Prevalence of Antiretroviral Drug Resistance in Patients Who Are Not Responding to Protease Inhibitor–Based Treatment: Results From the First National Survey in South Africa

Abstract

Limited data exist on human immunodeficiency virus type 1 (HIV-1) resistance in patients who are not responding to protease inhibitor (PI)–based regimens in resource-limited settings. This study assessed resistance profiles in adults across South Africa who were not responding to PI-based regimens. pol sequencing was undertaken and submitted to the Stanford HIV Drug Resistance Database. At least 1 major PI mutation was detected in 16.4% of 350 participants. A total of 53.4% showed intermediate resistance to darunavir/ritonavir, whereas high-level resistance was not observed. Only 5.2% and 32.8% of participants showed high-level and intermediate resistance to etravirine, respectively. Although the prevalence of major PI mutations was within previously reported ranges, most patients will likely experience virological suppression during receipt of currently available South African third-line regimens.

HIV-1 antiretroviral drug resistance, PI-based treatment failure, South Africa

Antiretroviral treatment (ART) has reduced morbidity and mortality in human immunodeficiency virus type 1 (HIV-1)–infected patients [1]. However, the number of patients needing PI-based second-line ART has increased as programs mature [2]. Moreover, knowledge of clinical outcome and antiretroviral drug resistance profiles in patients who are not responding to PI-based ART in South Africa is limited.

Globally, South Africa has the largest ART program, with approximately 3.1 million people receiving ART [3], of whom 144 000 are receiving second-line treatment (March 2014, personal communication, Clinton Health Access Initiative report on pharmacy records). Before 2010, South African Public Sector ART guidelines recommended second-line therapy of zidovudine (AZT), didanosine (ddI), and ritonavir-boosted lopinavir (LPV/r), and recommendations changed to AZT (after first-line TDF) or tenofovir (TDF; after first-line AZT/stavudine), lamivudine (3TC), and LPV/r after 2010.

In 2013, public sector HIV drug resistance testing and salvage drugs for patients with PI-resistant virus became available. However, representative HIV drug resistance data on PI-based ART failure in South Africa is limited. Three small studies (with 33, 33, and 75 participants and performed during 2009–2010) showed that 0%–7% of patients had no response to a LPV/r-based regimen, with ≥1 major PI mutation after a median of 10–16 months of PI-based ART [4–6]. A larger cross-sectional study (with 490 participants; 2006–2012) reported a prevalence of 11% among patients who were not responding to PI-based ART [7]. Two recent studies (with 58 and 65 participants and performed during 2012–2013) noted ≥1 major PI mutation in 24.1%–26% of patients who were not responding to PI-based regimens [8, 9]. Therefore, the first national cross-sectional survey was undertaken to assess HIV drug resistance profiles in patients across South Africa who were not responding to PI-based regimens.

MATERIALS AND METHODS

Survey Design

A cross-sectional survey was designed using probability-proportional-to-size sampling per province. The number of patients receiving ART in facilities with ≥2000 patients in ART care was considered a proxy of the relative number of patients receiving PI-based treatment per region. Patients aged ≥18 years who were not responding to a boosted PI-based regimen and provided signed informed consent were eligible for the study. Virological failure was defined as ≥1 HIV-1 RNA measurement of >1000 copies/mL after exposure to a PI-based regimen for ≥6 months.

Data and Sample Collection

Healthcare professionals selected eligible participants from 72 facilities across 38 districts in all 9 provinces (Supplementary Table 1 and Figure 1) between February 2013 and October 2014. A study questionnaire was completed, and specimens were processed at 1 of 3 HIV drug resistance testing laboratories in South Africa (each successfully participates in ≥1 international external quality assessment panel per annum or an interlaboratory comparison panel). Each laboratory generated pol sequences, using their validated population-based in-house genotyping method [10–12].

