The rapid scale-up of highly active antiretroviral therapy (HAART) for patients infected with HIV in Southern Africa has been based on a pragmatic public health approach.1-2 Treatment comprises standardized first-line and second-line regimens, and viral load testing is not generally recommended for guiding decision-making on switching regimens due to treatment failure. Data on epidemiological associations with virologic failure in routine scale-up settings in Southern Africa are therefore not generally available.
South Africa is one of the few countries in the region to routinely provide regular viral load monitoring for patients on HAART and to base regimen switching on confirmed virologic failure.3 This provides an opportunity to explore these associations in the context of a treatment program that otherwise follows the public health approach.
There are many unanswered questions about the antecedents of treatment failure in this setting, in which patients initiate treatment with very advanced disease,4-5 nonnucleoside reverse transcriptase inhibitors (NNRTIs) are universally used in first-line therapy in treatment-naive patients, there may be biological differences in how some drugs are metabolized,6 women are frequently exposed to single agent antiretrovirals in the prevention of mother-to-child transmission (PMTCT) of HIV,7 there are gender differences in service utilisation,8 and retention in uninterrupted care is an increasing challenge.9
The aim of this study was to identify factors associated with virologically confirmed treatment failure in adults in the South African public sector antiretroviral treatment program, through a case-control study covering an entire province. The Western Cape Provincial HAART program is uniquely placed to explore these associations due to the program having started as early as 2001 and due to the comprehensive monitoring system in place providing a sampling frame for a case-control study.5
The first clinics to provide HAART in the public sector in the Western Cape province were opened in May 2001,4 and enrollment was accelerated from 2004 with the launch of the national treatment program. The provincial population is largely concentrated in the Cape Town metropolitan district and includes densely populated townships such as Khayelitsha and Gugulethu.
The South African national and Western Cape provincial treatment guidelines provide for HAART to be prescribed to adults with a CD4 count below 200 cells per microliter or with a World Health Organization (WHO) stage IV illness other than extra-pulmonary tuberculosis. There is a fairly uniform treatment readiness program that involves 3 patient education sessions before HAART commencement, either individually or in groups, and typically in consecutive weeks. There is variable use of patient-selected treatment assistants, community health worker patient supporters, support groups, pillboxes, and pill counts. The starting regimen comprises 2 nucleoside reverse transcriptase inhibitors (NRTIs) and 1 NNRTI. The majority of patients start therapy on stavudine (d4T), lamivudine (3TC), and either efavirenz (EFV) or nevirapine (NVP), as individual drugs rather than fixed-dose combinations. The choice of NNRTI is informed by gender (NVP is suggested for women of child-bearing age not on reliable contraception), and EFV is suggested for patients concurrently on treatment for tuberculosis. Zidovudine (AZT) was used in place of d4T in the Khayelitsha district until the end of 2003.
Guidelines provide for 6 monthly viral load and CD4 cell count monitoring. Viral load tests were initially done using the NucliSens HIV-1 QT assay and later the NucliSens EasyQ HIV-1 assay (bioMérieux, Boxtel, the Netherlands). Patients with viral load values above 400 copies per milliliter are provided with a stepped-up adherence package comprising counselling, and in some instances pill counts and the reintroduction of pillboxes. The viral load is repeated after 3 or 6 months based on the viral load result (≥5000 copies/mL and 400-4999 copies/mL, respectively), and if above 5000 copies per milliliter, second-line therapy is recommended. One facility, representing less than 20% of patients included in the study, used a threshold of 1000 as opposed to 5000 copies per milliliter for the repeat viral load measurement. Baseline viral load measures were no longer routinely offered in the Western Cape after mid-2005.
PMTCT has been available since 1999, initially in Khayelitsha with the maternal component based on late antenatal (from 34 weeks gestation) and intrapartum AZT,10 followed from 2001 onwards by the progressive introduction of single-dose NVP at the onset of labor, and from late 2003 a combination of both antenatal AZT and NVP at the onset of labor.7
Study Design, Population, and Data Sources
We did a matched case-control study based on medical records of patients attending free public sector HAART clinics in the Western Cape Province.
To limit travel across the province, we included only those clinics that reported 4 or more eligible cases at the start of the study in June 2006. Cases and controls were identified from clinic registers for the period May 2001 to June 2006. We first ascertained that the registers provided a comprehensive list of clinic patients by randomly sampling 169 patient folders from participating clinics; 165 (98%) of these patients were listed in the registers. All patients in the study had started HAART at the clinic, and at the time of starting, they were naive to HAART and aged 14 years or more.
