Background.Episodes of viremia are common in African antiretroviral therapy (ART) programs. We sought to describe viremia, resuppression, and accumulation of resistance during first-line combination ART (cART) in South Africa.
Methods.Retrospective analysis of a cohort receiving zidovudine, lamivudine, and either efavirenz or nevirapine with human immunodeficiency virus (HIV) RNA monitoring every 6 months. We assessed viremia (HIV RNA >1000 copies/mL after initial HIV RNA response) and resuppression (HIV RNA <400 copies/mL after viremia). Genotypic resistance testing was performed using stored plasma on a subset of patients at first detection of viremia and subsequently among patients with persistent viremia.
Results.Between 2002 and 2006, 3727 patients initiated cART (median CD4, 147 cells/mm3). Of 1007 patients who developed viremia, 815 had subsequent HIV RNA assays, and 331 (41%) of these resuppressed without regimen switch. At identification of viremia, 45 (66%) of 68 patients had HIV-1 drug resistance, 42 (62%) had nonnucleoside reverse-transcriptase inhibitor (NNRTI)-resistance, 25 (37%) had M184V/I, and 4 (6%) had multi-nucleoside analogue drug mutations. By 12 months of persistent viremia among a subset of 14 patients with resistance testing to 12 months, 11 (78%) had nonnucleoside reverse-transcriptase inhibitor (NNRTI)-resistance, 8 (57%) had M184V/I, and 2 (14%) had multi-nucleoside analogue drug mutations. Resistance was associated with a reduced probability of resuppression; however, 50% of patients with NNRTI resistance resuppressed while receiving an NNRTI.
Conclusions.The majority of patients had NNRTI resistance mutations at detection of viremia. However, 41% resuppressed without regimen switch. Our findings support maximizing first-line use while minimizing risk of significant cross-resistance by implementing intensive adherence support and repeat HIV RNA testing 3–6 months after detecting viremia, with regimen switch only if viremia persists.
The recent availability of antiretroviral therapy (ART) in Africa has had a dramatic impact on suppressing human immunodeficiency virus (HIV) and reducing mortality [1, 2]. Within 12 months of initiating combination ART (cART), 65% to 90% of individuals achieve HIV RNA levels below the limit of detection [3–7]. However, not all individuals maintain virologic control ; those who do not are at risk for generation of drug-resistant HIV and eventual decrease in CD4 count. To identify patients who are not maintaining virologic control, treatment guidelines used in the United States recommend HIV RNA monitoring 2–4 weeks after treatment initiation and every 4–8 weeks until the HIV RNA level is <50 copies/mL . Guidelines used in resource-limited regions recommend monitoring of HIV RNA and/or CD4 every 6 months and make provision for situations in which no laboratory monitoring is available [9, 10].
As a consequence of the monitoring strategies available in resource-limited countries, identification of HIV viremia may be delayed. Furthermore, because of limited cART options, relative immune preservation despite viremia, and limited data to guide decision making, many providers are hesitant to switch therapy when viremia is identified. When HIV viremia is detected or decisions to switch therapies are made, multiple HIV resistance mutations may be present [11, 12]. However, minimal data are available on patterns of HIV viremia, resuppression, and the accumulation of resistance during persistent viremia; therefore, we evaluated these issues in an HIV treatment cohort in South Africa.
Patients included in this study were enrolled in a workplace HIV program in South Africa and met the following criteria: (1) initiated cART from November 2002 through May 2006, (2) had ⩾1 HIV RNA assay while receiving therapy, (3) were ⩾18 years old, (4) were ART-naive at cART initiation, and (5) received a cART regimen of zidovudine, lamivudine, and a nonnucleoside reverse-transcriptase inhibitor (NNRTI), either efavirenz or nevirapine. The observation period for each patient was from cART initiation until the earlier of June 2008 or last follow-up HIV RNA assay.