Sequence Analysis

The Stanford HIV Drug Resistance Database, version 7.0.1 (available at: http://sierra2.stanford.edu/sierra/servlet/JSierra), was used to identify HIV drug resistance mutations and generate resistance profiles, categorized as susceptible, intermediate, or high level. Sequences were aligned using Muscle, and a phylogenetic tree was constructed (PhyML and FastTree) to detect cross-contamination and duplicates. Subtypes were assigned using the Rega HIV subtyping tool, version 2.0 (available at: http://bioafrica.net/rega-genotype/html/subtypinghiv.html); the National Center for Biotechnology Information genotyping tool (available at: http://www.ncbi.nlm.nih.gov/projects/genotyping/); and the jumping profile Hidden Markov Model (available at: http://jphmm.gobics.de/submission_hiv.html).

Statistical Analysis

Statistical analyses were performed using GraphPad Prism 6 and SAS, version 9.3. Differences between groups in nonparametric data, such as age, HIV-1 RNA level, and duration of treatment, were calculated using the 2-sided Wilcoxon–Mann–Whitney test. Two-sided Fisher exact tests were used to compare mutation prevalence between treatment groups and sex categories, with a P value of <.05 considered statistically significant. The modified Wald method was used to calculate 95% confidence intervals (CIs). Log-binomial regression analysis tested the association between different mutations and antiretroviral drug exposure. Relative risks (RRs) and corresponding 95% CIs are presented. Mutations identified as being significantly associated with drug exposure in the univariate analysis (P < .2) and a priori variables of importance were included in multivariate models. Models were adjusted for CD4⁺ T-cell count, HIV-1 RNA levels, total duration of ART, and duration of current ART regimen, where data were available.

Sequence Data

Pol nucleotide sequences were submitted to GenBank (Sequin v9.50; available at: www.ncbi.nlm.nih.gov/Sequin), with accession numbers KU252880-KU253229.

Ethics Statement

The University of the Witwatersrand Research on Human Subjects (Medical) Committee provided ethics clearance (M120254). Written informed consent was obtained from all participants.

RESULTS

Demographic and baseline clinical characteristics of participants are presented in Table 1 and Supplementary Table 2. The median duration of ART exposure was 62 months, with a median duration of PI exposure of 25 months. The median HIV-1 RNA load was 4.8 log₁₀ copies/mL, which increased by 0.2 log₁₀ copies/mL between consecutive measurements (P < .001). Approximately 80% of participants were not responding to the recommended PI-based second-line regimens (Table 1). At the time of HIV drug resistance testing, 158 of 350 participants received a combination of TDF-3TC-LPV/r, and 120 participants received AZT-3TC-LPV/r. Six participants received an atazanavir/ritonavir (ATV/r)–based regimen, of whom 5 were previously exposed to LPV/r, and 3 received a LPV/r-based regimen with prior exposure to ritonavir-boosted saquinavir (n = 2) or ATV/r (n = 1). No exposure to an unboosted PI regimen was recorded. For further analysis, participants were classified as those who were not responding to TDF-3TC-LPV/r (hereafter, the “TDF group”; n = 158), those who were not responding to AZT-3TC-LPV/r (hereafter, the “AZT group”; n = 120), and those in a group referred to hereafter as the “other group” (n = 72). No differences were observed in HIV-1 RNA loads and CD4⁺ T-cell counts between treatment groups. However, significant differences were observed for the median total duration of ART and duration of PI exposure between different groups (Supplementary Table 3).

Table 1.

Demographic and Baseline Clinical Characteristics Among Participants Throughout South Africa Who Were Not Responding to Protease Inhibitor (PI)–Based Regimens