Defining Cases and Controls
Cases were defined as patients switched to second-line treatment due to virologic failure, defined as having had 2 consecutive viral loads greater than 1000 copies per milliliter. All cases meeting the criteria were included whether or not they were alive at the time of ascertainment. Controls were on first-line treatment at the time of case incidence and had viral load <400 copies per milliliter at that time. Patients were listed in the registers in chronological order of starting HAART. The fifth patient before a case was selected as the first control and the fifth patient after the case as the second control. If a selected patient did not meet the eligibility criteria, the previous or following patient was assessed for suitability. Duration of follow-up for controls was censored to match the duration of follow-up to treatment failure for the matched case for the purposes of risk factor assessment. Controls were therefore matched to cases by clinic, calendar period of starting treatment, and duration of treatment.
Data Extraction and Coding
Data were extracted chronologically from the medical records, thereby concealing patient outcomes until the end of capture in each instance. Quality of extraction was monitored by auditing samples of 5-10 folders at 3 time points in each clinic.
All variables except age were categorical and included an “unknown” category for missing data. Visit-level data were categorized separately before HAART and on HAART. Variables were organized into 7 groups as follows: “demographic and social”: age, sex, spoken language, partnership status, and HIV status of partner; “adherence-related behavior”: frequency of being late (≥1 day) for clinic appointments, travel to another province, and having interrupted treatment (≥1 week) by reason for the interruption; “baseline socioeconomic”: employment status, type of dwelling, number of occupants in the household; “baseline laboratory measures”: CD4 count, interval between measuring baseline CD4 and starting HAART, and log10 viral load; “Health status”: PMTCT exposure by type of PMTCT, WHO clinical stage, pregnancy during HAART, tuberculosis treatment during HAART, and hospitalizations during HAART; “drug regimen”: AZT or d4T as the second NRTI with or without substitution and NVP or EFV as the NNRTI with or without substitution; and “clinical disease mechanisms that could lead to failure”: oral pathology (eg, candidiasis) which could impact on swallowing; gastrointestinal pathology such as diarrhea and vomiting which could affect drug absorption, disfiguring conditions which could lead to stigma such as visible Kaposi sarcoma, and presence of comorbidities with and without additional pill burdens.
Sample Size and Statistical Analysis
At the planning stage of this study, we anticipated identifying 150 cases and set out to select 2 controls per case, that is, 300 controls in all. These numbers would have given 80%-84% power (at a 2-sided significance level of 0.05) to detect an odds ratio of 1.8, 2.0, 2.3, or 2.8 for risk factors with a prevalence in controls of 40%, 20%, 10%, or 5%, respectively.
Data were analyzed using SPSS v. 15.0 and Stata v. 9.0. Using conditional logistic regression, crude odds ratios were calculated for each of the variables. Variables in which any category had a P value <0.10 were entered into a full conditional logistic regression analysis using a forward stepwise procedure (P < 0.05 for entry). In this final analysis, age was treated as a continuous variable. To confirm the robustness of the model, we also examined manually the effects of possible confounders and carried out backward stepwise selection. All P values across categorical variables were based on likelihood ratio tests.
We obtained ethical approval for assembling and analyzing routinely collected clinical data for this study from the University of Cape Town Research Ethics Committee.
Eight clinics, treating 9249 of the 18,703 patients on HAART at the time, were included (Fig. 1). Five were primary care clinics and 3 hospital outpatient clinics; all were in the Cape Town metropolitan area.
Number of Cases and Controls and Duration of Treatment
Of the 139 identified cases in 8 clinics, 130 (94%) were included in the study (clinical records could not be retrieved for 8, and for 1 patient controls meeting the matching criteria could not be identified). The included cases represented between 0.4% (Mitchell's Plain) and 2.6% (Nolungile) of all patients who had started HAART in the study clinics (Fig. 1). Data were available for 238 (92%) of the 260 identified controls-this meant that for 22 cases, only 1 control was included.
All cases and controls had started HAART between 2001 and 2005, with more than half starting in 2004 or 2005 [76 of 130 cases (59%) and 146 of 238 controls (61%)]. Median (interquartile range) duration of treatment for the 130 case-control triplets or pairs was 16.6 months (12.2-24.6).