This ART program has been described elsewhere . In brief, cART eligibility was based on modified World Health Organization (WHO) criteria: CD4 count <250 cells/mm3, WHO stage 3 with CD4 count <350 cells/mm3, or WHO stage 4. CD4 count and HIV RNA levels were determined before cART initiation, after 6 weeks on cART, and then every 6 months. Program guidelines recommended a switch to second-line therapy if 2 sequential HIV RNA assays were >1000 copies/mL and if the provider assessed that good adherence was achievable. Regimen changes were recorded on a standardized clinic visit form. In addition, central pharmacy distribution records were used to verify medications. The second-line regimen was abacavir, didanosine, and lopinavir-ritonavir.
The adherence strategy was based on individual structured counseling with 1 to 2 visits at each of the following time points: cART preparation, initiation, maintenance, and treatment failure and suspected poor adherence. Routine adherence assessment included number of pills missed in the preceding 3 days, time of day pills were missed in the preceding 7 days, and the reason pills were missed.
All patients included in this study signed informed consent, and ethical approval for this study was obtained from the research ethics committees of the Anglo-Gold Health Service and the London School for Hygiene and Tropical Medicine, the University of KwaZulu Natal, and the Institutional Review Board at Johns Hopkins University School of Medicine.
HIV viremia. HIV RNA >1000 copies/mL while on first-line therapy after an initial drop >1 log10copies/mL from the pre-cART level.
Persistent viremia. HIV viremia >1000 copies/mL detected on ⩾2 consecutive HIV RNA assays while receiving first-line therapy.
Resuppression. HIV RNA to <400 copies/mL while continuing to receive first-line cART after ⩾1 HIV RNA assay that demonstrated viremia.
WHO CD4 failure criteria. Decrease of CD4 count to pre-cART baseline, 50% decrease from on-cART peak value, or persistent CD4 levels <100 cells/mm3for >12 months . We modified these criteria and did not use persistent CD4 levels <100 cells/mm3as a criterion for treatment failure, because of the number of individuals in this study with virologic failure before 12 months.
Resistance mutation. Resistance mutations were defined on the basis of the International AIDS Society-USA 2008 resistance guide . Thymidine analog mutations (TAMs) were M41L, D67N, K70R, L210W, T215Y/F, and K219Q/E, and multiple-nucleoside reverse-transcriptase inhibitor (multi-NRTI) mutations included TAMS, K65R, and L74V.
HIV-1 Drug Resistance Testing
In 2007 we selected every third individual with viremia on the basis of study number. We selected specimens for genotyping from first detection of viremia and, for those patients with persistent viremia, from subsequent time points for which samples were available. Subsequently, additional follow-up data were obtained. To assess if patients whose samples were genotyped were representative of the larger population, we compared patients with genotype results with patients who also developed viremia but did not have genotyping, by means of the χ2test or t test.
HIV RNA was assayed with the Amplicor HIV-1 Monitor Test (Roche Diagnostics). Sequencing was performed by extracting viral RNA from stored plasma and amplifying a 1.7Kb fragment spanning the pol gene by nested polymerase chain reaction (PCR) assay with the Thermoscript RT-PCR System (Invitrogen). PCR products were sequenced using BigDyeTerminators and an ABI 310 DNA Sequencer (Applied Biosystems) . Consensus sequences from all genotyped individuals were aligned and manually edited using the Sequencher program, version 4.5 (GeneCodes). Multiple alignments were performed using Clustal X, version 2.0 (Conway Institute). Phylogenetic analysis of nucleic acid sequences was performed with Mega, version 4.0 . Reference sequences were downloaded from the Los Alamos database .
Logistic regression was used to identify variables associated with the presence of NNRTI resistance and with the M184V/I mutation at first detection of viremia. Logistic regression was also used to evaluate associations with resuppression within 6 months of detecting viremia among subjects with an HIV RNA result within this time frame. We calculated 95% confidence intervals (CIs) for proportions with resistant virus with exact binomial methods. Bar graphs were used to display the accumulation of mutations (any NNRTI, M184V/I, or any TAM) at detection of viremia and at 6, 12, and 18 months of persistent viremia while receiving first-line cART for patients with multiple resistance tests ⩾1 of which was ⩾12 months after first detection of viremia. All HIV RNA values were log10transformed for analysis.
