Virological failure rates and HIV-1 drug resistance patterns in patients on first-line antiretroviral treatment in semirural and rural Gabon

Introduction As antiretroviral treatment (ART) continues to expand in resource-limited countries, the emergence of HIV drug resistance mutations (DRMs) is challenging in these settings. In Gabon (central Africa), no study has yet reported the virological effectiveness of initial ART given through routine HIV care. Methods Following the World Health Organization (WHO) recommendations, a cross-sectional study with a one-time HIV-1 RNA viral load (VL) measurement was conducted in Gabon to assess virological failure (VF) defined by a VL result ≥1000 copies/ml and DRMs among adult patients living with non-B HIV-1 strains and receiving first-line non-nucleoside reverse transcriptase inhibitor (NNRTI)-based antiretroviral therapy for at least 12 months. Risk factors associated with VF and DRMs were assessed. Results Between March 2010 and March 2011, a total of 375 patients were consecutively enrolled from two decentralized (one semirural and one rural) HIV care centres. Median time on ART was 33.6 months (range, 12–107). Overall, the rate of VF was 41.3% (36.4–46.4). Among viremic patients, 56.7% (80/141) had at least one DRM and 37.6% had dual-class resistance to nucleoside reverse transcriptase inhibitors (NRTIs) and NNRTIs. The most frequent DRMs were K103N/S (46.1%) and M184V/I (37.6%). Thymidine analogue mutations were found in 10.6%. Independent risk factors associated with VF were being followed up at the semirural centre (P=0.033), having experienced unstructured treatment interruptions (P=0.0044), and having low CD4+ counts at enrolment (P<0.0001). A longer time on ART (P=0.0008) and being followed up at the rural centre (P=0.021) were risk factors for DRMs. Conclusions This is the first study conducted in Gabon providing VF rates and DRM patterns in adult patients receiving first-line ART. In sub-Saharan Africa, where NNRTI-based regimens are recommended as the standard for first-line ART, strengthening virological monitoring together with preventing unplanned treatment interruptions are a global public health priority.


Introduction
Over the past decade, the roll-out of antiretroviral therapy (ART) has dramatically increased in low-and middle-income countries. More than six million people were receiving ART in these countries at the end of 2010 [1,2]. As ART continues to expand in resource-limited settings, the emergence of HIV drug resistance mutations (DRMs) is challenging. Due to the absence of virological monitoring in routine clinical care together with the use of antiretroviral (ARV) drugs with low genetic barriers (such as 3TC and non-nucleoside reverse transcriptase inhibitors [NNRTIs]), concerns remain regarding the emergence of high-level resistance during first-line therapy. If ART regimens are not effectively delivered, DRMs could become widespread, leading to an increase in therapeutic failures, transmission of drug-resistant viruses, and a decrease in therapeutic options, treatment program effectiveness, and survival. Thus, as recommended by the World Health Organization (WHO), besides determining rates of transmitted DRMs, monitoring acquired DRMs emerging during treatment is crucial in resource-limited countries [3,4].
In Gabon (a country of 1.4 million inhabitants located in central Africa) where HIV-1 seroprevalence is high (approximately 6%) [5], ART became available in 2001 through the gradual implementation of ARV centres. To date, there is a network of 10 centres located throughout the nine provinces of the country [6]. Of the approximately 53,000 Gabonese subjects living with HIV-1, 10,000 were receiving ART by the end of 2009. Currently, ART is initiated when CD4 counts are 5350 cells/ml, according to the WHO 2010 guidelines [7]. Due to financial and technical constraints, biological monitoring is restricted to CD4 counts, and remains associated with limited, if any, plasma HIV-1 RNA viral load (VL) measurements [8]. In Gabon, no study has yet reported the effectiveness of initial ART given through routine HIV care. Here we assessed virological failure (VF) rates and HIV DRMs among approximately 400 patients receiving ART for at least 12 months in two (semirural or rural) ARV centres in Gabon.

