a year period is >40%, whose risk of showing a LLV in a subsequent sample is low (20%) and in whom the risk of virologic failure, in the same period of time, is <0.5%.
In conclusion, according to current treatment guidelines, a HIV RNA level <50 copies per milliliter is the goal of HAART. Our data suggest that this goal may need to be revised to a lower cut-off value. A low viremia >3 copies per milliliter is linked to a significant increment of the risk of virological failure and predispose to genotypic resistance. Clinical management of patients with measurable LLV should be managed to better evaluate, over time, the risk of failure and to limit its consequences.
1. The Antiretroviral Therapy Cohort Collaboration. Rates of diseases progression according to initial highly active antiretroviral therapy regimen: a collaborative analysis of 12 prospective cohort studies. J Infect Dis. 2006;194:612–622.
2. Gilbert PB, DeGruttola V, Hammer SM, et al.. Virologic and regimen termination surrogate end-points in AIDS clinical trials. JAMA. 2001;285:777–784.
3. Kirk O, Pedersen C, Law M, et al.. Analysis of virological efficacy in trials of antiretroviral regimens: drawbacks of not including viral load measurements after premature discontinuation of therapy. Antivir Ther. 2002;7:271–281.
4. Gazzard BG, on behalf of the BHIVA Treatment Guidelines Writing Group. British HIV Association guidelines for the treatment of hIV-1-infected adults with antiretroviral therapy. HIV Med. 2008;9:563–608.
6. Thompson MA, Aberg JA, Cahn P, et al.. Antiretroviral treatment of adult HIV infection-2010. Recommendations of the IAS-USA Panel. JAMA. 2010;304:321–333.
7. Palmer S, Maldarelli F, Wiegand A, et al.. Low-level viremia persists for at least 7 years in patients on suppressive antiretroviral therapy. Proc Natl Acad Sci U S A. 2008;105:3879–3884.
8. Bonora S, Nicastri E, Calcagno A, et al.. Ultrasensitive assessment of residual HIV viremia in HAART-treated patients with persistently undetectable plasma HIV-RNA: a cross-sectional study. J Med Virol. 2009;81:400–405.
9. Antinori A, Marcotullio S, Ammassari A, et al.. Italian guidelines for the use of antiretroviral agents and the diagnostic-clinical management of HIV-1 infected persons. New Microbiol. 2011;34:109–146.
10. Ruelle J, Jnaoui K, Lefevre I, et al.. Comparative evaluation of the VERSANT HIV-1 RNA 1.0 kinetic PCR molecular system (kPCR) for the quantification of HIV-1 plasma viral load. J Clin Virol. 2009;44:297–301.
11. Troppan KT, Steltzl E, Violan D, et al.. Evaluation of the new VERSANT HIV-1 RNA 1.0 assay (kPCR) for quantitative detection of human immunodeficiency virus type 1 RNA. J Clin Virol. 2009;46:69–74.
12. Mackie N, Dustan S, McClure MO, et al.. Detection of HIV-1 antiretroviral resistance from patients with persistently low but detectable viremia. J Virol Methods. 2004;119:73–78.
13. Van Lunzen J, Maggiolo F, Phing B, et al.. Rapid, robust and sustained antiviral response with once-daily dolutegravir, a next generation integrase inhibitor in combination therapy in antiretroviral-naïve adults: 48 week results from SPRING-1. Presented at: 6th IAS Conference; July 17–20, 2011; Rome, Italy. Abstract TUAB01.
14. Lundgren JD, Babiker A, El Sadr WM, et al.. Inferior clinical outcome of the CD4+ cell count-guided antiretroviral interruption strategy in the SMART study: role of CD4+ cell counts and HIV-RNA levels during follow-up. J Infect Dis. 2008;197:1145–1155.
15. Mellors JW, Munoz A, Giorgi JV, et al.. Plasma viral load and CD4+ lymphocytes as prognostic markers of HIV-1 infection. Ann Intern Med. 1997;126:946–954.
16. Shiu C, Cunningham CK, Greenough T, et al.. Identification of ongoing human immunodeficiency virus type 1 (HIV-1) replication in residual viremia during recombinant HIV-1 poxvirus immunizations in patients with clinically undetectable viral loads on durable suppressive highly active antiviral therapy. J Virol. 2009;83:9731–9742.
17. Van Sighem A, Zhang S, Reiss P, et al.. Immunologic, virologic and clinical consequences of episodes of transient viremia during suppressive combination antiretroviral therapy. J Acquir Immune Defic Syndr. 2008;48:104–108.
18. Ho DD, Neumann AU, Perelson AS, et al.. Rapid turnover of plasma virions and CD4 lymphocytes in hIV-1 infection. Nature. 1995;373:123–126.