A total of 3727 patients met our inclusion criteria, of whom 3479 (93%) were male; the median age was 42 years. Median CD4 count at cART initiation was 147 cells/mm3(interquartile range [IQR], 80–216 cells/mm3) (Table 1). The median duration between HIV RNA assays was 4 months (IQR, 2.2–5.6 months). The median follow-up was 17.4 months (IQR, 6.0–31 months) with 6118 person-years of follow-up.
A total of 3432 subjects (92%) had >1 log10decrease in HIV RNA from the time of cART initiation to the end of the observation period, of whom 3,159 suppressed to <400 copies/mL. Of the 3432 subjects with >1 log10HIV RNA decline, 1007 (29%) developed HIV viremia, which was detected a median of 11 months (IQR, 6.2–18 months) after cART initiation. A total of 815 (81%) of those patients with HIV viremia had follow-up HIV RNA laboratory results while receiving first-line therapy (Figure 1). Reasons for absence of repeat HIV RNA assay results while receiving first-line therapy among the other 192 patients were the following: cART was stopped or switched, they left the cART program, they died, or they had no additional HIV RNA assays during the observation period. Of the 815 patients with subsequent HIV RNA results, 331 (41%) resuppressed (HIV RNA <400 copies/mL) without a regimen switch; of these, 224 achieved an HIV RNA <50 copies/mL. Only 46 of the 1007 patients with viremia were switched to second-line therapy during the observation period.
At the time of first detection of HIV viremia, 211 (21%) of the 1007 patients with HIV viremia met modified WHO CD4 criteria for treatment failure. Another 110 patients with viremia fulfilled the modified criteria at a later time point. Of these 321 patients (211 plus 110) with modified WHO CD4 failure and viremia, 111 (34%) resuppressed without a change in regimen.
We selected 138 patients for genotyping; samples from 18 could not be amplified, and insufficient sample was available from 30 patients. As a result, 90 patients underwent genotyping. Individuals with genotype results were similar in terms of sex, age, weight at cART initiation, median CD4 count at cART initiation, median HIV RNA level at cART initiation, WHO stage, and type of NNRTI received to those viremic individuals who did not undergo genotyping (all, P>.1). We used a neighbor joining tree to assess subtype clustering. One patient clustered with subtype A HIV-1, and the rest clustered with subtype C.
HIV Resistance at First Viremia
Sixty-eight subjects had genotypic resistance tests done at first detection of HIV viremia (the remaining 22 had insufficient sample at this time point). Forty-five subjects (66%) had ⩾1 HIV resistance mutation in reverse transcriptase. NNRTI mutations were most common (42 subjects [62%]), followed by the M184V/I mutation (25 subjects [37%]) (Table 2). Only 4 (6%) had ⩾1 multi-NRTI mutation at detection of viremia.
In unadjusted analysis, the presence of an NNRTI mutation and the M184V/I mutation at first detection of viremia were both associated with higher HIV RNA at cART initiation (Table 3). However, neither was associated with HIV RNA level at the time viremia was identified. Sex, age, CD4 count, NNRTI agent, modified WHO CD4 failure criteria, and WHO stage at cART initiation were not associated with the presence of either NNRTI mutations or the M184V/I mutation (all, P>.1). Higher weight at cART initiation (P=.02), modified WHO CD4 failure criteria (P=.06), and HIV RNA at cART initiation (P=.03) were associated with reduced odds of the M184V/I mutation. Results from multivariate modeling for the M184V/I for weight and baseline HIV RNA follow. For HIV RNA at baseline, compared with <4.6 log10copies/mL, 4.6–4.9 copies/mL had an odds ratio (OR) of 2.7 (95% CI, 0.64–11) and >4.9 copies/mL had an OR of 5.9 (95% CI, 1.4–26) (P=.04for trend). For weight per 10-kg increase, the OR was 0.44 (95% CI, 0.23–0.83; P=.01).