Materials and Methods
Study design and settings According to the WHO recommendations [4], we conducted a cross-sectional study by performing a one-time HIV-1 RNA VL measurement free of cost to patients meeting the following criteria: (i) being confirmed HIV-1-positive; (ii) ]18 years old; (iii) being on first-line ART for at least 12 months; and (iv) and consenting to participate in the study. The study was approved by the ethical biomedical research committee of Franceville (No. 023/2010/MESRS/CERB). After information about the study, an oral informed consent was obtained from all patients. All data were anonymized for further analysis.
Patients were consecutively enrolled from two HIV care centres. One ARV centre, located at Franceville (a city of 70,000 inhabitants), was a semirural site managing a total of approximately 1000 patients on ART. The second centre, located at Koulamoutou (a small town of 15,000 inhabitants) was a rural setting with approximately 400 treated subjects. A standardized form was filled out by medical doctors to record the following variables: sex, age, ARV use (date, name, switch), and previous CD4 counts obtained at ART initiation. Previous history of unstructured treatment interruptions (TIs) was also extracted from clinical records. Treatment interrupters were defined as subjects who defaulted and returned to care during their follow-up. When available, the duration of unplanned TIs was recorded. Patients who were transferred to other care centres were not considered as interrupters. If patients had a treatment interruption at the time of blood sampling, they were excluded from this survey.

Laboratory testing
Whole blood specimens were collected in EDTA tubes at each centre. For Franceville, they were processed within six hours at the Centre International de Recherches Médicales de Franceville (CIRMF) Retrovirology Laboratory. After centrifugation, plasma was frozen at (808C. For Koulamoutou, plasma specimens were processed similarly in the centre, stored at (208C, and transported by road in a cool box once per week to the CIRMF Retrovirology Laboratory. HIV seropositivity was confirmed using an in-house serotyping V3 immunoassay [9]. CD4 counts at enrolment were determined with flow cytometry (Fascount, Becton Dickinson, San Jose, CA). Plasma HIV-1 RNA levels were determined using the HIV Generic Viral Load assay (Biocentric, Bandol, France) [10]. This HIV-1 RNA test has a sensitivity threshold of 300 copies/ml by using 200 ml of plasma. Genotypic DR tests were carried out among viremic patients with VF (defined as a HIV-1 RNA viral load result ]1000 copies/ml, according to the WHO 2010 recommendations) [11], using a previously described in-house assay [12]. ARV drug resistance mutations (DRMs) were identified and interpreted using the Agence Nationale de Recherches sur le SIDA (ANRS) algorithm (May 2011) (http://www. hivfrenchresistance.org). Mutations were also studied using the Stanford HIVdb genotypic resistance algorithm and then coded as major resistant mutations using the International AIDs Society (IAS) list from January 2012 (http://hivdb. stanford.edu). HIV-1 subtyping was performed by phylogenetic analysis of pol region sequences. Our laboratory participated in the 2011 quality control assessment of HIV-1 drug resistance sequencing implemented by the ANRS [13].

Statistical analysis
All statistical analyses were performed using Stata 10 TM software (Houston, TX). First, a comparison of patients' characteristics between the two care centres was made using the x 2 for categorical variables and the Mann-Whitney-U test for continuous variables. Second, VF (]1000 copies/mL) and DRM prevalence rates with 95% confidence intervals (95% CIs) were determined. Third, logistic regression models were used to examine the following explanatory variables as potential factors associated with VF: gender, age, centre, time on ART, unstructured TIs, and CD4 count at ART initiation (month 0) and at the time of VL assessment. The same seven variables plus HIV-1 RNA levels were further examined as predictors of at least one significant DRM among patients with VF. Finally, all variables with P-values B0.2 in univariate analyses were entered in multivariate analyses.