Characteristics of Control Population
Among controls, the mean (SD) age at starting treatment was 34.7 (7.7) years; the majority (72%, 172 of 238) were female. Most were Xhosa speakers (79%, 189 of 238) and only 27% (64 of 238) were employed, the remainder having no employment income or receiving disability grants. Baseline (at HAART initiation) CD4 count was low; the median (interquartile range) was 76 cells per microliter (28-138). The 4 most frequent starting regimens accounted for 97% of starting regimens for controls: d4T/3TC/EFV (40%), AZT/3TC/EFV (23%), d4T/3TC/NVP (21%), and AZT/3TC/NVP (13%). For the 134 female controls for whom PMTCT exposure could be determined, 21 (16%) had received PMTCT interventions.
Associations With Virologic Failure
Tables 1 and 2 show the variables selected for the final model-building process after preliminary analysis. Spoken language, type of dwelling, interval between baseline CD4 count and starting HAART, log10 viral load, WHO clinical stage, hospitalizations during HAART, choice of NRTI, disfiguring HIV-associated conditions, and the presence of other comorbidities were not associated with the outcome in univariate analysis.
In the final model (Table 3) treatment interruption, NVP-based PMTCT, low baseline CD4 count, and NVP as the NNRTI were independently associated with virologic failure. Patients who interrupted treatment through defaulting or nonadherence increased their odds of treatment failure more than 8-fold [odds ratio (OR): 8.6, 95% confidence interval 3.6 to 20.8], whereas prior NVP-based PMTCT exposure was also associated with greatly increased odds (OR: 9.6, 2.9 to 32.2). A baseline CD4 count less than 50 cells per microliter or from 50-150 cells per microliter increased the odds of failure more than 5-fold (OR: 6.6, 2.3 to 18.8 and OR: 5.8, 2.1 to 16.3, respectively) compared with a baseline CD4 count of more than 150 cells per microliter. Using NVP as the only NNRTI more than doubled the odds of failure (OR: 2.5, 1.4 to 4.7).
Data on the timing of PMTCT exposure relative to starting HAART were available for 12 of 13 cases exposed to NVP-based PMTCT and 2 of 6 controls. The start of HAART was within a year of NVP-based PMTCT exposure for 6 of these cases and none of the controls.
This study has provided a unique opportunity to explore correlates of virologic failure in what is today globally the largest national antiretroviral treatment program. The increased risk of virologic failure in patients on NVP, in women exposed to NVP through PMTCT and in patients who have interrupted their NNRTI-based regimen without tail protection are important findings in a context of almost universal reliance on NNRTI-based first-line regimens.
There is a growing body of evidence from observational studies that in routine settings, the use of NVP is associated with a greater risk of virologic failure,11-16 including in studies from Africa.17-19 This is in keeping with the findings from this study, as are the data from the South African patients participating in the 2NN trial.20 Together with the disproportionate toxicity burden resulting from d4T use,21 affordable access to alternative fixed-dose triple therapy combinations is an urgent priority for the region where the majority of countries are completely reliant on coformulated d4T, 3TC, and NVP.
Clinicians have long been concerned that abruptly stopping NNRTI-based therapy places patients at risk of a period of NNRTI monotherapy due to the relatively long half-life of NVP and EFV and it is common practice, including in the Western Cape, for patients to be given a week's supply of the 2 NRTIs in their regimen for planned interruptions to therapy. The very high risk of virologic failure in this study in patients with unplanned interruptions to therapy raises a concern that the failure to stagger ART cessation may be contributing to subsequent virologic failure. Previous cohort studies have reported similar findings.22,23 Many patients who score poorly on pharmacy refill measures that correlate with virologic outcomes17,24 may also be interrupting therapy in this manner.
Most studies looking at the efficacy of NVP-based HAART regimens post NVP-based PMTCT exposure, look at the probability of nonsuppression,25-27 whereas this study has demonstrated an association with confirmed virologic failure. The data in this study were, however, insufficient to examine the effect of the time between NVP exposure and the start of HAART on virologic failure.27 When considering the effects of NVP exposure as part of PMTCT, longer term follow-up of mothers who initially respond to NNRTI-based therapy is warranted, especially in settings where the routine viral load tests do not reliably detect suppression to below 50 copies per milliliter as is the case in South Africa. In 2008, South Africa adopted new PMTCT guidelines which are closely aligned with WHO recommendations but which do not provide for tail protection after the maternal NVP dose.28 Adding additional coformulated antiretroviral agents to the antenatal regimen or early postnatal period might reduce NVP resistance,29-30 and would not necessarily compromise PMTCT coverage, for as long as postnatal HAART to cover the breastfeeding period is not a reality.