Characteristic	Participants, No. (%)	Median	IQR	Range
Female sex	242 (69.1)
Age, y	350 (100.0)	38	32–45	18–69
Time between HIV diagnosis and study sample collection, mo	298 (85.1)	79	58–107	11–248
CD4+ T-cell count, cells/mm3
Baselinea	269 (76.9)	99	37–170	1–931
Latest	330 (94.3)	201	99–295	1–984
Time between latest CD4+ T-cell count and study sample collection, d	329 (94.0)	112	34–217	0–1066
HIV-1 RNA level, log copies/mL
Latest	343 (98.0)	4.8	4.2–5.3	3.0–6.5
Previous	325 (92.9)	4.6	3.9–5.1	1.6–6.9
Time between latest HIV-1 RNA and study sample collection, d	343 (98.0)	50	28–111	0–538
Time between previous and latest HIV-1 RNA levels, d	324 (92.6)	190	120–280	23–2522
Duration of current regimen, mo	335 (95.7)	24	13–38	1–105
Total duration of PI treatment, mo	335 (95.7)	25	14–43	7–138
Total duration of ART, mo	336 (96.0)	62	46–85	10–165
ART use at time of HIV drug-resistance testing
TDF + 3TC + LPV/r	158 (45.1)	…	…	…
AZT + 3TC + LPV/r	120 (34.3)	…	…	…
AZT + ddI + LPV/r	23 (6.6)	…	…	…
ABC + 3TC + LPV/r	13 (3.7)	…	…	…
d4T + 3TC + LPV/r	10 (2.9)	…	…	…
Other triple combinationb	13 (3.7)	…	…	…
Dual combinationb	5 (1.4)	…	…	…
Quadruple combinationb	8 (2.3)	…	…	…

Abbreviations: ABC, abacavir; ART, antiretroviral therapy; AZT, zidovudine; ddI, didanosine; d4T, stavudine; HIV, human immunodeficiency virus; IQR, interquartile range; LPV/r, ritonavir-boosted lopinavir; TDF, tenofovir; 3TC, lamivudine.

^a Earliest available CD4⁺ T-cell count in the patient's medical record.

^b Details of these regimens are listed in Supplementary Table 2.

Overall, 350 sequences were available for HIV drug resistance analysis, of which 348 were HIV-1 subtype C. Fifty-eight participants (16.4%; 95% CI, 12.9%–20.6%) presented with virus containing ≥1 major PI mutation. Although participants who were not responding to an AZT-based regimen showed the lowest prevalence of PI mutations, likely due to shorter treatment duration, PI resistance prevalence between treatment groups did not differ significantly. Common major PI mutations were I54V (12.7%), V82A (12.1%), and M46I (10.2%; Supplementary Figure 2). Participants with virus exhibiting ≥1 major PI mutation were 68% less likely to have an HIV-1 RNA load of >5 log₁₀ copies/mL (adjusted RR, 0.32; 95% CI, .15–.65; P = .002). Differences in median PI exposure between participants with (35 months) and those without (24 months) PI mutations were not statistically significant. However, PI exposure for ≥8 years increased the risk of PI resistance 5.62-fold relative to exposure for <2 years (adjusted RR, 5.62; 95% CI, 2.25–14.0; P < .001).

Of the 58 participants harboring HIV-1 with PI mutations, 16 and 2 presented with a combination of nucleoside reverse transcriptase inhibitor (NRTI)/PI resistance or non-NRTI (NNRTI)/PI resistance, respectively, while 40 had triple-class resistance (Table 2 and Supplementary Table 4). Based on current South African guidelines, these 58 participants were eligible for third-line ART. Forty-seven and 11 patients presented with high-level resistance and intermediate resistance to LPV/r, respectively. High-level resistance to DRV/r was not observed, but 53.4% (31) had intermediate resistance. Only 3 participants (5.2%) and 19 (32.8%) had detectable high-level and intermediate resistance to etravirine.

Table 2.