19. Wei X, Ghosh SK, Taylor ME, et al.. Viral dynamics in human immunodeficiency virus type 1 infection. Nature. 1995;373:117–122.
20. Dornadula G, Zhang H, VanUitert B, et al.. Residual HIV-1 RNA in blood plasma of patients taking suppressive highly active antiretroviral therapy. JAMA. 1999;282:1627–1632.
21. Sedaghat AR, Siliciano JD, Brennan TP, et al.. Limits on replenishment of resting CD4+ T cell reservoir for HIV patients on HAART. PLoS Pathog. 2007;3:1165–1174.
22. Sedaghat AR, Siliciano RF, Wilke CO. Low-level HIV-1 replication and the dynamics of the resting CD4+ T cell reservoir for HIV-1 in the setting of HAART. BMC Infect Dis. 2008;8:1–14.
23. Siliciano JD, Kajdas J, Finzi D, et al.. Long-term follow-up studies confirm the stability of the latent reservoir for HIV-1 in resting CD4+ T cells. Nat Med. 2003;9:727–728.
24. Buzon MJ, Massanella M, Llibre JM, et al.. HIV-1 replication and immune dynamics are affected by raltegravir intensification of HAART-suppressed subjects. Nat Med. 2010;16:460–465.
25. Archin NM, Cheema M, Parker D, et al.. Antiretroviral intensification and valproic acid lack sustained effect on residual HIV-1 viremia or resting CD4+ cell infection. PLoS One. 2010;5:1–4.
26. McMahon D, Jones J, Wiegand A, et al.. Short-course raltegravir intensification does not reduce persistent low-level viremia in patients with HIV-1 suppression during receipt of combination antiretroviral therapy. Clin Infect Dis. 2010;50:912–919.
27. Widdrington J, Payne B, Medhi M, et al.. The significance of very low-level viremia detected by sensitive viral load assays in HIV infected patients on HAART. J Infect. 2011;62:87–92.
28. Doyle T, Smith C, Vitiello P, et al.. Plasma HIV-1 RNA detection below 50 copies/ml and risk of virologic rebound in patients receiving highly active antiretroviral therapy. Clin Infect Dis. 2012;54:729–737.
29. Gianotti N, Galli L, Racca S, et al.. Residual viremia does not influence 1 year virological rebound in HIV-infected patients with HIV-rna persistently below 50 copies/ml. J Antimicrob Chemother. 2012;67:213–217.
30. Lima VD, Bangsberg DR, Harrigan PR, et al.. Risk of viral failure declines with duration of suppression on highly active antiretroviral therapy irrespective of adherence level. J Acquir Immune Defic Syndr. 2010;55:460–465.
31. Gandhi RT, Deeks SG. Plasma HIV-1 RNA levels during antiretroviral therapy: how low is low enough? Clin Infect Dis. 2012;54:1–3.
32. Geretti AM, Smith C, Haberl A, et al.. Determinants of virologic failure after successful viral load suppression in first-line highly active antiretroviral therapy. Antivir Ther. 2008;13:927–936.
33. Nicastri E, Palmisano L, Sarmati L, et al.. HIV-1 residual viremia and proviral DNA in patients with suppressed plasma viral load (<400 HIV-RNA cp/ml) during different antiretroviral regimens. Curr HIV Res. 2008;6:261–266.
34. Maggiolo F, Ripamonti D, Arici C, et al.. Simpler regimens may enhance adherence to antiretrovirals in HIV infected patients. HIV Clin Trials. 2002;5:371–378.
35. Cohen CJ, Colson AE, Sheble-Hall AG, et al.. Pilot study of a novel short-cycle antiretroviral treatment interruption strategy: 48-week results of the five-days-on, two-days-off (FOTO) study. Antivir Ther. 2010;8:19–23.
36. Prosperi MCF, Mackie N, Di Giambenedetto S, et al.. Detection of drug resistance mutations at low plasma HIV-1 RNA load in a European multicenter cohort study. J Antimicrob Chemother. 2011;66:1886–1896.
37. Mackie NE, Phillips AN, Kaye S, et al.. Antiretroviral drug resistance in HIV-1-infected patients with low-level viremia. J Infect Dis. 2010;201:1303–1307.
38. Nettles RE, Kieffer TL, Simmons RP, et al.. Genotypic resistance in HIV-1-infected patients with persistently detectable low-level viremia while receiving highly active antiretroviral therapy. Clin Infect Dis. 2004;39:1030–1037.
39. Ostrowski SR, Katzenstein TL, Thim PT, et al.. Low-level viremia and proviral DNA impede immune reconstruction in HIV-1-infected patients receiving highly active antiretroviral therapy. J Infect Dis. 2005;191:348–357.
© 2012 Lippincott Williams & Wilkins, Inc.