The associations in this study between disease advancement and virologic failure are consistent with a number of studies in multiple settings.11,17,27,31 There is extensive pressure in South Africa to raise the threshold at which treatment should be started. As long as access to treatment is simultaneously expanded, this should further improve the clinical status of patients at the start of ART, as has already been observed in the initial scale-up.5
The strength of this study is that it was conducted in a context where HAART had already been available for 5 years with a clear virologic treatment failure guideline in place throughout. The unique monitoring system ensured that the study could encompass a large and geographically diverse routine service and is therefore fully reflective of the treatment program in which it was embedded. The only other case-control study of virologic failure in the region we are aware of was a smaller study in Botswana,24 in which cases were defined based on a single viral load above 1000 copies per milliliter. Weaknesses of our study included the lack of power to fully explore some of the associations such as the timing of PMTCT exposure relative to HAART, the absence of adherence measures besides treatment interruptions of a week or more, and the absence of resistance data, which if available, might have offered many additional insights.
Selection or recording bias are risks with any case-control study. To minimize selection bias, we selected cases and controls from study registers after ascertaining that they provided reliable coverage of clinic patients and we included almost all (94%) known cases in our study. We also minimized recording bias through matching cases and controls for clinic and for the calendar period when treatment was started. Each case and control triplet should therefore have experienced the same type and quality of care and their medical records should have been completed with the same level of diligence. Although it was impossible to mask patient outcome in the medical record during data extraction, the method of extracting data chronologically should have avoided extraction bias.
This study has yielded valuable insights into the correlates of virologic failure on HAART in South Africa. Although most cohort studies do not collect the same level of detail on exposures, the key independent associations described here are amenable to more detailed study in cohorts. It may also be feasible to alter laboratory procedures in the high throughput laboratories that perform the viral loads for the South African National Health Laboratory Service, to ensure that residual specimens in patients established on treatment with raised viral loads are systematically frozen and stored for future studies.
We are grateful to the staff and patients at all the study facilities. Meg Osler, and Catherine Orrell facilitated access to electronic records, I.D. Araoyinbo, John Ataguba, Sumaya Mall and Alfeous Rundare helped with data collection and Graeme Meintjes provided clinical coding advice.
1. Gilks CF, Crowley S, Ekpini R, et al. The WHO public-health approach to antiretroviral treatment against HIV in resource-limited settings. Lancet
2. World Health Organisation. Scaling up Antiretroviral Therapy in Resource-Limited Settings: Treatment Guidelines for a Public Health Approach
. Geneva, Switzerland: World Health Organisation; 2004:1-68.
3. National Department of Health. South African National Antiretroviral Treatment Guidelines
. 1st ed. Pretoria, South Africa: National Department of Health; 2004.
4. Coetzee D, Hildebrand K, Boulle A, et al. Outcomes after two years of providing antiretroviral treatment in Khayelitsha, South Africa. AIDS
5. Boulle A, Bock P, Osler M, et al. Antiretroviral therapy and early mortality in South Africa. Bull World Health Organ
6. McIlleron H, Meintjes G, Burman WJ, et al. Complications of antiretroviral therapy in patients with tuberculosis: drug interactions, toxicity, and immune reconstitution inflammatory syndrome. J Infect Dis
. 2007;196(Suppl 1):S63-S75.