Accumulation of Resistance Mutations with Persistent Viremia
We restricted analysis of the accumulation of mutations during HIV viremia to the 30 patients with multiple genotypes (with the latest genotype ⩾12 months from first detection of viremia) while continuing to receive first-line cART. At detection of viremia 5 (71%) of 7 tested patients had NNRTI resistance mutations; this increased to 8 (89%) of 9 patients by 6 months, 11 (78%) of 14 patients by 12 months, and 15 (94%) of 16 patients by 18 months (Figure 2). For lamivudine, at detection of viremia 43% (3/7) had the M184V/I mutation, by 6 months 44% (4/9), by 12 months 57% (8/14), and by 18 months 80% (12/15). At detection of viremia none had a TAM, but by 6 months 1 (11%) of 9, by 12 months 2 (14%) of 14, and by 18 months 4 (31%) of 13 had developed ⩾1 TAM. Of subjects with a TAM by 18 months, 2 had 1 or 2 TAMs and 2 had 4 or 5 TAMs. Among all subjects with resistance testing, median time to 1 TAM was 8.4 months (range, 5.6–11.3 months), median time to 2 TAMs was 2.7 months (range, 0.9–18.3 months), and median time to ⩾3 TAMs was 10.3 months (range, 5.5–18.4 months).
Predictors of Resuppression
Assessing the 655 patients (299 of whom resuppressed) with HIV RNA testing within 6 months of first detection of viremia (among the total 815 subjects with viremia and any subsequent test results), we found a trend toward resuppression and lower log10HIV RNA at cART initiation (compared with log10HIV RNA ⩽4.5 copies/mL, log10HIV RNA 4.6–4.9 copies/mL had an OR of 0.98; 95% CI, 0.65–1.47; and log10HIV RNA ⩾5 copies/mL had an OR of 0.67; 95% CI, 0.44–1.0; P=.08). We found associations with higher baseline CD4 count (per 50-cell increase, the OR was 1.1; 95% CI, 1.0–1.2; P=.01) and WHO stage at cART initiation (compared with stage 1 or 2, the stage 3 OR was 0.62; 95% CI, 0.42–0.90; and the stage 4 OR was 0.63; 95% CI, 0.39–1.0; P=.03). In multivariate analysis, WHO stage was not included because of colinearity with CD4 count. In this analysis, baseline CD4 remained associated with resuppression (adjusted OR per 50-cell/mm3increase, 1.1; 95% CI, 1.0–1.2; P=.01), but baseline log10HIV RNA lost significance (P=.5). In addition, among patients with resistance testing and follow-up HIV RNA testing (n=74), the absence of resistance mutations at first detection of viremia was associated with a greater chance of resuppression (OR, 4.7; 95% CI, 1.5–15; P=.009). There was no evidence of an association between resuppression and sex, age, weight at cART initiation, HIV RNA at detection of viremia, or modified WHO CD4 failure criteria (all, P>.1).
We also evaluated the impact of resistance on resuppression by comparing the level of resistance among patients who did resuppress with the level of resistance among patients who did not resuppress. We identified 11 (50%) of 22 patients (95% CI, 28%–72%) who resuppressed on the first-line cART and who had genotyping that had NNRTI resistance. Among 52 patients with genotyping who did not resuppress, 42 (79%; 95% CI, 65%–89%) had detectable resistance at first detection of viremia (Figure 1). Of the patients achieving resuppression despite detectable resistance, all had NNRTI resistance mutations: K103N (8), V106M (5), P225H (1), or G190A (1).
This study provides important information on the accumulation of drug resistance during HIV viremia while continuing to receive first-line cART. Unlike previous studies of cART in Africa, our study featured serial resistance testing during HIV viremia. In addition, we describe resuppression after viremia while continuing to receive first-line cART.
At first detection of viremia, the majority of patients had NNRTI resistance; lamivudine resistance was the next most common. TAMs and multi-NRTI mutations accumulated after longer duration viremia. The burden of resistance at detection of viremia is consistent with previous studies in which prevalence of NNRTI mutations ranged 45%–100%, the M184V/I mutation ranged 24%–95%, and any TAM ranged 0%–75% [5, 11, 12, 19–23]. We hypothesize that the greater proportion of TAMs reported by some cohorts reflects longer delays in detecting viremia. Our rate of TAM accumulation is consistent with clinical trials of dual NRTI therapy from industrialized countries [24–26]. Of note, we used a cutoff point of 400 copies/mL as a definition of resuppression for consistency with other studies and to avoid including blips as failures. Future studies evaluating lower-level viremia (50–400 copies/mL) are important for resource-limited settings.