Overall Drug-Class Resistance Among Participants Throughout South Africa Who Were Not Responding to Protease Inhibitor (PI)–Based Treatment

Variable	Overall, No. (%) (n = 350)	TDF Group,a No. (%) (n = 158)	AZT Group,b No. (%) (n = 120)	Other,c No. (%) (n = 72)
Wild-type virusd	87 (24.9)	38 (24.1)	25 (20.8)	24 (33.3)
NNRTI resistance only	52 (14.9)	23 (14.6)	18 (15.0)	11 (15.3)
NRTI resistance only	25 (7.1)	9 (5.7)	10 (8.3)	6 (8.3)
NNRTI + NRTI resistance	128 (36.6)	59 (37.3)	52 (43.3)	17 (23.6)
PI + NRTI resistance	16 (4.6)	9 (5.7)	4 (3.3)	3 (4.2)
PI + NNRTI resistance	2 (0.6)	0 (0.0)	0 (0.0)	2 (2.8)
Triple-class resistance	40 (11.4)	20 (12.7)	11 (9.2)	9 (12.5)

Abbreviations: AZT, zidovudine; LPV/r, ritonavir-boosted lopinavir; NNRTI, nonnucleoside reverse transcriptase inhibitor; NRTI, nucleoside reverse transcriptase inhibitor; TDF, tenofovir; 3TC, lamivudine.

^a Patients who were not responding to TDF + 3TC + LPV/r.

^b Patients who were not responding to AZT + 3TC + LPV/r.

^c Patients who were not responding to other PI-based regimens.

^d Viruses with only polymorphisms (T74S, Q58E, and E138A)

No clinically relevant mutations were detected among viruses in 87 participants (24.9%; 95% CI, 20.6%–29.6%). Despite ongoing NRTI exposure, ≥1 major NRTI mutation was detected in 59.7% of participants, with M184V the most common mutation (52.3%), followed by D67N (13.3%), M41L (10.7%), and K70R (10.2%). Although 47.7% of participants were not responding to a thymidine-containing regimen, thymidine analogue mutation prevalence remained relatively low: 26.3% and 9.0% of participants presented with ≥1 and ≥3 thymidine analogue mutations, respectively.

There was no NNRTI exposure at time of HIV drug resistance testing for 347 participants, but 65.0% had ≥1 NNRTI mutation. K103N (n = 120) was most common, followed by G190A (n = 45) and V106M (n = 43). Efavirenz or nevirapine resistance was not detected in 38.7% and 36.4% of participants, respectively. Similarly, no etravirine or rilpivirine resistance was detected in 68.9% and 56.5% of participants, respectively.

Participants with detectable PI mutations were followed up for third-line ART referral and virological suppression after survey completion (Supplementary Figure 3). Follow-up HIV-1 RNA levels were accessed up to December 2015; the median time between sample collection and the last available HIV-1 RNA measurement was 15 months (interquartile range, 11–21 months). Information on treatment adherence and date of switch to a third-line regimen was unavailable. Twenty-one of 58 participants switched to third-line ART, of whom 16 were virologically suppressed, 13 participants were not referred for third-line treatment, and 6 experienced virological suppression. No follow-up records were available for 24 participants.

DISCUSSION

The South African HIV national treatment program faces unique challenges because of the large number of individuals requiring treatment and conditions that inevitably strain a resource-limited public healthcare system. Assessing the success of the ART program has been a priority; thus, the first national survey among adults who were not responding to PI-based ART was conducted. Findings agreed with the national ART guideline changes: on the duration of treatment in the TDF group was longer than that in the AZT group. The longer treatment duration observed in the other group might reflect earlier unconventional drug switches owing to a lack of third-line ART at the time or exposure/access to these drugs in the private sector.

The prevalence of PI resistance (16.5%) has increased as compared to values of 0%–7% in earlier reports from South Africa [4–6], but it is comparable to findings ranging from 11.2% to 26.0% in recent reports [7, 13, 14]. This increase over time is attributed to maturation of the ART program and the longer treatment duration among patients. Additionally, the prevalence of PI mutations increases with on the duration of PI-based ART. Based on the HIV drug resistance profiles, DRV/r remains a feasible option for third-line regimens.