Results
From March 2010 to March 2011, a total of 375 adult patients on first-line ART were consecutively included (Table 1). In both sites, more women were enrolled and most patients were thus women (74%). The median age of women was slightly lower than for men (40.8 years versus 47.7 years). Except one woman who tested positive to subtype O strain, all subjects harboured HIV-1 variants belonging to group M. A large majority (357/375, 95.2%) of patients were receiving two nucleoside reverse transcriptase inhibitors (NRTIs) plus one NNRTI (combination of D4T/AZT/ABC'3TC'NVP/EFV) whereas approximately 4% (n 014, including 11 women) started a protease inhibitor (PI)-based regimen. Four (approximately 1%) additional patients (including three women) were receiving TDF'3TC'EFV. The switch mainly concerned replacement of D4T by AZT or NVP by EFV. At ART initiation (M0), the overall median CD4 ' count was 159 cells/ml (IQR, 70Á248), with no significant difference between women (158 cells/ml) and men (164 cells/ml). At the time of enrolment, the median time on treatment was 33.6 months (range, 12Á107) and was slightly higher for males (35.4 months) compared with females (32.3 months). The median CD4 ' count reached 337 cells/ml (IQR, 215Á492), with very similar values for women (334 cells/ml) and men (343 cells/ml). Previous unstructured TIs were recorded in more than 40% of patients (in women and men), with an overall median duration of seven months. There was no significant difference in the patients' characteristics between the two care centres except the median time on ARV therapy, which was slightly longer in the semirural site than in the rural centre (P 00.03). Also unstructured TIs were more frequently recorded in patients Liégeois F et al By using the ANRS algorithm, among the 141 sequenced specimens, 80 (56.7%) harboured ]1 DRM (Figure 1). Overall, 53 (37.6%) showed dual-class resistance to NRTIs and NNRTIs, 24 (17.0%) harboured resistance to NNRTIs only, and three (2.1%) were only resistant to NRTIs. Around twothirds of viral strains from patients with DRMs were resistant to at least two or three of the drugs constituting their ARV regimens. No clinically relevant protease resistance mutation was found amongst the four viremic patients with first-line PI-based regimens. Identical results were obtained using the Stanford HIVdb algorithm.