7. Coetzee D, Hilderbrand K, Boulle A, et al. Effectiveness of the first district-wide programme for the prevention of mother-to-child transmission of HIV in South Africa. Bull World Health Organ
8. Braitstein P, Boulle A, Nash D, et al. Gender and the use of antiretroviral treatment in resource-constrained settings: findings from a multicenter collaboration. J Womens Health (Larchmt)
9. Brinkhof MW, Dabis F, Myer L, et al. Early loss of HIV-infected patients on potent antiretroviral therapy programmes in lower-income countries. Bull World Health Organ
10. Abdullah MF, Young T, Bitalo L, et al. Public health lessons from a pilot programme to reduce mother-to-child transmission of HIV-1 in Khayelitsha. S Afr Med J
11. Phillips AN, Pradier C, Lazzarin A, et al. Viral load outcome of non-nucleoside reverse transcriptase inhibitor regimens for 2203 mainly antiretroviral-experienced patients. AIDS
12. Cozzi-Lepri A, Phillips AN, d'Arminio Monforte A, et al. Virologic and immunologic response to regimens containing nevirapine or efavirenz in combination with 2 nucleoside analogues in the Italian Cohort Naive Antiretrovirals (I.Co.N.A.) study. J Infect Dis
13. Keiser P, Nassar N, White C, et al. Comparison of nevirapine- and efavirenz-containing antiretroviral regimens in antiretroviral-naive patients: a cohort study. HIV Clin Trials
14. Matthews GV, Sabin CA, Mandalia S, et al. Virological suppression at 6 months is related to choice of initial regimen in anti retroviral-naive patients: a cohort study. AIDS
15. Phillips AN, Ledergerber B, Horban A, et al. Rate of viral rebound according to specific drugs in the regimen in 2120 patients with HIV suppression. AIDS
16. Smith CJ, Phillips AN, Hill T, et al. The rate of viral rebound after attainment of an HIV load <50 copies/mL according to specific antiretroviral drugs in use: results from a multicenter cohort study. J Infect Dis
17. Nachega JB, Hislop M, Dowdy DW, et al. Adherence to nonnucleoside reverse transcriptase inhibitor-based HIV therapy and virologic outcomes. Ann Intern Med
18. Boulle A, Van Cutsem G, Cohen K, et al. Outcomes of nevirapine- and efavirenz-based antiretroviral therapy when coadministered with rifampicin-based antitubercular therapy. JAMA
19. de Beaudrap P, Etard JF, Gueye FN, et al. Long-term efficacy and tolerance of efavirenz- and nevirapine-containing regimens in adult HIV type 1 Senegalese patients. Aids Res Hum Retroviruses
20. van Leth F, Phanuphak P, Ruxrungtham K, et al. Comparison of first-line antiretroviral therapy with regimens including nevirapine, efavirenz, or both drugs, plus stavudine and lamivudine: a randomised open-label trial, the 2NN Study. Lancet
21. Boulle A, Orrel C, Kaplan R, et al. Substitutions due to antiretroviral toxicity or contraindication in the first 3 years of antiretroviral therapy in a large South African cohort. Antivir Ther
22. Fox Z, Phillips A, Cohen C, et al. Viral resuppression and detection of drug resistance following interruption of a suppressive non-nucleoside reverse transcriptase inhibitor-based regimen. AIDS
23. Bansi LK, Benzie AA, Phillips AN, et al. Are previous treatment interruptions associated with higher viral rebound rates in patients with viral suppression? AIDS
24. Bisson GP, Rowh A, Weinstein R, et al. Antiretroviral failure despite high levels of adherence: Discordant adherence-response relationship in Botswana. J Acquir Immune Defic Syndr
25. Coovadia A, Hunt G, Abrams EJ, et al. Persistent minority K103N mutations among women exposed to single-dose nevirapine and virologic response to nonnucleoside reverse-transcriptase Inhibitor-Based Therapy. Clin Infect Dis
26. Jourdain G, Ngo-Giang-Huong N, Le Coeur S, et al. Intrapartum exposure to nevirapine and subsequent maternal responses to nevirapine-based antiretroviral therapy. N Engl J Med
27. Lockman S, Shapiro RL, Smeaton LM, et al. Response to antiretroviral therapy after a single, peripartum dose of nevirapine. N Engl J Med
29. McIntyre J, Martinson N, Gray GE, et al. Addition of short course Combivir to single dose Viramune for the prevention of mother to child transmission of HIV-1 can significantly decrease the subsequent development of maternal and paediatric NNRTI-resistant virus [TuFo0204]. Presented at: 3rd International AIDS Society Conference on HIV Pathogenesis and Treatment; July 24-27, 2005; Rio de Janeiro, Brazil.
30. Chi BH, Sinkala M, Mbewe F, et al. Single-dose tenofovir and emtricitabine for reduction of viral resistance to non-nucleoside reverse transcriptase inhibitor drugs in women given intrapartum nevirapine for perinatal HIV prevention: an open-label randomised trial. Lancet
31. Deeks SG. Determinants of virological response to antiretroviral therapy: Implications for long-term strategies. Clin Infect Dis
Keywords:© 2010 Lippincott Williams & Wilkins, Inc.
case-control study; HAART; nevirapine; PMTCT; South Africa; Southern Africa; virological failure