A significantly higher proportion of patients who were not responding to PI-based regimens (24.9%) had no HIV drug resistance detected as compared to those who were not responding to first-line NNRTI-based regimens (3.7%) [8]. The low HIV drug resistance prevalence and high genetic barrier regimen suggests that adherence contributes to virological failure in PI-based ART. This is supported by a recent South African study showing that the viral load in up to 64% of patients with no virological response to PI-based treatment can be resuppressed after intensified adherence counseling [13].

The finding that 1 in 3 participants presented with at least 1 NNRTI mutation is in line with previous studies showing that NNRTI mutations persists in two thirds of patients after withdrawal of NNRTI pressure [9]. However, substantial person-to-person variability in mutation replacement rates has been observed [15]. Notwithstanding NRTIs included in the PI-based regimens, the frequency of NRTI mutations in this cohort (59.7%) was substantially lower than in an NNRTI-based treatment failure cohort (91.0%) studied during the same period [8]. Results highlight the importance of HIV drug resistance genotyping and obtaining an accurate patient ART history at the time of PI regimen failure when selecting a third-line regimen, because of the dynamic evolution of HIV-1 quasispecies containing NRTI and NNRTI mutations over time. This is especially relevant when considering the inclusion of NRTI/NNRTIs in third-line regimens.

This cross-sectional observational survey has limitations: it relied on healthcare workers to obtain written consent from participants, complete study questionnaires, and collect specimens. This additional burden on understaffed clinics may have contributed to the extended sample collection time and undersampling in some areas. Although participants were not systematically sampled at the clinic level, all had to have at least 1 measurement in which HIV-1 RNA was detectable and to have received a PI-based regimen for ≥6 months, thereby reducing selection bias. A national distribution of participant samples was obtained, and although the study did not reach planned sample size, it had sufficient numbers to estimate with adequate precision the proportion of patients with HIV drug resistance mutations who were not responding to a PI-based regimen.

Data quality depended on the completeness/accuracy of questionnaires, and despite detailed questions, data capturing may have been incomplete at the source clinic because transfers between healthcare facilities are common. Finally, HIV drug resistance mutations were detected by the gold standard, population-based Sanger sequencing, which underestimates the prevalence of minority variants harboring HIV drug resistance. The clinical relevance of undetected HIV drug resistance minority variants is unknown, and although population-based sequencing may suggest susceptibility to particular therapeutics, minority drug-resistant HIV variants could emerge with drug pressure, leading to treatment failure. Next-generation/high-throughput sequencing could answer these questions, but no validated assay using these technologies is currently used for diagnostic purposes in the South African public sector. Despite limitations, the national representation and sample size presented here compose the best available data on HIV drug resistance profiles after PI-based failure in the South African public sector to date.

Overall, our findings confirm that there is currently no need to change the PI-based second-line regimens in the South African national ART program. The importance of adherence support and counseling in patients who are not responding to PI-based treatment is highlighted. Despite detection of moderate levels of PI resistance in the present study, the numbers will increase as ART programs mature, inevitably leading to an increased demand for access to third-line regimens. Thus, increased advocacy is needed to ensure HIV drug resistance testing for PI-based treatment failures is implemented, and third-line regimens in resource-limited settings are easily accessible.

Notes

Acknowledgments. We thank all study participants, healthcare workers, and regional ART program coordinators, who made this study possible; E. Letsoalo, A. Lukhwareni, C. Mutuku, L. Skhosana, H. Zwane, A. Bester, and M. Claassen, the staff at the HIV drug resistance testing laboratories, who ensured that all samples were processed in a timely manner; and L. Noble, for assisting with the data capturing.

Disclaimer. The views described herein do not represent the views or opinions of the Global Fund to Fight AIDS, Tuberculosis and Malaria, nor is there any approval or authorization of this material, express or implied, by the Global Fund to Fight AIDS, Tuberculosis and Malaria. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Financial support. This work was supported by The Global Fund to Fight AIDS, Tuberculosis and Malaria.

Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed. None of the authors declare any commercial or other association that might pose a conflict of interest.

Author notes

Presented in part: International AIDS Conference, Vancouver, Canada, July 2015. Poster TUPEB289.