An unusual contribution of our study is the repeated resistance testing on patients with persistent viremia while receiving first-line cART. This enabled us to assess the accumulation of resistance. This information is useful for informing clinical decisions, such as how rapidly to switch to second-line therapy after detecting viremia. We found that, at the time of detection of viremia, a patient has a high probability of NNRTI resistance. When viremic for an additional 6 months on first-line therapy, most patients had an NNRTI resistance mutation and approximately two-thirds had the M184V/I mutation. Importantly, only ∼10% had any TAM at 6 months. As multiple TAMs are required for significant resistance to thymidine analogues (zidovudine and stavudine) and for cross-resistance to other NRTI agents, we found that cross-resistance was minimal 6 months after viremia was detected.
Importantly, among patients who developed viremia, 41% resuppressed while continuing to receive first-line therapy. However, had they been rapidly switched to a second-line regimen, benefit from the first-line regimen would have been prematurely abbreviated and the option of having an additional therapy available for the future would have been eliminated. High rates of resuppression after developing viremia have been reported, in abstract form, from other African cohorts [27, 28].
We further found that even when NNRTI resistance mutations were present, successful resuppression occurred among some patients who were continuing the first-line regimen. This finding may seem surprising. However, resuppression on an NNRTI-based regimen despite NNRTI resistance has been noted in a treatment interruption study  and among individuals with transmitted resistance in a clinical trial that used NNRTI-based regimens .
Several limitations of our study are worth noting. First, we did not sequence HIV samples at the time of cART initiation. As a result, patients may have had resistance mutations before cART. We believe this is unlikely for the following reasons: (1) no patients reported prior ART; (2) 95% of the cohort was men, and thus potential for exposure through efforts for prevention of mother-to-child transmission was minimal; (3) our cohort started cART in 2002, two years before the South African national cART rollout; and (4) the levels of transmitted resistance in South Africa at this time were low . A second weakness is that we lacked resources to sequence specimens from all patients who developed viremia.
Additional limitations reflect the nature of our cohort. Our cohort was mostly men who worked in mines and received HIV care through workplace HIV programs. As a result, aspects of the cohort may differ from other cohorts. However, we believe that overall accumulation of resistance during viremia is likely to be similar across cohorts with subtype C HIV. In addition, this was an observational cohort, and decisions regarding when to switch therapy were likely affected by perceptions of patient adherence. Thus, it is possible that individuals who developed viremia and were thought to have good adherence were switched more rapidly than those with perceived poor adherence. However, the impact of such a bias is likely to be minimal, because only 46 of the 1007 patients who developed viremia were switched to second-line therapy. Furthermore, as an observational cohort that used local standard of care, we lacked the frequency of HIV RNA monitoring to accurately describe the time to resuppression after developing viremia.
In ART programs with access to intermittent HIV RNA assays and limited treatment options, our results support added emphasis on adherence when viremia is identified, followed by regimen switch only if viremia persists for 3–6 months. This strategy leads to limited risk of cross-resistance while maximizing the opportunity for resuppression on first-line therapy. This contrasts starkly with management in high-income countries, where testing resources and the availability of many ART options allow rapid regimen switch and resistance testing when viremia is first identified. However, our proposed strategy fits resource-limited settings, because it helps to avoid potentially premature switches to costlier and less tolerable regimens while minimizing the risk of the emergence of high-level cross-resistance.
We thank Drs Rami Kantor and David Katzenstein for constructive critiques.
Financial support. The Aurum Institute and the National Institute for Communicable Diseases, South Africa; National Institutes of Health (DK074348 to C.J.H. and AI5535901 and AI016137 to R.E.C.); and United Kingdom Department of Health Public Health Career Scientist Award (to A.D.G.).
Potential conflicts of interest. All authors: no conflicts.