Discussion
Here, for the first time in Gabon, we provided VF rates and DRM patterns in adult patients living with non-B HIV-1 strains and receiving mostly (!95%) first-line thymidine analogue, 3TC, and NNRTI-based ART for a median period of 34 months. Our survey, conducted in semirural and rural settings, first revealed a high rate of VF, with 41% of patients showing VL results ]1000 copies/ml. In our survey, a history of unplanned treatment interruptions and CD4 ' counts at inclusion were strong predictors of VF. Also, a higher VF rate was found for the semirural care centre. Second, among participants who virologically failed ART, 56.7% carried ]1 DRM and almost 40% had dual-class resistance to NRTIs and NNRTIs. DRMs were consistent with expected patterns given first-line NNRTI-based therapy predominantly used in Gabon. The proportion of patients with DRMs significantly increased with time on ART and was higher for the rural site.
Comparison of VF (or success) and DRM profiles rates obtained in different African populations must be made cautiously because study design, VL assays and threshold used, and duration of ART at time of failure may vary. In a review, Barth and colleagues reported virological success (defined by a VL B400 cp/ml) rates in sub-Saharan Africa of 78%, 76%, and 67% after 6, 12, and 24 months of ART, respectively [8]. Compared with those pooled data, the success rates found herein by using an identical sensitivity threshold ( B400 cp/ml) were consistent for the rural centre (68% for a median of 28 months of treatment) but potentially lower for the semirural centre (53.0% for a median of 35 months of treatment).
In addition, the burden of DRMs found in our survey was consistent with previous studies in which prevalence of DRMs showed highly contrasted results, with NNRTI mutations ranging from 47 to 100%, the M184V/I mutation from 24 to 81%, and any TAM from 0 to 63% [14Á22]. However, in our study, it must be noted that patients from the semirural site had infrequent DR mutations, albeit with a higher rate of VF as mentioned here. One reason for explaining varying patterns of VF and drug resistance may be related to transient episodes of viremia due to patient-initiated unstructured treatment interruptions that are a reality of routine clinical care in African ART programs [23]. In our survey, they were more frequent at the semirural site in comparison with the rural centre. They implied patientrelated factors (such as incomplete adherence, difficulties with payment for travelling to the HIV care centre and/or for routine biological tests, and toxicity and side effects of ART regimens), as well as program-related factors (such as pharmacy stock-outs). The impact of structured treatment interruptions (STIs) on virological outcomes has been previously reported through randomized STI trials [24Á26]. For instance, in the Trivacan trial conducted in Ivory Coast, Danel et al. obtained 24% of drug-resistant viruses in a twomonths-off, four-months-on ART arm, compared with 9% in a continuous strategy arm, with a highly significant difference for resistance to mostly NNRTIs. By contrast, very few studies ''in real world settings'' investigated the impact of unplanned treatment interruptions on the development of DRMs [23,27]. In our setting, we think that early transient viremia may occur prior to the accumulation of DRMs or may be associated with a reversion to wild type viruses. As a consequence, among patients having experienced unstructured treatment interruptions, resistance would not be identified as frequently.
Our study is limited by its cohort cross-sectional design and does not provide any information on the overall efficiency of the national ART program. Even if it is known that rates of lost to follow-up are higher in the first year and tend to decrease thereafter, no information was available on the dropout/death rates of the clinics and on the presence (or not) of DRM before ART initiation. In addition, subjects with a VL result ranged between 300 and 1000 copies/ml may harbour mutant HIV strains with DRMs. Therefore, the drug resistance rate that we obtained in our study is most likely a minimal estimate. Another limitation of our study is represented by the fact that no data on baseline prevalence of HIV drug resistance was available for our studied population. However, all studies recently conducted among untreated Gabonese subjects revealed low levels ( 51%) of transmitted HIV DR [5,28,29]. Finally, we were not able to fully investigate gender-related differences in virological outcomes given the relatively small sample size for men and the lack of information for a history of ARV prophylaxis to prevent mother-to-child transmission (MTCT) among women enrolled in our study.

Conclusions
In summary, this is the first study conducted in Gabon providing VF rates and DRM patterns in adult patients receiving first-line ART. In sub-Saharan Africa, where NNRTIbased regimens are recommended as the standard of care for first-line ART, strengthening virological monitoring aimed at early failure detection, together with preventing unintended treatment interruptions, are a global public health priority.
Access to virological monitoring should be also increased to limit the level and complexity of drug resistance observed in non-B HIV-1 strains circulating in resource-limited settings. Increase in VL monitoring will prevent inappropriate switches and preserve susceptibility to second-line regimens in Africa [30Á32]. As reported recently by Wallis and colleagues in South Africa [33], it is feasible to obtain low ( 510%) rates of VF and less complex drug resistance profiles using the D4T'3TC'EFV/NVP regimens.
Finally, the WHO/HIVResNet Global HIV drug resistance (HIVDR) prevention, surveillance, and monitoring strategy must be strengthened worldwide at sentinel sites to inform national health authorities on the efficiency of first-line ART and to allow recommendations on future ART strategies. Our study showed that, in the vast majority of our patients, the second-line regimen recommended by WHO would be active in our settings. Nevertheless, a minority showed complex drug resistance profiles predictive of resistance to the usual second-line regimens. Emerging multi-NRTI resistance in sub-Saharan Africa would not only compromise second-line treatment options and the success of antiretroviral roll-out but could also contribute to the spread of drug-resistant variants worldwide [34]. Thus frequent VL monitoring is definitively required to limit the threat of increased transmitted resistance in the face of continuing incident infection as observed in Gabon.